Author: Sven van der Voet

DOES SIZE MATTER? PERFORMANCE IMPLICATIONS OF

PARTICIPATION IN MULTI-PARTNER R&D ALLIANCES: ROLE OF

NUMBER OF PARTNERS AND FIRM’S RELATIVE SIZE

Author: Sven van der Voet

University of Groningen

Faculty of Economics and Business

MSc Business Administration, Strategic Innovation Management

June, 2014

1

_{Supervisor: Dr. I. (Isabel) Estrada Vaquero}

3

Does size matter? Performance implications of participation in

multi-partner R&D alliances: Role of number of multi-partners and firm’s relative

size.

Van der Voet, S Word count: 11.211

Faculty of Economics and Business, University of Groningen, Netherlands

Abstract

Multi-partner R&D alliances emerge rapidly in technology driven industries to cope with high levels of risk, short product life cycles and high demands for R&D investments, however it is unclear under which conditions these alliances succeed. This study examines the relationship between the firm’s number of partners and the firm’s relative size on its performance in a Multi-partner R&D alliance. Resource-based view and transaction cost perspective are used to explore whether there is a trade-off between increasing resource benefits as the number of partners grow, and increasing collaboration and coordination complexity. There are indications for a trade-off in firm’s relative size as well, where arguments are based on a higher bargaining power of larger firms and a better adaptability of smaller firms. Hypotheses are developed and tested on a sample of 141 firms involved in Multi-partner R&D alliances in High-Tech manufacturing, High-Tech services, and Biotechnology. The results suggest no significant relationship between number of partners and firm’s performance while relative smaller firms are performing better than relative larger firms.

Keywords:

Multi-partner R&D alliance, Firm’s relative size, Number of partners, Firm’s performance.

1. Introduction.

Alliances have become a popular strategy to strengthen a firm’s ability to compete and to innovate. More and more often these alliances are not dyadic but are formed with three or more partners in multilateral value chain activities such as in production, research, marketing and development. In technological development, Multi-partner R&D (MR&D) alliances are becoming increasingly important as a result of the globalization of competition (Lavie et al., 2007; Das and Teng, 2002; Thorgren et al., 2011). A MR&D alliance is defined as ‘collective voluntary inter-organizational

agreements that interactively engages its multiple partners in multilateral R&D activities’ (Lavie,

4

MR&D alliances tend to have the same value creation logic as dyadic R&D alliances but MR&D alliances differ in complexity from dyadic alliances on the level of control, knowledge sharing and collaboration (Hwang and Burgers, 1997). In addition, MR&D alliances feature no direct reciprocity, and social exchanges are generalized for all the partners. This result in more opportunities for a member of free-riding and the likelihood of a negative group process, in comparison to a dyadic alliance (Yin et al., 2012; Das and Teng, 2002; Lavie et al., 2007). Research suggested that the outcome of a MR&D alliance is heterogeneous for private and common benefits. This means that some partners can benefit more than others even when the alliance as a whole is the winner of the technology driven competition (Lavie et al., 2007).

Literature suggested a couple of factors why some partners can benefit more than others like; timing of entry, external involvement (Lavie et al., 2007), formal control (García-Canal et al., 2003), network configuration (Thorgren et al., 2009), technological diversity (Sampson, 2007), and internal capabilities (Mothe and Quelin, 2003). Authors found contradicting results on the number of partners in an alliance. Beamish and Kachra (2004) found no relationship with number of partners and firm performance. However, Park and Russo (1996), and Hennart and Zeng (2002), found evidence that an increase in the number of partners predicts a negative influence on the firm performance. On the one hand when the number of partners increases more resources and knowledge are added to the alliance which can accelerate novel innovations, at the same time firms can share risks and costs (Thorgren et al., 2012). On the other hand factors such as resource compatibility, structure, trust building and partner diversity become more complex if the number of partners in an alliance becomes larger. At the moment the complexity grows, achieving goals become increasingly difficult (Thorgren et al., 2012; Valdes-Llaneza and García-Canal, 2006). Furthermore, firm’s relative size deserves more attention as well, because authors argue that both smaller and larger firms can be beneficial while participating in a MR&D alliance. Larger firms have several advantages for higher collaborative performance: They are able to exploit economies of scale, they have access to wider markets, they have more bargaining power and they can take more risk (Rogers, 2004). Smaller firms use MR&D alliance more for information exchange, resource acquisition and technology transfer (Nieto and Santamaria, 2010).

More research is needed on these issues to explore if the benefits of the increase in number of partners are outweighing the disadvantages, and if relative smaller firms are able to reap the benefits of participation with larger firms. This research is important in order to fill the gap which kind of alliances are more beneficial for firm performance and which kind of collaborations can better be avoided. Therefore both the focal firms’ relative size as the number of partners should play a key role in explaining why some partners benefit more than others and achieve better performance. The research purpose of this study is to examine the implications of the focal firm’s relative size and the number of partners in a MR&D alliance on the firm performance. This leads to the main research question: What are the implications of firm’s relative size and/or number of

partners in MR&D alliance on a focal firm’s performance?

5

to research this phenomenon. The Orbis Database is used to gain data on the performance of a firm. The results of the analysis suggest that the number of partners in a MR&D alliance is not a predictor for the performance of the firm. However, when a firm is relative smaller in an alliance, it has a significantly better performance than larger firms.

This research is important for theoretical and managerial reasons. The theoretical contribution to the literature is on the performance implications of the alliance composition. This is essential to know in the MR&D alliance formation phase, because if firms know which composition has the highest likelihood of achieving good performance, then they are able to influence the MR&D alliance into their favor. This study contributes to the research of Doz et al. (2000) and Moth and Quelin (2000) on the formation processes of MR&D alliance and choice of partners by examining the best alliance composition. Also on the level of performance implications will it contribute by complementing the findings of Thorgren et al. (2012) who investigated how smaller firms are benefiting in a MR&D alliance. At last, this paper sheds light on the contradicting results of Beamish and Kachra (2004) and Hennart and Zeng (2002). Those authors found, respectively, no relationship and a negative relationship between number of partners and firm’s performance.

This study has some implications for managers in MR&D alliances. First, this paper illustrates what the effect of a higher number of alliance partners has on the firm performance. Is it for managers valuable to invest in a MR&D alliance with a large number of partners? Second, this paper demonstrates what the best variety of sizes is in an alliance. Managers are able to better analyze in which kind of MR&D alliance they should cooperate. Is it for a relative larger firms better to cooperate with innovative smaller firms or should they avoid such collaborations? The paper is structured as follows. The first section develops hypotheses based on arguments of leading papers in this field of research. Followed by a description of the methods and the results. This paper ends with a discussion, main conclusions and future research.

2. Theory and hypotheses.

This section provides some theory building and hypotheses. In the theory section this paper elaborates on a multi-partner R&D alliance and how the number of partners and firm’s relative size are outlined in existing research. The hypotheses focus on the theoretical arguments how these factors affect the performance of the firm.

2.1 Multi-partner R&D alliances.

6

Alliance Constellations (Das and Teng, 2002), Multi-Party Joint Ventures (García-Canal et al., 2003), and Multifirm Alliances (Hwang and Burgers, 1997). In this paper I follow the definition of multi-partner R&D alliance according to Estrada Vaquero (2012): ‘a Multi-partner R&D alliance is formed with three or more partners as a collective voluntary arrangement to engage in R&D activities’.

