The Determinants of the Holy Grail in Interorganizational Relationships: Cooperation.

The Determinants of the Holy Grail in

Interorganizational Relationships:

Cooperation.

A contingency perspective.

Master Thesis MSc Business Administration:

Organizational and Management Control

June 2016

Faculty of Economics and Business

Supervisor: A. Rehman Abbasi Co-assessor: Dr. S. Girdhar

by

M.M. Oude Sogtoen

Student number: 2021927

Word count: 16,981

2

Abstract

Given its relatively high failure rate, cooperation turns out to be neither automatic nor easily fostered in interorganizational relationships (IORs). Hence, effective IOR governance is sought after by both academics and practitioners. In this vein, the importance of contractual and relational governance, and more specifically of formal contracts and trust, has been stressed. Yet, their interrelationship remains a matter of ongoing debate in the extant literature. In the quest for empirical evidence, the multidimensionality of trust and the influence of boundary spanners have often been neglected by scholars. Using survey data of 83 respondents from the technological and health industry in the Netherlands, the relationship between trust and cooperation was examined as well as the interactions of trust with formal contract and contact intensity with the other party’s boundary spanner. The findings confirm the effectiveness of trust as governance mechanism in IORs. In this sense, competence trust was found to be particularly effective in promoting a cooperative orientation. Moreover, the impact of formal contract and contact intensity on the trust-cooperation linkage was found to be contingent on the trust dimension involved. Overall, trust emerged as dominant influence on the level of cooperation in IORs.

Keywords: contractual governance, relational governance, formal contract, goodwill trust,

3

Table of Contents

1 Introduction ... 5

2 Literature review ... 9

2.1 Interorganizational Relationships (IORs)... 9

2.2 Governance Mechanisms: Contractual and Relational... 9

2.3 Formal Contract... 11

2.4 Trust... 12

2.4.1 Dimensions of trust... 13

2.5 The Interplay between Formal Contract and Trust... 13

2.6 Conceptual Model and Hypotheses Development... 15

2.6.1 Cooperation as IOR outcome... 15

2.6.2 Different dimensions of perceived risk in IORs... 16

2.6.3 The influence of trust... 16

2.6.4 The influence of formal contract... 18

2.6.5 The influence of boundary spanners... 21

3 Methodology... 24

3.1 Design and Sample ... 24

3.2 Data Collection... 25 3.2.1 Survey design ... 26 3.3 Measurement... 28 3.3.1 Dependent variable... 28 3.3.2 Independent variables... 28 3.3.3 Moderators... 29 3.3.4 Control variables... 29 3.4 Data Analysis ... 31 4 Results…... 33 4.1 Factor Analysis... 33 4.2 Reliability Analysis... 34

4.3 Hierarchical Quantile Regression Analyses... 35

4.3.1 The effect of trust on the level of cooperation... 35

4.3.2 The joint effect of formal contract and trust on the level of cooperation... 36

4.3.3 The effect of contact intensity on the level of cooperation... 39

4

4.3.5 The effect of control variables... 40

5 Discussion and Conclusion... 42

5.1 Theoretical Implications... 44

5.2 Managerial Implications... 45

5.3 Limitations and Future Research Directions... 46

References... 48

5

1. INTRODUCTION

The last two decades have been evidence for the omnipresence of interorganizational relationships (IORs) in both theory and practice. Not only the formation rate of IORs skyrocket in recent years, it also became a research topic of substantial interest in a variety of research disciplines (Barringer & Harrison, 2000; Coletti, Sedatole, & Towry, 2005; Dekker, 2004; Mellewigt, Madhok, & Weibel, 2007; Vlaar, Van den Bosch, & Volberda, 2007). As a result of this proliferation in both the business and academics world, a renewed interest has been born for the control-trust nexus and this relationship returned to the forefront of debate (Cao & Lumineau, 2015; Dekker, 2004; Reuer & Ariño, 2007).

This extensive attention devoted by scholars seems justly given the critical role that formal control – and more specifically formal contracts – and trust play in the success or failure of IORs (Cao & Lumineau, 2015; Dekker, 2004; Vélez, Sánchez, & Álvarez-Dardet, 2008). According to the corresponding literature these concepts represent two main types of governance involved in IORs. One is known as contractual governance, i.e. formal contracts, while the other one is entitled as relational governance, i.e. trust (Cao & Lumineau, 2015; Poppo & Zenger, 2002). It has been widely acknowledged that these types of governance are interrelated in managing and supporting IORs. Moreover, scholars have argued that both are of substantial importance for the performance of IORs (Blomqvist, Hurmelinna, & Seppänen, 2005; Cao & Lumineau, 2015; Dekker, 2004). Yet, effective IOR governance seems to be elusive given the relatively high failure rate of these interfirm collaborations (Barringer & Harrison, 2000; Das & Teng, 1998a; Dekker, 2004; Faems, Janssens, Madhok, & Van Looy, 2008).

6 van der Meer-Kooistra, 2009; Woolthuis, Hillebrand, & Nooteboom, 2005; Yang, Zhou, & Jiang, 2011; Zhou & Xu, 2012). More specifically, these researchers have argued that both complementary and substitute propositions are possible depending upon other contingent and contextual factors.

A contributor to this controversy in the academic literature is the mixed empirical evidence. The empirical results about the relationship between formal control and trust are inconsistent and therefore inconclusive (Bijlsma-Frankema & Costa, 2005; Cao & Lumineau, 2015; Mellewigt et al., 2007; Vlaar et al., 2007). As a result, “empirical research has not yielded decisive support for one stance over another” (Bijlsma-Frankema & Costa, 2005, p. 270). In order to comply with the expressed need for clarification it has been deemed necessary by many scholars to ascertain the relationship between formal control and trust. However, in this spirit many scholars have focused on the constructs themselves and thereby overlooked the multidimensionality of trust. According to Lui and Ngo (2004) this negligence has nourished the ambiguous empirical findings. Hence, disentangling the distinct forms of trust in the control-trust nexus seems to be a fertile research avenue (Cao & Lumineau, 2015; Das & Teng, 2001; Zaheer & Harris, 2006).

7 opportunism. Contracts as formal control mechanism might only have an indirect impact on opportunism via trust. Since opportunism can be regarded as the opposite of cooperation investigating the trust-cooperation link will refine the current understanding of the effectiveness of trust as governance mechanism (Das & Teng, 1998a). Finally, the effect of contact intensity with the partner firm’s boundary spanner on this trust-cooperation linkage will be probed. Boundary personnel are the representatives of the parties involved in interorganizational exchanges and are claimed to play a prominent role in managing IORs (Schilke & Cook; Zaheer, McEvily, & Perrone, 1998). Nevertheless, the literature regarding the governance of IORs has often neglected their impact (Ferguson, Paulin, & Bergeron, 2005; Haytko, 2004).

Besides the above-mentioned theoretical contributions, this study will inform and guide managerial practice as well. The governance mechanisms chosen by practitioners are important determinants for the success of the relevant IOR (Faems et al., 2008). Yet, the current equivocal findings are likely to confuse managers in their governance of IORs (Cao & Lumineau, 2015). Since this research aims to broaden the understanding of IOR governance, specific practical guidance important for the effective management of interfirm collaborations will be provided. In light of the increasing number of IORs and the accompanied increased potential for misuse and failure, this seems to be particularly valuable (Barringer & Harrison, 2000).

The study’s aim gives rise to the following research questions:

What is the influence of both types of trust on the level of cooperation in IORs?

What is the influence of formal contract on the effect of trust on the level of cooperation in

IORs?

