Trust and Contracts in Strategic Alliances

(1)

UNIVERSITY OF GRONINGEN FACULTY OF ECONOMICS AND BUSINESS

Trust and Contracts in

Strategic Alliances

An alternative approach to substitute governance forms

Master Thesis of the Double Degree Program in International Economics and

Business at University of Groningen and Corvinus University of Budapest

Thesis supervisor: Dr. Gjalt de Jong

Nora Balogh (s2251787) n.balogh@student.rug.nl

(2)

2 Abstract.

This study analyses the role of interorganizational trust and contracts in strategic alliances. The thesis proposes that conflicts stemming from organizational culture differences and the relative dependence on the alliance project (“interdependence”) are an underrepresented area of research concerning alliance governance (interorganizational trust and contractual complexity). By analyzing a database of 391 Dutch high-tech alliances, this study finds support for the importance of organizational culture differences and interdependence in shaping alliance governance. The study also introduces several indirect and direct methods (such as interaction terms regressions and multiple equation techniques) of analyzing the substitution between interorganizational trust and contractual complexity in strategic alliances.

(3)

3

I. Introduction

The high number of interfirm ventures in the global economy (for example, between Sony and Ericsson in developing mobile phones or the Star Alliances collaboration in the aviation industry) has drawn the attention of academic research to firms collaborating with each other. As companies are facing fierce competition in the global market, one way for them to survive is cooperating with other firms (Das & Teng, 2001). Strategic alliances are “arrangements between two or more firms that establish an exchange relationship without joint ownership” (Lowensberg, 2010). The strategic alliance management research field is concerned with how interfirm collaborations such as international joint ventures and R&D alliances are formed, executed and managed.

This study is based on two general assumptions about strategic alliances. First, all strategic alliances are similar from the aspect that they contain firms who depend on each other and the alliance project in achieving their strategic goals. This characteristic of alliances is referred to as “interdependence”. Second, all alliances entail differences on the level of the collaborating companies, because firms differ from each other in their organizational culture. This aspect of strategic alliance research is labeled “organizational culture differences”.

Partners to strategic alliances have to adapt a governance mechanism for their activities. According to the relevant literature, there are two different governance mechanisms, corresponding to two considerable research areas. One research stream states that strategic alliances should be organized by developing a contract that, for example, covers possible future contingencies as much as possible. The other research stream states that relational governance (i.e., management through building a trust relationship with partners) is a more sufficient way of dealing with the obstacles in the collaboration process. However, the interaction of the two governance forms is subject to heavy debate that has resulted in studies offering contradicting findings.

(6)

6

organizational culture affect interorganizational trust (Model I) and contractual complexity (Model II), respectively. Since prior research has not integrated these constructs in a single analysis, the study contributes to the strategic alliance management research field by opening up an alternative method of analyzing the determinants of alliance governance; as well as understanding the interaction between the two governance forms (trust and contracts) in the context of interdependence and organizational culture differences.

This research examines the following questions:

 What is the effect of organizational culture differences on alliance governance?  What is the effect of interdependence on alliance governance?

 Is interorganizational trust and contractual complexity complementary or substitute governance forms in the context of organizational culture differences and interdependence?

The specific research question of this thesis is the following: How do organizational culture differences and relational dependence influence interorganizational trust and contractual complexity?

(7)

7

II. Literature review

Embededness of the study in prior research

The strategic alliance management field, as it has been stated earlier, is concerned with a wide range of issues regarding collaborative ventures. The scope of analysis among these studies differs quite substantially, resulting in a growing research area that sheds light on multiple aspects of firm cooperation. In order to understand the theoretical foundations of the model developed and tested in the thesis, three areas of research have to be analyzed. The issues concerning the governance of strategic alliances are the point of departure in this section, as contractual complexity and interorganizational trust form the backbone of Model I and Model II. The studies that elaborate on organizational culture concerns in strategic alliance are the second level of analysis. The third key area of prior research that is examined in this section is the extent to which companies are dependent on their partners as well as the alliance project.

Contractual complexity

Strategic alliances “[…] include voluntarily collaborating firms who engage in exchange, sharing, or co-development, and include contributions by partners of capital, technology, or firm-specific assets” (Gulati & Singh, 1998). In principle, all strategic alliances can be thought of as co-alignments between two or more firms (Lei et al., 1997), which means that their interests are aligned in achieving a common goal. However, many scholars argue that the risk of opportunism is inherent to transactions. Opportunism refers to incomplete or distorted disclosure of information, and especially to calculated efforts to mislead the partner (Hill, 1990).

(8)

8

output, processes of their exchange and dispute resolution mechanisms” (Reuer & Ariño, 2007).

Since the interpretation of the contract influences the behavior of the parties to the alliance (De Jong & Klein Woolthuis, 2008a) and contracts are costly to renegotiate, the drafting of a proper alliance contract is of high importance in the alliance process. Ariño & Reuer (2004) state that the high level of relationship-specific assets associated with the alliance and the threat of (unsupervised) spillover are two cases in which detailed contracts can be beneficial. However, they argue that alliance contracts should be applied by carefully balancing safeguarding clauses with flexibility, and by putting the contract aside once the alliance is set into motion.

The function of contracts in strategic alliances has been studied by several scholars and resulted in different classification methods. In the drafting process of the contract, specific contract clauses correspond to three different contract functions: they can refer to the traditional safeguarding function of contracts; they can serve in line with the coordinating and commitment function; or they can serve as means to mitigate external contingencies (De Jong & Klein Woolthuis, 2009). With respect to the application of the alliance contract once the collaborative venture has been set into motion, contracts have a dual function: they may serve as either control mechanisms that create the basis of authority or power mechanisms; or they can serve as coordinating devices that make sure that the different entities subject to the alliance develop appropriate linkages (Mellewigt et al., 2007).

(9)

9

among the partners (Artz & Brush, 2000). Although transaction costs theory acknowledges the “atmosphere” as a relevant factor in exchanges, it neglects sociological terms such as trust and power (Nooteboom, 1996).

Though contractual complexity is arguably an important facet of alliance relationships, empirical testing of contractual complexity can be difficult as the characteristics of the contract are matter of high confidentiality in business relationships (De Jong & Klein Woolthuis, 2009). Moreover, even in case information about these clauses is accessible, the development of a general measure for contract complexity is rather difficult. In theory, contract designs correspond to the characteristics of specific business relationships that they are to govern, thus rendering a generalized measure of contractual complexity difficult to construct. Some papers analyze the length of the contract or, along with present study, develop indicators that measure whether a specific provision has been employed in the drafting process (Reuer & Ariño, 2007). Although these methods provide partial remedy for the measurement issue, it is important to note that the instrumentalization of contractual complexity remains an important aspect of contract research.