Sakakibara (2003) revealed that participation in a MR&D alliance contributes to the knowledge sharing and R&D spending and simultaneously stimulates the R&D productivity of the partners, thus enhancing economic welfare. However there are also a lot of risks, Katz (1986) found that the motive of participating in a MR&D alliance can be the substantial spillover of knowledge. However MR&D alliances can eliminate wasteful duplication of effort and can also increase the efficiency in the overall R&D stage by disseminating knowledge and by the coordination of individual projects. A MR&D alliance allows participating firms to gain access to resources which are hard to obtain on an arm’s length market. The purpose of a MR&D alliance is to do pre-competitive research, which aims to deliver knowledge which is not marketable yet (Mothe & Quelin, 2001). Because of the characteristic that a MR&D alliance is often a pre-competitive research it is not suitable for every market. MR&D alliances have emerged mainly in technology-driven industries to cope with the high R&D expenses, the race for shorter product life cycles, the high levels of environmental uncertainty, and technical risk. Typical MR&D markets are those with a high pace of innovation like the Chemical, High-Technology and Biotechnology. Next to the fact that participating firms can obtain spillovers, there is often a true creation of value in a MR&D alliance. For this reasons firms can gain an incentive to participate and contribute, because they want to appropriate the results of the common efforts.

Partner diversity in MR&D alliances hampers alliance activities such as establishment of domain consensus, alignment of interests and cognition, and integration of resources (Estrada Vaquero, 2012), however it can also evolve in a non-imitable combination of resources. Nevertheless, MR&D alliance have many similarities with dyadic R&D alliance, such as they are both established to reduce transaction costs associated with bearing risk and accessing resources (Doz and Hamel, 1998). Also the benefits of dyadic R&D alliances apply for MR&D alliances such as: access to market and technology information, investment opportunities, sharing in innovation, and influence over the evolution of industry standards. Meanwhile both alliance forms face the similar challenges such as asymmetric bargaining power, incompatibility and partner opportunism (Li, 2013).

7

participate it is more likely a large variety of firm sizes emerge in the alliance. These factors form the basis for the hypotheses in this research about MR&D alliances.

2.2 Number of partners.

The importance of the number of partners is elaborated in a few papers, mainly, for providing evidence for the difference between dyadic and multi-partner alliances. As García-Canal et al. (2003) pointed out that the important difference between dyadic R&D alliance and MR&D alliance is whether social exchanges are direct or generalized. Due to the fact that more firms have to cope with knowledge and resources contributions these exchanges are generalized in a multi-partner setting. Most of the time these contributions are on a voluntary basis to the group and when more members are involved in the MR&D alliance, more information must be generalized. Further, because exchanges are generalized, the firm who contributes the resources and information does not necessarily have to be the one who benefits the most from the contribution (Thorgren et al., 2011; Takahashi, 2000).

Another characteristic for different numbers of partners is that it represents different levels of complexity, not only on the coordination part but prominently on the contributions in the alliance (García-Canal et al., 2003). Many scholars point at the risk of free-riding, because with a higher number of partners, some firms can benefit without ever contributing anything. While free-riding is also a potential problem in dyadic alliances the level of transparency is higher in such collaborations (Das and Teng, 2002; Thorgren et al., 2011).

Gong et al. (2007) found that an increase of the number of partners will decrease the goal performance due to divergence in goals and behavioral uncertainty. Also will an increase in the number of partners, decrease the partner cooperation due to cohesiveness reduction and increasing information cost. However when more partners are contributing to the MR&D alliance there is more opportunity that complementary resources are developed into novel innovations as Makri, Hitt and Lane (2010) pointed out. The increase in the number of firms in an alliance, decreases the firm’s risk of abnormal returns including the market risk. Beside it is more likely that the development becomes an industry standard (Aggarwal et al., 2011). Summarized, the number of partners may determine the level of generalized exchange, cost, trust, complexity and coordination in a MR&D alliance, which can have consequences for the MR&D alliance outcome.

2.3 Firm’s relative size.

The relevance of size for exchange in multi-partner alliance is outlined in several ways in research. Hannan and Freeman (1977) discovered that smaller firms tend to collaborate with larger partners because they possess more slack in resources and economies in scale. Larger firms may team up with smaller firms because they are more flexible which leads to higher ability to innovate and to adapt, including less bureaucratic procedures in accepting and implementing change (Mintzberg, 1980). Through this flexibility they tend to initiate competitive actions more rapidly (Storey, 1998).

8

smaller firms are less attractive partners (Sharfmann et al., 1988; Ahuja, 2000). This is complemented with Pfeffer and Salancik’s (1987) findings that larger partners are attractive partners, because of a better reputation and more market power. Concluding that the relative size of a firm is crucial in a MR&D alliance.

2.4 Hypotheses.

Number of partners and firm performance.

The number of partners represents a dimension of complexity affecting the extent to which focal firm’s goals for the alliance are fulfilled. It is earlier mentioned that the difference in dyadic and multi-partner alliances is whether social exchanges are direct or generalized, and also the immediacy of reciprocity (Das and Teng, 2002; García-Canal et al., 2003; Valdes-Llaneza and García-Canal, 2006). However, this raises the question if this phenomenon also have influence in a multi-partner setting and if this has consequences for performance.

Beamish and Kachra (2004) concluded that transaction cost reasons argue not in favor of an increase in number of partners, because of inherent complexity and the myriad of transactions. However, the authors present an opposite argument that the benefits of resource complementarity in multi-partner alliances outweigh the cost of management and operations. Despite the fact that Beamish and Kachra (2004) only researched the hierarchical governance mode, joint ventures, they did not find a negative relationship between performance and number of partners, but nor a positive effect. Further, I will present the transaction cost logic and resource based arguments on the increase in number of partners and the implications on firm performance.

From a resource-based perspective, the involvement of multiple partners, presumably, increases the diversity of resources and capabilities including more complementary resources (Doz and Hamel, 1998). This increase reduces the borne risks of each partner which encourage knowledge sharing (Hwang and Burgers, 1997). An increase in the number of partners causes more resource and capability contributions. Vanhaverbeke and Noorderhave (2001) found that alliances are predominantly formed by firms who have complementary capabilities. These firms can create unique outcomes and gain competitive advantage over other firms (Makri et al., 2010). This is why companies rely heavily on joint R&D activities for gaining competitive advantage (Das and Teng, 2002). Such a competitive advantage can create extraordinary performances. Increases in the number of partners can be very beneficial for firms because of the wide-spread knowledge in several areas. Also, a large number of partners can positively affect the longevity of a stake in an alliance, when all the partners are competitors, it can create high levels of coopetition and a special creation of value (Valdes-Llaneza and García-Canal, 2006).

9

Complex negotiation processes and finding common ground for collaboration are challenging with increasing partners (Yin et al., 2012). Also Valdes-Llaneza and García-Canal (2006) argue that increases in the number of partners reduce the longevity of stakes in the MR&D alliance. They conclude that having more partners in an alliance is an obstacle, because it can hinder good functioning and development of the agreement. The incentive for freeriding is also bigger when there are more partners, which hinders value creation in a MR&D alliance (Das and Teng, 2002). Furthermore, the number of dyadic relationships becomes larger in a multi-partner setting which can cause more conflicts (Carcía-Canal et al., 2003). For example, when the number of partners is higher, knowledge and economic exchanges between partners are more difficult because explicit contracts are more challenging and harder to develop. This raises the cost of an alliance because a more hierarchical governance mode is needed (Li et al., 2012). Finally when social exchanges become more generalized the knowledge sharing could be less useful for the partners in the alliance which can affect the firm performance.