What is the influence of contact intensity with a firm’s boundary spanner on the relationship between trust in the firm and the level of cooperation in IORs?

9

2. LITERATURE REVIEW

In this section the most important concepts within this research will be explained. First, a theoretical background will be provided by means of addressing the current literature regarding these concepts. Moreover, the applied concepts will be defined accordingly. The section ends with the development of the hypotheses and the ensuing conceptual model.

2.1 Interorganizational Relationships (IORs)

Aforementioned, the unit of analysis of this study concerns the IOR. As strategically important relationships, IORs can be defined as “cooperative relationships between a focal organization and one or more other organizations to share or exchange resources with the goal of improved performance” (Parmigiani & Rivera-Santos, 2011, p. 1109). These collaborative arrangements exist in a variety of forms and are initiated for many reasons (Dekker, 2004; Mellewigt et al., 2007; Parmigiani & Rivera-Santos, 2011). The most prevalent forms of IORs are known as (strategic) alliances, joint ventures, buyer-supplier agreements, licensing, co-branding, franchising, cross-sector partnerships, networks, trade associations, and consortia (Barringer & Harrison, 2000; Parmigiani & Rivera-Santos, 2011). As noted by Parmigiani and Rivera-Santos (2011), scholars tend to view particular forms from certain theoretical perspectives, e.g. agency theory with franchising. As a result, the formation of IORs has been explained by a broad array of theoretical motivations ranging from an economic rationale towards a more behavioral rationale (Barringer & Harrison, 2000; Parmigiani & Rivera-Santos, 2011). Hence, both economic and sociological forces help explain the existence of IORs (Parmigiani & Rivera-Santos, 2011).

2.2 Governance Mechanisms: Contractual and Relational

10 have devoted substantial attention to the governance of IORs, a variety of terms has been used in the respective studies to refer to these prevailing types of governance. For instance, the synonyms “legal safeguards” (Achrol & Gundlach, 1999), “contractual safeguards” (Lui & Ngo, 2004), “formal control” (Li, Xie, Teo, & Peng, 2010; Yang et al., 2011), “explicit contract” (Zhou & Poppo, 2010), and “formal contract” (Burkert et al., 2012; Mellewigt et al., 2007) all refer to contractual governance. Relational governance on the other hand, is also called “social safeguards” (Achrol & Gundlach, 1999), “social control” (Li, Xie, et al., 2010), “informal control” (Dekker, 2004), “social mechanisms” (Zhou & Xu, 2012), or “relational mechanisms” (Li, Poppo, & Zhou, 2010).

Regardless of how researchers have named contractual and relational governance, their definitions all point in similar directions. This study will adopt the definitions emerged from the qualitative review and meta-analysis of Cao and Lumineau (2015). Accordingly, contractual governance refers to “the extent to which roles, obligations, responsibilities, contingency adaptation, and legal penalty are specified or well-detailed in formal agreements” (Cao & Lumineau, 2015, p. 24) whereas relational governance is defined as the degree to which trust, shared norms, informal rules and procedures govern the relationship (Cao & Lumineau, 2015; Poppo & Zenger, 2002; Zhou & Xu, 2012). Although interrelated, the aforementioned definitions reflect clearly the distinction between the types of governance that are at play in IORs. As such, contractual and relational governance represent a dichotomy, e.g. reliance on formal structure versus reliance on informal structure, and third-party enforcement versus self-enforcement (Dyer, 1997; Malhotra & Murnighan, 2002).

11 latter theoretical lens have conceptualized relational governance with shared norms, i.e. social norms (e.g. Jap & Ganesan, 2000; Lusch & Brown, 1996). Still others have used a more umbrella conceptualizing of relational governance in their empirical studies implying that the researchers did not explicitly state whether the construct is trust-oriented or norm-oriented but instead simply referred to the concept itself (e.g. Zhou & Xu, 2012).

Nonetheless, trust seems to be the principal governance mechanism associated with relational governance in the extant literature (Cao & Lumineau, 2015). Poppo and Zenger (2002) claim that empirical studies predominately associate relational governance with trust. In line, Dekker (2004) states that trust is the principal mode of social control, i.e. relational governance. Hence, the debate essentially boils down to the relationship between these two prevailing governance mechanisms: formal contracts and trust.

2.3 Formal Contract

Due to the ongoing control-trust debate in the IOR literature, formal contract has received as distinct concept a fair amount of scholars’ attention. Different from trust, formal contract is considered to be a control mechanism and more specifically a formal control mechanism (Dekker, 2004; Vlaar et al., 2007; Woolthuis et al., 2005; Yang et al., 2011). According to Li et al. (2010), and Malhotra and Murnighan (2002), the use of contracts is the most prevalent and pervasive among the formal control mechanisms. Likewise, Dyer (1997) states that the formal contract is the predominantly safeguard employed in Western countries.

12 Aforementioned reveals two distinctly held perspectives in the extant literature: the structural and functional approach. The former views contracts primarily as a safeguard mechanism and emphasizes the structure, i.e. design of contracts (Schepker, Oh, Martynov, & Poppo, 2014). Conversely, as the name suggests the functional approach stresses the specific functionality of contracts which can be distinguished in safeguarding, coordination, and adaptation (Schepker et al., 2014; Woolthuis et al., 2005). Scholars holding this latter perspective contend that the predominantly conceptualization of formal contracts as safeguarding mechanism instrument is short-sighted because contracts serve more than one function (Malhotra & Lumineau, 2011; Mellewigt et al., 2007; Schepker et al., 2014; Woolthuis et al., 2005). For instance, the specification of roles and responsibilities serves the coordination function of contracts whereas the provision for contingency adaptation supports the adaptation function (Schepker et al., 2014). The so-called functional approach is not only reflected in some of the above-mentioned definitions, it is also present in this study’s definition of contractual governance. Since formal contract has been treated as both synonym and conceptualization of contractual governance they are used interchangeable in this study.

2.4 Trust

Next to contracts, managers tend to rely on trust to regulate a partner’s behavior in interfirm collaborations (Lui & Ngo, 2004). In the IOR context, trust has been argued to be “the glue that keeps the partners of IORs together” (Mellewigt et al., 2007, p. 837). As such, it is seen as a particularly important factor for the performance of IORs, i.e. its continuation and/or success (Ashnai, Henneberg, Naudé, & Francescucci, 2016; Child, 2001; Zaheer et al., 1998). Despite the importance of trust and the tremendous attention received from scholars, an agreement on the exact definition of trust is lacking (Bijlsma-Frankema & Costa, 2005; Pavlou, 2002). A diversity of trust definitions has been provided in the literature varying across research disciplines and levels of analysis (Das & Teng, 2001; Faems et al., 2008; Malhotra & Murnighan, 2002; Rousseau, Sitkin, Burt, & Camerer, 1998). In spite of these different definitions abound in the literature, scholars tend to agree on two elements critical for defining trust: positive expectations and the willingness to become vulnerable (Bijlsma-Frankema & Costa, 2005; Malhotra & Murnighan, 2002; Rousseau et al., 1998).

13 trustor, irrespective of the ability to monitor or control that other party” (p. 712), i.e. the trustee. Hence, the definition covers both main elements. Considering its popularity and frequent usage in the literature, this definition seems to be proper.