Overall, prior research concerning the governance of strategic alliances points out that the extent to which partners develop a detailed alliance contract is likely to influence the alliance process on multiple levels. Contracts containing, for example, several safeguarding clauses might curb potential losses from opportunistic behavior; however, detailed contracts simultaneously imply that opportunism is regarded as a real threat and therefore hinder the development of a strong, stable and trusting business relationship. Complex contracts also entail ex post costs as they are more difficult to renegotiate and might therefore decrease the flexibility of the alliance. This study proposes that contract detail can be explained by, among other factors, differences on the organizational level and differences in the relative dependence as perceived by the alliance partners.

Trust

(10)

10

challenged the findings of transaction cost economics and shed light on informal governance mechanisms inherent to cooperation, such as trust. Their studies point out that the development of a trust relationship can be an effective way of decreasing the risk of opportunism in alliance relationships.

When firms engage in alliance activity, they develop a business relationship with another company. Many studies point to numerous examples of “preferential, stable, obligated and bilateral trading relationships” to illustrate that firms develop close bonds with each other through their transactions (Gulati, 1995). Das & Teng (1998) state that, since it is impossible to monitor every single detail of their exchange, firms must always have a minimum level of trust in their partners. According to behavioral research, trust emerges in exchange relationships without legal protection (Barney & Hansen, 1994), in the process of cooperation as firms try to circumvent the difficulties of formal contracting (Baker et al, 2002).

Trust is “[…] a belief that one party can subordinate its own, selfish interest to the interest of the cooperative relationship” (Ariño et al., 2001). It emerges in the context of “behavioral uncertainty and vulnerability associated with social exchange” (Parkhe, 1998a). Trust also represents “one partner’s expectation that the other will not expropriate their vulnerabilities, even when presented with opportunity to do so” (Krishnan et al, 2006). Moreover, trust is unlikely to be one-sided, because in the assumption of mistrust, a party will not develop a trusting behavior (Anderson & Weitz, 1989).

Though trust primarily exists between individuals, it can be extended to collaborating organizations, since interorganizational relationships are managed by individuals (Aulakh et al, 1996). In this sense, interorganizational trust is the form of trust that develops on the organizational level. This research applies “trust” and “interorganizational trust” as synonyms as its level of the analysis is interorganizational relationships.

(11)

11

An important aspect of the collaborative relationship that has been studied in great detail in alliance research is the sources or foundations upon which trust develops. According to Parkhe (1998a), there are three bases for trust to develop. Process-based trust evolves in the process of the collaborative relationship or as a result of good reputation. Characteristics based trust emerges between organizations who share similar –cultural or corporate cultural- characteristics. Institutional-based trust stems from formal mechanisms put to place, such as guarantees that decrease the possibility of opportunism. Trust may also arise from “macro sources” which apply generally and impersonally (such as the institutional environment of laws, norms, values, standards, and agencies for their enforcement); or from “micro sources” – referring to the personalized nature of trust as it develops in specific relationships (Nooteboom, 2006).

Several scholars argue that trust is a multidimensional construct with multiple antecedents (Nooteboom, 2002; Aulakh et el, 1996; De Jong & Klein Woolthuis, 2008b; Mayer et al, 1995). These studies highlight the importance of, for example, open information sharing and flexibility in determining interorganizational trust. Furthermore, these scholars argue that, as trust develops in the process of collaboration, it is likely to be different in the initial stage of collaboration than in latter phases of the business relationship. This reasoning has led scholars to creating models in which different dimensions of trust are linked to the stages of the collaborative venture in which they emerge (such as initial conditions and negotiation stage) (Ariño et al., 2001; De Jong & Klein Woolthuis, 2008b). Nooteboom (2006) refers to trust in people or organizations as “behavioral trust” and highlights various aspects of the phenomenon: trust in competence (competence trust), intentions (intentional trust), honesty or truthfulness, resource availability, and robustness, (i.e. limited sensitivity to outside disturbances). Trust may also comprise different antecedents such as continuity expectations, flexibility, information exchange and different control mechanisms (output, process and social control) (Aulakh et al., 1996).

(12)

12

– an unobservable construct that cannot be directly measured or a sentiment that is unobservable to the researcher but enters the decisions (Hill et al., 2011). The multidimensionality of trust, therefore, has resulted in the application of factor analysis as the appropriate method to reveal the antecedents and create instruments to measure interorganizational trust. Further elaboration on this methodological process can be found in the Methodology section.

Overall, interorganizational trust is a facet of strategic alliances that has received much attention in the literature. This study expands the existing research in two ways. First, by examining the effect of organizational culture differences and interdependence on trust, the thesis introduces an alternative to the performance-orientated scope that has dominated prior analyses of interorganizational trust in strategic alliances. Second, by highlighting the interaction between the two governance forms – interorganizational trust and contractual complexity -, this study aims to expand the discussion on the substitution or complementary relationship between the two alliance governance forms. However, before developing the theoretical model further, some attention should be drawn to the body of research dedicated to the interaction of interorganizational trust and contractual complexity.

The link between contractual complexity and interorganizational trust

Considerable body of research in the strategic alliance management field is concerned with the interaction between interorganizational trust and contractual complexity. Generally, there is wide acknowledgement among strategic alliance management scholars that interorganizational trust and contracts coexist and interact with each other in the context of alliances (Contractor, 2005; Baker et al, 2002; Lyons & Mehta 1997. etc.); however, there are major differences among studies in the assessment of the relationship between the two constructs.

(13)

13

“family jewels” such are core technologies, since partners lack the appropriate contractual clauses to guard against, for example, unsupervised spillover.

The arguments on the complementary relationship between trust and contracts stem from the key characteristic of contracts that they are inherently incomplete. As it is impossible to foresee all future contingencies and incorporate them into the contract, in the presence of incomplete contracts, at some point partners have to seek resource to trust. In this sense, trust becomes important where contracts no longer suffice (Nooteboom, 2006). This argument corresponds to several arguments among studies that trust and contracts act as complements in alliances (Poppo & Zenger, 2002; Klein Woolthuis et al., 2005). These scholars state that a complex contract is associated with the expectation that the other party will behave cooperatively, as a specified contract the manifestation of long-term commitment to the alliance project. Therefore, trust complements the contract in the governance of the alliance.

