Strength in Numbers

(1)

Strength in Numbers

The Probability of Forming a Multi-Partner Alliance

By

Vincent Alexander Koster

University of Groningen Faculty of Economics and Business

MSc BA Strategic Innovation Management 19-06-2015

Het Sticht 74 9405NR Assen

Supervisor: d.r. I. Estrada Vaquero Co-assessor: d.r. K.J. McCarthy

v.a.koster@student.rug.nl

(2)

(3)

Abstract

This paper investigates how a firm’s exogenous and endogenous factors relate to the likelihood of forming a Multi-Partner Alliance. The research is conducted with a sample of 198 firms, of which 96 engage in dyadic alliances and 102 in multi-partner alliances. For analysis, Logistic regression is used. Emprical findings showed that both technological uncertainty and firm age have significant negative relationships to the probability of forming a multi-partner alliance. There is support that firm size has a positive relationship to the probability of forming a multi-partner alliance. No support was found for a relationship between industry competition and the probability of forming a multi-partner relationship. Moreover, there is support that having a marketing agreement results in lower probabilities of having a multi-partner alliance. This research provides insight into how dyadic and multi-partner alliances differ from each other and this paper might form the basis for further research into the field of multi-partner alliances.

(4)

Hypothesis Development ... 10 Industry competition. ... 10 Technological Uncertainty. ... 11 Firm age. ... 13 Firm Size. ... 14 Methodology ... 15 Sample ... 15 Dependent Variable ... 16 Independent Variables ... 16 Industry competition. ... 16 Technological Uncertainty. ... 16 Firm size. ... 17 Firm age. ... 17 Control Variables ... 17 Techniques ... 18 Results ... 19

Discussion & Conclusion... 26

(5)

Introduction

Multi-Partner (MP) alliances are becoming more popular and more important in today’s economic environment. An example is the study of Makino & Beamish (1999), they discovered that 55% of their sample of joint ventures was in fact a multi-partner one. Moreover, the use of multi-partner alliances is a viable strategy with its own advantages (Beamish & Kachra, 2004). Studies like these indicate that more specific studies into multi-partner alliances are highly relevant. Strengthening the call for more research on multi-partner alliances is the fact that to this day, the number of research in multi-partner alliances is overshadowed by research in dyadic (two partner) alliances or alliances in general. For example, it is already researched that firms form alliances for multiple reasons: there are market entry related, product related, resource related and skill related motives for a firm to enter an alliance (Varadarajan & Cunningham, 1995). Previous research also already shows that not every firm is willing to enter an alliance to the same extent. The probability of forming an alliance differs among firms. It is argued this is constituted by firm-endogenous and firm-exogenous factors (Sakakibara, 2002; Li, 2013; Stuart, 1998; Hitt, Dacin, Levitas, Arregle & Borza, 2000; Eisenhard & Schoonhoven, 1996).

However, most of these previous studies only researched the dichotomous outcome of having an alliance or not, or they are solely focused on dyadic alliances. It did not clarify whether there is a difference in the probability of forming dyadic alliances or the probability of forming multi-partner alliances. In other words: there is no clear distinction made between dyadic alliances and multi-partner alliances within previous research. The question remains: what alliance size is then preferred in certain conditions? One could wrongfully imply from the previous studies that the factors constituting the likelihood of forming an alliance in general have no consequences for the type of alliance (dyadic or multi-partner) and for example, that there is no difference between dyadic or the larger multi-partner alliances.

(6)

Research that did focus on the number of partners, and thus to a lesser extend to multi-partner alliances, did not look at formation factors but solely at the performance relationship. Hu & Chen (1996) found a positive relationship between number of partners and performance. The same was argued by Park & Russo (1996), when they found the same relationship, but used a different measure. To date, empirical research on whether the factors relating to the propensity of forming an alliance in general also relate to the probability of forming a MP alliance versus a dyadic alliance is scarce. There exists a research gap between the possible determinants of entering or forming an alliance and the possible determinants of forming a MP or dyadic alliance specifically. Moreover, while the above stated relation remains unclear, the relation between number of partners and performance has already been researched. However, the relation between these two fields remains rather unclear. So there exists a second research gap, namely: the gap between forming the MP alliance and alliance performance. Is knowing the probability of forming a MP alliance a link between the field of alliance formation and the field of alliance performance? It is clear what the relationship between number of partners and performance is and thus to a lesser extend the difference between dyadic and MP alliances, but it is still not clear what the specific relationship is between formation factors and the likelihood of forming a MP alliance.

This paper will contribute to existent research by closing the first above stated research gap and to become the basis for the second research gap. It will focus on dyadic and multi-partner alliances. It will use two exogenous and two endogenous variables relating to the propensity of entering or forming an alliance in general and will link them to the likelihood of choosing and forming a MP alliance, to be more specific: this paper will investigate whether industry and firm characteristics relate to the formation of a MP alliance.

Using a Resource-Based View (RBV), it will build on the results and theories of previous research and tries to extent it to MP alliances. It will have a focal firm perspective and level of analysis. The model will be restricted to two endogenous and two exogenous factors since, in multi-partner alliances implementing all endogenous factors and firm perspectives into the model becomes too complex due to the number of firms and all their different needs, strategies and characteristics. The consequential research question can therefore be stated as:

How do a firm’s exogenous and endogenous factors relate to the likelihood of forming a Multi-Partner Alliance?

(7)

(8)

Literature Review

This part will discuss the most important aspects of the paper based on previous research and important theories. This will include the discussion of alliances in general, theoretical perspective, the factors used in this paper, and hypotheses.

Theoretical Perspective

Inside the field of alliances there exist a number of major theoretical perspectives (Williamson, 1985; Barney, 1991; Dyer & Singh, 1998). The one that this paper will use is the resource-based view (RBV). The RBV sees the firm as a “bundle of resources” (Wernerfelt, 1984) meaning that a firm is dependent on the resources it possesses. Without resources the firm is nothing. It is argued from a RBV perspective that, assuming firms are heterogeneous and resources are immobile, firms can achieve sustained competitive advantages. Firms can create this superior value by having unique resources that adhere to the VRIN framework, Valuable, Rare, imperfectly immmitable, and non-substitutionable (Barney, 1991). Since some firms do not possess these resources they may try to access it in three different ways. A firm can create the resources on its own, it can buy them on the market, or it can choose a hybrid form of cooperation (van den Bosch, Elfferich, 1993). That last one is the one this paper will focus on, since it captures the phenomenon of alliances.

In the coming sections the discussed perspective will be linked to the use of Alliances.

Motivations for Alliances

One can define strategic alliances as an intra-firm bond to cooperate with another entity on a voluntary basis (Deeds & Rothaermel, 2003; Gulati, 1995). With regard to the RBV, one can argue that an alliance can give vast amount of resources that can be acquired from a partner. Hamel (1991) argues that alliances are a well-suited opportunity for one partner to acquire knowledge from the other. The alliance gives an easy access to knowledge that the firm lacks or that would otherwise be secret, since some information is stored in a tacit form (Polanyi, 1973). Acquisition processes will be faster in alliances, especially for this tacit knowledge the alliances is a great transfer vehicle (Kogut, 1988). Alliances are thus primarily suited when knowledge has been present in a short time-span, too short to develop it in one firm (Madhok, 1997). This study will link the RBV with the four factors researched in the hypothesis development section.

