Empirical and qualitative analysis of conflict in inter-organizational relationships

(1)

Characteristics of partners and alliances shaping the likelihood and size of inter-organizational conflict

Master Thesis

MSc Business Administration – Strategic Innovation Management

by

Dion van den Bosch S2173700 Commissarislaan 91

8016 LL Zwolle dionvdbosch@gmail.com

26th_{August, 2016}

Supervisor: Dr. Isabel Estrada Second Supervisor: Dr. Aneta Oleksiak

(2)

ABSTRACT

This study examines why firms face inter-organizational conflicts in general and between alliance partners in the biotechnology industry. Different characteristics between firms but also between alliance partners increase the likelihood of conflict. Previous research showed that inter-organizational conflict has several antecedents and poses different challenges for firms. However, the field of inter-organizational conflict lacks empirical evidence and additionally it is also in need for detailed descriptions of inter-organizational conflict cases. This research consists of two parts which will explore empirically and qualitatively in more detail which characteristics of firms and of alliance partners increase the likelihood but also the size of the conflict. The first part explains the characteristics and likelihood of conflict in general and is tested with a sample of 86 cases of conflict. This part shows that only industry diversity is a characteristic that determine the likelihood of conflict. The second part of the research consisted of a sample of 43 cases of conflict between alliance partners. The analysis resulted in the fact that the later the timing of the conflict (so after the termination of the alliance), the size of the conflict increased significantly. Additionally, a qualitative part in which extraordinary cases are explored in more detail is provided in this part of the research. The results of this paper are influenced by limitations like a small samples sizes and industry specificity. Furthermore, this paper puts forward important directions of future research in the field of inter-organizational conflict.

(3)

During the last decades it has become more and more important to be innovative and survive by the help of complex relationships with external partners. These relationships often encompass a strategic alliance with competitors. With the growing numbers of patent applications and granted patents, the number of patent litigations have more than doubled over the last decade (Barry, Arad, Ansell, & Clark, 2014). Inter-organizational conflict comes in many forms (Connelly, 2007) including the incongruence of interests (Hardy & Phillips, 1998) and disputes in knowledge management (Tan, Pan, Lim, Chan, & Al-Hawamdeh, 2005). Therefore, managing conflicts becomes an increasingly important topic in the literature field of alliances (Booth & Wang, 2012). Conflicts have a very large impact on the company and should be minimized. As these conflicts are inevitable and often result in patents infringements, the issue of inter-organizational conflict is becoming more and more apparent. These recent developments in the business world makes it interesting to look further into these conflicts and establish how these conflicts arise.

Within the literature of conflict several definitions are formulated. In the light of previous works in the research field of conflict, I will determine the definition to define inter-organizational conflict (Deutsch, 1973; Lumineau, Eckerd, & Handley, 2015; Rex, 1981). I will define inter-organizational conflict as the misalignment of goals, activities, characteristics and outcomes but also the tension of knowledge misappropriation between the different organizations of an alliance.

However conflict is not a phenomenon that just occurs. In this paper I will explore the different characteristics that determine the likelihood of conflict in general and, additionally, explore which characteristics of alliances and the timing of the conflict determine the size of conflict. Within the theory of inter-organizational conflict there is not much discussion about the relationship between different characteristics of alliances, size of conflict and the likelihood of conflict and this is therefore an opportunity to investigate (Lumineau et al., 2015). Moreover, existing literature like Sampson (2007) investigates alliance performance based on firm level constructs. By using alliance level constructs and qualitative cases to explore inter-organizational conflict, this research is addressing an important gap within the literature. Existing literature was characterized by literature reviews and theories based upon several literature fields, but lacks the empirical and in-dept qualitative analysis of inter-organizational conflict. Therefore the goal of this study is to investigate what characteristics of the alliance increase the likelihood of conflict in general and which characteristics increase the size of conflict between alliance partners, either before, during or after the termination.

(5)

between firms differ for conflict in general. Additionally, a more important question can be formulated about which characteristics of alliances increase the size and likelihood of conflict? Alliance performance is often measured by patents, which give a good indication of the success of the alliance. As inter-organizational conflict will not mean that the alliance has failed, patent litigations could give a good indication of how the conflict arose. Therefor patent litigations can be a good way to indicate conflict as litigations provide information on the conflicting elements between the partners.

In order to answer these research questions, I aim to do a database research based on alliance information from the SDC database and patent litigation information from the Patexia database. I will base my research on bio-technology alliances as these alliances apply for a lot of patents and therefore the chances of patent litigation are the highest. Furthermore, the biotechnology industry consists of many laboratories and firms which increases the variety in the sizes of conflict. The analysis consists of two parts in which the first part is answering the first research questions by using an independent sample t-test to compare the samples of having an alliance and having no alliance. This is concluded by a logit regression to check which characteristics have a significant contribution to the dependent variable. The second part of the analysis explores the different characteristics between firms that had an alliance before, during or after the inter-organizational conflict and analyze the database based on a simple hierarchical linear regression analysis. An additional analysis is done based upon crosstabs and an in-dept analysis of specific cases which increase the implacability for managers.

The results of the research show that only industry diversity is a characteristic that has a significant contribution to the prediction of inter-organizational conflict in general. This shows that firms should take into account the industry diversity when choosing a partner, because lower diversity leads to a higher likelihood of conflict due to easier assimilation of the industry specific knowledge. Furthermore, analysis showed that the timing of the conflict significantly increases the size of the conflict. As partners collaborate longer with each other, the amount of knowledge that is shared increases, which will increase the size of the conflict. This was also shown in the linear regression analysis which indicated a significant positive effect of the timing of the conflict and the size of the conflict. The last result is further strengthened when analyzing the crosstabs which shows that for every characteristic that is tested in the research, the majority of the conflicts happen after the alliance.

(6)

I will contribute to the theoretical field as follows, firstly I will look at the differences between firms that had alliance experience in combination with a conflict and firms without this alliance experience but still were involved in a conflict. This contributes to the literature field as it indicates the different characteristics firms possess with alliances or without alliances and their likelihood of conflict. Furthermore, I will explore in more detail the cases with alliances which will indicate what characteristics are most valuable in decreases the likelihood of conflict and when firms need to be “on their toes” to expect conflict in the form of patent litigations. These basic contributions will construct a foundation for further research.

Managerial interest will come in the way that managers could see what combinations of characteristics work and which combinations will are problematic. This will save companies and organizations time and money (Crocker & Masten, 1988) when searching for an appropriate alliance partner. Additionally, the exploration of the specific cases within the sample provides managers with real-life examples of how inter-organizational conflict can manifest itself. Furthermore, these cases show managers were to look for in a strategic alliance and what to watch out for to minimize the chance of inter-organizational conflict.