The studies listed above not only differ in the predicted sign of the relationship between trust and contractual complexity, but they also analyze the line of causality between the two constructs along contradicting assumptions. These analyses offer different findings on whether trust determines contract detail, or conversely, the characteristics of the contract shape the trust relationship. Klein Woolthuis et al. (2005) examine this ambiguity and find that, although theoretical papers assume the legal framework of the alliance to predict the level of interorganizational trust, empirical findings show opposite - that interorganizational trust is the determinant of contractual complexity. An interesting argument on the line of causality between trust and contracts is offered by Nooteboom (2006), who indicates that contracts and legal enforcement are one of the sources of “reliability” in exchanges, with “reliability” referring to trustworthiness in a broad interpretation. Klein Woolthuis et al. (2005) also argue that in case trust and contracts are complementary governance tools, the contract is the basis for trust to develop. These findings are not only interesting because of their implications for contract research, but also raise methodological issues as the apparent path dependence of the two constructs might bring about modeling issues.

(14)

14

This study aims to reveal how interorganizational trust and contractual complexity are affected by organizational difference and interdependence and how these effects might change when moving from one model to the other. The hypotheses concerning these relationships form the basis for determining whether trust and contracts act as substitutes in the alliance.

The second issue discussed in this study concerning the link between the two governance forms is the inconsistent arguments and findings on the line of causality between trust and contracts; and the endogeneity problem relating to the contemporary presence of trust and contracts in alliances (Poppo & Zenger, 2002). A possible methodological remedy for this issue is the seemingly unrelated regressions (SUR) method, which enables the simultaneous testing of regressions in which the error terms are correlated. Further elaboration on this method is done in the Results section of the thesis.

The gap in strategic alliance research

As stated in the introduction of this study, multiple research papers have highlighted the importance of firm-level differences (Beugelsdijk et al., 2009; Denison et al., 2005), and interdependence between alliance partners (Guidice & Cullen., 2007; Sambasivan et al., 2011) in strategic alliances. These papers offer important arguments; however, their scope of research and findings do not touch upon the dimensions of alliance governance. After careful examination of the literature, I suspect two possible reasons for the lack of detailed analysis regarding the effect of firm level differences and interdependence on the governance of the alliance.

(15)

15 Organizational culture

Organizational science has devoted much attention to how organizations are bound together by the norms and shared values of the individual members. Beugelsdijk et al. (2008) describe organizational culture as “shared values and beliefs that help individuals understand organizational functioning”, and also “[…] help managers direct the course of the organization more effectively”. This concept of organizational culture has two interesting elements that are important for present research. First, as organizational culture is based on a sentiment, similarly to trust, it is likely to influence the behavior and beliefs of individuals. Second, although it is a sentiment, it can be applied as a managerial tool to develop the organization – that is, it can be applied as a coordination device, such as the alliance contract. It follows from this line of reasoning that organizational culture of the focal companies is likely to be an influential factor in the governance of the alliance they develop.

Organizational culture may also be defined as “the set of integrated concepts that become manner or strategies through which an organization achieves its specific goals” (Marcoulides & Heck, 1993). Organizational culture comprises the “shoulds and ougths of organizational life” (Veiga et al., 2000) and influences, both directly and indirectly, the achievement of organizational outcomes (Marcoulides & Heck, 1993). Hofstede et al. (1990) associate organizational culture with the management of organizational life and differentiate six core aspects, along which the management of organizations may differ: process vs. result; employee vs. job; parochial vs. professional; open vs. closed system; loose versus tight control; normative versus pragmatic. These differences are relevant as they influence the behavior of the employees as well as the performance of the organization (Marcoulides & Heck, 1993).

(16)

16

culture differences not only in context of operating alliances, but also as influential factors in the partner selection process for cooperation. Moreover, this “distance” perspective raises interesting questions about the possible dynamic nature of organizational distance. Strategic, structural and technical distances might change through the life cycle of the alliance, and if so, might affect the quality of the relationship differently in different stages of the alliance process. The dynamics of organizational culture differences are an intriguing aspect of strategic alliances that remains for future research to explore in detail.

To sum up, the organizational culture of the individual alliance partners is an important aspect of collaboration, because management styles applied in the individual companies are reflected in the assessment of the management of the alliance. Rationale suggests that firm-level differences might hinder cooperation as firms might not be able to create compatible goals along different strategies and incentive systems. Moreover, if organizational distances create ambiguous expectations about the behavior of the partner firm, trust might be difficult to develop and maintain. Furthermore, the less similar the organizational culture of the partner firm, the more guarantees a firm might aim to include in the alliance contract. Detailed analysis of the effect of organizational culture differences on the quality of the trust relationship and the extent of contract detail is conducted in the Hypothesis development section.

Dependence

(17)

17

& Boon (2008) examine supplier-buyer relationships involving limited alternative sources of a critical resource and find that, in case a buyer has to depend heavily on a supplier of the critical resource, high level of trust in the supplier is likely to exhibit low uncertainty associated with the expected outcomes of the relationship. This argument implies that relational uncertainty relating to the level of dependence can be eased by developing a trust relationship.

The term “dependence” refers to a firm’s need to maintain an exchange relationship with other firms to achieve desired goals (Razzaque & Boon, 2008). Dependence can also be understood as “a link, tie or bond of one company in relation to another” (Svensson, 2004). This relationship can be asymmetric, if one firm is more dependent on the relationship than the other; this type of dependence is labeled as “unilateral, unbalanced or asymmetric” (Jiang et al., 2008). Dependence can also be symmetric, in which case it is labeled “mutual interdependence”, and in which case it refers to the extent to which partners value one another`s resources (Razzaque & Boon, 2008). Kumar et al. (1995) analyze the level of dependence in channel relationships and differentiate between dependence (the level of dependence of the individual companies), total interdependence (the sum of the partner firms’ dependence) and interdependence asymmetry (the difference of the firms’ dependence). “Interdependence” generally refers to the belief that the outcome of a relationship depends on the separate as well as joint actions of the members of the cooperation (Sambasivan et al., 2011). Interdependence also comprises the extent to which the partners view their relationship not to be readily replaced (Smith & Barclay, 1999). Sambasivan et al. (2011) differentiate three types of interdependence in strategic alliances and define them as extent to which firms have to rely on their partners to complete their task (task interdependence) or reach their goals (goal interdependence) in the alliance; moreover, how their individual success influences the success of the collaboration (reward interdependence).

(18)

18

losses by a detailed contract is likely to be affected by the level of interdependence among the partner firms.

Four constructs have been introduced and analyzed in this section: contractual complexity, interorganizational trust, organizational culture and dependence. The theoretical foundations and the perspectives of prior studies have been examined with respect to how these elements of the collaborative venture might be connected to each other. Though some indication has been made as to how these constructs can be implemented in a single analysis, the exact nature of these relationships is revealed in the next section.