In conclusion, the literature agrees that when the number of firms in a MR&D alliance becomes higher, the complexity of coordination, cooperation and trust building increases. Further, it becomes more difficult to divide the benefits of R&D. However, the combination of complementary resources causes a high potential that a unique competitive advantage emerge. It seems to be that when the level of firms is manageable, collaboration in MR&D alliance can be beneficial. Therefore, I hypothesize that:

H1: The relationship between the firm’s number of partners and the firm’s performance in a MR&D alliance is inverse U-shaped.

Firm’s relative size and firm performance.

Firms in MR&D alliances tend to form heterogeneous collaboration in terms of contributions and investments (Mothe and Quelin, 2011; Hagedoorn, 2002). However, smaller firms always have been negatively associated with prospering in alliances that include larger partners (Thorgren et al., 2012). Smaller firms have some characteristics which reduce bargaining power such as capabilities that fill a niche, and fewer resources to contribute. They also have difficulties to reach technological agreements and set up lines of communication (Nieto and Santamaria, 2010). Inkpen and Beamish (1997) argue that differences in bargaining power cause alliance instability which has negative implications for firm’s performance. These results are supported by Nieto and Santamaria (2010) who found that R&D collaboration contributes to the innovativeness of smaller and medium sized firms, at least more than for larger firms. However, they note that it is less likely for smaller firms to be invited for a MR&D alliance than larger firms, because smaller firms are less attractive partners (Ahuja, 2000).

10

goals than smaller sized firms.

Some studies argue from the perspective of trust on performance (Dyer and Chu, 2000; Mohr and Puck, 2013). According to Thorgren et al. (2009) varying firm size does have effect on trust between partners, because different sized firms have resource inequity. Thorgren et al. (2011) found an inverted U-shape in the relationship between trust and size of the firm. The trust between partners is higher between medium sized firms than between either smaller and/or larger firms. Larger firms also have low perceptions of their partner’s ability to keep promises and show commitment to exchanges. However, the authors found evidence that smaller firms make adjustments, after being in an alliance, to remain aligned with environmental changes and reap the benefits in terms of innovative outcomes. On the other hand, larger firms often take advantage by smaller firms because they are more powerful and independent. This implies an imbalance of power and dependence. Accordingly, this would suggest that medium sized firms in MR&D alliances are performing better than larger and/or smaller firms in MR&D Alliances. Therefore, I hypothesize that:

H2: The relationship between firm’s relative size and the firm’s performance in a MR&D alliance is inverse U-shaped.

Figure 1 shows the conceptual model. This conceptual model has two independent variables, Firm’s Number of Partners and Firm’s Relative Size. The effect of these independent variables is tested on the dependent variable Firm Performance. Control variables are added to analyze if those variables influence the firm performance.

Figure 1: Conceptual model

3. Methods.

3.1 Sample and data.

11

Biotechnology industries that are involved in multi-partner alliances in the years 2005, 2006, 2007, 2008 en 2009. Firms in these industries are suitable for research in MR&D alliances, because the ability of the firm, in these particular industries, to innovate and commercialize quickly is important for a firm’s survival and profitability to stay ahead of the competition. The definition of a High-Technology industry comes from the America Electronic Association (AeA), who was the largest association of High-Tech companies in America. They recently merged with Information Technology Association of America (ITAA) to form TechAmerica. The AeA defines two High-Tech categories; the High-High-Tech manufacturing (SIC codes 283, 357, 365, 367, 381, 382, and 384) and High-Tech services (SIC Codes 481, 482, 484 and 489) (Li, 2013). In order to create a sufficient sample for testing, this paper adds a third category of firms in Biotechnology. Those firms operate in Commercial Physical and Biological Research (SIC code 873). The SIC codes are used to search for multi-partner alliances in the SDC Alliance Database. The Securities Data Corporation (SDC) database on alliances and joint ventures contains information on several types of alliances and is composed from public sources as industry and trade journals, U.S. Securities and Exchange Commission (SEC) filings and news reports. The SDC database tracks a wide range of agreements like sales, R&D, marketing and manufacturing. Firms are not required to report alliance activities, coverage is inevitably complete. The source is not the only alliance database available, however the database is the most used database in empirical studies and the most comprehensive database on alliance activity (Schilling, 2009).

The search criteria to construct the sample of alliances are (1) the above mentioned SIC codes, (2) more than two participant in an alliance, and (3) the alliance was announcement in the years from 2005 till 2009. This research has his focus on the recent years, because the expectations are that MR&D alliances are more common in these years. I strive to measure the effect on firm performance including financials from one, two and three years after the alliance announcement. Due to this measure I cannot take more recent years because the financial data are not available yet. The alliance sample consists of 144 MR&D alliances with 463 participating firms. The SDC database provides essential information on alliances however the database does not provide financial data at the firm level. Therefore I make use of the Orbis database which contains comprehensive information on companies worldwide. Orbis has an emphasis on private company information as companies’ financials, ratings, contacts, and industry research from the last 10 years. Therefor the data on alliances is supplemented with the financial data from Orbis. The database gains its information from i.e. the national Chamber of Commerce.

12

firm would appear two times in the sample and performance implications cannot be effectively measured. In order to prevent overlapping data, only the first announced alliance is included. Finally, (3) firms with highly odd financial data are excluded because these could bias the sample. The final sample is reduced to 141 firms. In this sample 114 firms participate in a 3-partner alliance, 11 firms in a 4-partner alliance and 16 firms in a 5-partner alliance.

3.2 Measures.

Dependent variable.

Firm Performance. As mentioned earlier, this work strives to measure the performance of the firms

who participate in a MR&D alliance. This cannot simply be measured with the profit because firms in the sample largely differ in size. The implications for firm’s performance are measured using the focal firm’s profit margin. Past alliance literature suggests that profit margin is an indicator for firm’s performance (Faems et. al., 2010; Jiang et. al., 2010). This measure is a reliable indicator for the performance of the firm, based on the accounting literature (Dhaliwal et al., 1999). Profit margin reflects the profit divided by the turnover, where an increase in profit should increase the profit margin when the turnover stays the same.

To measure the firm’s performance of the focal firm I calculate the average fluctuation in profit margin over the years. The variable is built by the percentage difference between (1) the average of the year before the alliance and the start of the alliance, and (2) the average between the three years after an alliance. This variable must be able to filter out high fluctuations in the numbers, for example caused by the crisis.

𝐹𝑖𝑟𝑚 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 =PM(Year 1 + 2 + 3)

3 −

PM(Year before + Starting Year)

2

PM = Profit margin

Independent variables.

Firm’s Relative Size. This variable reflects the size of the firm compared with the other firms in

the alliance. In order to measure the size of the firm this paper uses the number of employees. This measure captures the argumentations used in the theoretical section. The number of employees is widely used in the alliance literature to measure the firm’s size (Thorgren et al., 2011; Goerzen and Beamish, 2005). A weakness of this measure is that the number of employees in the firm is not always representative for the number of employees who work in the R&D alliance. However it is still a suitable proxy for the firm size, because it is not biased to industry or type of organization. The variable Firm’s Relative Size is constructed as follows; the variable takes the number of employees of the focal firm divided by the average number of employees of that particular alliance.