(7)

2. LITERATURE REVIEW

The emergence of global competitors and the increasing speed and cost of technological development provide an increasing uncertain environment for firms (Hagedoorn & Schakenraad, 1992). These dynamic and changing markets call for organizations to be innovative, efficient and flexible (Duncan, 1976). These statements hold particularly well in the fast-paced industries or hypercompetitive markets which are characterized by fast technological change, short product life-cycles, and global competition and intense rivalry (Volberda, 1996). Due to the greater environmental uncertainty, firms want to avoid long entanglements, like merger and acquisitions, which could be detrimental in the long run. Instead, firms rather favor less binding and more flexible relationships (Crocker & Masten, 1988), like inter-organizational relationships. An inter-organizational relationship is defined as a formal cooperative agreement between two or more firms. Examples of these agreements are strategic alliances. Strategic alliances provide the firm with a greater ability to pursue new developments within the market and it allows it to initiate or adapt to competitive change (Volberda, 1996). These relationships are essential in overcoming difficulties in the market and can improve competitiveness. This is the reason that the amount of inter-organizational relationships is growing but this is not without consequences as shown in the next paragraphs.

2.1 Inter-organizational conflict

(8)

Inter-organizational conflict comes in many forms. The most common used conflict type is competence-based versus integrity-based failures whereas the competence-based failures relate to the skills and knowledge of the partner firm, while for integrity-based failures, inherent behaviors and norms of the partner firm are questioned (Lumineau et al., 2015). Next to this distinction of conflict, a few other forms of inter-organizational conflict can be identified. For example, Frazier and Rody (1991) distinguish between latent and manifest conflict in which the first addresses the ‘underlying state of incompatibility between two firms’ while the latter involves overt behaviors that keep the partner from achieving his goals (Frazier & Rody, 1991). Another distinction which is equally common is made between constructive and destructive conflicts (Deutsch, 1973; Dwyer, Schurr, & Oh, 1987; Hibbard, Kumar, & Stern, 2001), and functional and dysfunctional conflict (Koza & Dant, 2007; Rose & Shoham, 2004).

The size of conflict is also an important aspect of inter-organizational conflict. As the size of the conflict increases, the risks and impact of the conflict increases as well, which leads to negative consequences in respect of profit and reputation. However, conflict is not necessarily evil. Recent research shows that conflict is natural and an inevitable outcome in any organization which potentially can pose a positive force in the contribution of alliance performance (Chen, 2004). But this does not mean that all levels of conflict are beneficial. Robbins and Decenzo (2001) propose that managers need to maintain an optimal level of conflict in order to keep their organizations self-critical and creative (Robbins & DeCenzo, 2001; Yu, 1998). Therefore it is good to look at what characteristics increase the size of conflict in alliances and not just whether or not the conflict happens.

As discussed above, conflict can be characterized by very different forms and sizes. In this research I will focus on the different characteristics of an alliance like the numbers of partners, the national diversity of the partners but also industry diversity and governance form of the alliance. Furthermore, I will use whether the firms are direct competitors as a construct. Additionally, the timing of the conflict, so before, during or after the alliance will be used in this research as well. I will use these six different characteristics of firms and their inter-organizational relationship based on previous findings within the literature field of inter-organizational conflict. These characteristics are relevant for the research of inter-organizational conflict as these characteristics are widely used in the literature as constructs for performance. (Das & Teng, 2002; Hagedoorn & Schakenraad, 1994; Jiang, Tao, & Santoro, 2010; Sampson, 2007). Performance is closely linked with conflict and therefore these characteristics are capable of explaining issues related to inter-organizational conflict.

(9)

general. This section will end with a conceptual model based upon these three hypotheses and summarizing the reasoning of how these characteristics relate to the likelihood of inter-organizational conflict. The second section will be exploring which characteristics relate to the size of conflict. The hypotheses proposed in this section will further explain, once a conflict is present, how certain characteristics increase or decrease the size of conflict.

2.2 Partner characteristics and the likelihood of inter-organizational conflict

In this first section, three characteristics at the firm level will be discussed and explored how they relate to the likelihood of conflict. These characteristics are based upon diversity between the alliance partners or between the firms that had a conflict between them without having an alliance.

National diversity of the partners

In the research of Hagedoorn and Schakenraad (1994) they found that the characteristics of the alliance partners are more important than the absolute number of alliances (Hagedoorn & Schakenraad, 1994). In this way diversity can be a big issue when choosing your partner. In the last years of globalization, more and more firms look for alliances from different nation states. These cross-border alliance have several advantages like the facilitation of market entry (Glaister & Buckley, 1996), it could provide complementary capabilities (Lane, Salk, & Lyles, 2001) and enhance different knowledge bases (Lubatkin, Florin, & Lane, 2001) but it can also pose a high potential of conflicts (Jiang et al., 2010). In this research I will focus on the national diversity between the different partners. National diversity will be defined as the difference in nationality between the firms within the relationship. Jiang et al. (2010) state that due to the different macro-level factors (political and economic systems, societal and cultural institutions, government policies and national industry structure), the corporate culture, strategic direction, and management practices there is a higher chance for conflicts and an increase alliance costs. Furthermore, different communication patterns between two countries lead to difficulties in the interaction and acculturation process within the alliance (Tung, 1993). Due to these challenges, managing and maintaining alliances with national diverse partners is rather difficult which will in turn increase the likelihood of conflict between the partners. The impact of the challenges concerned with cross-border alliances is larger than the benefits so this would mean that national diversity has a positive effect on the likelihood of conflict. This results in the following hypothesis:

H1: There is a positive relationship between the national diversity of the partners and the

(10)

Industry diversity

Not only do firms look for alliances from other nation states, they are also looking for partner firms from other industries. Industry diversity will be defined in this research as the difference between the industries the partners operate in. Partners from the same industry, achieve the greatest learning through imitation and greater absorptive capacity. This is because they have an overlap in background, experiences, knowledge and technological bases (Cohen & Levinthal, 1990). But partners from the same industry also have incentives to misappropriate knowledge and use it for themselves which negatively influences the cooperative behavior between the firms. When looking at partners from different industries, other issues arise. These issues reside in the misalignment of processes and routines that can make collaboration difficult (Jiang et al., 2010). Again, partners from other industries do not have the incentives to use the knowledge obtained directly in their industry as most knowledge that could be misappropriated is not applicable in their industry. Along this reasoning, I propose that having more diversity, so partners from other industries, does not necessarily mean that there is more conflict. The likelihood of conflict between partners from the same industry is higher because the knowledge obtained from the partner can be directly used in the industry and negatively affecting the innovating firm’s performance. Therefore I hypothesize that:

H2: There is a negative relationship between the industry diversity and the likelihood of

inter-organizational conflict. Competition

(11)

H3: There is a positive relationship between the degree of competition between the firms and

the likelihood of inter-organizational conflict.