III. Hypothesis development

As it has been stated above, economic models of transaction costs stem from the general idea that organizations emerge to curb transaction costs (Gulati, 1995). Following their reasoning, firms formally organize their transactions with other companies in order to monitor the hazards associated with exchange such as opportunism. It has also been argued that Williamson, the scholar defining the general principles of transaction cost economics, or “new institutionalism”, focuses on the institutional arrangements, more specifically, formal contracts, of cooperation and denies the importance of trust in the governance of exchange. However, this paradigm has been challenged greatly by “neo-institutionalists”, who expand the theory of Williamson with arguments taken over from sociology. Neo-institutionalist scholars focus on explaining the institutional environment of firm collaboration such as norms and habits and argue that trust develops in the process of collaboration (Nooteboom, 2002). These two theories provided the basis for the division of Model I and Model II.

(19)

19

fluent cooperation. “Interdependence” and “organizational culture differences” are the two phenomena that are introduced in both models.

Figure 1.Theoretical model (Expected signs in parentheses).

Model I

Model I is concerned with interorganizational trust, defined as “one partner’s expectation that the other will not expropriate their vulnerabilities, even when presented with opportunity to do so” (Krishnan et al, 2006). This model examines how firm differences and differences in relative dependence on the alliance affect the level of trust among alliance partners.

Organizational culture has been defined as the norms and beliefs that bind organizations together. These values help managers communicate the goals and strategy through the organization as the norms become the integrated strategy of the firm (Marcoulides & Heck, 1993). Several scholars found that when companies engage in collaboration, the outcome of their cooperation and the satisfaction associated with their partnership depends on the similarities in their organizational culture (Beugelsdijk et al, 2009, Pothukuchi et al, 2002).

(20)

20

similarity positions the firms for a less conflicted interaction. They argue that similarities in organizational practices foster knowledge-driven trust stemming from repeated interaction.

Partners with dissimilar organizational cultures spend valuable time and energy on arranging common managerial practices, and may experience lower level of trust than culturally similar partners (Potchuchi et al, 2002). Organizational differences may also be associated with lower commitment and cooperation in the context of mergers and acquisitions (Veiga et al., 2000). Building trust might be more difficult when cultures are highly dissimilar, since homogenous expectations and shared assumptions about the alliance may not exist readily (Parkhe, 1998b). These arguments lead to

Hypothesis 1a: Organizational culture differences have a negative effect on interorganizational trust.

Interorganizational trust has been defined as a firm’s expectation that the other partner will not expropriate their vulnerabilities, even when presented with opportunities to do so (Krishnan et al, 2006). It has also been argued that, in the presence of trust, parties can cooperate with each other more fluently, as trust lowers the transaction costs inherent to exchanges (Aulakh et al., 1996). Trust can also be understood as the willingness to rely on the behavior of others and the “willingness of a party to be vulnerable to the actions of another party, based on the assumption that the other party will perform a particular action important to the trustor” (Mayer et al, 1995). This definition of trust implies that a trusting relationship is associated with the acceptance of the vulnerability stemming from the dependence on the partner. Similar argument is expressed by Razzaque & Boon (2008), who state that a trusting relationship deems the inherent dependence of the alliance more acceptable to partners, since they feel assured that the desired outcome of the partnership will be achieved. Moreover, a symmetric high-dependence relationship engenders valuable resources as well as cooperation between partners (Razzaque & Boon, 2008). The authors also state that if mutual dependence in the partnership is high, satisfaction with the alliance is more likely achieved in the presence of trust. These arguments shed light on the relationship between trust and interdependence and imply that trust is necessary for interdependence to be acceptable in the relationship between the alliance partners.

(21)

21

in achieving their goals, they experience “high exit barriers”, which is a strong motivation for them to build a strong relationship. Trust and commitment emerge when the interdependence structure between the partners leads the interests of the participants to converge (Kumar et al., 1995). Ariño et al. (2001) define mutual interdependence as an element of interorganizational trust and state that trust develops in the presence of vulnerability from dependence inherent to cooperation. These arguments indicate that trust emerges as a result of the high interdependence inherent to the relationship, because partners have valuable resources to loose in case they behave opportunistically; therefore interdependence fosters trust between the partner firms. These arguments lead to

Hypothesis 1b: Interdependence has a positive effect on interorganizational trust. Model II

In Model II, the center of analysis is how organizational culture differences and interdependence influence the extent of contractual complexity in strategic alliances. Contractual complexity, as it has been pointed out, is an increasing function of the risk of opportunistic behavior (Reuer & Ariño, 2007). However, monitoring, enforcing and negotiating a complex legal document is difficult, that is why it is important to develop a contract that is applicable to the transactions and tasks inherent to the alliance.

Relational risk in alliances, as it has been shown, can be mitigated in alliances by adapting a formal contract (Mellewigt et al, 2007). Relational risk relates to whether cooperation will go smoothly between alliance partners (Das & Teng, 1996). However, it has also been argued previously that organizational culture differences may lead to conflicts in cooperation. It follows from this line of reasoning that conflicting management practices might create obstacles in the cooperation process, and if so, such relational risk can be mitigated by developing a detailed contract that creates framework for cooperation.

(22)

22

companies experience differences in their assessment of contract detail and the function of the contract in their business relationships in general, they could experience difficulties in reconciling these competing views, and might develop a complex contract in the process of collaborating. These arguments lead to

Hypothesis 2a: Organizational culture differences have a positive effect on contractual complexity.

It has been stated previously that interdependence is inherent to strategic alliances as partners to the cooperation rely on their partners and the alliance project to achieve their goals. The extent to which parties are dependent on their partner and the alliance project correspond to different dimensions and levels of interdependence that have been analyzed in the previous section. Guidice & Cullen (2007) examine the relationship between interdependence, control and performance and find that in alliances with high strategic interdependence, experienced alliance partners who use less formal controls achieve higher performance than those who rely more on formal control mechanisms. Although the authors imply that the use of formal control decreases with interdependence only in case alliance partners have experience in alliance building, they offer an interesting view on the control-interdependence relationship.

(23)

23

Hypothesis 2b.Interdependence has a negative effect on contractual complexity.

As pointed out in the literature review, the interaction between interorganizational trust and contractual complexity has received much attention in strategic alliance research; however, the exact nature of the relationship is subject to intensive academic debate with conflicting views on substitution or complementary relationship between the two governance forms. It has also been argued that prior research also entails inconclusive findings on the line of causality between trust and contracts – i.e. whether trust determines the level of contract detail or whether the alliance contract acts as a basis for trust to develop.