𝐹𝑖𝑟𝑚′_{𝑠 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑆𝑖𝑧𝑒 =} Number of employees in the firm

13 Number of Partners. This variable consists of the number of firms participating in the particular

alliance including the focal firm. This independent variable is used by Valdes-Llaneza and García-Canal (2006) to measure the number of partners in the analyzed alliances.

Control variables.

The literature on multi-partner and dyadic alliances suggests several factors that influence firm performance. I control for variables that may influence the focal firm performance at the alliance, firm, and industry level as age, previous alliance experience, industry, joint venture, crisis, resource diversity and R&D intensity. Table 1 gives an overview of the variables.

Age. This variable equals the number of years since the foundation of the business and is modeled

as a control variable. Older firms are likely to exploit existing competencies and capabilities than to explore new terrains (Cui and O’Connor, 2012). That could mean that younger firms are more eager to learn and to absorb more knowledge from an alliance than older firms, which can create more sustainable performance in the long term (Thorgren et al., 2012).

Previous Alliance Experience. Firms with frequent alliance activities are better in managing

alliances. They learn how they can capture and transform more value from an alliance which can improve the performance of the firm (Li et al., 2012; Valdes-Llaneza and García-Canal, 2006). I measure this by counting the number of alliances from the past 5 years of the focal firm. A dummy variable is created which takes value 0 when a firm is inexperienced. This is the case when a firm has less than, or equal to, 4 alliance activities in the past 5 years. The firm is experienced when it has more than 4 alliance activities in the past 5 years and in this case the variable takes value 1.

Industry. Prior research suggests that manufacturing industries are associated with higher levels of

fixed investments which results in higher exit cost than service industries (Li et al., 2012). Accordingly, I expect that profit margins could be more under pressure and varies between industries. The control variable Industry includes three dummy variables for every sector. A dummy takes has value 1 for Commercial Physical and Biological Research (SIC code 873) otherwise 0. A dummy variable takes value 1 for High-Tech manufacturing (SIC codes 283, 357, 365, 367, 381, 382, and 384) otherwise 0, and a dummy variable takes value 1 for High-Tech services (SIC Codes 481, 482, 484 and 489) and otherwise 0.

Joint Venture. In case of creating a joint venture, firms form a separate legal entity, while they

14

venture in which the assets are pooled, adapted and integrated for a common purpose. This could give performance differences between a joint venture and a strategic alliance in a multi-partner R&D setting. A dummy variable is created where a joint venture has value 1 and a strategic alliance has value 0.

Crisis. The performance is measured during the years in which the financial crisis occurs. Since

sales of firms could drop dramatically, the focus of the firm could change, and the firm could lose his interest in the alliance. I control if the financial crisis has influenced the performance of the firm. A dummy variable has value 1 when the alliance is created before the financial crisis, the years; 2005, 2006 and 2007. The dummy has value 0 when the alliance is announced after the start of the financial crisis; 2008 and 2009.

Resource Diversity. The paper of Lavie et al. (2012) suggests that organizational differences in the

operational domain influence firm performance due to different routines and mechanisms. Diversified alliances also have access to resources and knowledge in other industries from which novel ideas can be generated (Hitt et al., 1997). This suggests that firms creating MR&D alliances with partners from different industries could have different performance outcomes than alliances in the same industry. I control for this difference by creating a dummy variable. When the first two digits of the focal firm’s SIC code match the first two digits of the Alliance SIC code I suggest that the firms operate in the same industry and the dummy takes value 1. Otherwise the dummy variable takes value 0.

R&D Intensity. Since the focus of this research is on MR&D alliances it seems to be essential to

measure the R&D activities. R&D investment is generally seen as a crucial determinant of innovation. It helps firms to absorb, create, exploit and transform knowledge (Cohen and Levinthal, 1990). However, 52 firms from the total sample of 141 didn’t provide R&D information or the data is inconclusive on Orbis. For this reason, I leave the control variable out for the main analysis and run a separate analysis with the R&D control variable. This measure is built in the same way as the dependent variable firm’s performance; the percentage difference from the average of R&D expenditure to the total sales between (1) the average of the year before the alliance and the year during the alliance and (2) the average of the 3 years after the alliance started.

Table 1: Variables, Measures and Data sources

Variables Measures Data source

Firm Performance

A combined variable between (1) the difference of the average profit margin between one year before- and starting year of the alliance and (2) the average of 3 years after the alliance.

Orbis

Firm’s Relative Size

The number of employees of the firm divided by the average

number of employees of the partners in the alliance. Orbis

Number of Partners Number of Alliance partners SDC Platinum

15 Previous Alliance

Experience

A dummy which takes value 1 when a firm has more than 4 alliance activities in the past 5 years, otherwise it is inexperienced and takes value 0.

SDC Platinum Industry A dummy variable for every industry. SDC Platinum Joint Venture A dummy variable which takes value 1 when it is a JV,

otherwise 0. SDC Platinum

Crisis

A dummy variable has value 1 when the alliance is created before the financial crisis (Years: 2005, 2006 and 2007), otherwise it takes value 0 (Years: 2008 and 2009).

SDC Platinum

Resource Diversity

A dummy variable which takes value 1 if the firm’s two digits SIC Code are similar to the others in the alliance, otherwise it takes value 0.

SDC Platinum

R&D Intensity

A combined variable between (1) the difference of the average R&D expenditures between one year before and starting year of the alliance and (2) the average of 3 years after the alliance.

Orbis

3.3 Methods.

With the purpose to test the two hypotheses, I create OLS multiple regression models. A general model should accept or reject the two hypotheses, in order to test for an inverted U-shape squared terms are included in the models. After testing the general models the sample will be split to explore more about the independent variables. The independent variable, Number of Partners, is tested in the general linear regression models, however during the data collection, I discovered that there are very limited firms who are participating in a 4-partner alliance or more. In the sample of 141 only 11 firms where participating in a 4-partner alliance and 16 firms in a 5-partner alliance. In order to explore this independent variable better, this variable will be split into two groups; 3-partner alliance and more than 3-3-partner alliances. In addition a Kruskal-Wallis test should give evidence if there is any significant difference in performance of alliances with different number of

partners.

The second independent variable, Relative Firm Size, will be split into three groups; relative small-, medium- and large sized firms. A firm is absolute a medium sized firm when the variable has value 1. The total sample has an average with a value 1,3. The criteria to split the subsample are, for the small subsample, that the variable must have the value between 0 and 0,5; the medium subsample has the value between 0,5 and 1,5 and the large subsample is larger than 1,5. This creates, around, the same amount of firms in each group, respectively 41, 39 and 61 firms in each subsample. From this subsample, it can be observed if there are any linear relations and what effect the control variables have on firm’s performance in that particular subsample. In order to see if there is difference between the groups, first, a Kruskal-Wallis test is conducted to measure the mean ranks between the subsamples to test if they differ significantly from each other. Second, curve estimation should predict what the path is that firm performance has on relative firm size and an attempt is made to capture this path in a figure.