Alliance Formation

(9)

previous research has shown that many factors influence the formation rate both positively and negatively. One can define two main classes for factors that influence alliance formation, these are: exogenous factors and endogenous factors (Gulati, 1995).

Exogenous factors. Exogenous factors for alliance formation are factors which a firm cannot influence itself. To be more specific, firm exogenous factors like industry competition was found to have a significant impact on the probability of forming an alliance. A high degree of competition, an emergent market stage and in the industry means a higher rate of alliance formation (Eisenhardt & Schoonhoven, 1996), while in other studies competition has a negative relationship (Sakakibara, 2002). Appropriability conditions (Sakakibara, 2002) and market uncertainty (Li, 2013) were found to relate to the likelihood of entering an alliance as well. Li (2013) found a positive relationship between uncertainty and probability of alliance formation. Appropiability conditions are argued to influence the rate of formation in a negative manner (Sakakibara, 2002). Hit et all (2000) argued that the focus of the alliance will be constituted by the stage the market is in. Firms from emerging markets will have a different focus then firms from developed markets.

Endogenous factors. Endogenous factors for alliance formation are factors which a firm can influence itself. There is also support that firm endogenous factors influence this likelihood of participating in alliances. For example, Stuart (1998) found that a firm’s competitive positioning was a significant factor. Eisenhard & Schoonhoven (1996) found that experience and composition of top management teams had an influence on the rate of formation. More specifically, firms with teams that were large experienced and had many connections often form alliances more easily. R&D capabilities, sales, budget, previous experience in alliances and firm age were also found to be significant factors that influence the propensity of entering into an alliance (Sakakibara, 2002). Furthermore, there is support that a firm’s top management team’s social capital is positively related to alliance formation, the same holds for a firm’s technological capability (Li, 2013).

(10)

partner takes marginally more investments (García-Canal, Valdés-Llaneza & Ariño, 2003). However, when agreements can be met, MP alliances might perform better (Park & Russo, 1996).

Based on these theories this study will further focus on two exogenous and two endogenous factors and their relationship to multi-partner alliance formation, namely industry competition, technological uncertainty, firm age and firm size. One can argue that these four factors are often only researched in dyadic settings (Sakakibara, 2002; Eisenhard & Schoonhoven, 1996), or in Li’s (2013) case, in a multi-partner setting, but solely focusing on new ventures. However, as stated earlier, multi-partner alliances are different from dyadic alliances (García-Canal, Valdés-Llaneza & Ariño, 2003). One cannot and should not assume that the same results from research in dyadic alliance apply in multi-partner alliances. The choice for these particular four factors is that they all four have had extensive research in previous fields other than multi-partner alliances. There remains a lack of understanding of these factors in multi-partner alliances. Moreover, the combination of exogenous and endogenous factors may prevent assumptions that choice for alliance is solely based on endogenous or exogenous factors. These four factors are not exhaustive, but due to limits on time and resources not more factors are being researched.

Hypothesis Development

Industry competition. Competition can be defined as: ‘actors that produce and market similar products’ (Bengtsson & Kock, 2000). Industry Competition for the firm is therefore: all the actors that produce and sell the same product as the firm, within that firm’s particular industry.

Eisenhard et al (1996) found that the more competitive the industry the more likely a firm will form an alliance. Li (2013) however, found no significant relationship between industry competition and the formation of a MP alliance, while hypothesizing an inverted u shape relation. Since the study of Li (2013) only focused on new ventures, one might assume outcomes can be different for already established firms. Therefore this paper still finds it relevant to hypothesize and research the relationship between industry competition and the formation of MP alliances. Both Li (2013) and Eisenhard et al (1996) base their hypotheses on the fact that competitive environments cause a firm’s resources to be lessened and generate pressure on the strategic position of the firm. Profits are smaller in more competitive industries. As stated earlier, alliances can offer cost reductions and relief some stress on the profit margins, improving the position of a firm. Moreover, an alliance can give some advantages as well. Some firm’s might combine resources and arrive at better strategies and products, or use the alliance as a tool for visibility and legitimacy (Hitt et al, 2000; Li, 2013), giving an edge over the competitors.

(11)

environment is rather risky, firms need to have an eye on more actors within the industry (Porter, 1986). Splitting this risk with alliance partners will make the risk relatively smaller for a firm. Thus increases the chance of survival (Lavie, Lechner & Singh, 2007). Multi-partner alliances can of course split the risks with multiple participants, while a dyadic alliance can only split the risk under two participants, making the MP alliance a better option.

When putting the previous research in perspective. From a RBV: firms might try to strengthen their market position by improving their capabilities and ease market pressure in high competitive environments. A firm’s capabilities might be improved through forming an alliance. A multi-partner alliance may be even stronger, since it might form a powerful block within the industry, by pooling resources from multiple firms each with its own specialization and by splitting risks and costs (Sakakibara, 2002; Lavie et al, 2007). Thus, it might be likely that alliances in more competitive industries will tend to have more partners and thus becomes a multiple partner alliance. Alliances in general are more likely to form in more competitive industries (Eisenhard et al, 1996). Therefore, this paper argues that industry competition has a positive relationship with the possibility of forming a MP alliance and hypothesizes:

H1a: the more competitive a firm’s industry, the more likely firms are to enter a multiple partner alliance.

On the other hand, Studies of García-Canal, et al (2003), and Hagedoorn & Schakenraad, (1994) show that larger alliances are more complex and vulnerable, a MP alliance is thus more complex than a dyadic one. Moreover, Sakakibara (2002) who explicitly researched alliance formation found that firms in oligopolistic industries tend to form more alliances, implying a negative relationship.

Following simple logical reasoning again using the RBV, one might argue that: competitive industries are already complex, since firms have to keep a close eye on many other firms within the industry (Porter, 1986). The likelihood of forming or joining an alliance in general is already lower in competitive industries (Sakakibara, 2002). One does not want to unnecessary increase complexity by adding to many partners to a possible alliance (Hagedoorn & Schakenraad, 1994). Firms might want to protect their market share in an already competitive market and might become more cautious when considering multi-partner alliances, since they might lose competencies to the partners (Brouthers, Brouthers & Wilkinson, 1995). Therefore the following is hypothesized:

H1b: the more competitive a firm’s industry, the less likely firms are to enter a multiple partner alliance.

(12)

uncertainty and technological uncertainty. Such an all-encompassing concept is not suitable (Song & Montoya-Weiss, 2001). Therefore, this study defines uncertainty as technological uncertainty, this is most relevant for this study and more specific. Technological uncertainty can be defined as: ‘uncertainty stemming from technological change, product obsolescence and complexity’ (Quinn & Hilmer, 1994).