These two limitations of prior studies concerning the relationship between trust and contracts are the reasons this study takes an alternative approach in determining the nature of this relationship. The arguments of the thesis propose that conflict from high level of organizational dissimilarity can be solved by developing a complex contract (Hypothesis 2a); however, organizational culture differences decrease interorganizational trust (since Hypothesis 1a indicates a negative relationship between organizational culture differences and trust). This means that, since organizational culture differences decrease interorganizational trust, a detailed formal contract can substitute for trust as a governance mechanism in case of high organizational dissimilarity. Similar arguments can be drawn upon the hypotheses concerning interdependence. Mutual dependence increases trust in the alliance (Hypothesis 1b); however, interdependence decreases the level of contract detail (Hypothesis 2b). Since partners who mutually depend on each other do not find it necessary to develop a detailed contract, in case of interdependence, trust may substitute the contract as a governance mechanism. This reasoning, and therefore the third hypothesis is based on the assumption that the hypotheses of Model I and Model II are accepted. These arguments lead to

(24)

24

IV. Methodology

Survey methodology

The majority of empirical research concerned with the facets of firm cooperation such as governance of strategic alliances, firm-level differences and dependence are conducted by structuring a survey or questionnaire (Dyer & Chu, 2003; Gulati & Sytch, 2008; Poppo & Zenger, 2002; Reuer & Ariño, 2007; Sambasivan et al., 2011; Smith & Barkley, 1999). The benefit of this method is that it provides insight on a group by obtaining information from a relatively large sample (Leeuw, Hox & Dillman, 2008). The survey method is especially useful in the analysis of socioeconomic constructs such as trust and interdependence, as these constructs represent sentiments that can best be assessed upon the experience and insight of a member of the alliance project.

Sample and data collection1

The focus of this study is business relationships between two or more firms and/or research institutes that operate in high-tech industries (biotechnology, new material development, information technology, maritime technologies and environmental technologies) in the Netherlands. The life cycle of R&D in these industries is very short, as much of the new technological knowledge quickly becomes outdated, often even before it has been incorporated in new products and/or services. In addition, R&D services induce investment costs that are difficult for a firm to endure independently. This is one reason why high level of cooperation between firms can be seen in this industry, even among rival firms. Furthermore, given environmental uncertainty, interorganizational trust is presumed to operate in this context.

In the preparatory phase of the fieldwork, the research team conducted 25 semi-structured interviews with consultants who had been involved in R&D alliances. These interviews provided a wealth of information on characteristics of high-tech industries, relationships between firms and new technological knowledge development. This information

1_{This sections draws upon de Jong & Klein Woolthuis, 2008b. The author thanks Gjalt de Jong (University of}

(25)

25

was used to design the survey and select the respondents. The survey was then field-tested on a sample of 10 companies involved in alliances in research and development. After some modifications in the questionnaire, the team conducted interviews with 572 business managers. Prior to the interviews, all managers received an explanatory letter inviting them to participate. The answers were all measured on 5-point Likert scales. During the interview, main topics such as the history and purpose of the alliance as well as contracts, power, investments, industry dynamics and third party mediation were discussed. Some open questions were added to enliven the interview and to enable the respondents to tell their own story to some extent. In total, 50 questions (often divided into several sub-questions) were asked, which resulted in the half hour design of the interview to often expand to one hour, depending on the respondent.

The research team made three attempts to identify and interview the selected respondents. The case companies were selected from the database of the Dutch Ministry of Economic Affairs. This database enabled the team to contact business managers who were responsible for interfacing with the partner firms and were hence considered to be informants of upmost knowledge about the interfirm relationships. One of the first questions required the respondent to identify their partner in the alliance in question. This information provided the basis for cross-validating the information obtained from the database. As high-tech alliances are typically concerned with specific projects and goals, the respondents were also asked to identify a project that was the most important in the interfirm alliance. By focusing on interfirm collaboration within one sector (high-tech industries), the team reduced the range of extraneous variations that might have influenced the constructs under study. The telephone interviews resulted in 391 usable responses, providing an effective response rate of 68.5 per cent. This response rate is considerably higher than those of prior research on interfirm ventures.2

2_{The average response rate for business surveys via e-mail or the Internet in the Netherlands is 5-8 per cent. This}

(26)

26 Issues relating to survey methodology

The application of a survey as data collection method entails certain limitations and possible issues which, if not accounted for properly, might hinder the reliability of the estimation method based on a survey dataset. Three of the potential issues relating to survey methodology are briefly discussed in this section: non-response bias, the issue of missing values and common method variance.

Non-response bias

Many scholars studying interfirm relationships identify non-response bias as one of the most relevant issues of survey methodology. This problem relates to the failure of obtaining a response from a sample unit (Leeuw et al., 2008). The most common method across the studies concerning the governance of strategic alliances is to account for non-response bias by comparing early and late respondents as late respondents share similar characteristics to non-respondents (Aulakh et al., 1996; Poppo & Zenger, 2002; Reuer & Ariño, 2007; Sambasivan et al., 2011).

The non-response in the data collection process applied in this case was relatively low (31.6 per cent), especially considering that only 10.5 per cent actually refused to be interviewed. To rule out potential bias resulting from non-respondents, the non-cooperative respondents were asked to give their reason for not participating. These reasons were two-fold: either the lack of time and interest (I have no time; I am not interested; I do not feel like it; I am too busy), or the irritation from recently cooperating in a telephone interview was mentioned. Although these reasons might hide the true motivation for not participating (such as the reluctance to discuss an unsuccessful alliance), they did not raise serious concerns of non-response bias (De Jong & Klein Woolthuis, 2008b).

Missing values

(27)

27

information on conflicts in their working relationship or whether they achieved a satisfactory outcome from collaborating, thus creating a pattern of missing values corresponding to specific questions.

The adjustment techniques for missing values range from relatively simple methods (as complete-case analysis, available-case analysis or replacing missing observations with the mean value) to more sophisticated techniques (such as imputation, multiple imputation) (Leeuw et al., 2008). (The missing observations for variables of the thesis are presented in Table 1 in the Appendix). In the statistical software Stata, most commands ignore observations that are missing in the variables referred to in a specific command. For example, the OLS regression disregards all observations that have a missing value for the dependent variable for any of theindependent variables. This method is commonly referred to as "list-wise deletion" or "complete cases only" (Leeuw et al., 2008). For the estimation of the hypotheses, I relied on the list-wise deletion technique incorporated to Stata in managing the missing observations. However, in order to rule out biased estimates as a result of missing values (Harrington, 2009), I reconstructed the dataset by mean substitution as part of the robustness analysis of the results.