2.2.1 Conceptual model on likelihood of conflict

According the hypotheses discussed above, a conceptual model can be created and is illustrated below. This conceptual model describes the interaction between the characteristics of the involved firms and inter-organizational conflict in general. Furthermore, the proposed relationships are illustrated between the dependent and independent variables.

2.3 Characteristics of alliances and the size of inter-organizational conflict

For the second part in this research, I will use specific alliance level constructs in order to test which characteristics of alliances affect the size of the conflict and explore when the conflict is happening based on the timing before, during or after the alliance. This part is an extension upon the first as this will further explore alliance characteristics while the first part is a more general research on inter-organizational conflict.

Multi-partner alliances

(12)

the alliance (Beamish & Kachra, 2004; Doz & Hamel, 1998; Hwang & Burgers, 1997). On the other hand, increasing the number of partners increases the managerial complexity and coordination costs (Garcı ́a-Canal, Valdés-Llaneza, & Ariño, 2003; Gong, Shenkar, Luo, & Nyaw, 2007; Hennart & Zeng, 2002; Hu & Chen, 1996; Park & Russo, 1996; van de Ven, 1976). Due to the high complexity and managerial costs, the size of the conflict will be higher in the case of a multi-partner alliance than for a dyadic alliance. Garcia-Canal et al (2003) state that in a multi-partner alliance “there are fewer incentives to behave cooperatively . . . [and] the incentives for free-riding behavior are greater when partners are more numerous” (Garcı ́a-Canal et al., 2003). This means that in a multi-partner alliance the incentives for free riding are larger than for dyadic alliances. Additionally, in a multi-partner alliance, the chance of conflicting interests will be larger as more companies are involved. This leads to larger conflicts as the larger the amount conflicting interests and incentives for free riding, the harder it gets to solve the conflict, thus increasing the size of the conflict. The size of the conflict is restricted in dyadic alliances due to the fact that the alliance consists of only two parties and therefore the conflicting interests are restricted to only two parties. Furthermore, the incentives for free riding are smaller as in dyadic alliances the control over the partner firm is larger. Therefor I will formulate the following hypothesis. H4: The size of inter-organizational conflict will be bigger in the case of a multi-partner alliance than for dyadic alliances.

Governance form

Besides the number of partners, alliances can take very different forms like joint ventures, bilateral contracts or licensing agreements. In this research the focus will be two different forms, namely equity joint ventures and bilateral contracts. A bilateral contract, according to Sampson (2007) is “a contractual arrangement in which partners pool their capabilities for the purposes of collaborative R&D but do not form a separate legal identity for the alliance” (Sampson, 2007). An equity joint venture is defined by several authors as “the pooling of capabilities under a newly created entity that is jointly owned and operated by two or more collaborating firms” (Oxley, 1997; Pisano, Russo, & Teece, 1988). Sampson (2007) argues that equity joint ventures facilitates communication and coordination (Sampson, 2007). On the other hand, management under bilateral contracts is based on independent decision making in the best interest of the alliance. This means that management under bilateral contracts remain nimble and result in more timely decision making in comparison with equity joint ventures.

(13)

ventures and bilateral contracts, the impact of these issues will be larger in bilateral contracts as the protection by coordination and communication is less than for joint ventures.

Furthermore, the size of the conflict is increased when the incentives for opportunism increases (Garcı ́a-Canal et al., 2003). This means that, the larger the pool of knowledge to misappropriate, the larger the incentives for opportunisms which increases the size of the conflict. Therefore, as the incentives of opportunism are higher in the case of bilateral contracts, so will be the chance of a bigger conflict. This is the opposite for joint ventures as the communication and coordination advantages decreases the chance for opportunism, reducing the size of conflict (Pangarkar, 2003). Therefore, I propose that alliances governed by a joint venture due to the communication and coordination advantages are less likely to end up in large conflicts. In this line of reasoning the following hypothesis is proposed:

H5: The size of inter-organizational conflict will be smaller in the case of an equity joint venture than for alliances managed by a bilateral contract.

Timing of conflict

(14)

H6: The size of conflict is bigger for conflicts that happen after the alliance than before or during the alliance due to the degree of knowledge sharing.

2.3.1 Conceptual model on size of conflict

To summarize the reasoning of the hypotheses in an illustration, the following conceptual model can be created. This model shows the interaction pattern between the alliance characteristics and the size of inter-organizational conflict. Additionally, the model shows the proposed directions of the relationships, so whether the relationship between the independent and dependent variable is positive or negative.

3. METHODOLOGY

This paragraph will elaborate how the research is set up and in what way the hypotheses are tested. This part will consist of three parts namely the data collection, measurements and data analysis. The first will deal with how the data is collected and organized to reach a sample that can be tested. In the measurements section the constructs are described in respect of how they are measured throughout the research. The last part consists of an elaboration which analysis techniques are used to test the proposed hypotheses.

3.1 Data Collection

(15)

collects data from the U.S. Securities and Exchange Commission (SEC) filings (and their international counterparts), trade publications, wires, and news sources. A key advantage of this database is extensive search ability. SDC offers over 200 data elements including the name, SIC code, and nationality of participants, the terms of the deal, and deal synopsis for each alliance agreement (Schilling, 2009). As the SDC database is sometimes missing information on competitors or size of the company I used the LexisNexis database to complete the data I retrieved from SDC. Additionally, Patexia’s patent litigation database is used to combine the information from the SDC and patent litigations to link alliances with inter-organizational conflict. Initially 1.313 patent litigations are analyzed in the search of a combination between litigation and alliances. These litigations were filed by the 18 biggest biotechnology firms according to Forbes ranking of public firms (Forbes, 2015). I narrowed my search in the Patexia Database as the data on patent litigation was most accurate as of 2005. Therefor the choice was made to take a time span of 10 years starting from 1st_{of January 2005}

and ending 31st_{of December 2015. After this selection and forming the litigation database for these}

companies, the alliance database was created for these 18 companies. For SDC a different time span was used as I wanted to know specifically when the litigation was filled, meaning before, during or after the alliance. Therefor a time span starting from 2001 and ending in 2012 was used. In order to reach conclusions I need another sample of patent litigations in order to compare the combinations to this sample to reach conclusions based on how these two samples differ. I will collect this sample by randomly selecting a comparable sample size out of the other 1272 litigations in Excel. After combining the two different samples, so the one with alliances and the one without alliance, the sample resulted in 86 cases in which further analysis is conducted, 43 cases with an alliance and 43 without an alliance. This sample will be used to test the first three hypotheses of this research. In order to test the last three hypotheses, the sub-sample of 43 cases that involved some sort of alliance are further investigated to see which characteristics are increasing the size of conflict.

3.2 Measurements

To effectively measure the constructs used in this research existing literature is used in order to make sure that the measurements of the constructs are valuable and validated. This makes sure that the measured constructs are measured in the right way and lead to interpretable results.