In times of environmental uncertainty, having a more flexible organizational structure is preferred, since one can make shifts quickly compared to more rigid organization structures. An organic structure can include alliances (Li, 2013). Thus logically, when uncertainty is high, the rate of formation of alliances is high. This is also supported by Li’s study (2013). An important thing to note is that Li (2013) did not use the same sort of uncertainty; Li’s study was mainly focused on market uncertainty, measured by stock price, while this study will use technological uncertainty. The same logic will still be relevant, since it is stated in Li’s (2013) paper that market uncertainty itself is also more stemming from technological changes and Li explicitly focused on multi-partner alliances. Moreover, when knowledge lifespan is short, alliances are primarily suited to develop new knowledge (Madhok, 1997). Alliances may also enhance technological certainty (Lavie, et al 2007). Contrasting to the view of Li (2013), a firm can lose its core competencies to the alliance’s partners within the alliance, something which is not preferred in technological uncertainty (Brouthers, Brouthers & Wilkinson, 1995). Next to that, strategic convergence with uncertainties is harder, implying firms will be less likely to form an alliance when uncertainties are high (Doz, 1987) something which is even harder when there are multiple players within the alliance. Thus, there also exists an incentive to not pursue alliances let alone multi-partner alliances. This paper will use the same logic as previously used for the hypotheses between industry competition and participation rate which is based on previous research and theory; that firms tend to split risks and costs within alliances (Lavie, et al 2007).

Logically reasoning from previous research and from a RBV: it might be likely that formed alliances in more uncertain industries have more partners and thus will be a MP alliance, because within larger alliances risks and costs can be spread more than in dyadic alliances. (Lavie, et al 2007). Alliances in general are good vehicle for development when the knowledge is only there for a short time span (Madhok, 1997). One might extend this argument, that MP alliances have the ability to quickly pool multiple resources from multiple firms and thus can offer new technologies more rapidly. This is highly needed in a market where technology becomes obsolete quickly. Therefore it is hypothesized:

H2a: the more technological uncertain a firm’s industry, the more likely firms are to enter a multiple partner alliance.

(13)

vulnerable and complex (García-Canal, et al, 2003; Hagedoorn & Schakenraad, 1994), and thus to reduce the risk of losing competencies alliances tend to be dyadic, especially in technological uncertain industries.

Therefore it is hypothesized:

H2b: the more technological uncertain a firm’s industry, the less likely firms are to enter a multiple partner alliance.

Li (2013) found a u shaped relationship in his research when looked at the probability of forming an alliance. When converting these results from probability to MP alliances using the theory that alliances tend to split costs (Levie et al, 2007), one can reason the following: It can be that firms try to enter MP alliances in extreme values of technological uncertainty. For example, extreme low industry technological uncertainty allows for larger and more complex alliances due to the less risky environment. Extreme high technological uncertainty makes large alliance a necessity to ensure advantages and a strong strategic position. Middle values host no ideal environment for MP alliances. However, as discussed later, the proxy used for this variable does not allow for curvilinear terms, therefore this research will not focus on this relationship. It is still important to note that this relationship might exist, since otherwise one can come to biased conclusions.

Firm age. Firm age is used as a control variable in previous studies in the field of alliances (Sakakibara, 2002; Li, 2013). However, in other fields, like the one of internationalization, it is argued to be an important factor. Older firms often have many routines in place and possess already quite an extensive knowledge base. Acquiring new knowledge or other routines when internationalizing might be restricted by these former routines and knowledge (Autio, Sapienza & Almeida, 2000). The study of Sorensen and Stuart (2000) however, both a negative relation between firm age and firm innovativeness and a positive one is found. The negative relation is similar as the one discussed by Autio et al (2000). The positive relation is based on the fact that the efficiency of older firms increases due to experience. Of course, this is in the field of internationalization and firm innovativeness. However, one might argue that this still holds true when working with alliance partners. A study that used firm age as a control variable in the field of network formation is the one from Sakakibara (2002). In this study it is found that firm age has a positive influence on the formation rate. However, one can argue that the development of their hypothesis is not very extensive (e.g. firm age as a barrier is neglected), therefore this study will not blindly follow Sakakibara (2002) and will try to clarify the effect of firm age on the formation rate. This means that both a positive as a negative relationship is hypothesized.

(14)

the partner´s knowledge. It is locked in its own mature routines and has more difficulty to adapt to the partner’s routines, acquire these routines or resources bounded to these routines Autio et al (2000). Therefore it is not likely that an older firm will participate in a multi-partner alliance. Since the addition of multiple partners and having to deal with multiple new routines might form a barrier (García-Canal, et al, 2003; Hagedoorn & Schakenraad, 1994; Autio et al, 2000), a multi-partner alliance is therefore not a benefit. It is hypothesized that:

H3a: the older a firm the less likely a firm is to enter a multiple partner alliance.

However, on the other hand, the higher efficiency of general routines and acquisition routines due to experience of older firms (Sorensen & Stuart, 2000) might enable these older firms to actually deal with the increased complexity of multi-partner alliances (García-Canal, et al, 2003; Hagedoorn & Schakenraad, 1994). Therefore this paper will also hypothesize a positive relationship:

H3b: the older a firm the more likely a firm is to enter a multiple partner alliance.

Firm Size. Splitting costs between participants (Levie et al, 2007) can either be done by a few wealthy participants or many less wealthy participants. In both scenarios the ratio of alliance costs to revenue per firm may be the same. It is interesting to see how firm size relates to the formation of a MP alliance, since no consensus exists in previous research of what the effect of firm size is. Kleinknecht & Reijnen, (1992) argue that firm size has little effect on the decision to cooperate, while Shan (1990) found a significant negative relationship implying that larger firms tend to cooperate less. However, studies of Hagendoorn et al (1994) and Gulati (1999) found opposite results and find that larger firms tend to have a higher propensity in alliance formation and have the capabilities to run complex alliances. While small firms do not possess these capabilities, but what does this imply for the probability of multi-partner alliances?

Logically reasoning from the above stated research and from a RBV: alliances can combine and create the same amount of resources by many small firms bringing many small chunks of resources together or a small number of large firms bringing large chunks of resources together (Madhok, 1997). So for small firms to still be competitive, it must cooperate with many others. Therefore it is hypothesized:

H4a: the smaller a firm, the more likely it will enter a multiple partner alliance..

However, large firms have the resources and capabilities to operate in large alliances (Hagendoorn et al, 1994; Gulati, 1999), implying their alliance can become larger with more partners involved, than those of smaller firms. Therefore this paper hypothesizes that:

(15)

Methodology

Sample

Sampling was done in a two-step process. First, data was collected on firms that engage in multi-partner alliances. Secondly, the data was supplemented with data on firm that engage in dyadic alliances. This sampling method is in line with research done by Li (2013). The sample will be mainly collected from the Securities Data Company (SDC) Platinum database by Thomson Reuters. This database offers information about most of the world’s existing alliances and mergers. It hosts a rather complete set of variables which can be used for analysis (Thomsonreuters.com, 2015). Variables from the SDC database that are particularly interesting for this research are: the number of partners, SIC codes of the firms, SIC codes of the alliances, agreements, and year of alliance announcement. Additional data on industry competition, firm age and firm size will be deducted from the Orbis database, by Bureau van Dijk. This database hosts many financial and non-financial (e.g. number of employees) variables for nearly 150 million companies worldwide (Bvdinfo.com, 2015). For information about employees, firm age and financial numbers that the Orbis database cannot provide, firms websites and 10k filings of the respective firms are used. As previously stated the sample will compass multiple industries (128 different industries, high-tech or not) within the US. This research handles US Alliances, since most data is available for this country and sector.