Common method variance

Data obtained by survey method might also be biased in case the survey responses are obtained from a single source. Since the measures come from the same source, “any defect that source bares, contaminates the measures, presumably in the same fashion and in the same direction”. This problem is referred to as common method variance or common method bias (Podsakoff & Organ, 1986). If we presume common-method bias to be present, than a factor analysis of all variables representing the question items would yield a single factor (Aulakh et al, 1996). Exploratory factor analysis yielded 12 factors with eigenvalues higher than 1. Principal component factor resulted in 16 factors passing the Kaiser criterion of eigenvalue higher than 1 and hence confirmed the assumption that in case of this study, the dataset is highly unlikely to suffer from common method variance.3

(28)

28 Instrument development

Most of the constructs that have been introduced in Model I and Model II are, from a methodological view, latent variables – that means they are unobservable constructs that cannot be measured directly but can rather be traced in different constructs relating to or loading on the same, unobserved phenomenon (Hill et al., 2011). Prior research concerning trust and contracts apply factor analysis in order to confirm the multidimensionality of the constructs applied in their theoretical models (Aulakh et al., 1996; Denison & Mishra, 1995; De Jong & Klein Woolthuis, 2008a; De Jong & Klein Woolthuis, 2008b; Sambasivan et al., 2011). Factor analysis refers to “a set of statistical techniques that represent a set of variables in terms of smaller number of hypothetical variables” (Lewis-Beck et al., 1994). The aim of factor analysis is two-fold, corresponding to two distinctive analysis processes. Exploratory factor analysis (EFA) is applied in case the facets of a variable are unknown to the researcher, and the aim of the analysis is to explore the dimensions of the variable. Confirmatory factor analysis (CFA) is used to confirm hypotheses based on the assumption that the applied variables contain several different elements (Lewis-Beck, 1994). The general idea behind applying factor analysis in case of surveys is that the instruments used in the estimation of the theoretical models can be done by factoring the values of multiple question items into a single, latent variable. Therefore, the measurements representing trust, contractual complexity, interdependence and organizational culture differences (and all control variables) are developed upon particular question items (henceforth referred to as “primary variables”).

All question items correspond to the research context and relate directly to the construct they are intended to measure (de Jong & Klein Woolthuis, 2008b). The reliability of the question items is reported by the Cronbach’s alpha coefficient. Prior studies applying survey methodology in strategic alliance management apply Nunnally’s recommendation of the threshold value of 0.7 for determining the reliability of the question items (Sambasivan et al, 2011; Reuer & Ariño, 2007; Dyer & Chu, 2003; Peterson, 1994). However, it is important to note that this coefficient acts more as a guideline than a rule that every scale has to pass (Dyer & Chu, 2003). The variables, corresponding question items and Cronbach’s alpha are shown in Table 2 in the appendix.

(29)

29

analysis were obtained in two steps. The most common procedure to determine the number of initial factors is the Kaiser criterion, which advises to retain factors whose eigenvalues are higher than or equal to 1. The resulting initial solutions were then, as a second step, rotated using orthogonal varimax rotation in order to improve the degree of fit between the data and the factor structure (Lewis-Beck, 1994). Following the recommendation of Harrington (2009), the question items whose rotated factor loadings are greater than 0.6, and who indicate a standardized scale reliability coefficient (Cronbach’s alpha) of at least 0.7 confirm the assumption that they relate to a single underlying construct and can be used as factor scores for the latent variable. For those constructs which are based on only two question items, the pairwise correlation coefficients are also reported in order to justify the unidimensionality of the items (Hulin et al., 2001). The question item scores that met these criteria were, building on the general method applied in Stata, stored as linear functions of the latent variable. Dependent variables

As it has been indicated in the Hypothesis development section, the division of Model I and Model II was based on the two research streams concerning the governance of strategic alliances. Both models in the thesis analyze the effect of organizational culture differences and interdependence on alliance governance. However, in Model I, interorganizational trust (TRUST) is the dependent variable; whereas in Model II, contractual complexity (CC) is dependent construct.

The variable interorganizational trust (TRUST) is based on five question items. These question items aim to assess whether the relationship between the alliance partners was characterized by, for example, open, honest information sharing and communication. The respondents were asked to indicate their assessment on 5-item Likert-scales. Factor analysis showed that these five questions load on the same single factor with factor scores higher than 0.6. Moreover, Cronbach’s alpha for the scales is 0.73, which indicates satisfying reliability.

(30)

30

contract is detailed, these primary variables have been transformed so that values for non-application are indicated by zero. The sum of these transformed variables resulted in a composite index of contractual complexity ranging from 0 to 13. The Cronbach’s alpha coefficient for the 13 transformed variables representing the contract clauses is 0.83, indicating high scale reliability. Factor analysis of the 13 clauses showed three underlying factors with eigenvalues greater with one. These three factors are in line with the findings of de Jong & Klein Woolthuis (2009), who associate these three factors with three specific contract functions. The function of contracts is analyzed in further detail as part of the additional analysis of the dataset.

Independent variables

Organizational culture differences (OCD) are measured by two question items, which assess whether the respondent experienced any conflict resulting from potential differences in the organizational culture of the focal companies. Principal component factor analysis of the two items resulted in a single factor with eigenvalue higher than 1; moreover, the Cronbach’s alpha coefficient of 0.72 and the two-item correlation coefficient of 0.57 indicate that these two question items provide reliable basis for the variable OCD.

Interdependence (ID) is measured by the two items: the first question item assesses the level of dependence of the respondent firm; whereas the second question aims at revealing how the respondent perceives the dependence of its partner. Although the Cronbach’s alpha of 0.41 is below the satisfactory level of reliability, the factor analysis revealing the single factor with eigenvalue higher than 1, the satisfactory factor loadings and the correlation coefficient of 0.26 as well as the theoretical arguments have led me to retain these questions for developing the measurement for interdependence.

Control variables

The studies concerned with the governance of strategic alliances, organizational culture differences as well as interdependence apply several control variables to ensure the validity of the hypotheses and to rule out other potential ones. In this study, I applied four control variables that might influence the variables and relationships proposed in the theoretical models.

(31)

31

Moreover, smaller firms may lack the experience, resources, or staff to construct more sophisticated alliance agreements (Reuer & Ariño, 2007). Firm size is measured by two items: the number of employees and the turnover of the focal firm in 1997. Both items are measured on reliable scales (Cronbach’s alpha: 0.86) and exceed the threshold value of 0.6 for rotated factor scores. The two-item correlation coefficient is also of satisfactory value (0.76).