16

participation in MR&D alliances. This should suggest that the control variable R&D intensity has some predictive power for firm’s performance. However due to the fact that the sample including the R&D data is considerably small (N=89) it is not included in the main analysis. I run a separate OLS regression to see if this R&D data has predictive power for the firm’s performance in the different subsamples.

Second, differences in industry performance could influence the outcome of the model as well. I perform a check for industry to investigate whether the results were consistent across the industries (High-Tech manufacturing, High-Tech services and Biotechnology) (Thorgren et al., 2012). Because the sample is distributed normally but the groups are unequal divided I use a non-parametric test for independent samples. Finally, I check if there is an interaction effect between the two independent variables, number of partners and firm’s relative size. This could give more information about the two variables.

Overall I use a minimum confidence level of 90% to conclude if there are any implications for firm’s performance. In general a higher confidence level has more predictive power but due to the fact that I have a small dataset (N=141) the probability value of p <0.1 is included in the results. Although this probability is not significant, Fisher (1925) argued that in case a p-value ranges between 0.05 and 0.1 it provides some reason to predict that the relation is not totally random and shows a trend towards significance.

4. Analysis.

4.1 Descriptive statistics.

Table 2 displays Spearman’s correlations and significance for each of the variables used in the analysis. I found that the independent variable Relative Firm Size is correlating, not surprisingly, with Age and Previous Experience. However this correlation is not really strong (both lower than 0,24). The other independent variable, Number of Partners, has very weak correlations with all the other variables (all lower than 0,18) and only has a significant correlation with Joint Venture. This could suggest that the Number of Partners is a weak measure in this analysis. Furthermore, the variable Joint Venture has surprisingly some interesting results, because it has a negative correlation with Firm Performance. It seems to be that the MR&D alliances based on Strategic

Alliance are performing better. A positive significant correlation between Joint Venture and Number of Partners suggests that firm's with a higher number of partners are more likely to form

a Joint Venture than a Strategic Alliance.

Interesting in this correlations matrix is to see that the findings of Chen and Chen (2003) are confirmed. Firms that face more Resource Diversity are likely to seek more hierarchical control in the alliance and will chose a joint venture over a strategic alliance. The variable R&D Intensity has a negative and positive correlation with respectively Previous Experience and Resource

Diversity. In which the negative correlation is surprising, because it may suggests that firms with

a lot of previous experience in alliances have less R&D expenditures. The dependent variable Firm

17

I carefully examine the data for potential multicollinearity problems. The variance inflation factors (VIF) for all variables are below 5 and besides are the average VIF’s for all the regression models below 2. Both indicate that the figures are within the accepted threshold (Neter et al., 1990). Therefore, I’m confident that the variables are not threatened by multicollinearity problems. 4.2 General models.

I test hypotheses 1 and 2 trough linear regression in a general model, because inverted U-shapes in the relationships are hypothesized, the models 4, 5 and 6 test for potential curvilinear effects. Model 1 tests the control variables; Model 2 and Model 3, include respectively the independent variables, Number of Partners and Relative Firm Size. In order to test my prediction that Number

of Partners would have an inverted U-shape relation to firm’s performance (hypothesis 1), I

introduce the Number of Partners Squared term in Model 4 of the regression equation. The same transition applies for Relative Firm Size (hypothesis 2), which is tested in Model 5; finally, Model 6 shows all variables in one regression model. Table 3 shows the standardized coefficient and level of significance with the dependent variable Firm Performance.

Table 2: Descriptive statistics

Variable Firm Performance Relative Firm Size Number of Partners Age Previous Experience Resource Diversity Joint Venture Crisis R&D Intensity Firm Performance (N=141) 1.000

Relative Firm Size (N=141) -.073 1.000

18

Table 3: General models

As can be seen from table 3, the results of the regression don’t confirm any significant relationship between Number of Partners to Firm Performance. The regression models 2 and 4 lacks a significant relationship for Number of Partners (β = -0.059, p < 0,503) and Number of Partners

Squared (β = -0.478, p < 0,797). Since no relationship between Number of Partners and Firm Performance can be confirmed, hypothesis 1 is rejected. The same applies to the relationship

between Firm’s Relative Size to Firm Performance. No significant relationship is present in models 3 and 5; Firm’s Relative Size (β = -0.016, p < 0,857) and Firm’s Relative Size Squared (β = -0.312, p < 0,342). Thus hypothesis 2 is rejected as well.

The control variable Joint Venture has a negative, significant correlation (β = -0.195, p < 0,05) to Firm Performance in all the models. Resource Diversity has a slightly positive correlation (β = -0.160, p < 0,1), however this relationship is not applicable in all the models. In order to explore the data in depth the sample is split in different groups. First, the variable Number of

Partners is split in two groups; 3-partner alliance and more than 3 partners in an alliance. Second,

the variable Relative Firm Size is split in three groups; small, medium and large.

Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Constant 4.16 (0.17) 9.13 (0.26) 9.40 (0.25) -12.52 (0.88) 9.96 (0.23) 12.66 (0.89) Age -0.09 (0.30) -0.09 (0.28) -0.09 (0.30) -0.09 (0.31) -0.08 (0.36) -0.08 (0.36) Previous Experience -0.04 (0.64) -0.04 (0.67) -0.034 (0.70) -0.03 (0.70) -0.02 (0.83) -0.02 (0.83) Resource Diversity 0.15 (0,1†) 0.14 (0.12) 0.14 (0.12) 0.14 (0.13) 0.16 (0,08†) 0.16 (0,09†) Joint Venture -0.21 (0.02*) -0.20 (0.03*) -0.20 (0.03*) -0.20 (0.03*) -0.20 (0.03*) -0.20 (0.03*) Crisis -0.04 (0.62) -0.05 (0.54) -0.05 (0.54) -0.05 (0.53) -0.06 (0.51) -0.06 (0.51) Number of Partners -0.06 (0.50) -0.060 (0.50) 0.42 (0.82) -0.05 (0.59) -0.11 (0.96)

Relative Firm Size -0.02 (0.86) -0.02 (0.84) -0.32 (0.34) -0.32 (0.35)

Number of Parters² -0.48 (0.80) 0.06 (0.98)

Relative Firm Size² 0.31 (0.34) 0.32 (0.36)

R2 0.06 0.06 0.06 0.06 0.07 0.07

F 1.71 (0.14) 1.50 (0.18) 1.28 (0.27) 1.12 (0.35) 1.23 (0.29) 1.09 (0.38)

N=141 Two tailed test † P<0,1

19

4.3 Sample split: Number of partners. To explore the data in more depth, I

split the sample in table 4. These results explain more about the findings of the

general model. The positive

relationship of Resource Diversity, and the negative relationship of Joint

Ventures to firm performance only

apply for 3-partner alliances.

Furthermore, it is more likely that firms with more previous experience are gaining better firm performance in larger MR&D alliances (β = 0.418, p < 0,1).

Due to the limited size of the second subsample a regression analysis may not be the best option. Therefore a Kruskal-Wallis test is performed to test if there is a difference in the performance between firms with different number of partners in an alliance. The sample is split in 3-, 4- and, 5-partner alliances and the outcome suggests no significant differences in the firm performances (p<0.787). Concluding that the variable

Number of Partners is not a predictor

for the firm performance.