2.4.1 Dimensions of trust

In the perspective of the control-trust nexus, the solid majority has treated trust as a unidimensional construct (Cao & Lumineau, 2015; Jiang et al., 2013; Lui & Ngo, 2004). Only a few scholars have made the explicit distinction between different types, i.e. dimensions. In the extant literature two dimensions of trust have predominantly been acknowledged, known as competence trust and goodwill trust (Das & Teng, 2001; Lui & Ngo, 2004; Sako, 1992; Şengün & Wasti, 2007). This study follows the definitions of Lui and Ngo (2004, p. 474) who refer to competence trust as “the expectation that partners have the ability to fulfill their roles” and is also named capability trust (Dekker, 2004). Goodwill trust on the other hand is defined as “the expectation that another intends to fulfill their role in the relationship” (Lui & Ngo, 2004, p. 474) and is also called intentional trust (Woolthuis et al., 2005).

The omission to account for this multidimensionality of trust has led to ambiguous conclusions about the relationship between trust and formal contracts (Lui & Ngo, 2004). Therefore, this study accounts for both types of trust implying that whenever the term trust is used, this denotes both competence and goodwill trust unless otherwise explicitly stated.

2.5 The Interplay between Formal Contract and Trust

A considerable amount of research (Burkert et al., 2012; Cao & Lumineau, 2015; Das & Teng, 1998a; Li et al., 2010; Li, Xie, et al., 2010; Mellewigt et al., 2007; Poppo & Zenger 2002; Woolthuis et al., 2005; Yang et al., 2011; Zhou & Poppo, 2010) has been devoted to formal contract and trust, and their interplay in IORs. Yet, their relationship remains a matter of dispute among academics.

14 the relationship: “replacing” and “dampening” (Huber et al., 2013). The replacing mechanism implies that both instruments of governance replace each other due to the fact that they are functional equivalents (Huber et al., 2013). To illustrate, as soon as trust is well developed, contractual governance will be redundant given that trust as relational governance mechanism is able to solely govern the relationship in an effective manner (Dyer, 1997; Gulati, 1995; Wang et al., 2011). The dampening mechanism on the other hand explains the occurrence of substitution by means of the pernicious effects of one governance mechanism on the other’s strengths or bases (Cao & Lumineau, 2015; Huber et al., 2013). For instance, it has frequently been argued that contracts harm the development of trust due to its signal of distrust (e.g. Huber et al., 2013; Poppo & Zenger, 2002; Yang et al., 2011).

Conversely, the other stream of research pursues the complementary perspective and contends a positive relationship between the two types of governance. Proponents of this complementary view have argued that these main types of governance complement each other by means of the “enabling” mechanism. This mechanism suggests that each creates favorable conditions that facilitate the other type of governance (Huber et al., 2013). Another line of reasoning used to justify the complementary nature of the relationship is based on the “compensating” mechanism. This mechanism relates to the ability of both types of governance to address the limitations of the other, i.e. to compensate for each other weaknesses, given that each type of governance has its own unique strengths (Huber et al., 2013). Nourished by the inconclusive evidence provided by empirical research, these two opposing views have led to an ongoing debate regarding the control-trust nexus in IORs.

15 Based on the above, this study follows the latter stream of studies by holding a nuanced view regarding the nature of the nexus in the development of the hypotheses.

2.6 Conceptual Model and Hypotheses Development

Now that the study’s main constructs have been clarified and defined, the section will proceed with the development of the hypotheses and closes with the resulting conceptual model.

2.6.1 Cooperation as IOR outcome

A broad array of studies has effectively shown the effects of formal contract and trust on IOR performance, and more specifically on opportunism (Liu, Luo, & Liu, 2009; Lui, Wong, & Liu, 2009; Yang et al., 2011), performance satisfaction (Lui & Ngo, 2004; Lui et al., 2009; Poppo & Zenger, 2002), innovation performance (Wang et al., 2011), relationship outcomes (Woolthuis et al., 2005; Yang et al., 2011) and relationship performance (Chen et al., 2013). However, in this light scholars tend to overlook the fundament of the cooperative arrangements itself as another, important, and distinct “performance” outcome, namely cooperation. In essence the motive for entering into an IOR in the first place is to reap the benefits of cooperation (Coletti et al., 2005; Das & Teng, 1998b).

Cooperation is “the willingness of a partner firm to pursue mutually compatible interest” (Das & Teng, 1998a, p. 492). In the IOR context, the presence of cooperation is apparent from the coordinated actions performed by the exchange partners with the purpose of achieving shared goals or mutual outcomes (Doucette & Wiederholt, 1997; Lui et al., 2009; Pearce, 2001). In this vein, Heide and Miner (1992), Lui et al. (2009), and Pearce (2001) contend that the coordinated actions taken in terms of flexibility, information exchange, and shared problem solving are important behavioral manifestations of cooperation in IORs.

16 since the lack of cooperation and the failure to establish a cooperative relationship due to opportunistic behavior has often been cited as causes of IOR’s premature ending, i.e. its failure (Das & Teng, 1998a).

2.6.2 Different dimensions of perceived risk in IORs

As the term suggests, interorganizational relationships (IORs) involve a relationship between two or more partner firms. The key attribute that set these strategically important relationships apart from single-firm strategic manoeuvres is the reliance on partner cooperation (Das & Teng, 1996, 1998a). Hence, a major concern in IORs is the absence of such desired cooperation, i.e. the potential of opportunistic behavior by the partner firm. This so called relational risk is related to goodwill trust and is unique to IORs (Das & Teng, 1996).

However, not merely the exposure of relational risk makes the IOR vulnerable to failure (Coletti, 2005). Another uncertainty faced by the exchange partners is performance risk. Different from the former, performance risk is related to competence trust and is not unique to IORs but inherent in all other types of arrangements and strategies (Das & Teng, 2001). Performance risk refers to the usual risk of unsatisfactory business performance and as such relates to the probability and consequences that the partner firm is not able to achieve mutual goals, given full compliance (Das & Teng, 2004).

From the above it becomes apparent that risk in IORs consists of two types: relational and performance risk. Das & Teng (2001) argue that both trust and formal control are determinants of the perceived level of risk which is the estimated level of (objective) risk by the exchange partner. This implies that both governance mechanisms – trust and formal contract – might impact the perceived risk in interfirm relationships, although in different manners. To elaborate, trust might lead to low risk perceptions by influencing the partner’s expectations whereas formal contracts influences the partner’s behavior and as such leads to a perception of lowered risk (Das & Teng, 2001).

2.6.3 The influence of trust

17 neither of them are complete governance mechanism in their own right. Nevertheless, the underlying mechanisms and their impact on cooperation is expected to differ.

Empirically, studies have pointed to the trust-cooperation link. For instance, trust in IORs increases relational behaviors of information sharing (Ashnai et al., 2016; Sezen & Yilmaz, 2007), communication (Anderson & Weitz, 1989), flexibility (Johnston, McCutcheon, Stuart, & Kerwood, 2004), and solidarity, e.g. shared problem solving (Sezen & Yilmaz, 2007). Since scholars (Heide & Miner, 1992; Pearce, 2001) have asserted that these relational behaviors are key manifestations of cooperation in interfirm exchanges, it is expected that trust has a positive effect on the level of cooperation. To elaborate, trust leads to confidence in partner cooperation (Das & Teng, 1998a), reduces outcome uncertainties (Mayer et al., 1995), increases predictability (Sako, 1994), and fosters risk taking-behavior (Das & Teng, 2004; Morgan & Hunt, 1994; Şengün & Wasti, 2007; Yilmaz, Sezen, & Ozdemir, 2005). Hence, trust seems to stimulate to act in the spirit of cooperation. As a matter of fact, trust is claimed to promote a cooperative orientation (Mayer et al., 1995; Sezen & Yilmaz, 2007). Since scholars have often treated trust as a unidimensional construct, it is unclear which type of trust is the underlying driver. It is in this spirit that the importance of the multidimensionality of trust and accordingly its consideration comes into play.