Previous alliance history (PREVALL) might influence the governance of the alliance, as the past relationship can serve as basis for mutual understanding and hence drives cognition-based trust (De Jong & Klein Woolthuis, 2008b). Furthermore, previous experience with the alliance partner might be an influential factor regarding the complexity of the contract governing the common activities, as firms familiar with their partners might not find it necessary to develop highly detailed contracts. Both items measuring previous alliance history resulted in factor loadings over 0.6 and a satisfactory coefficient for scale reliability (alpha: 0.72, two-item correlation coefficient: 0.57).

The control variable transaction specific investments (TRANSSPEC) is measured by two items, and is regarded an influential variable because transaction specific, non-recoverable investments to the alliance project might act as means to deteriorate from opportunistic behavior (Parkhe, 1993). It follows from this argument that transaction specific investments might substitute complex contracts in curbing opportunistic behavior, indicated by a negative regression coefficient. The satisfactory Cronbach’s alpha coefficient (alpha = 0.80) and factor scores showed that these items are adequate measurements for transaction specific investments. The two-item correlation coefficient of 0.67 ascertained the assumption of unidimensionality.

The reputation (REPUT) of the partner firm is an important facet of firm cooperation as reputation substitutes for costly mechanisms that act to verify the intentions and monitor the actions of business partners (Arend, 2009). This variable is measured by two items. Although the scale reliability coefficient is lower (0.53) than the threshold value, factor loadings are high and the correlation coefficient of the two items (0.37) is satisfactory. These results and the direct assessment of reputation in the wording of the question items have led me to accept these items as measurements of reputation.

Regressions

(32)

32

Model I: TRUST = βo + β1*OCD + β2ID + β3SIZE + β4REPUT + β5PREVALL + β6TRANSPEC + e

Model II: CC = βo + β1*OCD + β2ID + β3SIZE + β4REPUT + β5PREVALL + β6TRANSPEC + e

As indicated in the description of Model I, I expect to see a negative coefficient for β1

in the first regression which indicates that organizational culture differences decrease interorganizational trust (Hypothesis 1a). The coefficient for interdependence (β2) is expected

to be positive, as interdependence increases interorganizational trust (Hypothesis 1b). In case of Model II regression, the coefficient for OCD (β1) is expected to be positive with contractual

complexity increasing in organizational culture differences (Hypothesis 2a). Interdependence is expected to have a negative coefficient (β2) in Model II since mutually dependent partners

are argued to adapt less detailed contracts (Hypothesis 2b). Moreover, I expect to see positive coefficients for all control variables in the Model I regression, and negative regressors in the Model II regression. The control variables therefore serve as additional means to verify the presumed substitution between trust and complex contracts as the pairwise (Model I-Model II) coefficients of the variables are expected to have opposite signs.

Estimation of the models

Estimation of Model I: Assumptions of least squares estimation

Building on prior research concerning interorganizational trust in strategic alliances, Model I is estimated by ordinary least squares procedure (Gulati & Sytch, 2008; Dyer & Chu, 2003; Kumar et al., 1995). The application of the OLS technique, however, requires specific criteria to be met so that the resulting estimates are reliable. In order to validate the application of the OLS procedure for Model I concerning the effect of organizational culture differences and interdependence on interorganizational trust, four assumptions or ordinary least squares have to be discussed: homoscedasticity, endogeneity, collinearity, and normality (Hill et al., 2011).

Homoscedasticity

(33)

33

observations do not lead to biased estimates, the standard errors computed for the estimations are usually incorrect. Inaccurate standard errors lead to incorrect confidence intervals, t-values and ultimately, might allow for accepting hypotheses that are not true and rejecting ones that are correct in case heteroskedasticity is accounted for (Hill et al, 2011).

In the dataset, the variable SIZE could raise concerns of heteroskedasticity, as bigger firms are usually more complex, have more organizational units involved in their project and inevitably, have several factors influencing the quality of their business relationships, which is captured by the different variance across the error terms. The test for heteroskedasticity in case of variable SIZE was two-fold. Figure 1 in the Appendix presents the graphical test for heteroskedasticity, showing the residuals of the Model I regression plotted against the two primary variables measuring firm size (the number of employees and turnover) and the latent variable (SIZE). From these graphs, I did not obtain certain evidence for heteroskedasticity, and decided to verify the assumption of non-constant variances by the Breusch-Pagan test.

I decided to run the BP test for the latent variable SIZE as well as the underlying question items separately to validate the assumption of non-constant variances. From these tests I concluded that the dataset is heteroscedastic on the 5 per cent confidence level with respect to firm size (measured by the number of employees and also by the multidimensional variable “SIZE”). The Breusch – Pagan test for all variables in the Model I regression also indicated high significance of the heteroskedasticity concern (p = 0.000). In order to circumvent the possible accuracy issues stemming from the difference in the variance of the error terms, I decided to employ White’s robust standard errors in the regression representing Model I.

Endogeneity

The second general assumption of the least squares estimation is that the error terms are not correlated with any of the explanatory variables, which means the explanatory variables are exogenous. Conversely, the endogeneity problem occurs when an independent variable is measured with error and indices correlation with the error term of the regression. In this case, it is very difficult to assess the true impact of the independent variable on the

Variable chi2(1) Prob > chi2

Number of employees 3.71 0.0541

Turnover 6.93 0.0085

SIZE (factor scores) 5.75 0.0165

All variables 83.09 0.000

(34)

34

dependent variable, as its influence is overstated due to an effect that has not been accounted for. For example, if an important variable is omitted from the regression, the estimation can suffer from endogeneity. The issue of endogenous variables leads to inconsistent estimators as the estimation overestimates the effect of the explanatory variables on the variation of the dependent variable values (Hill et al., 2011).

Statistical remedies for the endogeneity problem include instrumental variables estimation (or two-stage least squares estimation) and generalized least squares estimation. However, both of these estimation techniques require an instrumental variable; a variable that is not included in the equation itself but has a strong correlation with the dependent variable (in case of this estimation, interorganizational trust) and can act as an instrument for the variables presumed to be endogenous (Hill et al., 2011). Since such additional variables were not available due to the time-bound nature of the dataset inherent to the data collection process, the assumption of exogenous explanatory variables had to be verified on the basis of the theoretical model. Interdependence and organizational culture differences are facets of cooperation that are themselves affected by multiple factors that cannot be measured directly; therefore, the model might leave room for several unobserved factors affecting interorganizational trust, although this effect cannot be accounted for in this study. The goal of this study is not to reveal all influential factors affecting interorganizational trust but rather grasp a specific, previously overlooked facet of how interorganizational trust develops in alliances in the presence of interdependence and organizational culture differences. Hence, I have decided to maintain the OLS estimation technique, and propose for future research to develop direct measures for these constructs that can rule out any potential of endogeneity. Collinearity

(35)

35

least squares estimation procedure. (Hill et al, 2011). The analysis of this table has lead me to conclude that exact collinearity is not present among the independent variables.