Inter-organizational conflict (Alliance/No alliance)

(16)

the firms have had an alliance, either before, during or after the time of conflict. Furthermore, I will code items with a “0” when there were no previous ties between the firms. In this way I can conclude when conflict is more likely to happen in general and I can compare the two different samples. In order to effectively measure the size of conflict for the other three hypotheses, I will make use of patent litigations. Patent litigations are a good indication for conflict as firms can only litigate once they have the knowledge of this patent. As patents disclose the information concerning the discovery, other firms may want to profit from this (Nunnally, Webster, Brown, & Cohen, 2005). As the biotechnology industry is vulnerable to free riders, patent litigations indicate free riding and therefor potential conflict (Grabowski, 2002). In order to use this construct in the statistical part of the research, I will look at the size of the conflict as well. I will use the size of conflict by the number of departments or different companies involved in the patent litigation case. These departments could be laboratories of the companies involved, or firms that are a part of the parent firm after merger or acquisitions.

National diversity

As we want to see how diverse the inter-organizational relationship is, I will use the description of the construct based on previous research. I will code the national diversity in the following way: a “0” for an alliance with no foreign partners, “1” for an alliance with partners from another country (Jiang et al., 2010; Sampson, 2005). By using the information from Jiang et al. (2010) and validating this scale with the article of Sampson (2005) I can confidently use this measure of the construct national diversity.

Industry diversity

(17)

Competitors

The degree of competition was found using the LexisNexis database which indicates who the main competitors are of the focal firm. When direct competitors were indicated, I coded the combination of the firms with a “1” for direct competition and a “0” for firms that are not direct competitors, irrespectively of the industry they are operating. By using this measure, the distinction between direct competitors or not will indicate whether direct competitors have an increasing likelihood of conflict.

Multi-partner alliance

In the paper of Li et al. (2012) they measure the construct of number of partners and I will use the same method for this construct. If the alliance is based on a dyadic relationship, I will code the variable as “0”. If the alliance is formed as a multi-partner alliance I will code it “1” (Li, Eden, Hitt, Ireland, & Garrett, 2012; Sampson, 2005). In this way I differentiate whether the alliance consists of two partners (dyadic) or 3 or more partners (multi-partner).

Governance form

Organizational diversity will be measured by using the information in the SDC database and according to the construct provided by previous research. I will create a dummy variable to capture the governance form (Heimeriks & Duysters, 2007; Sampson, 2007). I will code the variable with a “1” in the case that the alliance is a joint venture and a “0” when it is organized otherwise. In the case of joint ventures, protection and mutual collaboration is higher and should therefor relate to less conflict.

Timing of conflict

The timing of conflict variable will be a dummy variable based upon the time of the conflict. Due to the configuration of the sample I decided to code the variable of timing of conflict as follows: I will code this variable with a “0” when the conflict happens before or during the alliance and a “1” when the conflict happens after the alliance. This distinction is made due to the nature of the sample in which only 2 cases of conflict were during the alliance, these cases will be further explored qualitatively.

Control variables

(18)

conflict. The second control variable in this case is firm age. This is measured by the log of the amount of years between founding and the patent litigation case (Shu et al., 2014). Again, using the paper of Baum et al. (2000), the liability of newness is applied. The newer the firm is in the industry, the less experience it has and is therefore more susceptible for patent litigations than experienced firms. The last and final control variable is the patent experience of the firm. This will be measured by the log of the amount of patents acquired between the start of the study period and the moment of litigation (Al-Laham, Amburgey, & Bates, 2008; Trajtenberg, 1990). The amount of patents relates to the experience the particular firms has in the filing and acquiring of patents. Furthermore it shows how capable the company is in filing patents and whether it possesses the sufficient resources to use patents as a protection method. Also the likelihood of litigating the patent is related to the amount of patents owned by the firm at that moment which is the reason why I control for this in this case. To control for the hypotheses based on the size of conflict, the following control variables based upon alliance level constructs are used. The first control variable will be the total number of alliances formed during the period of study. This shows how active the search for alliances is and could therefor influence the likelihood of conflict (Al-Laham et al., 2008; Sampson, 2005). The second control variable is firm size which is measured the same way as for the test concerning the first three hypotheses. Furthermore, I propose that the scope of the alliance has an influence on the likelihood of conflict as the more activities pursued in the alliance leads to more coordination which in turn could lead to more problems and confusion and therefor influencing the likelihood of conflict (Di Guardo & Harrigan, 2016; Shu et al., 2014). Lastly, I will use the construct of experience in inter-organizational relationships as a control variable. I will measure this construct by the number of previous alliances before the time of conflict. Research has suggested that the number of alliances that the firm has entered in can be used as proxy for alliance experience (Heimeriks & Duysters, 2007; Schilke & Goerzen, 2010; Tzabbar, Aharonson, & Amburgey, 2013).

3.3 Analysis

(19)

t-test (Li, 2013). Additionally, a hierarchical logit regression analysis is used to find out which of the variables have a significant influence on the likelihood of a conflict despite having an alliance or not. This will be done in a comparable way as used in the article of Estrada et al. (2010) by the use of simple, reduced and full models. This means testing each predictor individually in simple models, then adding the control variables to the predictors to obtain the reduced models. The final conclusions regarding the hypotheses will be made by the use of the full models in which all the variables are entered at once.

The next step of the research is to further investigate the 43 cases in which the combination was found between litigation and alliance. Again based on the article of Estrada et al. (2010), a linear hierarchical regression analysis will be done to determine the quality of the predictors. Comparable to the first part of this research, the analysis in done in three steps: the simple models, the reduced models and lastly the full models. As the predictors in this research are chosen based on theory and past research, a hierarchical linear regression analysis is more suitable (M. Lewis, 2007). Lewis (2007) also states inhis paper that, in social sciences, this approach is more appropriate than the stepwise regression as the hierarchical regression can cope better with high correlation between the predictors. Therefor I will first test the model with the control variables by adding them one by one. Next is adding the predictor variables one by one to the control variables. Finally adding all the variables at once to obtain evidence in order to reach validated conclusions. After the empirical test I will further explore these 43 cases with the use of crosstabs. This will provide more information on how the sample is divided which characteristics are most prevalent in this research. Next to the crosstabs, I will further explore five individual cases that were extraordinary and therefor valuable to explore in more detail.

4. RESULTS 4.1 Results (Full-Sample)

(20)

4.1.1 Descriptive statistics and Correlation matrix (Full-Sample)

The table below shows the descriptive statistics of the complete dataset of the two samples. One striking result is the fact that “industry diversity” correlates significantly with all the other variables except for the variable “competitors”. Furthermore, “firm size” correlates with “national diversity” and “patent experience”. Next to these correlations, the dependent variable correlates significantly with “patent experience”, “national diversity” and “industry diversity”. This is something to take into consideration when analyzing the result of the t-test.