Table 4: Splitting the sample with the variable ‘number of partners’

Variable model 1 model 2 model 3

Constant 4.98 (0.17) 4.99 (0.22) 6.25 (0.15) Age -0.09 (0.34) -0.09 (0.35) -0.08 (0.39) Previous Experience -0.08 (0.42) -0.08 (0.43) -0.07 (0.48) Resource Diversity 0.19 (0,06†) 0.19 (0,06†) 0.21 (0.047*) Joint Venture -0.22 (0.03*) -0.22 (0.03*) -0.21 (0.041*) Crisis -0.05 (0.59) -0.05 (0.59) -0.05 (0.61) Relative Firm Size -0.01 (0.99) -0.30 (0.44) Relative Firm Size² 0.31 (0.42)

R2 0.07 0.07 0.08

F 1.68 (0.14) 1.39 (0.23) 1.28 (0.27)

Variable model 1 model 2 model 3

Constant -0.36 (0.95) 3.31 (0.56) 7.49 (0.24) Age -0.07 (0.79) 0.026 (0.92) 0.01 (0.95) Previous Experience 0.23 (0.30) 0.28 (0.21) 0.42 (0,09†) Resource Diversity -0.13 (0.56) -0.26 (0.27) -0.22 (0.33) Joint Venture -0.19 (0.54) -0.25 (0.40) -0.22 (0.45) Crisis 0.05 (0.86) 0.23 (0.50) 0.13 (0.70) Relative Firm Size -0.35 (0.15) -1.34 (0,074†) Relative Firm Size² 1.05 (0.16)

R2 0.11 0.20 0.28

F 0.51 (0.77) 0.82 (0.57) 1.06 (0.43) Two tailed test

† P<0,1 *P<0.05

3-partner Alliance (N=114)

20

4.4 Sample split: Firm’s relative size. The size difference of relative small and relative large firms could be enormous. For deeper exploration I split the cases in three groups in table 5; relative small, medium and large sized firms. Every group has three different models: Model 1 includes only the control variables; Model 2 includes the number of partners; and lastly, Model 3 tests for a potential curvilinear relationship.

With respect to the control variables, it is relevant pointing out that the Joint

Venture variable has a negative and

significant influence on the variable Firm

Performance in relative larger firms (β =

-0.303, p < 0,05). This provides evidence that a strategic alliance is performing better than a joint venture in a MR&D alliance. Further, the variable Resource

Diversity has a positive and significant

correlation with Firm Performance (β = 0.294, p < 0,063). This result suggests that relative smaller sized firms are benefiting from higher resource diversity in a MR&D alliance, however both results are rather weak.

In spite of the fact that linear relationships in all of the relative firm size groups is not found, I try to analyze the path relative firm size takes on firm performance in my data. First a Kruskal-Wallis test is conducted to analyze the mean ranks. The test for independent samples has mean ranks of 82,34, 59,09 and 71,57 for respectively relative small-, medium- and large firms and their means differ significantly (p = 0,028).

Variable Model 1 Model 2 Model 3 Constant 4.66 (0.52) 5.30 (0.79) 207.13 (0.33) Age -0.13(0.38) -0.14 (0.39) -0.15 (0.36) Previous Experience 0.00 (0.96) 0.01 (0.97) 0.03 (0.85) Resource Diversity 0.33 (0.04*) 0.33 (0.04*) 0.41 (0.03*) Joint Venture -0.13 (0.41) -0.13 (0.41) -0.09 (0.58) Crisis 0.02 (0.88) 0.02 (0.89) 0.055 (0.74) Number of partners -0.01 (0.97) -3.30 (0.34) Number of partners² 3.30 (0.34) R2 0.14 0.14 0.16 F 1.16 (0.34) 0.94 (0.47) 0.94 (0.49)

Variable Model 1 Model 2 Model 3 Constant -1.77 (0.80) 7.93 (0.59) 2.03 (0.81) Age -0.16 (0.39) -0.16 (0.40) -0.16 (0.40) Previous Experience 0.078 (0.66) 0.10 (0.58) 0.10 (0.58) Resource Diversity 0.07 (0.74) 0.03 (0.89) 0.03 (0.89) Joint Venture -0.11 (0.59) -0.03 (0.88) -0.03 (0.88) Crisis 0.05 (0.79) 0.03 (0.86) 0.03 (0.86) Number of partners -0.15 (0.45) -0.15 (0.45) Number of partners² 0.15 (0.45) R2 0.04 0.06 0.06 F 0.26 (0.93) 0.31 (0.92) 0.31 (0.92)

Variable Model 1 Model 2 Model 3 Constant 5.15 (0.16) 12.66 (0.22) 16.63 (0.86) Age 0.01 (0.93) -0.01 (0.99) -0.01 (0.98) Previous Experience -0.07 (0.60) -0.06 (0.65) -0.06 (0.65) Resource Diversity 0.03 (0.85) 0.00 (0.98) 0.00 (0.98) Joint Venture -0.30 (0.04*) -0.28 (0,06†) -0.28 (0,07†) Crisis -0.14 (0.30) -0.16 (0.24) -0.16 (0.25) Number of partners -0.11 (0.44) -0.21 (0.93) Number of partners² 0.11 (0.97) R2 0.11 0.12 0.12 F 1.35 (0.26) 1.21 (0.31) 1.03 (0.43) Two Tailed test

† P<0,1 *P<0.05

Relative Large Firms (N=61) Relative Small Firms (N=41)

Relative Medium Firms (N=39)

21

Second, to analyze the shape I conduct a Curve Estimation with a quadratic function. This curve estimation, shown in figure 2, has an R² of 0,04 with a significance of p < 0,1. Respecting the fact that these results are weak, it gives

a small implication about the

performance of firms in MR&D alliances and how the relative size could influence the degree of performance. Figure 2 shows that relative smaller firms are benefiting from participation in a MR&D alliance. The minimum is positioned right after the absolute

medium sized firms who are

participation in MR&D alliances,

thereafter the performance rises again for the relative larger firms. Overall, hypothesis 2 (inverted U-Shape) cannot be confirmed. However, despite the weak analytical grounds, it is interesting to see that figure 2 looks like a normal U-shape.

4.5 Post-hoc analysis.

This section includes three post-hoc analyses. First, given the fact that this research is about R&D alliances, it could be relevant to explore if the R&D expenditures have explanatory power on the firm performance. However, in the sample of 141 firms only 89 firms have R&D data available, for this reason I couldn’t include it in the regression analysis. An additional regression should give an answer if the variable R&D Intensity has any influence as a control variable on the firm performance in the data. I conduct the linear regression model as performed in table 5, model 1 including the new variable. The results suggest that the R&D Intensity has no significant influence in the relative small (β = 0.122, p = 0,671) and medium (β = 0.429, p = 0,257) sized groups. However in the relative large group R&D Intensity has a significant positive relationship on Firm

Performance (β = 0.358, p = 0,037). It suggests that large firms are performing better when they

increase R&D expenditures. When R&D intensity is included in the general model (table 3, model 1) and the sample split in number of partners model (table 4, model 1) this variable gives no significant results.