18 Yilmaz et al., 2005), a sense of trust is needed to exert risk-taking behaviors such as cooperation. More formally stated:

H1a: Trustor’s goodwill trust in the trustee’s firm has a positive effect on the level of trustor’s cooperation IORs

H1b: Trustor’s competence trust in the trustee’s firm has a positive effect on the level of trustor’s cooperation in IORs

Yet, relational risk is unique to interfirm cooperation (Das & Teng, 1996, 2001; Gallivan & Depledge, 2003) and thus of particular concern in this context. More specifically, relational risk is created by and a consequence of the IOR formation whereas performance risk arises from factors extraneous to cooperation (Das & Teng, 1998b). As a result, the latter is seen as a usual and unavoidable type of risk and therefore largely taken as given by exchange partners (Das & Teng, 2001; Şengün & Wasti, 2007). As such, a reduction in perceived relational rather than performance risk is posited to have a particular strong impact on trustor’s cooperative orientation and behavior. To that end, goodwill trust is claimed to be a critical means in mitigating perceived relational risk (Das & Teng, 1996, 2001, 2004; Lui & Ngo, 2004). The findings of Liu, Li, Tao, and Wang (2008) empirically affirm this claim. Moreover, Şengün & Wasti (2011) have found a positive relationship between goodwill trust and risk-taking tendency whereas no significant effect was found between competence trust and risk-taking tendency. Hence, goodwill trust as relational governance mechanism is supposed to be more effective than competence trust in promoting cooperation. In light of the above, the following is hypothesized:

H1c: The effect of trustor’s goodwill trust in the trustee’s firm on the level of

trustor’s cooperation in IORs is stronger than the effect of trustor’s competence trust

in the trustee’s firm

2.6.4 The influence of formal contract

19 studies found no direct and significant impact of contractual governance on opportunism and/or cooperation. To elaborate, the meta-analysis of Cao and Lumineau (2015) and the findings of Lui et al. (2009) demonstrate that formal contracts have no significant, direct impact on opportunism. Accordingly, Yang et al. (2011) have shown that formal contract not significantly reduces opportunistic behaviors in IORs with strong ties. Also Zhou and Xu (2012) claim that contracts alone are not an effective governance mechanism to curb opportunism. Since opportunism is defined by Williamson (1975) as pursuing self-interest with guile, it can be conceived as the opposite of cooperation (Das & Teng, 1998a). Therefore, forbearance of opportunism can be viewed as an indication for a partner’s willingness to serve mutual interest, i.e. to cooperate. Moreover, Lusch and Brown (1996) found no significant effect of the extent of contractual governance on relational behavior, i.e. cooperative behavior, in the wholesaler-supplier relationship.

The above findings demonstrate the need for trust as complementary governance mechanism to promote cooperation and/or to dampen opportunistic behavior in IORs. Scholars have claimed not only that trust curtails opportunism but also showed its positive effect on cooperative behavior in for instance procurement relationships (Lui et al., 2009) and relationship marketing (Morgan & Hunt, 1994). As noted by Rindfleisch (2000), trust has emerged as the central means of establishing and sustaining cooperation in all types of interfirm relationships. As such, trust rather than formal contract is posited to have a direct effect on the level of cooperation. More specifically, formal contract is seen as moderator on the trust-cooperation link. In this vein, the interaction with the distinguished trust dimensions is expected to differ and hence the interaction effect on the level of cooperation in accordance.

20 ability to reduce both perceived relational and performance risk in IORs. Therefore, it is posited that formal contracts might substitute and/or complement trust depending on the trust dimension involved. To elaborate, this implies that in terms of goodwill trust, formal contracts have the ability to compensate for the former’s weakness of merely reducing relational risk. Moreover, scholars have contended that from the coordination perspective formal contracts might be considered as a sign of commitment rather than a sign of distrust (Cao & Lumineau, 2015; Woolthuis et al., 2005). Additionally, it has been claimed that due to the explicit statement of expectations and punishment in contracts, the effects of goodwill trust on reducing relational risk and curtailing opportunism are improved (Liu et al, 2009; Zhou & Xu, 2012). Hence, formal contract complements goodwill trust in fostering a cooperative orientation in IORs. More formally stated:

H2a: Formal contract positively moderates the relationship between trustor’s goodwill trust in the trustee’s firm and the level of trustor’s cooperation in IORs

21

H2b: Formal contract negatively moderates the relationship between trustor’s competence trust in the trustee’s firm and the level of trustor’s cooperation in IORs

2.6.5 The influence of boundary spanners

Boundary spanners are known as the key individuals in an interfirm relationship that provide the linkage between the focal and partner organization (Vanneste, 2016). These agents acting on the boundary of organizations are also called “boundary-spanning personnel” (Huang, Luo, Liu, & Yang, 2013), “liaison personnel” (Anderson & Weitz, 1989), and “trust guardians” (Child, 2001). These key representatives might vary from top executives and department heads at higher levels to project managers, supplier representatives, salespersons et cetera at lower levels (Huang et al., 2013; Schilke & Cook, 2013).

22 According to Gulati (1995), repeated interaction between firms can incrementally build interfirm trust. In line, Shapiro et al. (1992) argue that improved trust emerges as firms repeatedly interact due to the many opportunities for dialogue inherent in regular communication. As a result, firms develop “knowledge-based trust” which is based on the premise of trust emerging from predictability (Shapiro et al., 1992). In addition, Nicholson, Compeau and Sethi (2001) state that frequent contact results in more opportunities to observe the boundary spanner’s behavior which leads to an improved ability of the other firm to predict outcomes or behaviors in future interactions. Similarly, Doney and Cannon (1997) suggest that frequent interactions with a firm’s boundary spanner foster trust as it provides the other party with information which facilitates the ability to predict. In other words, frequent contact with the trustee’s boundary spanner enables the trustor to collect evidence about the trustee’s firm trustworthiness and competence across a variety of situations. Consequently, this might results in reduced relational and performance risk perceptions.

Hence, frequent communication with the trustee’s boundary spanner – as primary point of contact (Doney & Cannon, 1997) – can foster both goodwill and competence trust. Subsequently confidence in trustee’s cooperation is increased. This certainty about reciprocate cooperative behavior strengthens the firm’s willingness to cooperate (Luo, 2002; Shapiro et al., 1992). As stated by Fryxell, Dooley, and Vryza (2002) confidence in the partner’s cooperation enhances honest and open information exchange and cooperation. Moreover, due to regular contact exchange partners might signal future intentions and commitment (Mohr & Spekman, 1994), create mutual understanding (Doney & Cannon, 1997), and learn about each other (Gulati, 1995). In addition, as noted by Nicholson et al. (2001) frequent contact is an important relationship signal that demonstrates the importance of the trustor to the trustee and vice versa. Aforementioned benefits are likely to spur a cooperative orientation. In fact, Heide and Miner (1992) empirically found a positive association between frequency of contact and the level of flexibility and shared problem solving in the buyer-seller relationship. Based on the above discussion, it is hypothesized that:

23 CONTROL VARIABLES  Trust propensity  Interdependence  Asset specificity  Long-term orientation  Industry TRUST IN PARTNER FIRM

FORMAL CONTRACT

H3b: High contact intensity with the trustee’s boundary spanner positively moderates the relationship between trustor’s goodwill trust in the trustee’s firm and the level of

trustor’s cooperation in IORs

H3c: High contact intensity with the trustee’s boundary spanner positively moderates the relationship between trustor’s competence trust in the trustee’s firm and the level of trustor’s cooperation in IORs

Goodwill trust Competence trust LEVEL OF COOPERATION H1a,b,c H2a H3b,c

+

+

+

+

Figure 1. Conceptual Model

Note. The broken line indicates that based on the extant literature no direct relationship was hypothesized

for formal contract on the level of cooperation.