For the first step:

All alliances announced in the year 2007 till 2012 will be taken in the preliminary sample. Then this sample will be filtered for alliances within the US and that solely has US firms in it, eliminating any cultural diversity. Moreover this study will solely encompass public firms and its subsidiaries, since private firms are not obliged to disclose the information needed. All multi-partner alliance firms that are left will be used. This comes down to a preliminary size of 140 firms operating within MP alliances. These 140 cases will be supplemented with other data. Due to some missing data on the variables of employees and year of foundation, the number of cases declined to 96. These 96 cases were all public firms. For subsidiary alliances the research took the data of the parent, this is due to the fact that for subsidiaries many accounts are consolidated making financial info hard to retrieve for one specific subsidiary. Moreover, it is highly likely that these subsidiaries receive much support from the parent firm, therefore focusing on the parent firm is still relevant. The data will be complemented with firms that do not engage in multi-partner alliances

For the second step:

(16)

alliance that also operates within the US. Furthermore, these firms are all public. After this filter, 2141 firms remain. This sample is too large to supplement with other data, since it will take too much time and resources. This study therefore uses a random sample of 150 out of these 2141 firms, generated by a random algorithm in the statistical software package STATA. The second part finally consists out of 102 firms with dyadic alliances.

Altogether, the total sample has a size of 198 firms. Moreover, this sample size is adequate since Hosmer & Lemeshow (1989) state that for logistic regression a minimum of 10 events per predictor is necessary to come at a sense-making model. The model used in this paper will contain 8 independent variables, thus a minimum of 80 observations required.

Dependent Variable

The dependent variable (DV) in this study will be whether an alliance is a multi-partner alliance or a dyadic one, thus the outcome will be dichotomous. The value of 1 is given to cases in which the alliance takes the form of a multi-partner alliance, and 0 when the alliance is dyadic. This is according to the measures used by Li (2013).

Independent Variables

The independent variables in this study are industry competition, Technological uncertainty, firm age and firm size. All independent variables are measured in one year prior to the alliance announcement. This due to the fact that the decision to ally is probably made on basis of numbers in the year before the actual announcement.

Industry competition. Industry competition will be measured in line with previous research. To be more specific, the number of firms within an industry will be the proxy for competition; this is in line with Sakakibara (2002). Of course there are many other measures, like the reverse concentration ratio (Li, 2013) and concentration per market segment (Eisenhardt & Schoonhoven, 1996). However there exist heavy data constraints, and thus the more simple number of competitors within an industry is used. Specific Industries are defined by using the SIC index of industries. The SIC codes categorizes industries by using numerical codes. By using these codes in combination with the Orbis database, specific number of firms within the precise industry can be recovered.

Technological Uncertainty. Technological uncertainty will be measured by using a classification developed by the AEA1 (

Georgia State University, 2000)

. This classification identifies a number of high tech industries by using US SIC codes2. High-tech is often related to high technological uncertainty (Bauer, Lang & Schneider, 2012). Moreover, this paper assumes that the high technology industries will have higher R&D intensity. Reason for this assumption lies in the fact

1_{American Electronics Association} 2

(17)

that the OECD made a similar list for OECD countries, their methodology states that the R&D intensity is the sole factor for establishing their classification (Oecd-ilibrary.org, 2015). Since the classes are partly overlapping between the two classifications, one might assume that R&D intensity is an important antecedent in the list of AEA (2000) as well. R&D intensity is argued to be associated with faster technology cycles and higher levels of uncertainty (Albors-Garrigos, Zabaleta & Ganzarain, 2010). Therefore we use the AEA (2000) classification as a measure for average technological uncertainty. Of course, this classification is less specific then when using other known measurements, like actually measuring the R&D intensity for every firm. However, this information is often not public or not separated from other costs. When this research would have chosen such a measure, the sample would be drastically cut down in size.

Firm size. Firm size will be measured by using the total number of employees within a firm. Firms size is also often measured by total assets (Eisenhard & Schoonhoven, 1996).

Firm age. Firm age will be measured as the number of years a company or business unit is in function since its foundation.

Control Variables

This research controls for earlier experience with alliances (Eisenhardt & Schoonhoven, 1996; Gulati, 1999), whether there is a R&D, marketing, manufacturing agreement or none at all, the relatedness of the firm’s primary industry to the alliance’s primary industry (Doz & Hamel, 1998; Cohen & Levinthal, 1990; Jiang, Tao & Santoro, 2010), and the fact whether in the same year a firm had both forms (dyadic or multi-partner) of alliances.

One might argue that when a firm has more experience with alliances in general, the management of a multi-partner alliance might become easier and thus the probability of forming one might increase. Earlier alliance experience will measured by the number of previous alliances held by a firm up till 5 years prior to the alliance in the sample (Eisenhardt & Schoonhoven, 1996).

(18)

Firms that share their primary industry with the alliance might be more likely to form an alliance since greater absorptive capacity is achieved when there is industrial overlap (Cohen & Levinthal, 1990), and thus might also be prone to multi-partner alliances. On the other hand diversity and little overlap increases complexity, constraining firms to form alliances (Doz & Hamel, 1998), let alone multi-partner alliances. The relatedness will be captured by a dummy variable that takes the value of 1 when the first two digits of the firm’s industry SIC code is similar to the alliance’s industry SIC code and 0 if otherwise, used before by Doukas & Kan (2004).

Lastly, this study controls for the fact whether firms that engage in MP alliances also engage in dyadic alliances in the same year and vice versa. It could be that the decision to form a MP alliance might have something to do with the formation of a dyadic alliance earlier in the same year and vice versa. In order to make sure outcomes are not distorted by this the research controls for it.

Techniques

The basis for the results of this study is logistic regression, since the dependent variable is

dichotomous (Peng, Lee & Ingersoll, 2002). In contrast to linear regression models, logistic regression models do not need the assumption of normality and homogeneity of variance (Meshbane & Morris, 1996). However, there are some assumptions underlying logistic regression that do need to be checked3. This paper will discuss the results for the assumption in the result section, and if necessary transform the variables into more suitable ones. After that, the paper will continue establishing the logistical regression model. As described by Field (2005), this research will first run a hierarchical analysis with different models. In the end the model with the best fit will be considered as the final model.

3_{The assumptions are: Assumption of linearity, Independence of errors, Assumption of mutual exclusive}

(19)

Results

First of all, descriptive statistics are used to describe the data in general terms. In table 1, one can see an overview of the statistics. From these statistics it becomes clear that there are no missing cases, since every variable has exactly 198 cases. An remarkable fact is that the minimum and the maximum values for all continues variables: Competition (minimum: 93, maximum: 7139999), Firm size (minimum: 2, maximum: 1800000), Firm age (minimum:7, maximum: 187), and Previous Experience (minimum: 0, maximum: 138) are rather far apart, considering their mean and standard deviation. This might give substantial outliers which could bias the research. This paper therefore looks deeper into the distribution of these variables.