Second, I test if interaction effects could exist between the independent variables Number

of Partners and Firm’s Relative Size. In order to test this interaction effect an additional analysis

is conducted with the whole sample. This analysis includes the control variables, the independent variables and the interaction term Number of Partners x Relative Firm Size. The result of this analysis suggests that there is no significant interaction effect (β = -0.139, p = 0,792). Therefore I cannot conclude any presence of interaction effects between the independent variables. Third, MR&D alliances can be different across industries, for example High-Tech

22

manufacturing could focus more on new production processes while High-Tech services could have his focus on quality improvements or efficiency improvements, which could influence the firm performance. With the purpose of checking if the results are similar between industries an additional analysis is conducted. The Kruskal-Wallis test with independent samples has mean ranks of 71,74 (Commercial Physical and Biological Research) 68,59 (High-Tech Manufacturing) and 75,48 (High-Tech Services). However, the result has no significant differences between the industry groups (p = 0.789). Concluding that the results of this analysis are consistent across industries.

5. Discussion.

As a result of the globalization of competition MR&D alliances are increasingly important. Where MR&D alliance emerge rapidly in technology driven industries to cope with high levels of risk, short product life cycles and high demands for R&D investments. Previous research suggested that the composition of a MR&D alliance in terms of number of partners, and relative size plays an important role for the outcome of the alliance (Thorgren et al., 2011; Li, 2013; Nieto and Santamaria, 2010; Doz, 2000; Valdes-Llaneza and García-Canal, 2006). However, scholars have found mixed results about the performance implications of alliance composition in terms of number of partners and firm’s relative size. The purpose of this study is therefore to find empirical evidence for the best performing alliance compositions, through a resource-based view and transaction cost perspective. Using data from 141 firms in the Biotechnology, High-Tech manufacturing and High-Tech services industry this paper explores the implications of the focal firm’s relative size and the number of partners in a MR&D alliance on the firm performance. 5.1 Number of partners and firm performance.

Based on previous publications, this study hypothesize that the relationship between the number of partners and firm performance should have an inverted U-shape relationship, however the results suggest that the number of partners is not a predictor for the firm performance in a MR&D alliance. No distinction in firm performance can be made between a 3- ,4- or 5-partner alliance. The results are similar to the research of Beamish and Kachra (2004), they did not find a clear relationship as well between the number of partners and firm’s performance in multi-partner joint ventures. This is contradictory to the research of Park and Russo (1996), and Hennart and Zeng (2002), because these authors found evidence that an increase in the number of partners predicts a negative influence on the performance and the duration of an alliance.

23

additional partner in a multi-partner R&D setting.

Beside the lack of evidence for the transaction cost arguments, there is also a lack of evidence to use the resource based view to explain the relationship between the number of partners and firm performance. An increase in number of partners presumably increases the diversity in resources and complementary capabilities in a MR&D alliance. This can create unique outcomes and gain competitive advantage over other firms. Such a competitive advantage can create extraordinary performances (Doz and Hamel, 1998; Makri et al., 2010; Vanhaverbeke and Noorderhave, 2001). However, the results on the number of partners in this research doesn’t support these theoretical argumentations. The lack of benefits could be due to rising coordination cost and challenging ability to achieve strategic fit (Beamish and Kachra, 2004).

The inconclusive data could give an indication that a trade-off takes place between transaction costs and additional resource benefits, that a firm achieves due to increasing number of partners. The results that the number of partners in a MR&D alliance is not a predictor for firm performance could imply that this trade-off is in balance and shows that firms should focus primarily on the quality of partners and not be derivative from the number of partners.

5.2 Relative firm size and firm performance.

The literature provides large amounts of contradicting argumentation, on the one hand relative smaller firms should benefit from a MR&D alliance; more flexibility, easier to innovate and higher ability to adapt and improve. However, on the other hand relative larger firms should benefit because of more market power, more resources and more bargaining power. The purpose in this research is to seek for evidence that the relationship between firm’s relative size in a MR&D alliance and firm’s performance is inverted U-shaped. The findings in this research suggest that the performance of relative smaller firms is better than relative medium sized and relative larger firms after participation in a MR&D alliance. The trend in the data tends to be U-shaped but the evidence is very weak. Although exploration of linear or curvilinear relationships do not uncover much, it is clear that the relative smaller firms are performing much better than the relative larger firms.

According to Thorgren et al. (2012) smaller firms have always been negatively associated with prospering in alliances with a larger partner, because they have lesser resources to contribute, and have lesser bargaining power. This can cause for alliance instability. Nonetheless, the data is contradicting to these argumentations, smaller firms are precisely those who benefit. It could be the case that a classic Schumpeterian debate about firm size and innovation is playing a large role here, especially in R&D alliance were innovative outcomes are the goal. Nieto and Santamaria (2010) found that the marginal effect of technological collaboration is significantly higher for smaller firms than for larger firms. This difference in effect of technological collaboration could be the reason that smaller firms have better performance because better innovation performance drives firm’s performance (Lahiri and Narayanan, 2013).

24

that routine rigidity could be an underlying problem in an investment. Routine rigidity is an inability to change the logic and patterns of an investment. When the MR&D alliance is seen as an investment smaller firms have lesser problems with routine rigidity than larger firms. Thus, this rigidity could be the reason why some larger firms lack the effectiveness to be really productive in a MR&D alliance, and which consequently could lead to a lower performance in comparison with smaller firms. Especially larger firms with strong technological capabilities are more likely to use their resources for exploitation instead of exploration in a MR&D alliance (Zhou and Wu, 2010). In general, an alliance has smaller implications on the resources for larger firms, because it has relatively less impact on the firm. Larger firms have often many alliance activities, which may also cover some explanation in performance difference.

This paper has some theoretical contributions and contradictions to other researchers. If I assume that there is a slightly U-shaped pattern in the data than this results contradicts the literature on trust in alliances from Thorgren (2012), because they found that the trust between medium sized firms is higher than between larger and/or smaller firms. This contradiction is a bit odd, however it could be that trust doesn’t translate directly to better performance of the firm, but to a longer duration of the alliance (Krishnan et al., 2006). Nevertheless, there are scholars that find the same results as I do, Feldman (1994) suggests that small firms appear to benefit from innovative collaboration because of the presence of external resources and institutions. Moreover, my findings are consistent with Lin et al.’s (2009) paper which discovers that firms with low status, within the alliance, can benefit the most from the resources of the partner.

5.3 Control variables.

This analysis uncovers two interesting relationships in the model with respect to the control variables; the resource diversity predicts a positive relationship on the performance in the relative smaller firms, and joint ventures have a negative relationship on performance in the relative larger firms. Relative small firms in MR&D alliances benefit from a great diversity of resources. Reasons for these benefits could be that smaller firms get acquainted with relatively new technologies and capabilities in a MR&D alliance and thus increase their knowledge fast. Cui and O’Connor (2012) argue that diversity of resources in an alliance gives access to new markets and opportunities. The authors found evidence that diverse resources are only beneficial under the conditions of effective information and resource sharing across the alliance, and the ability to manage them. My findings complement the paper of Cui and O’Connor (2012) with a nuance that resource diversity is probably more beneficial for relative smaller firms.

25

development is very difficult in a multiparty case to implement, because there are fewer incentives to make the relation-specific investments which is needed for trust building. For this reason they suggest that joint ventures are adopted more frequently in a multi-partner alliance, which is not supported by my data (58 JV, 83 SA). The evidence that being in a joint venture is more negatively associated, than being in a strategic alliance, is rather weak, however it raises some questions. 5.4 Managerial implications.