To test the possibility of multicollinearity further, I calculated the variance inflation factor (VIF) coefficient for each explanatory variable. This index provides a measure of “[…] how many times larger the variance of the regressors will be for multicollinear data than for orthogonal data (where each VIF is 1.0)” (Mansfield & Helms, 1982). If the computed value of the VIF index is not unusually higher than one, than multicollinearity is not a problem. The computed VIF values presented in Figure 3 do not significantly exceed the threshold value of 1, implying that multicollinearity is not present and therefore the OLS assumption of non-collinear explanatory variables is satisfied.

Normality:

The normality assumption of the OLS estimation is a generally optional criterion that dictates that the values of the dependent variable should be normally distributed around their mean value. Normality also implies that the values of the error term are characterized by normal distribution, since the error term is a linear function of the dependent variable (Hill et al, 2011). The normality assumption is important as it influences the p-values indicating the significance of the regressors. However, since the variable TRUST represents factor scores of the principal component factor analysis, the distribution of this latent construct does not necessarily satisfy the normality assumption (see Figure 4 and 5 in the Appendix for graphical representation of the normality test). Nevertheless, I concluded that the major criteria of the OLS estimation have been satisfied, and the OLS method is suitable for estimating the Model I equation.

Estimation of Model II: Ordered logistic model

The estimation of the model concerning the effect of organizational culture differences and interdependence on contractual complexity (Model II) is conducted by ordered logistic

Variable VIF 1/VIF

ID 1.04 0.958263 OCD 1.04 0.961081 SIZE 1.04 0.962234 TRANSPEC 1.03 0.969518 PREVALL 1.07 0.934739 REPUT 1.07 0.938655 Mean VIF 1.05

(36)

36

regression. This estimation technique is in line with Reuer & Ariño (2007), who employ ordered logistic regressions in models relying upon a non-weighted measure of contractual complexity as the dependent variable. The single requirement for the application of the ordered logistic model is that the dependent variable represents a categorical construct whose values are ordinal in nature (Hill et al., 2011). This criterion has been met by constructing a single index for contractual complexity, which takes values 0 to 13, indicating the extent of detail in the formal contract.

V. Results

Descriptive statistics

The previous section described the methodological issues and processes that have led to the construction of the variables subjected to the two regressions corresponding to the two models. However, since the final dependent, independent and control variables represent factor scores of the principal component factor analysis, the description of the dataset along these constructs does not allow for as tentative conclusions about the sample than the values of the primary variables corresponding to the question items. Therefore, this section (based on Table 1) is dedicated to the brief description of the primary variables that represent the coded responses of the survey.

Table 1. Descriptive statistics

Latent variable name Primary variable name Obs. Mean Std. Dev. Min Max

TRUST Trust_care 388 4.480 0.955 1 5

Trust_misled 390 4.395 1.096 1 5

Trust_open 390 4.605 0.797 1 5

Trust_info 391 4.662 0.740 1 5

Trust_criticism 389 4.769 0.490 2 5

CC Composite index of contractual complexity 234 9.338 3.269 0 13

ID ID_wegave 370 2.932 1.683 1 5 ID_partnergave 382 2.382 1.544 1 5 OCD OC_culturaldiff 390 1.628 1.139 1 5 OC_ineffcomm 391 1.491 1.015 1 5 SIZE Size_employee 389 2.843 1.319 1 5 Size_turnover 333 3.868 1.273 1 5 PREVALL Prevall_continuity 389 3.625 1.576 1 5 Prevall_friendly 389 3.689 1.562 1 5 REPUT Reput1 379 4.087 1.293 1 5 Reput2 384 3.594 1.474 1 5 TRANSSPEC Transspec1 381 2.420 1.700 1 5 Transspec2 359 2.230 1.659 1 5

(37)

37

mean values for each question item are fairly close to 5, the assessment of high level of trust. The composite index for contractual complexity (CC) represents a mean value of 9.34 with standard deviation of 3.27, indicating a wide array of fairly detailed contracts among the observed alliance relationships.4 These statistics have led me to conclude that the alliances under observation are characterized by high interorganizational trust and relatively detailed contracts. These characteristics of the sample are favorable as they allow for managerial implications as to how high level of interorganizational trust develops and is maintained in alliances. Furthermore, it allows for the practical assessment of the transaction cost paradigm and the characteristics of complex contracts in high-tech alliances.

The primary variables measuring organizational culture differences (OCD) indicate that the alliances under study experienced relatively low level of conflict stemming from organizational culture differences as the variables indicating conflict ranked around the mean value of 1.5. I decided nevertheless to retain these question items for representation of organizational culture differences in the primary analysis of the theoretical models. However, in order to assess further the importance of organizational culture differences in strategic alliances, I constructed an alternative variable for the measurement of organizational culture differences by factoring the log of the respective question item values items into a single variable. This variable represents how the increase in organizational culture differences affects alliance governance. The analysis of this alternative construct is conducted as part of the robustness analysis of the first results.

The primary variables assessing the power balance in the cooperative relationship indicate moderate level of dependence among the alliances under study (mean values of the two items are 2.93 and 2.38). However, pairwise interpretation of these variables shows that there is little difference between the extent to which a firm depends on the partner (the first interdependence variable), and the extent to which the partner is perceived to depend on the respondent firm (represented by the second variable), as the mean values of the two variables are fairly close. This characteristic of the sample can be interpreted as the firms under study experiencing dependence to be fairly mutual in their alliance. Further analysis of the pairwise interpretation of interdependence is conducted in the next section, where I develop an

(38)

38

alternative measure for interdependence assessing the pairwise difference between the question items.

The control variable SIZE consist of two items; one item measures the turnover of the firm in 1997, and one question indicates the number of employees in 1997. From the descriptive statistics I concluded that, although both question items indicate average size for firms across the sample (mean value 2.84 and 3.87), the standard deviations (1.32 and 1.27, respectively) ascertain the concern about diverse firm size discussed in the previous section and validate the application of heteroskedasticity-consistent standard errors in the OLS regression. According to the summary statistics of the question items representing reputational concerns (REPUT) and pervious relationship history (PREVALL), the alliances under study are characterized by good reputation (mean value 4.09 and 3.59); moreover, the majority of the partnerships stemmed from previous relationships (the mean value of the primary variables representing previous acquaintance faired around 3.5). Transaction specific investments (TRANSSPEC) faired around the mean value of 2.5, indicating moderate importance of relationship-specific investments in the context of Dutch high-tech alliances.

Regression results

Trust and Contracts in Strategic Alliances