Table 1: Descriptive Statistics

N Minimum Maximum Mean Std. Deviation Multi-Partner 198 0 1 0,48 0,501 Industry competition 198 93 713999 24891,43 72180,45 Technological uncertainty 198 0 1 0,26 0,441 Firm Age 198 7 187 50,16 41,315 Firm Size 198 2 1800000 38758,79 139906,5 Previous Experience 198 0 1 0,14 0,344 Multiple Forms 198 0 1 0,14 0,344 R&D Agreement 198 0 1 0.10 0.295 Marketing Agreement 198 0 1 0.10 0.295 Manufacturing Agreement 198 0 1 0.07 0.257 Industry Relatedness 198 0 1 0,36 0,482 Valid N (listwise) 198

(20)

not be a solution, since this will reduce sample-size, which will costs even more statistical power. To oversee any changes due to the transformation, to add robustness regarding the results, and increase validness this research will both run models with the raw variables and the transformed variables. For most assumptions this paper does not have to test for4. There are however two assumptions that are more complex and thus need to be tested, these are: the assumption of linearity and the assumption of no multicollinearity. This research will test these assumptions using a method described by Field (2005).

The first assumption that is tested is the assumption of linearity of the data. According to Field (2005), to analyze this linearity, all the continuous variables should be put in a regression analysis including the variables’ logarithms as an interaction effect. It is important to not have a hierarchical method while regressing (Field, 2005). The regression results can be found in Figure 2

Table 2: Regression result for Linearity Analysisab Predictors Significance Industry competition 0,793 Firm age 0,454 Firm size 0,692 Pre_exp_total 0,924 Industry competition * LN_Industry competition 0,761

Firm age * LN_firm age 0,417

LN_Firm size * Firm size 0,757

LN_Previous_Exp*Previous_Exp 0,958

Constant 0,826

a: Dependent Variable: Multi-partner (1= yes; 0= no)

b: For lay-out purposes, some statistics are left out and are available at request.

As described by Field (2005), one should look only to the interactions terms when determining whether the assumption of linearity holds. When these interactions are significant (P< 0.1), then the main effect does not hold the assumption of linearity. None of the interaction terms have significant values. Both the interactions of Industry competition, Firm age, Firm size in employees, and the Total previous experience all have significance values above 0.1 indicating no significance. One can assume that for this dataset that the assumption of linearity holds. Please note that applying the same method on the already log transformed variables seems inappropriate, since taking the logarithm of a

4_{The assumptions are: Independence of errors, Assumption of mutual exclusive categories and Assumption of a}

(21)

logarithm will distort the data, an may be considered transformation overkill. Therefore in order to come to conclusions about linearity of the log transformed model, this study will compare the model fit of the transformed variable model to the fit of the original scale model later on in the regression analyses (Restore.ac.uk, 2015).

The second assumption that needs to be tested is the assumption of no multicollinearity. The correlation matrix (See Appendix F) shows some significant values, however all correlation coefficients between the two independent variables are rather low. To make sure the significant values will not form a problem, this paper uses the method of Field (2005). According to Field (2005), collinearity statistics can only be obtained through running a linear regression using the same dependent and independent variables. This regression offers statistics such as the VIF and tolerance, which can be used to determine whether there is multicollinearity. The results for the collinearity diagnostics can be seen in Table 3a and 3b. When looking at the results, one can see that the tolerance levels of all variables are rather high. Most of the variables show tolerance levels around the 0.9 level. According to Menard (1995), tolerance values of less than 0.1 may indicate an issue of multicollinearity. Since none of the variables in this research exposes such a value, one may assume that there is no multicollinearity. Moreover, VIF values higher than 10 may also indicate multicollinearity (Myers, 1990). However, all values of VIF are around the value of 1, implicating that there is no multicollinearity.

Table 3a: Collinearity Diagnostics Original Scale Variablesab

Predictors Tolerance VIF

Main Effect Industry competition ,930 1,075 Technological Uncertainty ,859 1,164 Firm age ,816 1,225 Firm size ,929 1,076 Control Previous Experience ,903 1,108 Multiple Forms ,899 1,112 R&D Agreement ,942 1,062 Marketing Agreement ,946 1,057 Manufacturing Agreement ,956 1,047 Alliance industry relatedness ,921 1,086

Table 3b: Collinearity Diagnostic Transformed Variablesab

Predictors Tolerance VIF

Main Effect Industry competition ,849 1,178 Technological Uncertainty ,871 1,148 Firm age ,722 1,385 Firm size ,504 1,985 Control Previous Experience ,570 1,755 Multiple Forms ,855 1,169 R&D Agreement ,922 1,084 Marketing Agreement ,926 1,080 Manufacturing Agreement ,899 1,113 Alliance industry relatedness ,932 1,073

(22)

Since the assumptions are all met and possible outlier problems are decreased as much as possible, this research can continue with building a regression model. As stated earlier, this paper will both run models with raw variables and with transformed variables to increase robustness.

To remain oversight when discussing different models, the model with the raw variables will be designated as the ‘original scale model’, while the transformed model will be designated as the ‘log transformed model’. Both the original scale as well as the log transformed model will have a hierarchical regression strategy, beginning with all control variables and adding one main effect per regression. In table 4 all the regression results are showed for the original scale model, while in table 5 all regression results are showed for the log transformed model. Included in the regression tables are the respective model summaries. This research will first discuss the original scale model. It will start with the control variables, and will then discuss the hypotheses for the main effects, after which the model fit will be discussed. After discussing the original scale model, the log transformed model will be discussed in the same systematic way.

The first thing that draws attention is the fact that one of the control variables (marketing agreements) is already having a highly significant influence (P<0.05) in model 1, it remains this significant throughout the other models. The other control variables do not show any significance.

Another interesting observation is the fact that although not significant, the industry competition variable has a very stable coefficient (,000) throughout all models. It was previously stated that outliers could affect the model. Since there are some potential outliers and given the fact that the original scale model is not transformed, a logical explanation could be that there might be outlier influence that could create such an odd observation. Apart from the coefficient, the variable does not show any significance. There is no support for either hypothesis 1a or 1b.

The original scale model shows that having high technological uncertainty within an industry is a significant factor (P<0.1) in determining whether a firm pursues a dyadic or a multi-partner alliance. However, one should note that this significance is rather low, but stable throughout the different models. These results supports hypothesis 2b and reject hypothesis 2a.

For firm age, the original scale model does not show significance as well. However, since the firm age variable is an interval variable like industry competition, it might be influenced by the outliers. No support exists for hypotheses 3a and 3b.

The same reasoning counts when assessing firm size. Again there is no evidence of any significance for this variable. Giving no support for hypotheses 4a and 4b.

(23)

0.349), the highest percentage correct statistic (67.2%), and its Hosmer & Lemeshow statistic is not significant (p=0.720), something which is desired. This study takes model 5 as the final model for the original scale model.