The empirical results in this research have some implications for managers in MR&D alliances. This research has provided some evidence that it does not really matter how many partners are in the MR&D alliance. Therefore, when managers decide to form a R&D alliance they should look beyond the number of partners and make a decision based on the quality of the partner and not the quantity. Managers should weigh if every additional partner provide complementary contributions that are beneficial for the group process. They should not be dissuaded from the risk of decreased performance when adding partners.

Second, according to the results of this paper, managers of smaller firms should be aware that collaboration with experienced larger partners can be very beneficial for the firm due to knowledge and resource sharing. However, larger firms are not always willing to cooperate with smaller firms, because smaller firms have fewer resources to contribute. Therefore managers of smaller firms should find opportunities where their resources can be complementary for the larger incumbents to collaborate.

Third, according to this research, managers should avoid the creation of multi-partner R&D joint ventures, especially relative larger firms. Multi-partner R&D Joint ventures perform less well than multi-partner R&D collaborations in a strategic alliance. Joint ventures are more difficult to exit and it seems to be that larger firms find it more difficult to reap the benefits of these collaborations than smaller firms.

6. Conclusion, limitations and future research.

6.1 Conclusion.

Although a broad literature background is being developed and different scholars are examining the conditions influencing multi-partner R&D alliances, a few took the firm size into account. However, none of them measured the firm’s performance for relative difference in firm size in a MR&D alliance. Literature gave insights about the effect of the number of partners in an alliance, however with contradictory results. This paper examines the implications for firm’s performance in terms of firm’s relative size and the number of partners in a MR&D alliance. In doing so, this paper builds on theoretical arguments that relative larger firms have strong negotiation power, but lack the flexibility to effectively deploy R&D outcomes in the organization, however smaller firms face the opposite problem.

26

firms are benefiting more from participation in a strategic alliance than in a joint venture. Finally, the theoretical arguments on the number of partners explain that on the one hand the alliance can achieve a non-imitable combination of resources with a large number of partners, but on the other hand hinder the coordination of activities and development of relationships. The findings suggest that there is no significant relationship between the number of partners in a MR&D alliance and the firm’s performance. This research helps to explain why the alliance composition is especially important for smaller firms and that managers should focus on quality instead of quantity.

6.2 Limitation and future research.

Although I believe some interesting insights have been gained from this study, this work is not free from limitations. First, the generalizability of the results contains some concerns. The sample is based on MR&D alliances in the Biotechnology, High-Tech manufacturing and High-Tech services including both joint ventures and strategic alliances. Besides the advantage of this sample to test my hypotheses, generalization of the results for other areas, such as multi-partner alliances for marketing or supplier networks, is doubtful. Further the sample is mixed with joint ventures and strategic alliances which could be, according to theory, a critical distinction (Beamish and Kachra, 2004). Future research with more extensive data could analysis if all industries face the same results. In addition, a larger dataset opens the possibility to make more and better additional distinctions and explore the differences between joint ventures and strategic alliances. Similar studies in different settings could set the boundaries for my conceptualization.

27

28

References

Aggarwal, N. Dai, Q., and Walden, E.A. 2011. The more, the merrier? How the number of partners in a standard-setting initiative affects shareholder’s risk and return. MIS Quarterly, 35(2): 445-462.

Ahuja, G. 2000. Collaboration network, structural holes, and innovation: A longitudinal study. Administrative Science Quarterly, 45(3): 425-455.

Beamish, P.W., and Kachra A. 2004. Number of partners and JV performance. Journal of World Business, 39:107-202.

Chen, H., and Chen, T. 2003. Governance structures in strategic alliances: transaction cost versus resource-based perspective. Journal of World Business, 38(1): 1-14.

Cohen, W.M., and Levinthal. 1990. Absorptive Capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35(1): 128-152.

Cui, A.S., and O’Connor, G. 2012. Alliance portfolio resource diversity and firm innovation. Journal of Marketing, 76(4): 24-43.

Das, T.K., and Teng, B. 1998. Between trust and control: developing confidence in partner cooperation in alliances. Academy of Management Review, 23(3): 491-512.

Das, T.K., and Teng, B. 2002. Alliance constellations: A social exchange perspective. Academy of Management Review, 27(3): 445-456.

Dhaliwal, D., Subramanyam, K.R., and Trezevant, R. 1999. Is comprehensive income superior to net income as a measure of firm performance. Journal of Accounting and Economics, 26: 43-67.

Doz, Y.L., and Hamel, G. 1998. Alliance advantage. Boston, MA: Harvard Business School Press.

Doz, Y.L., Olk, P.M., and Ring, P.S. 2000. Formation processes of R&D consortia: Which path to take? Where does it lead? Strategic Management Journal, 21(3): 239-266.

Dyer, J.H., and Chu, W. 2000. The determinants of trust in supplier–automaker relationships in the US, Japan and Korea. Journal of International Business Studies, 31(2): 259–285. Dyer, J.H., and Singh, H. 1998. The relational view: cooperative strategy and sources of

interorganizational competitive advantage. Academy of Management Review, 23(4): 660-679.

29

Faems, D., De Visser, M., Andries, P., and Van Looy, B. M. 2010. Technology alliance portfolios and financial performance: Value-enhancing and cost-increasing effects of open innovation. Journal of Product Innovation, 27(6): 785-796.

Feldman, M.P. 1994. Knowledge complementary and Innovation. Small Business Economics, 4: 363-372.

Fisher, R. 1925. Applications of Student’s distribution. Metron, 590-124.

García-Canal, E. Valdés-Llaneze, A., and Ariño. A. 2003. Effectiveness of dyadic and multi-Party joint ventures. Organization Studies, 24(5): 734-770.

Gilbert, C.G. 2005. Unbundling the structure of inertia: Resource versus routine rigidity. Academy of Management Journal, 48(5): 741-763.

Goerzen, A., and Beamish, P.W. 2005. The effect of alliance network diversity on multinational enterprise performance. Strategic Management Journal, 26(4): 333-354.

Gong, Y., Shenkar, O., Luo, Y., and Nyaw, M. 2007. Do multiple parents help or hinder international joint venture performance? The mediating roles of contract completeness and partner cooperation. Strategic Management Journal, 28(10): 1021-1034.

Hagedoorn, J. 2002. Inter-firm R&D partnerships: an overview of major trends and patterns since 1960. Research policy, 31: 477-492.

Hannan, M.T. and Freeman, J. 1977. The population ecology of organizations. American Journal of Sociology, 82(5): 929-964.

Hennart, J., and Zeng, M. 2002. Cross-cultural differences and joint venture longevity. Journal of International Business Studies, 33(4): 699-716.

Hitt, M.A., Hoskisson, R.E., and Kim, H. 1997. International diversification: effects on

innovation and firm performance in product-diversified firms. Academy of Management Journal, 40(4): 767-798.

Hwang, P., and Burgers, W.P. 1997. The many faces of multipartner alliances: Lessons for managers. California Management Review, 39(3): 101-117.

Inkpen, A., and Beamish, P.W. Knowledge bargaining power and the instability of international joint ventures. Academy of Management Review, 22: 177-202.

Jiang, R.J., Qingjiu, T.T., and Santora, M.D. 2010. Alliance portfolio diversity and firm performance. Strategic Management Journal, 31:1136-1144.