24

3. METHODOLOGY

This section sheds light on the study’s methodology and starts with an explanation and justification of the research’s design and sample followed by a description of the data collection. The third part concerns the measurement of the variables. The section ends with an explanation of the data analyses. As such, the methodology section includes a detailed description of how the research was conducted thereby complying with a principal quality criterion in research: controllability (Van Aken et al., 2012).

3.1 Design and Sample

Due to the considerable popularity of interfirm collaborations in the business world, IORs are a phenomenon faced by many companies. Yet, given the relatively high failure rate of IORs this popularity seems to be somewhat paradoxical (Barringer & Harrison, 2000). Triggered by the above-mentioned business phenomenon, theory testing has been used as the study’s knowledge-generating research process. As a result, the study focused on the second part of the empirical cycle and followed in succession its latter steps of deduction, testing and evaluation (Van Aken et al., 2012). According to Van Aken et al. (2012) theory testing is the appropriate research approach in this case since the literature field is elaborated and mature yet the empirical evidence is still inconclusive. Moreover, the research questions concern a ‘what’-question which is perceived as particularly suitable for the theory testing research approach.

25 technical self-sufficiency (Fusfeld, 1986). Hence, both industries guarantee a broad application of IORs.

Due to the specific context of the study and the associated time constraints, convenience sampling has been used rather than random sampling. The multinational company L’Oréal is known as the worldwide market leader in both the cosmetic and dermo-cosmetic industry. Consequently, it is a main provider of dermo-cosmetics in Dutch pharmacies. Therefore, L’Oréal served as database to trace potential respondents, i.e. pharmacies. The Dutch FME functioned as another database since its members are organizations active in the technological industry of the Netherlands. The association is known to be the largest of its kind and represents a significant part of the technological industry in the Netherlands.

3.2 Data Collection

To test the hypotheses primary data had to be generated. Ilieva, Baron, and Healey (2002) have found that online data collection methods are superior to postal methods in several ways including higher data entry efficiency, shorter response times, higher response rates and lower costs. Also Wright and Schwager (2008) assert its superiority by highlighting the reduced involvement of the researcher. This increased researcher independency of results serves the reliability quality criterion of research (Van Aken et al., 2012). Moreover, Wright and Schwager (2008) have shown that the use of a hyperlink is dominant to an attached form, e.g. Word-document, in terms of survey participation. For these reasons, data has been collected through an online survey instrument sent by e-mail. Based on research by Faught, Green, and Whitten (2004), all invitations were sent on Wednesday morning and Tuesday afternoon because surveys conducted via e-mail yield a significant higher response rate on these moments. Moreover, a personalized reminder e-mail was sent approximately 1,5 week after the first wave of data collection.

26 of their pharmacy/firm. To minimize informant bias, two post hoc self-reports were included to ensure the competence of the key informants. Based on the findings of Kumar et al. (1993) two measurement items – “I am familiar with most aspects of the business relationship with the partner firm” and “I am involved in our business relationship with this partner firm” – were included in the survey designed to measure the informant’s knowledge of and involvement in the IOR under investigation. On the Likert scale from one (strongly disagree) to seven (strongly agree) the mean scores for informant’s knowledge and involvement were respectively 5.99 (SD=0.99) and 5.64 (SD=1.21), thereby guaranteeing adequate competence of the respondents to report on the dyadic relationship in question.

In total 1408 pharmacies and 303 firms were approached of which respectively 1387 and 292 were actually reached. Reasons for being unreachable were obsolete mail addresses, permanent closure of the pharmacies/firm, and technical errors, i.e. non-deliveries. Of these 1679 reached, 147 responded yielding a response rate of 8,8 %. Since out of these responses 17 concerned a notification of non-participation, the effective amount of submitted response reports yields 130 accounting for a response rate of 7,7 %. Based on the extrapolation method of Armstrong and Overton (1977), early and late responses (responses received in the second wave of data collection) were compared on the study’s major construct, i.e. cooperation, to test for possible non-response bias. The results indicated no significant differences at a 95 % confidence interval.

In the occasion that non-participation was notified, the provided reasons boiled down to a lack of time, structural abstention of participation, and the experienced work pressure in particularly the pharmacy sector. Moreover, those belonging to the pharmacy sector indicated to have difficulties with the English language. As such, the reminder mail contained an additional hyperlink to a Dutch version of the online survey. Incomplete responses were removed from the dataset and accordingly not included in the research’s analysis. As such, 83 responses were appropriate for further analysis representing 83 different interfirm relationships and organizations. Table 1 presents the descriptive data of the respondents.

3.2.1 Survey design

27 instructed to keep one specific IOR in mind during the completion of the entire questionnaire. In addition, every section was briefly introduced by means of an explanation of the questions that followed to ensure respondents’ understanding and the response quality. For instance, the respondents were provided with a definition of “interorganizational relationship”.

For all survey questions (except for most of the general, informative questions in the first section) a seven-point Likert scale has been used to preserve the balance of having sufficient points of discrimination without having to maintain too many response options (Nunnally, 1978; Preston & Colman, 2000).

Table 1. Characteristics of the Respondents (N=83)

N % Mean (SD) Age 80 96.4 47.50 (9.93) Gender Male Female 67 16 80.7 19.3 Function Owner Direct/(Co)-owner Director Manager Pharmacist Pharmacist/(Co)-owner Pharmacy assistant Other 5 8 5 9 33 17 1 5 6.0 9.6 6.0 10.8 39.8 20.5 1.2 6.0 n.a. n.a. Sector Metal Industry Engineering Pharmacy Other 13 12 2 52 4 15.7 14.5 2.4 62.7 4.8 n.a. n.a. Type of IOR Buyer-supplier agreement Strategic alliance Franchising Joint venture Trade association Licensing Co-branding Other 45 8 11 3 6 1 4 5 54.2 9.6 13.3 3.6 7.2 1.2 4.8 6.0 n.a. n.a.

Length of IOR (in years) 0-5 years 5-10 years 10-15 years 15-20 years > 20 years 16 26 16 3 22 19.3 31.3 19.3 3.6 26.5 n.a. n.a.

28 To check for the clarity of the survey and the accompanying e-mail, four reviewers with different backgrounds examined the questionnaire. Moreover, the comprehensibility and interpretability of the measurement items as well as the appropriateness of the survey’s length were checked. In the same vein, the Dutch translated version was reviewed by two of these persons having mastery in both languages. Based on these comments some minor modifications were made with respect to the e-mail invitation, the survey introduction(s), and measurement items resulting in the final survey of this study. See below for the description and elaboration of the measurement items.

3.3 Measurement

All study’s measures were adopted from well-validated measurement instruments in the extant literature. According to Swanborn (1996), the use of multiple items for the measurement of variables is equivalent to the use of different instruments, i.e. triangulation. Hence, to serve the reliability criterion of research, the variables were measured by means of multi-item scales using a seven-point Likert scale with anchors ranging from ‘strongly disagree’ (1) to ‘strongly agree’ (7). To reflect the specific context of the study some measurement items were slightly modified and redundant items were eliminated. Appendix A provides an overview of the study’s measurement items and its sources.