4.1.2 Multicollinearity (Full-Sample)

To ensure that problems concerning multicollinearity and therefor problems concerning the predictive relationship between the variables, the variance inflation factors (VIF) levels during all the first analyses will be examined. In the literature, several recommendations of acceptable levels are provided. The most common maximum value is 10 (Kennedy, 1985; Neter, Wasserman, & Kutner, 1989), however sometimes a maximum value of 5 is used (Rogerson, 2001). In this first analysis, the VIF values did not exceed 3 so this means that there are no problems concerning multicollinearity and the predictive value of the variables. The individual VIF values are provided table 3 and 4 in order to assess the individual models on multicollinearity issues. When several variables are tested in one model, the highest VIF value is reported. All the other VIF values can be found in appendix 1. Only models 7 until 14 are reported as these models involve multiple variables and therefor the VIF value is not equal to 1 due to interaction between the predictors.

4.1.3 Hypotheses tests H1-H3

(21)

Table 1

Descriptive statistics + Correlation matrix for explanatory variables

Note: Coefficients in bold and marked with * and ** are significant at 95% and 99% confidence level, respectively.

N Minimum Maximum Mean, % Std. Deviation 1 2 3 4 5 6 7

(22)

Table 2

Comparison of matching samples

Group Statistics Levene's Test for

Equality of Variances

t-test for Equality of Means Alliance N Mean Std. Deviati on Std. Error Mean F Sig. t Sig. (2-tailed) National diversity 1,0 43 ,884 ,3313 ,0517 43,956 ,000 -3,093 ,003*** ,0 43 ,605 ,4947 ,0754 Industry diversity 1,0 43 3,721 ,7824 ,1222 200,735 ,000 -6,275 ,000*** ,0 43 1,674 1,9967 ,3045 Competitors 1,0 43 ,302 ,4711 ,0736 ,905 ,344 -,476 ,635 ,0 43 ,256 ,4415 ,0673 Firm size 1,0 43 4,6307 ,70002 ,10933 ,928 ,338 -1,684 ,096* ,0 43 4,3344 ,88373 ,13477 Firm age 1,0 43 1,5412 ,42642 ,06503 3,843 ,053 ,065 ,948 ,0 43 1,5484 ,58462 ,08915 Patent experience 1,0 _,0 43 ₄₃ 2,9005 _2,5702 ,62191 _,66558 ,09484 _,10150 ,161 ,690 -2,377 ,020**

Note: Coefficient is bold and marked with ***, ** and * are significant at 99%, 95% and 90% confidence level, respectively.

(23)

mean difference is 0.007 which is almost the same and is therefore not significant at both levels. The last variable of “patent experience” is significant at a 0.05 level with a mean difference of 0.3303. The last step in this analysis is checking which of these variables actually have a significant influence on the likelihood of conflict. This is done by doing a binary logistic regression analysis based on the sample of 86 cases. According to table 1 “industry diversity” significantly correlates with all the other variables except with “competitors”. Additionally, “firm size” correlates significantly with “patent experience” and “national diversity” and the dependent variable also correlates significantly with “patent experience”, “national diversity” and “industry diversity”. This influences the outcomes of the regression analysis.That is why I choose to start with a “simple” regression model based on the paper of Estrada et al. (2010) in which merely the individual effects of the explanatory variables are investigated and their influence on the likelihood of conflict (Estrada, de la Fuente, & Martín-Cruz, 2010). Additionally, the reduced and, especially, the full models will be used to infer the relationships between the predictors and the dependent variable and are most important for reaching conclusions about the hypotheses. Below you will find the outcomes of the regression analysis.

In the first six models, the “simple” regression models are illustrated. The individual contribution of the variables is tested here and these outcomes are not enough to accept or reject hypotheses. These models suggest that “national diversity” (chi square = 9,180, p < .01 with df = 1) and “industry diversity” (chi square = 30,869, p < .01 with df = 1) have a significant influence in the prediction of the dependent variable. The Wald criterion shows that “national diversity” made a significant contribution to the prediction of the dependent variable (p = .005). The Wald statistic for “industry diversity” demonstrates a significant contribution of the prediction of the dependent variable (p = .003).

As shown in the table, “competitors” does not have a significant influence on the prediction of the dependent variable (chi square = 0.231, p > .01 with df = 1).

For the three control variables, the fourth model wherein “firm size” was tested not significant and is therefore unreliable in the prediction of the dependent variable (chi square = 2,934, p > .01 with df = 1). It is the same case for the fifth model where “firm age” was tested not significant, which results in unreliable predictions of the dependent variable (chi square = ,004, p > .01 with df = 1). Unlike the other control variables, “patent experience” is tested significantly and therefor individually contributes to the prediction of the dependent variable (chi square = 5,651, p < .05 with df = 1). The Wald criterion shows that “patent experience”, so the amount of patents acquired during the study period, contributes significantly to the dependent variable of having an alliance or not (p = .026).

(24)

Table 3

Estimates of binary logistic models (simple, reduced and full models)

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) National diversity 0,569*** (4,969) ,652*** (5,690) ,865 (2,880) Industry diversity ,189*** (2,268) ,278*** (2,931) ,278*** (2,710) Competitors ,482 (1,261) ,500 (1,165) ,641 (1,514) Firm size ,299 (1,618) ,440 (,932) ,477 (,598) ,658* (,328) ,442 (,928) ,693* (,268) Firm age ,427 (,972) ,438 (1,017) ,496 (1,286) ,639 (2,665) ,439 (1,025) ,670 (2,939) Patent experience ,373** (2,297) ,573 (2,463) ,660* (3,472) ,811** (5,713) ,573 (2,450) ,867** (6,267) Log Likelihood -55,021 -38,611 -59,495 -58,144 -59,609 -56,785 -56,771 -52,543 -37,559 -56,725 -39,553 N 86 86 86 86 86 86 86 86 86 86 86 Wald Statistic 7,943 13,624 0,231 2,598 ,004 4,970 2,476 7,105 14,916 0,093 4,616 VIF value 1,000 1,000 1,000 1,000 1,000 1,000 2,318 2,483 2,412 2,320 2,530

(25)

In model 8, the predictor variable “national diversity” is added to the control variables to get a first indication of whether “national diversity” has a significant predictive power of the dependent variable. As shown in the table, “national diversity” is able to predict a significant distinction between having an alliance or not (chi square = 14.135, p < .01 with df = 4). The Wald criterion shows that “national diversity” made a significant contribution to the prediction of the dependent variable (p = .008). This shows that “national diversity” indeed has a positive relationship with the dependent variable, only this model is only an indication. Furthermore, “patent experience” shows a significant contribution which indicates that “national diversity” has some influence on “patent experience”. This could relate to multicollinearity issues but the VIF value is lower than the proposed threshold so no problems should exist when interpreting the outcomes of this model.