Considering this final model and the hypotheses, this research concludes that there is solely evidence to assume that hypothesis 2B is significant. Uncertainty is the only main effect that has a significant value (P<0.1). Looking at the B-coefficient of this variable (-0.686) it becomes clear that the relationship is negative. In other words: One can assume that high technological uncertainty results in lower probability of entering a multi-partner alliance. All other hypotheses than 2B are assumed not significant under the original scale model.

Besides the significant main effect of technological uncertainty, having a marketing agreement also is a significant factor in entering a multi-partner alliance. To be more specific, marketing agreements (B-coefficient= -1.776) are assumed to be negatively related to the probability of having a multi-partner alliance.

Table 4: Regression results original scale modelabc

Model 1 Model 2 Model 3 Model 4 Model 5

Industry_Competition ,000 ,000 ,000 ,000 Tech_Uncertainty -,743* -,713* -,686* Firm_Age ,001 ,001 Firm_Size ,000 Previous_Experience -,007 -,007 -,002 -,003 -,004 Multiple Forms 21,599 21,584 21,547 21,508 21,501 R&D_Agreement ,288 ,254 ,122 ,117 ,157 Marketing_Agreement -1,787** -1,800** -1,820** -1,803** -1,776** Manufacturing_Agreement -,857 -,900 -,949 -,970 -,996 Industry_Relatedness -,004 ,032 -,024 -,021 ,004 Constant -,175 -,133 ,068 ,000 -,014 Model Fit -2 Log likelihood 219,947 219,296 215,717 215,616 214,258

Cox & Snell R Square ,240 ,243 ,256 ,257 ,262

Nagelkerke R Square ,320 ,243 ,342 ,342 ,349

Percentage Correct (%) 65,7 65,7 64,1 62,6 67,2

Hosmer & Lemeshow

Significange ,801 ,852 ,905 ,412 ,720

b: For lay-out purposes, some statistics are left out and are available at request. c: * = P < 0.1 ** = P < 0.05 *** = P <0.01

(24)

transformed. The transformation resulted in less obvious outliers and more normal distributed variables, even though logistic regression does not assume normality (Field, 2005).

First of all, it is interesting that a control variable for marketing agreements is again significant, even though on a lower scale (P<0.05 in model 3 and 4; P<0.1 in model 1, 2 and 5). The other agreement variables are not significant. Moreover, all other control variables do not show any sign of significance.

First of all, the coefficients of industry competition in the log transformed model, are actually fluctuating in contrast to the industry competition coefficients in the original scale model , which were constant (,000). However, there was no change in significance. Thus there is still no support for hypotheses 1a and 1b.

The main effect technological uncertainty is significant through model 3, 4 and 5. However, in contrast to the original scale model, Uncertainty has higher significance in the log transformed model (P<0.05). The coefficient is again a negative one, implying a negative relation. There is again support for hypothesis 2b. Hypothesis 2a is rejected.

For firm age (which is added in model 4), an interesting observation is the fact that it only has significant values in model 5 (P<0.05), but not in model 4. This study cannot provide a clear explanation why this occurred. However, the variable does differ from the original scale model, since in the log transformed model it actually is significant. Thus under the log transformed model there is support for hypothesis 3a, since the coefficient is a negative one.

Firm size employees shows significant values in model 5 (P<0.05). Remarkable, since the variable had no significance under the original scale model. Under the log transformed model there exists support for a positive relationship. In other words: hypothesis 4b is supported.

When looking into the model summary of the log transformed model, it can be seen that model 5 has the best fit. Model 5 has the lowest -2 log likelihood (142,901), the highest R squared values (0.363 & 0.484), the highest percentage correct (78.4%) and the Hosmer & Lemeshow statistic is not significant. Therefore for regression purposes model 5 is considered the final model for the log transformed model.

(25)

a negative relationship. In other words: One can assume that higher firm age results in a lower probability of entering a multi-partner alliance. For Firm size (in employees), the B-coefficient shows a positive value (0.284), and thus a positive relationship. One can assume that the larger a firm, the more likely it will enter a multiple partner alliance.

Besides the significant main effect of Uncertainty, Firm age and Firm size, having a marketing agreement also is a significant factor in entering a multi-partner alliance. To be more specific, marketing agreements (B-coefficient=-1.942) are assumed to be negatively related to the probability of having a multi-partner alliance.

Table 5: Regression results log transformed modelabc

Model 1 Model 2 Model 3 Model 4 Model 5

Industry_Competition -,030 -,042 -,038 -,201 Tech._Uncertainty -,991** -1,141** -1,017** Firm_Age -,319 -,690** Firm_Size ,284** Previous_Experience -,056 -,050 ,043 ,095 -,161 Multiple Forms 21,864 21,851 21,763 21,910 21,853 R&D_Agreements ,371 ,352 ,188 ,182 ,184 Marketing_Agreements -2,071* -2,080* -2,168** -2,274** -1,942* Manufacturing_Agreements -1,491 -1,508 -1,571 -1,397 -1,764 Industry_Relatedness ,190 ,190 ,119 ,072 ,196 Constant -,365 -,088 ,227 1,328 2,022 Model Fit -2 Log likelihood 154,938 154,904 150,072 148,895 142,901

Cox & Snell R Square ,310 ,310 ,332 ,337 ,363

Nagelkerke R Square ,414 ,414 ,443 ,450 ,484

Percentage Correct (%) 70,6 69,3 72,5 73,2 78,4

Hosmer & Lemeshow

Significange ,523 ,931 ,961 ,560 ,715

b: For lay-out purposes, some statistics are left out and are available at request. c: * = P < 0.1 ** = P < 0.05 *** = P <0.01

(26)

outliers, which might distort the results in the original scale variable, are mostly dealt with due to transformation in the log transformed model. Moreover, as stated earlier, model fit will be used to come to conclusions about the assumption of linearity. Since the model fit for the transformed variable model is higher, one may assume that the assumption of linearity is met (Restore.ac.uk, 2015).

Discussion & Conclusion

Discussion

After analyzing the results one can state that there is enough statistical evidence to assume that technological uncertainty, firm age and firm size have a relationship with the likelihood of forming a multi-partner alliance. Industry competition however does not show any significant relationship. First of all, regarding the results on industry competition, partly confirms the study of Li (2013). The insignificance of industry competition in this research is in line with Li (2013). There is now evidence that the relationship between industry competition and the probability of forming a multi-partner alliance is non-existent. However, Sakakibara (2002) and Eisenhard et al (1996) focused on the relationship between industry competition and forming an alliance in general. It could be that industry competition might prove significant when determining to form an alliance, but does not determine which form to use (multi-partner or dyadic). A possible explanation can be study itself. In other words: the variable might be measured too narrow and needs improvement. An extrinsic explanation, thus looking further than the study’s methodology, might be the following: competition raises a need for firms to develop or acquire new resources and routines. Alliances in general can satisfy such needs (Sakakibara, 2002; Eisenhard et al, 1996), therefore this relationship is significant. However, when determining whether to have a dyadic or a multi-partner alliance, industry competition is insignificant. This might be due to the fact that in both cases (dyadic or multi-partner) the need for having an alliance is already met. It does not matter which form an alliance takes, since both forms can offer the earlier stated needs when chosen the correct partner or partners. Industry competition thus does not dictate which form to take on.