3.3.1 Dependent variable

Cooperation. Cooperation is the trustor’s willingness to pursue mutually compatible interest

(Das & Teng, 1998a). According to Heide and Miner (1992), Lui et al. (2009), and Pearce (2001), the coordinated actions taken in terms of flexibility, information exchange, and shared problem solving are important behavioral manifestations reflecting cooperation in IORs. In accordance, to measure cooperation eight measurement items were adopted from Pearce (2001) with some minor modifications. The wording is such that higher scores reflect a higher level of trustor’s cooperation. Similar measurement items were used by Heide and Miner (1992), Johnston et al. (2004), and Lui et al. (2009).

3.3.2 Independent variables

Trust. Trust refers to the trustor’s willingness to become vulnerable to the actions of the

29 competence trust. Accordingly, based on the trust scale of Zaheer et al. (1998) a six-item scale was used to measure goodwill trust. Researchers Chen et al. (2013), Li et al. (2010), Lui and Ngo (2004), Lui, Ngo, and Hon (2006), Perrone et al. (2003), and Şengün and Wasti (2007) used these measurement items in their corresponding studies as well. Competence

trust, on the other hand, was measured by six measurement items adopted from Fryxell et al.

(2002), Mayer and Davis (1999), and Şengün and Wasti (2007). The same items were used in the studies of Doucette and Wiederholt (1997), and Şengün and Wasti (2009, 2011).

3.3.3 Moderators

Formal contract. In this study formal contract is defined as the degree “to which roles,

obligations, responsibilities, contingency adaptation, and legal penalty are specified or well-detailed in formal agreements” (Cao & Lumineau, p. 24). Based on measurement items of Cannon and Perreault (1999), and Lusch and Brown (1996) an eight-item Likert-type scale for formal contact was composed covering all the concept’s aspects. These and similar items were also used by other researchers such as Chen et al. (2013), Li et al. (2010), Lui et al. (2009), Yang et al. (2011), Zhou and Poppo (2010), and Zhou and Xu (2012).

Contact intensity. Contact intensity refers to the frequency of contact between the trustor on

the one hand and the trustee’s boundary spanner on the other hand and is irrespective of the purpose: personal versus business (Crosby, Evans, & Cowles, 1990). Three items based on Doney and Cannon (1997), and Nicholson, Compeau, and Sethi (2001) measured contact intensity.

3.3.4 Control variables

In this study four control variables were included that are generally believed to influence the research’s independent variable, i.e. cooperation. Moreover, the study controlled for the different industries in the study’s sample.

Trust propensity. Trust propensity pertains to a person’s “general willingness to trust others”

30 Davis (1999), and Yamagishi and Yamagishi (1994) a five-item scale was composed to measure the trustor’s trust propensity.

Interdependence. Interdependence relates to the importance of one party for the other’s

survival and more specifically their business continuation (Şengün & Wasti, 2007). As such a desire to maintain a smooth relationship exist which is expressed by means of strong relational behaviors including cooperative behavior (Lusch & Brown, 1996; Şengün & Wasti, 2007). Measurement items of Li, Xie et al. (2010), Lusch and Brown (1996), and Şengün and Wasti (2007) were used resulting in a five-item scale for interdependence.

Asset specificity. Asset specificity refers to investments made by the trustor in the exchange

relationship which are non-recoverable and/or idiosyncratic (Parkhe, 1993). It has been argued that the locked-in effect resulting from these transaction specific investments promotes cooperation and relationship continuation due to the financial disadvantage of relationship termination (Lui & Ngo, 2004; Lui et al., 2006). To measure asset specificity a five-item scale was composed based on Ganesan (1994), Lui et al. (2006), Lui et al. (2009), and Poppo, Zhou, and Ryu (2008).

Long-term orientation. Long-term orientation concerns the expectation of relational

continuity (Lusch & Brown, 1996) and is also referred to as “extendedness of the relationship” (e.g. Heide & Miner, 1992; Perrone et al., 2003). Accordingly, the expectation of relational continuity rather than relationship length has been shown to affect relational behavior, i.e. cooperative behavior (Heide & Miner, 1992; Lusch & Brown, 1996). Therefore, long-term orientation is included as fourth control variable. Based on Ganesan (1994), and Lusch and Brown (1996) a four-item scale was used to measure long-term orientation.

Industry. Since data was obtained in two industries the study controlled for their potential

31

3.4 Data Analysis

All statistical analyses were conducted using the software program IBM SPSS Statistics for Windows version 22.0 except for testing of the hypotheses. For this latter, Stata version 14.1. was used since the required analysis was not available in SPSS.

Since the study’s variables concern latent constructs, a construct validity check was performed prior to the hypotheses testing (Gefen & Straub, 2005). First, Exploratory Factor Analysis (principal component analysis) with direct oblimin rotation was conducted on multi-item scales in order to assess whether the measurement multi-items reflect the constructs they are purported to measure (Leech, Barrett, & Morgan, 2015). Since the study’s constructs are known in the literature the closed rather than free Exploratory Factor Analysis (EFA) was the preferred type of factor analysis. As such, the number of factors was determined by a theory-based a priori criterion. To confirm the adequacy of the sample for factor analysis, the Kaiser-Meyer-Olkin (KMO) test and Barlett’s test of sphericity were performed. Moreover, given that factor analysis is a large-sample procedure and the study’s sample is relatively small, two separate analyses were conducted to arrive at reliable factor solutions (Costello & Osborne, 2005).

In the literature a minimum factor loading of 0.30 has been acknowledged as the rule of thumb (O'Leary-Kelly & Vokurka, 1998). Yet, to guarantee strong factors despite of the study’s small sample size only items with a factor loading ≥ 0.50 were included in the scales (Costello & Osborne, 2005; Hair, Anderson, Tatham, & Black, 1992). Moreover, items with cross-loadings greater than the Tabachnick and Fidell (2007) threshold of 0.32 were deleted, as well as those that did not load on any of the study’s constructs or on the wrong one. Second, a reliability analysis was performed to examine whether the remaining items have a good construct reliability, i.e. Cronbach alpha of 0.70 or higher (Nunnally, 1978). By means of the above-mentioned analyses, this study strived to serve two principal quality criteria in research, namely validity and reliability respectively (Swanborn, 1996; Yin, 1994).

32 been used. All Variance Inflation Factors (VIFs) were well below the cut-off point of 10 (Kutner, Nachtsheim, & Neter, 2004). The largest VIF with regard to the variables of interest was 2.70. Hence, the multicollinearity diagnostic test revealed no serious multicollinearity problems in the mean-centered regression models. Moreover, as recommended by Aiken and West (1991) all variables were mean-centered to test interaction effects. Lastly, the hypotheses were tested by means of hierarchal quantile regression models.

33

4. RESULTS

This section presents the results from the analyses performed. First, the results from respectively the factor and reliability analysis are provided followed by the correlational and descriptive statistics. The section closes with the results of the hierarchical quantile regression models.

4.1 Factor Analysis

Initially, an EFA (principal component analysis) with oblique rotation and more specifically direct oblimin was carried out on 23 items representing the independent variables and moderators. The KMO measure was 0.831 which is well above the minimum acceptable value of 0.50 and as such considered to be meritorious (Kaiser, 1974). In addition, the Bartlett’s test of sphericity was significant (χ2 (190) = 1566.00, p <.001). Hence, both results demonstrate the suitability of the respondent data for factor analysis (Williams, Brown, & Onsman, 2010). Given the fact that the items were designed to index four constructs – goodwill trust, competence trust, formal contract, and contact intensity – four factors were requested. The resulting component correlation matrix demonstrated correlations among factors close to the Tabachnick and Fidell (2007) threshold of 0.32 which confirms the decision to prefer oblique above orthogonal rotation. Three items (G1, G2, G3) were eliminated because they loaded on the wrong factor. After rotation with the remaining 20 items, the four-factor solution accounted in total for 78,5 % of the variance. A second factor analysis was conducted for the dependent variable to guarantee the convergent validity of the construct. All items measuring cooperation converged into one factor explaining in total 57,2 % of the variance. The final factor loadings are provided in Table 2.