In model 9, the predictive power of “industry diversity” was tested and the Wald criterion shows a positive significant contribution to the prediction of the dependent variable (chi square = 38,371, p < .01 with df = 4).This indicates that there is a positive relationship between “national diversity” and the dependent variable of having an alliance or not, while I proposed the opposite (p = .000). Additionally, “industry diversity” correlates highly with all the control variables and this is shown in the table by the significant values of “firm size” and “patent experience”. This is also an indication of multicollinearity but again the VIF values meet the threshold so no problems of multicollinearity should be present in this model.

Model 10 shows the variable of “competitors” and shows positive but non-significant contribution to the prediction of the dependent variable (chi square = 5,772, p > .1 with df = 4). This models shows comparable values of the control variables like the values of model 7.

The last model, model 11, all the variables are added at once and will be used to either accept or reject my proposed hypotheses.

In the last model, I added all the variables at once and the only predictor that has a significant contribution to the prediction of the dependent variable is “industry diversity” (chi square = 40,115, p < .01 with df = 6). “National diversity” has a positive contribution to the prediction of the dependent variable, however it is not significant in the full model so hypothesis 1 is rejected (p = .221). This model shows that only “industry diversity” has a positive significant relationship with the dependent variable (p = .000). However, I proposed a negative relationship between “industry diversity” and the dependent variable and hypothesis 2 should therefore be rejected. Lastly, “competitors” has a positive relationship with the dependent variable but is again not significant which is in line with the other models and hypothesis 3 should therefore be rejected (p = .517).

(26)

performed three extra test in order to find how these variables relate and whether problems concerning multicollinearity exist as this could be the cause of the inconsistent values.

Table 4

Control models for multicollinearity issues

Variable (12) (13) (14) National diversity ,687*** (6,593) ,837 (2,595) Industry diversity ,279*** (2,943) ,277*** (2,717) Competitors ,560 (1,686) ,621 (1,326) Firm size ,482 (,576) ,658* (,330) ,693* (,269) Firm age ,511 (1,370) ,645 (2,728) ,656 (2,806) Patent experience ,659* (3,397) ,815** (5,457) ,863** (6,660) Log Likelihood -52,099 -40,321 -39,767 N 86 86 86 Wald Statistic 7,535 14,914 13,049 VIF value 2,504 2,416 2,511

Note: the standard errors, the odds ratio and the individual significance levels are provided. Coefficients in bold and marked *, ** and *** are significant at 90%, 95% and 99% confidence levels, respectively.

(27)

4.2 Results (Sub-Sample)

The second part of this research is to further explore the sub sample of 43 cases in which the firms involved were alliance partners, either before, during or after the time of inter-organizational conflict. This will be divided into a quantitative analysis based on a hierarchical linear regression model and a qualitative analysis with the use of cross tabulation and examples of the specific cases.

4.2.1 Descriptive statistics and Correlation matrix (Sub-Sample)

In the table 5 the descriptive statistics and correlation matrix are presented. The table demonstrates that some of the variables correlate significantly with each other which is a potential problem for doing the regression analysis. The correlation between “timing of conflict” and the dependent variable is significant at a 95% confidence level. Additionally, the correlation between “governance form” and “timing of conflict” and “alliance experience” is significant at a 95% confidence level. Furthermore, the “number of alliances” and “firm size” but also “number of alliances” and “alliance experience” correlate significantly at a 95% confidence level. Lastly, “scope” correlates significantly at a 90% confidence level with “firm size” and “number of alliances”.

4.2.2 Multicollinearity (Sub-Sample)

Also for this second part of the research, to ensure the predictive power of the variables, the variance inflation factor (VIF) levels are examined. Using the threshold level of 5, proposed by Rogerson (2001), the VIF values are below the threshold and therefore no problems concerning multicollinearity exist in this second part of the analysis. The individual VIF values for each model are given in table 6 and 7 in order to assess the individual model’s problems concerning multicollinearity. The largest VIF value is reported when a model consisted of several variables. Additionally, the other VIF values are provided in appendix 2. Again only the models with more than one variable, so models 5 until 9, are reported in the appendix as the VIF values will be higher than 1 as these models involve interaction between the variables.

4.2.3 Hypotheses tests H4-H6

(28)

Table 7 consists of models 6 till 9, including the full model, which in turn add the predictor variables to the control variables in which the control variables stay constant. In model 6, the variable “multi-partner alliance” is added to the control variables. A positive relationship between “multi-“multi-partner alliance” and the dependent variables is found but is not significant (β = .053, p > .1). Model 7 shows the reduced model in which “governance form” is added to the control variables. This model shows a negative relationship between “governance form” and the “size of conflict” but no significant relationship is founded (β = -.245, p > .1). In model 8 the last variable “timing of conflict” is added to the control variables. This model shows a positive and significant relationship between “timing of conflict” and the dependent variable (β = .507, p < .01). This shows some indication in whether the hypotheses will be accepted or rejected. In the last model, model 9, all the variables are added to the control variables at once to form the full model. This model provides the values that will determine whether to accept or reject the proposed hypotheses.

(29)

Table 5

Descriptive statistics + Correlation Matrix (Sample = 43)

Note: Coefficients in bold and marked with ** and * are significant at 99% and 95% confidence levels, respectively.

Mean, % Std.

Deviation Maximum Minimum 1 2 3 4 5 6 7 8

(30)

Table 6

Estimates of simple and reduced linear regression models

Variable (1) (2) (3) (4) (5) Firm size ,194 ,091 Number of alliances ,269* ,187 Alliance experience ,195 ,082 Scope -,028 ,066 Number of observations 43 43 43 43 43 R2 _,038 _,073 _,038 _,001 _,082 Adjusted R2 _,014 _,050 _,015 _-,024 _-,014 F 1,601 3,207* 1,627 ,032 ,852 VIF value 1,000 1,000 1,000 1,000 2,763

Note: Coefficients in bold and marked with *, ** or *** are significant at 90%, 95% or 99% confidence levels, respectively.

Table 7

Estimates of full linear regression models

Variable (6) (7) (8) (9) Firm size ,091 ,193 ,122 ,196 Number of alliances ,201 ,034 -,134 -,291 Alliance experience ,066 ,253 ,402* ,578** Scope ,076 ,036 ,008 -,056 Multi-partner alliance ,053 -,137 Governance form -,245 -,178 Timing of conflict ,507*** ,523*** Number of observations 43 43 43 43 R2 _,085 _,123 _,265 _,294 Adjusted R2 _-,039 _,005 _,166 _,153 F ,686 1,040 2,673** 2,084* VIF value 2,838 3,335 3,326 4,184

(31)

The next part of the analysis will be based on a cross-tabulation of the sample in which specific cases will be discussed in order to get a deeper understanding of the population. This will further explore when the timing of conflict is higher and in what cases it is more common to happen. I will start with describing the data with the use of cross tabulation tables. The second step will be analyzing specific cases and see why they are so different from the rest and what specific details were involved in the alliance and the patent infringement. Below in table 8 you will find the different variables and the summary of the findings concerning the relationship between these variables. In appendix 3 you can find the crosstabs results including the values of the measures of association.