(27)

relation between uncertainty and multi-partner alliances. Moreover there are differences in the way uncertainty is measured. This study uses a more technological sense, while Li (2013) looks more into the overall market uncertainty. However, the findings on technological uncertainty do support the line of argumentation that strategic convergence might be harder in multi-partner alliances when technological uncertainty is high (Doz, 1987), and that there might be more risk of losing core competencies (Brouthers, Brouthers & Wilkinson, 1995) within MP alliances in uncertain industries, and that therefore firms are more committed to dyadic alliances. This argumentation also lends support to the RBV, stating that a firm is reliable on the bundles of resources and competencies it possesses (Barney, 1991). Losing competencies not regarded as good from a RBV view, and thus firms tend to minimize this risk by engaging in less risky dyadic alliances.

Thirdly, the results of this study imply that higher firm age results in lower likelihood of forming a MP alliance. This is in line with Autio et al (2000) and Sorensen and Stuart (2000). In other words: it supports the argumentation that older firms have more mature routines in place which are difficult to unlearn and replace. Learning the routines from multiple partners might be more difficult than from one partner (dyadic). From a RBV, routines can be seen as a resource, since they can offer an advantage. It is interesting to see that firms do not automatically choose for the option with more routines, and thus more resources, but rather for a dyadic one, with probably less different routines. Besides the learning difficulty argument of Autio et al (2000) and Sorensen and Stuart (2000), one could argue that older firms simply do not feel the need to work with multiple-partners, because most resources, like legitimacy or unique routines (Hitt et al, 2000; Li, 2013) might already be possessed by older firms.

Fourthly, the last main effect of firm size implies a positive relationship with the likelihood of forming a MP alliance. The implication that larger firms are more likely to participate in MP alliances strokes with previous research. It is highly likely that larger firms have more resources than their smaller counterparts, to run alliances with multiple partners (Hagendoorn et al, 1994; Gulati, 1999), and thus therefore more likely to enter such an alliance. This does not imply that it is impossible for smaller firms to engage in MP alliance.

(28)

Implications

This research offers some implications for both scholars as well as managers.

For scholars, it implies that in future alliance research, both exogenous as well as endogenous factors should be taken into account, since both have significant effects on the decision whether to form a dyadic or multi-partner alliance. A simplified model which only looks at factors within firms might thus be inappropriate. Moreover, the results of this study show that factors, like firm age, technological uncertainty and firm size, which influence the decision to form an alliance in general also automatically, influence the decision whether this should be a multi-partner alliance or a dyadic alliance.

Furthermore, this study might be the basis for future research for forming a bridge between the formation field of multi-partner alliances and the performance field of multi-partner alliances. In other words: The relation between performance and multi-partner is already known. Due to this study, a couple of the formation factors are now known as well. Future research can and should now try to study new linkages. For example: whether multi-partner alliances in low uncertainty industries also perform better than the dyadic ones, since low uncertainty increases the odds of having a multi-partner alliance.

Moreover, the significance of having marketing agreements, implicated that for certain agreements multi-partner alliances are more suited. However since the focus of this study was not on agreements, future research should try to clarify this relation.

The research might also slightly extend knowledge in how different stages of life in a firm affect the influence of uncertainty in the need for multi-partner alliances. The contradicting results between established firms and new ventures (Li, 2013) imply such differences. However, to reach more grounded conclusion on this aspect, further research is warranted.

For managers, the study implies that when one wants to form an alliance, it is important to look at and understand the exogenous and endogenous situation a firm is in. The situation might predetermine the form of alliance. For example, when the firms wants to form an alliance, it might be that in more high uncertain industries there is less chances in forming multi-partner alliances. Another implication is that the results make managers more conscious about the alliance environment in certain industries, and thus whether to expect more dyadic alliances or more multi-partner alliances.

Limitations

(29)

other nations might be different. However, due to data and time limitations more extensive research was not possible. Thirdly, the research makes extensive use of proxies, it could be that these proxies are not exact representatives for what is meant to be measured. For example the high-technology industries are more technological uncertain. However, there are many differences between one high technology industry and the other, in terms of environmental uncertainty. The variables of technological uncertainty and industry competition, used in this research might be too superficial and narrow. Moreover Li (2013) found a quadratic relationship for uncertainty, but due the narrow proxies this study was unable to test for this form of relationship. Future research should focus on more detailed measures, focus more on inter-industrial differences and focus on possible quadratic and cubic relationships. Fifthly, although the results show statistical significance in several factors, the underlying motivations are not certain, nor tested for. Therefore, future research should become more in-depth regarding motivations. Lastly, there might be more variables that can influence the probability of forming a multi-partner alliance. Due to time and resource restrictions it was not possible to incorporate them all. However, future research should try to use more or other variables, like appropriability regimes, or strategic positioning, to enhance the research field.

Conclusion

(30)

References

Albors-Garrigos, J., Zabaleta, N., & Ganzarain, J. (2010). New R&D management paradigms:

rethinking research and technology organizations strategies in regions. R&D Management, 40(5), 435-454.

Audretsch, D. (1995). Innovation, growth and survival. International Journal Of Industrial

Organization, 13(4), 441-457.

Autio, E., Sapienza, H., & Almeida, J. (2000). EFFECTS OF AGE AT ENTRY , KNOWLEDGE INTENSITY, AND IMITABILITY ON INTERNATIONAL GROWTH. Academy Of Management

Journal, 43(5), 909-924.

Axelrod, R., Mitchell, W., Thomas, R., Bennett, D., & Bruderer, E. (1995). Coalition Formation in Standard-Setting Alliances. Management Science, 41(9), 1493-1508.

Barney, J. (1991). Firm Resources and Sustained Competitive Advantage. Journal Of

Management,17(1), 99-120.

Bauer, J., Lang, A., & Schneider, V. (2012). Innovation policy and governance in high-tech industries. Berlin: Springer.

Beamish, P., & Kachra, A. (2004). Number of partners and JV performance. Journal Of World

Business, 39(2), 107-120.

Belderbos, R., Carree, M., Diederen, B., Lokshin, B., & Veugelers, R. (2004). Heterogeneity in R&D cooperation strategies. International Journal Of Industrial Organization, 22(8-9), 1237-1263. Bengtsson, M., & Kock, S. (2000). Coopetition in Business Networks to Cooperate and Compete Simultaneously. Industrial Marketing Management, 29(5), 411-426.

Brouthers, K., Brouthers, L., & Wilkinson, T. (1995). Strategic alliances: Choose your partners. Long

Range Planning, 28(3), 2-25. Burns, R., & Burns, R. (2008). Business research methods and statistics using SPSS. Los Angeles: SAGE.

(31)

Celly, K., Spekman, R., & Kamauff, J. (1999). Technological Uncertainty, Buyer Preferences and Supplier Assurances: An Examination of Pacific Rim Purchasing Arrangements. J Int Bus Stud,30(2), 297-316.

Chandler, A. (1994). The Competitive Performance of U.S. Industrial Enterprises since the Second World War. Business History Review, 68(01), 1-72.