Note. The bold numbers indicate that the item loaded to the factor.

4.2 Reliability Analysis

The results of the reliability analysis demonstrate that all multiple-item constructs have good reliabilities with coefficients αs between 0.75 and 0.96. To improve the initial reliability of the control variable trust propensity one measurement item (TP1) was eliminated. As a result, the construct’s reliability increased from α=0.600 towards α=0.786 thereby exceeding the 0.7 threshold (Nunnally, 1978). The Cronbach’s alpha coefficients as well as the descriptive and correlational statistics are provided in Table 3. In line with the VIF results, the correlation statistics give no rise to multicollineairity concerns as none of the predictor variables

Table 2. Factor Loadings from Factor Analyses

35 demonstrate pairwise correlation higher than the threshold of 0.90 (Kennedy, 1992). Furthermore, note that the mean of the dummy variables requires a different interpretation than the other continuous-valued data: it represents the proportion, i.e. percentage of cases, that belongs to that variable.

4.3 Hierarchical Quantile Regression Analyses

Table 4 reports the results of five hierarchical quantile regressions for the level of cooperation. As noted by Petscher and Logan (2014), and Yang et al. (2012) quantile regression allows for the possibility that the importance of predictors may differ among the quantiles of the response variable. As the study’s interest concerns the determinants of cooperation in general and that of high levels of cooperation in particular, a distinction is made between the default quantile, i.e. the median or Q(0.50), and quantile 0.9, i.e. Q(0.90). In other words, the 10 % performers have been distinguished from the median performers. Model 1 includes the four control variables as well as the dummy variables. Model 2 contains the control and dummy variables plus the study’s variables: goodwill trust, competence trust, formal contract, and contact intensity. Model 3 and 4 add the interaction variables for respectively formal contract and contact intensity into Model 2. Hence, the interaction model for each moderator is separately presented in respectively Model 3 and 4. The pseudo R-square associated with each model is presented in the bottom line. No formally stated test such as the incremental F-test could be conducted to examine whether each quantile regression model is a significant improvement over the other. Notwithstanding, the pseudo R-square exhibits an upward trend for both Q(0.50) and Q(0.90).

As stated by Fairchild and MacKinnon (2009) moderation effects are challenging to detect when the sample size is small. Therefore, the scholars’ suggestion was followed implying that the results were evaluated against a less stringent significance level of 0.10. As a result of the increased Type 1 error rate, the possibility of a Type 2 error – the failure to find a significant effect when one actually exists – was decreased (Yang et al., 2012). Table 5 summarizes the results.

4.3.1 The effect of trust on the level of cooperation

36 Therefore, H1a is partially supported for Q(0.50). With respect to quantile 0.9, goodwill trust

has a positive and significant effect in Model 2 (b=0.05, p<0.05), Model 3 (b=0.14, p<0.001), Model 4 (b=0.09, p<0.01), and Model 5 (b=0.06, p<0.10). Hence, concerning Q(0.90) the results provide empirical support for H1a. Yet, the effect of competence trust on the level of

cooperation is significant and positive for both quantiles and in all models. To specify, the coefficient is positive and significant for Q(0.50) in Model 2 (b=0.36, p<0.01), Model 3 (b=0.28, p<0.05), Model 4 (b=0.36, p<0.05) and Model 5 (b=0.24, p<0.01). With regard to Q(0.90) the coefficient is also positive and significant in Model 2 (b=0.30, p<0.001), Model 3 (b=0.10, p<0.05), Model 4 (b=0.21, p<0.01), and Model 5 (b=0.21, p<0.01). These results empirically support H1b for both quantiles.

H1c predicted that goodwill trust has a stronger effect than competence trust on the level of

cooperation. In order to determine which dimension of trust has more of an effect, the magnitude of the coefficients had to be compared. Since the study’s model includes interaction terms the unstandardized (b) rather than the standardized (beta) coefficients should be interpreted in this case (Aiken & West, 1991). In general, it appears that in all models (except from Model 3) and for both quantiles the coefficients of goodwill trust are lower than the coefficients of competence trust. For instance, concerning Q(0.90) the coefficient of competence trust is in the all including model, i.e. Model 5 (b=0.21, p<0.01) greater than the coefficient of goodwill trust (b=0.06, p<0.10). Hence, contrary to the expectations the results in Table 4 suggest that competence trust is likely to have a stronger impact and therefore do not support H1c.

4.3.2 The joint effect of formal contract and trust on the level of cooperation

The results in Table 4 reveal results consistent with the moderating predictions. First of all, as predicted by H2a formal contract has a positive moderating effect on the goodwill

trust-cooperation link. For the median performers, Q(0.50), the effect is solely significant in Model 5 (b=0.07, p<0.05) thereby providing partial support for H2a. In light of the 10 % top

performers, the effect is merely significant in Model 3 (b=0.03, p<0.05). Hence, H2a is partially supported for Q(0.90) as well.

As predicted by H2b, formal contract has a negative moderating effect on the competence

trust-cooperation link. The effect is strongly significant for the median in both Model 3

Notes. The coefficients on the diagonal in bold are the Cronbach’s alpha of each scale.

*Significant at p < .05; **Significant at p < .01.

38

Model 1 Model 2 Model 3 Model 4 Model 5

Q(0.50) Q(0.90) Q(0.50) Q(0.90) Q(0.50) Q(0.90) Q(0.50) Q(0.90) Q(0.50) Q(0.90) Intercept -3.11*** (0.79) -4.63*** (0.65) -1.15 (0.80) -3.02*** (0.28) -4.63*** (0.81) -4.66*** (0.31) -4.39*** (0.85) -5.11*** (0.44) -4.58*** (0.56) -4.81*** (0.41) Goodwill trust (GT) -0.11† (0.06) -0.05* (0.02) -0.08 (0.07) -0.14*** (0.03) -0.02 (0.07) -0.09** (0.03) -0.14** (0.05) -0.06† (0.04) Competence Trust (CT) -0.36** (0.11) -0.30*** (0.04) -0.28* (0.12) -0.10* (0.04) -0.36* (0.12) -0.21** (0.06) -0.24** (0.08) -0.21** (0.06) Formal contract (FC) -0.09 (0.06) -0.03 (0.02) -0.12† (0.06) -0.01 (0.02) -0.10 (0.06) -0.01 (0.03) -0.06 (0.04) -0.01 (0.03)