Table 8

Cross-tabulation analysis

Cross-tabulation analysis. Summary of the findings

Relation Remarks Measures of association

National diversity &

Conflict time

• When the conflict happens after the alliance, 86.1% of the cases is between firms from different countries

• Only 11.6% of the litigations happened between firms from the same country

• The two variables may be independent (r = -,160) • If there is association, it is

weak to moderate and negative.

(U = ,061, φ = -,160) Governance form

& Conflict time

• Again, when the conflict happens after the alliance, only 2.7% of the cases are governed by a joint venture • 93% of the litigations happened

under a bilateral contract

• The two variables are not independent (r = -,374) • The association is moderate

to strong and negative. (U = ,195, φ = -,374) Industry diversity

& Conflict time

• 86.1% of the conflicts after the alliance happen between firms with the same SIC-code

• Only 2.3% of the cases were between firms that involved completely different SIC-codes

• The two variables may be independent (r = ,170) • The association is moderate

to strong and positive. (U = ,178, φ = ,515) Competitors

& Conflict time

• When the firms are direct competitors, 100% of the litigations happen after the alliance

• 76% of the litigations happen after the alliance when the there is no direct competition

• The two variables may be independent (r = ,290) • If there is association, it is

moderate and positive. (U = ,106, φ = ,290) Firm size

& Conflict time

• 69,2% of the companies that are involved in patent infringement are companies with more than 40.000 employees

• 11.6% of the cases involve companies with less than 10.000 employees

• The two variables may be independent (r = ,203) • There is a weak to moderate

and positive association between the variables.

(λ = ,083) Number of alliances

& Conflict time

• When the conflict happens after the

(32)

a total number 50 alliances or more during the period of analysis

• 30.2% of the cases have had less than 50 alliances during the period of analysis

• If there is an association between the variables, it is weak to moderate and positive.

(λ = ,083) Alliance experience

& Conflict time

• In 51.2% of the cases, the firms have had less than 10 previous alliances • Only 33.3% of the cases in which the

conflict happened after the alliance, happened between firms with more than 20 previous alliances

• The two variables may be independent (r = -,284) • There is a weak to moderate

and negative association between the variables. (λ = ,065)

Scope & Conflict time

• In alliances with only 1 activity, 69.8% of the conflicts happened after the alliance

• Only 16.2% of the cases involved more than 1 activity within the alliance

• The two variables are independent (r = ,024)

• There is a weak positive association between the variables. (U = ,001, φ = ,024) Multi-partner alliance & Conflict time

• Of all the alliances, 83.7% conflict, happened after the termination of the alliances

• 91.7% of the alliances of which the conflict happened after the alliance were dyadic alliances

• The two variables may be independent (r = ,121) • There is a weak to moderate

positive association between the variables.

(U = ,051, φ = ,121)

Note: The contingency analysis was reinforced by series of measures of association. All the variables are tested for correlation with the use of Pearson’s correlation coefficient to assess their independence. For the dummy variables (national diversity, governance form, industry diversity, competitors, scope and multi-partner alliances), I included the uncertainty coefficient and Phi measure. For the nominal variables (firm size, number of alliances and alliance experience), I included Lambda.

(33)

4.2.4 Conflict during the alliance

The first of the two was based on a licensing agreement between Abbott and Orchid Chemicals & Pharmaceuticals Ltd. in which Orchid Chemicals & Pharmaceuticals received the license to develop and wholesale an anti-infective drug. This drug was patented by two different patents of which one expired in 2007, the year of the alliance and one in 2011 which were both owned and licensed out by Astellas to Abbott. The latter covered also most of the important parts of the first patent. The lawsuit was about the fact that Orchid Chemicals & Pharmaceuticals infringed the last patent while Orchid Chemicals & Pharmaceuticals stated that they used the expired patent for the use of their drug (Andersen, 2007). So in this case, the patents involved in the patent infringement case were also the patents and drugs involved in the strategic alliance, which is a striking discovery. Without the alliance, Orchid Chemicals & Pharmaceuticals could have never acquired this knowledge about this drug. The second case in which the conflict happened during the alliance was between Warner Chilcott Company and Watson Pharmaceuticals. Again, the Patexia database did not provide which patents are involved but it is less important in this case. Warner Chilcott Company formed a strategic alliance with Watson Pharmaceuticals to wholesale and market pharmaceuticals in the US. As Watson should receive information about the product to market it in the US and also receive the product itself in order to wholesale it, knowledge about the product is spilled over to Watson Pharmaceuticals. This could have enables Watson to use this knowledge in products of their own and that could be the reason that Warner Chilcott Company sued Watson Pharmaceuticals based on patent infringement. In a later stage of this litigation trial, Warner Chilcott Company and Watson Pharmaceuticals reached a settlement and licensing agreement for the drugs involved (Fuhrmann & Eisenhaur, 2009).

As common practice shows that in most cases the patent infringement case is filed after the alliance as this is the moment in which the firms can use the knowledge obtained from their partner for themselves. In these two cases it shows that even during the alliance, firms will infringe patents while still working together, and possibly harming the outcome of the strategic alliance.

4.2.5 Conflict before the alliance

Another small part of the sample consisted of 5 cases, of which will only discuss the most striking ones, in which the conflict between the firms happened before forming an alliance. This is quite striking because why should firms form an alliance with another firm with which they had a conflict? That is why these cases are interesting to look at how these cases differ from each other.

(34)

infringement was based upon another type of drug that the one that is developed in the strategic alliance. A further explanation of this phenomenon could relate to the fact that Merck & Co Inc. saw the potential of Ranbaxy in developing new drugs that they decided that Ranbaxy could be a great partner for developing new drugs.

The second case which is somewhat different from the first case is the one between Proctor & Gamble and Teva Pharmaceutical as in this case the firms are operating in different industries. The case was filed in 2008 and in 2011 they formed a joint venture based upon developing and marketing over-the-counter drugs. The patent infringement case was about the arrangement of atoms in a molecule which could be modified in a specific manner to obtain the claimed compound, which was patented by Proctor & Gamble (Farnan, 2008). Three years later, Proctor & Gamble decided upon a joint venture to develop and market over-the-counter drugs (SDC Database, 2011). These drugs are readily available without any prescription while the patent infringement case was about pharmaceuticals that are only available with a prescription of a doctor. This case shows the fact that having a conflict in one part of the industry does not mean that you cannot have a successful joint venture in another area of the industry.