Cohen, W., & Levinthal, D. (1990). Absorptive Capacity: A New Perspective on Learning and Innovation. Administrative Science Quarterly, 35(1), 128.

Deeds, D., & Rothaermel, F. (2003). Honeymoons and Liabilities: The Relationship between Age and Performance in Research and Development

Doukas, J. A., & Kan, O. B. (2004). Excess cash flows and diversification discount. Financial

Management, 71-88.

Doz, Y. L. (1987). Technology partnerships between larger and smaller firms: some critical issues. International Studies of Management & Organization, 31-57.

Doz, Y., & Hamel, G. (1998). Alliance advantage. Boston: Harvard Business School Press. Dyer, J., & Singh, H. (1998). The Relational View: Cooperative Strategy and Sources of Interorganizational Competitive Advantage. The Academy Of Management Review, 23(4), 660. Eisenhardt, K., & Schoonhoven, C. (1996). Resource-based View of Strategic Alliance Formation: Strategic and Social Effects in Entrepreneurial Firms. Organization Science, 7(2), 136-150. Field, A. (2005). Discovering statistics using SPSS. London: Sage Publications.

García-Canal, E., Valdés-Llaneza, A., & Ariño, A. (2003). Effectiveness of Dyadic and Multi-Party Joint Ventures. Organization Studies, 24(5), 743-770.

García-Canal, E., & Sánchez-Lorda, P. (2007). One more only if it is one of us. The number of partners and the stock market reaction to domestic and international alliance formation in EU telecom firms.International Business Review, 16(1), 83-108.

Georgia State University,. (2000). Fiscal Research Program. Atlanta.

(32)

Gulati, R. (1999). Network location and learning: the influence of network resources and firm capabilities on alliance formation. Strat. Mgmt. J., 20(5), 397-420.

Hagedoorn, J., & Schakenraad, J. (1994). The effect of strategic technology alliances on company performance. Strat. Mgmt. J., 15(4), 291-309.

Hall, B. (1994). Corporate Restructuring and Investment Horizons in the United States, 1976– 1987.Business History Review, 68(01), 110-143.

Hall, B. (1999, July). Mergers and R&D revisited. In Prepared for the Quasi Experimental Methods

Symposium, Econometrics Laboratory, US Berkeley.

Hamel, G. (1991). Competition for competence and interpartner learning within international strategic alliances. Strat. Mgmt. J., 12(S1), 83-103.

Hitt, M., Dacin, M., Levitas, E., Arregle, J., & Borza, A. (2000). PARTNER SELECTION IN EMERGING AND DEVELOPED MARKET CONTEXTS: RESOURCE-BASED AND

ORGANIZATIONAL LEARNING PERSPECTIVES. Academy Of Management Journal, 43(3), 449-467.

Hosmer, D., & Lemeshow, S. (1989). Applied logistic regression. New York: Wiley.

Hu, M., & Chen, H. (1996). An empirical analysis of factors explaining foreign joint venture performance in China. Journal Of Business Research, 35(2), 165-173.

Jiang, R., Tao, Q., & Santoro, M. (2010). Alliance portfolio diversity and firm performance. Strat.

Mgmt. J., 31(10), 1136-1144.

Kleinknecht, A., & Reijnen, J. (1992). Why do firms cooperate on R&D? an empirical study. Research

Policy, 21(4), 347-360.

Kogut, B. (1988). Joint ventures: Theoretical and empirical perspectives. Strat. Mgmt. J., 9(4), 319-332.

Logistic Regression - Complete Problems. (2015). University of Texas at Austin.

Li, D. (2013). Multilateral R&D alliances by new ventures. Journal Of Business Venturing, 28(2), 241-260.

(33)

Madhok, A. (1997). Cost, value and foreign market entry mode: The transaction and the firm. Strat.

Mgmt. J., 18(1), 39-61.

Makino, S., & Beamish, P. W. (1999) Characteristics and performance of international joint ventures with non-conventional ownership structures. Journal of International Business Studies, 29

Mairesse, J., & Mohnen, P. (2004). The Importance of R&D for Innovation: A Reassessment Using French Survey Data. The Journal Of Technology Transfer, 30(1-2), 183-197.

Myers, R. (1990). Classical and modern regression with applications (2nd ed.). Boston, Mass.: Duxbury Press.

Oecd-ilibrary.org,. (2015). Retrieved 18 March 2015, from

http://www.oecd-ilibrary.org/docserver/download/5lgsjhvj7nkj.pdf?expires=1426691967&id=id&accname=guest&chec ksum=5B36DA28B4BC29C80D2A905748E664B2

Oxley, J. (1999). Institutional environment and the mechanisms of governance: the impact of

intellectual property protection on the structure of inter-firm alliances. Journal Of Economic Behavior

& Organization, 38(3), 283-309.

Park, S., & Russo, M. (1996). When Competition Eclipses Cooperation: An Event History Analysis of Joint Venture Failure. Management Science, 42(6), 875-890.

Peng, C., Lee, K., & Ingersoll, G. (2002). An Introduction to Logistic Regression Analysis and Reporting. The Journal Of Educational Research, 96(1), 3-14.

Polanyi, M. (1973). Personal knowledge. Towards a post-critical philosophy. (Reprinted.). London: Routledge & Kegan Paul.

Porter, M. (1986). Competition in global industries. Boston, Mass.: Harvard Business School Press. Quinn, J. B., & Hilmer, F. G. (1994). STRATEGIC OUTSOURCING. Sloan Management

Review, 35(4), 43. Retrieved from http://search.proquest.com/docview/1302974737?accountid=11219

Restore.ac.uk,. (2015). 4.14 Model Diagnostics. Retrieved 11 June 2015, from

http://www.restore.ac.uk/srme/www/fac/soc/wie/research-new/srme/modules/mod4/14/

Sakakibara, M. (2002). Formation of R&D consortia: industry and company effects. Strat. Mgmt.

J.,23(11), 1033-1050.

(34)

Song, M., & Montoya-Weiss, M. (2001). THE EFFECT OF PERCEIVED TECHNOLOGICAL UNCERTAINTY ON JAPANESE NEW PRODUCT DEVELOPMENT. Academy Of Management

Journal, 44(1), 61-80.

Sorensen, J., & Stuart, T. (2000). Aging, Obsolescence, and Organizational Innovation. Administrative

Science Quarterly, 45(1), 81.

Stuart, T. (1998). Network Positions and Propensities to Collaborate: An Investigation of Strategic Alliance Formation in a High-Technology Industry. Administrative Science Quarterly, 43(3), 668. Thomsonreuters.com,. (2015). SDC Platinum. Retrieved 18 March 2015, from

http://thomsonreuters.com/en/products-services/financial/investment-banking-and-advisory/sdc-platinum.html

Varadarajan, P., & Cunningham, M. (1995). Strategic Alliances: A Synthesis of Conceptual Foundations. Journal Of The Academy Of Marketing Science, 23(4), 282-296.

van den Bosch, F., & Elfferich, P. C. (1993). Make, buy or cooperate decisions with respect to research and development in international business networks.

Strength in Numbers