Contact Intensity (CI) -0.09

(0.06) -0.02 (0.02) -0.09 (0.06) -0.08** (0.02) -0.11† (0.07) -0.07* (0.03) -0.18*** (0.04) -0.07* (0.03) GT x CI -0.04 (0.04) -0.04† (0.02) -0.07 (0.04) -0.06* (0.03) CT x CI -0.18** (0.06) -0.12*** (0.03) -0.06 (0.05) -0.15*** (0.03) GT x FC -0.05 (0.04) -0.03* (0.01) -0.07* (0.03) -0.02 (0.02) CT x FC -0.11** (0.04) -0.03* (0.01) -0.12** (0.03) -0.01 (0.02) Trust propensity -0.19† (0.11) -0.18* (0.09) -0.19† (0.11) -0.03 (0.04) -0.17 (0.10) -0.15*** (0.04) -0.18 (0.10) -0.09 (0.06) -0.10 (0.07) -0.11* (0.05) Interdependence -0.04 (0.11) -0.24* (0.09) -0.14 (0.10) -0.03 (0.04) -0.10 (0.10) -0.07† (0.04) -0.14 (0.10) -0.08 (0.05) -0.08 (0.07) -0.08 (0.05) Asset specificity -0.01 (0.10) -0.13† (0.08) -0.14 (0.09) -0.13*** (0.03) -0.14 (0.09) -0.10** (0.03) -0.12 (0.10) -0.07 (0.05) -0.10 (0.06) -0.09* (0.05) Long-term Orientation -0.31** (0.11) -0.37*** (0.09) -0.08 (0.12) -0.21*** (0.04) -0.05 (0.11) -0.23*** (0.04) -0.12 (0.12) -0.23*** (0.06) -0.09 (0.08) -0.24*** (0.06) Metal -0.04 (0.47) -0.79* (0.39) -0.36 (0.44) -0.42** (0.16) -0.15 (0.42) -0.57** (0.16) -0.12 (0.44) -0.49* (0.23) -0.04 (0.29) -0.40† (0.21) Industry -0.05 (0.47) -0.78* (0.39) -0.51 (0.43) -0.73*** (0.15) -0.41 (0.41) -0.64*** (0.16) -0.31 (0.44) -0.53* (0.23) -0.12 (0.29) -0.52* (0.21) Engineering -0.85 (0.70) -0.61 (0.58) -0.33 (0.65) -0.23 (0.23) -0.53 (0.62) -0.52* (0.23) -0.45 (0.65) -0.51 (0.34) -0.27 (0.43) -0.44 (0.31) Pharmacy -0.05 (0.42) -1.03** (0.35) -0.50 (0.39) -0.56*** (0.14) -0.42 (0.38) -0.84*** (0.14) -0.28 (0.40) -0.78*** (0.21) -0.16 (0.27) -0.79*** (0.19) Pseudo R2 0.17 0.30 0.33 0.46 0.38 0.48 0.39 0.49 0.43 0.51

Notes. Dependent variable: level of cooperation; N=83. Standard errors are reported in parentheses. † p < .10; * p < .05; ** p < .01; *** p < .001.

Regarding the upper decile, although not significant in the all including model, the effect is significant in Model 3 (b=-0.03, p<0.05). As such, H2b is partially supported for Q(0.90).

Based on earlier findings (Cao & Lumineau, 2015; Lui et al., 2009; Lusch & Brown, 1996; Yang et al., 2011; Zhou & Xu, 2012) no hypothesis was formulated for the direct effect of formal contract on the level of cooperation. In accordance, the results reveal that in general formal contract has no significant effect on the level of cooperation in any of the study’s regression models. An exception of aforestated is Model 3 in which the effect is positive (b=0.12) and modest significant (p<0.10) for Q(0.50).

4.3.3 The effect of contact intensity on the level of cooperation

The empirical results in Table 4 indicate that contact intensity has a positive and significant effect on the level of cooperation for quantile 0.5. Yet, aforementioned applies only to two of the five models. More specifically, the effect is modest significant in Model 4 (b=0.11,

p<0.10) and strongly significant in Model 5 (b=0.18, p<0.001). Therefore, H3a is partially

supported for Q(0.50).

Similarly, the coefficient of contact intensity is positive and significant for quantile 0.9 in Model 4 (b=0.07, p<0.01) and in the all including model, i.e. Model 5 (b=0.07, p<0.05). Moreover, the effect appears to be positive (b=0.08) and strongly significant (p<0.01) in the formal contract interaction model (Model 3). Nevertheless, no significant effects were found in Model 2. Hence, the study’s results partially support H3a for Q(0.90).

4.3.4 The joint effect of contact intensity and trust on the level of cooperation

H3b predicts a positive moderating effect of contact intensity on the goodwill

trust-cooperation link. However, no significant effects were found for Q(0.50). Different from the

median, the effect is positive and significant for the upper decile in both Model 4 (b=0.04,

p<0.10) and Model 5 (b=0.06, p<0.05). Hence, H3b is empirically supported by the study’s

findings, yet merely for Q(0.90).

40 than positive for both quantiles. Therefore, the study yields no empirical support for H3c.

4.3.5 The effect of control variables

As the quantile regression models show, trust propensity has in general a positive effect on the study’s dependent variable, i.e. the level of cooperation. However, for the median performers this effect is only modest significant and exclusively in Model 1 (b=0.19, p<0.10) and Model 2 (b=0.19, p<0.10). With respect to the top 10 % performers, the effect is significant in Model 1 (b=0.18, p<0.05), Model 3 (b=0.15, p<0.001), and Model 5 (b=0.11,

p<0.05).

Furthermore, no significant effects were found for interdependence in the all including model for neither of the quantiles. Yet, contrary to the expectations a negative rather than positive significant effect was found in Model 1 (b=-0.24, p<0.05) and Model 3 (b=-0.07, p<0.10). Regarding asset specificity, the results reveal solely for Q(0.90) a positive and significant effect in respectively Model 1 (b=0.13, p<0.10), Model 2 (b=0.13, p<0.001), Model 3 (b=0.10, p<0.01), and Model 5 (b=0.09, p<0.05). Lastly, the coefficient of long-term

orientation is strongly significant (p<0.01) and positive in all models for Q(0.90). For the

median, Q(0.50), long-term orientation seems to have no significant effect in general, except for Model 1 (b=0.31, p<0.01).

41 Note. Dependent variable: level of cooperation.

Table 5. Summary of the Results

Hypothesis Expectation Empirical evidence

Direct effect Moderation effect Q(0.50) Q(0.90)

H1a: Goodwill trust (GT) + Partially supported Supported

H1b: Competence trust (CT) + Supported Supported

H1c: Goodwill trust > CT Not supported Not supported

H2a: Formal contract x GT + Partially supported Partially supported

H2b: Formal contract x CT - Supported Partially supported

H3a: Contact intensity + Partially supported Partially supported

H3b: Contact intensity x GT + Not supported Supported

42

5. DISCUSSION AND CONCLUSION

The research’s findings provide both empirical evidence and theoretical insights. First of all, a significant and positive association between the different types of trust and the level of cooperation was found supporting the reasoning that a party is inclined to involve in risky and coordinated behaviors once trust is established (e.g. Das & Teng, 2004; Morgan & Hunt, 1994; Şengün & Wasti, 2007; Yilmaz et al., 2005). Hence, the empirical findings confirm the conviction that trust is a critical driver of relational behavior in IORs such as cooperation (Rindfleisch, 2000; Yilmaz et al., 2005; Zaheer et al., 1998). Yet, contrary to the expectations competence trust is likely to have more of an impact in this sense. As noted by Johnston et al. (2004) and Woolthuis et al. (2015) trust can be both the basis and outcome of an IOR. Accordingly, scholars (Blomqvist et al., 2005; Malhotra & Lumineau, 2011; Sako, 1992) have argued that competence trust is a prerequisite for the development of collaborative interfirm relationships. Hence, a possible explanation is that competence trust indeed served as a precondition for signing the formal contract and consequently was of greater importance in the examined interfirm collaborations. As noted by Malthora and Lumineau (2011) “it is perhaps axiomatic that parties prefer to do business with those they consider to be competent” (p. 984).