The third and final case I will discuss, in which the conflict happened between firms before they formed an alliance, is between Eli Lilly and Company and Lupin Pharmaceuticals Inc. The case was filed in 2008 and in 2011 these two firms formed a strategic alliance. In this case, the strategic alliance did not involve any development of drugs and was purely based on promotion and distribution (SDC Database, 2011). This is why this case is different from the other cases as the knowledge spillovers would be limited by a pure distribution or promotion strategic alliance. Furthermore, the patent that was litigated in 2008 was very different from the drug that was promoted during the alliance of 2011 (Wolfson, 2008).

This indicates that firms are not necessarily enemies after they are involved in a patent litigation case. Firms learn from each other and even identify things within the other firm that could be useful attributes when being alliance partners.

5. DISCUSSION

(35)

The first part of this research was to compare whether the characteristics differ for inter-organizational conflict between firms that had an alliance and firms that did not have a cooperative agreement. This resulted in a significant difference between the two samples in national diversity, industry diversity, firm size and patent experience. This was a first indication of how the characteristics differ between the two samples but this research was interested in which characteristics actually increased the likelihood of conflict which was the next step in this analysis. In this next step I found that only industry diversity had a significant influence in the likelihood of conflict, whether the firm had an alliance or not. Besides the fact that the hypothesis is rejected, my first research question is answered which resulted in the fact that only industry diversity as a characteristic of alliances influence the likelihood of conflict in general. The proposed hypothesis for industry diversity was negative but turned out to be a significant positive relationship which reinforces the theory of Cohen and Levinthal (1990) who state that low industry diversity relates to more conflict as the knowledge obtained through the alliance is easily assimilated and could therefor develop conflict. Although patent experience was a control variable, it still shows a significant positive relationship with conflict which indicates that the experience in filing and acquiring patents relates to conflict. Only one side note has to be made in the interpretation of the results. As the values were not consistent across the models, the results should be interpreted with caution.

(36)

research question based on the statistical analysis, the size of conflict is larger for alliances that have terminated their alliance. This is also in line with the qualitative analysis which shows high percentages of conflict after the alliance in the cross tabulation matrix. So it can be concluded that both the size and likelihood of conflict increase when the alliance is terminated. Additionally, further analysis was based on an in-dept qualitative research on seven specific cases which stood out from the sample. These cases were interesting to further explore as they showed content that were conflicting to the existing theories. These inter-organizational conflict cases mostly differed from the fact that the ground for the conflict was not the same as the purpose of the alliance. This resulted in strategic alliances between partners that faced inter-organizational conflict some years before the start of the strategic alliance. Besides having a conflict before starting an alliance, two cases demonstrated that it is also possible to incur inter-organizational conflict during the alliance. These two cases involved critical patents and could lead to negative consequences for one of the companies which decided to sue his alliance partner, jeopardizing the outcome of the alliance as a whole. These cases gave existing theories more inside of how firms work and why they are involved in inter-organizational conflict.

5.1 Implications

(37)

Managerial implications are shown in the way that managers could see what characteristics work and which will increase the chance of conflict. This will save companies and organizations time and money (Crocker & Masten, 1988) when searching for an appropriate alliance partner. Furthermore, the second part of the research shows “real-world” examples which are important learning possibilities for managers and policy makers. These examples show for example that conflicts do not necessarily have something to do with the purpose of the alliance. These conflicts arise when firms benefit from the unintended knowledge spillovers. So managers should be aware to minimize these spillovers to reduce the chance of conflict. Furthermore, previous conflict can be beneficial as this can be an indication of the capabilities of the other company as in the case of Eli Lilly and Company and Lupin Pharmaceuticals. Additionally, this research shows that managers should be on their toes when entering an alliance and should also be thoughtful when terminating the alliances as from then on the likelihood of a larger conflict is highest.

6. CONCLUSION

This paper aimed to shed light upon the inter-organizational conflict in general and specifically in the case of inter-organizational relationships. The results show that in general only industry diversity is the main characteristic between firms that increases the likelihood of conflict. Furthermore, in the case of alliances, the likelihood of conflict increases by the later timing of the conflict which implies that the likelihood of conflict is higher when the firms have terminated the alliance.

Next to the empirically tested hypotheses, this paper sheds light upon specific cases in which the conflict happened before or during the alliance. As these cases were small in number within the research sample, they gave a good insight in how these conflicts arose and why companies should be careful when setting up an alliance.

6.1 Limitations and future research

(38)

Furthermore, the logit regression analysis showed inconsistent values across the models which could indicate multicollinearity problems. As these were not present when analyzing the VIF values, cautious interpretation should be applied when reaching conclusions from these models. This limitation relates to the validity of the research and should be taken into account in further research.

Additionally, a deeper qualitative research can be done to demonstrate how different alliances have different purposes and how these relate to the construct of conflict.

(39)

7.REFERENCES

Al-Laham, A., Amburgey, T. L., & Bates, K. (2008). The dynamics of research alliances: Examining the effect of alliance experience and partner characteristics on the speed of alliance entry in the biotech industry. British Journal of Management, 19(4), 343-364.

Andersen, W. R. (2007). ABBOTT LABORATORIES ET AL V. ORCHID CHEMICALS PHARMACEUTICALS LTD. ET AL.

Barry, C., Arad, R., Ansell, L., & Clark, E. (2014). 2014 patent litigation study. ( No. NY-14-0601). New York: PWC.

Baum, J. A. C., Calabrese, T., & Silverman, B. S. (2000). Don't go it alone: Alliance network composition and startups' performance in canadian biotechnology. Strategic Management

Journal, 21(3), 267-294.

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

Business, 39(2), 107-120.

Booth, S. A., & Wang, Z. (2012). How do chinese firms deal with inter-organizational conflict? Journal

of Business Ethics, 108(1), 121-129.

Chen, C. (2004). The effects of knowledge attribute, alliance characteristics, and absorptive capacity on knowledge transfer performance. R & D Management, 34(3), 311-321.

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

Connelly, D. R. (2007). Leadership in the collaborative interorganizational domain. International

Journal of Public Administration, 30(11), 1231-1262.

Crocker, K. J., & Masten, S. E. (1988). Mitigating contractual hazards: Unilateral options and contract length. The Rand Journal of Economics, 19(3)

Das, T. K., & Teng, B. (2000). Instabilities of strategic alliances: An internal tensions perspective.

Organization Science, 11(1), 77-101.

Das, T. K., & Teng, B. (2002). Alliance constellations: A social exchange perspective. The Academy of

Empirical and qualitative analysis of conflict in inter-organizational relationships

TABLE OF CONTENTS