0
THE INFLUENCE OF TRUST AND CONTRACTS ON THE SUCCESS OF AN INTER-
ORGANIZATIONAL COLLABORATION: AN EMPIRICAL ANALYSIS.
Master thesis, Msc Business Administration: Organization Management & Control.
Rijksuniversiteit Groningen, Faculteit Economie en Bedrijfskunde
22 June, 2014
Ryan Langbroek Studentnumber: 1989103
Populierenlaan 111 9741 HC Groningen Tel: +316-15954332
Email: R.langbroek@student.rug.nl
Supervisor:
Abdul Rehman Abbasi
ABSTRACT:
This paper studies the relations between trust, contracts, relational risk and performance in inter-organizational collaborations. It is hypothesized that an increase in both trust and contracts reduce the relational risk, which in turn increases the performance of the
collaboration. After collecting data through the use of a survey, it has been found that trust indeed lowers relational risk. Contracts however increase relational risk. Finally, no strong evidence has been found for improved performance as a result of reduced relational risk.
Keywords: trust, contract, inter-firm collaboration, relational risk, collaboration performance.
Word count: 12.129.
The author would like to thank the organizations that participated in this study, his colleagues
Fedde Tilma and Remon Korteweg and his research team for their support.
1
1. Introduction
With increasing globalization due to innovations in information technology, competition between organizations increase at a high pace. As a result, organizations collaborate more to cope with increasing standards to remain profitable (Grant & Baden- Fuller, 1995). However, the question remains if the collaboration between organizations is indeed helpful, as the failure rate of collaborations remains between fifty and sixty percent for many years (Hughes & Weiss, 2007). This raises the question what makes these
collaborations succeed. A wide range of research has been done concerning the mechanisms that influence inter-organizational collaborations, ranging from governance (Williamson, 2005) and control (Dekker, 2004; Grandori & Soda, 1995) to trust (Coletti, Sedatole &
Towry, 2005; Zaheer, McEvily & Peronne, 1998) and information (Tomkins, 2001).
However, since inter-organizational collaborations crosses organizational borders, not all control mechanisms may be useful in improving the collaboration. In many cases, there is no overarching hierarchy to control the parties. This way, control in collaborations is limited to control mechanisms that can be used universalistic, such as the use of contracts (Nooteboom, 2002). Based on the transaction cost economics (TCE), one would expect the use of these control mechanisms in collaborations as a safeguard (Blumberg, 2001; Woolthuis, Hillebrand
& Nooteboom, 2005). This way, both parties reduce the risk of opportunism by restricting the range of the actions that the partner can take, and thus increases the likelihood of a desirable outcome (Williamson, 1975, 1985). In this case, an organization acts in self-interest to protect itself from harmful actions from the other party. However, more recent studies have found that the use of contracts does not always improve the collaboration, and in some cases may actually improve the chance of opportunistic behavior (Bradach & Eccles, 1989; Goshal &
Moran, 1996; Lyons & Metha, 1997). Sociology and psychology studies on the other hand argue that other forms of control such as trust are of greater importance in reducing
opportunistic behavior and thus improve the results of collaboration between parties (Cook &
Wall, 1980; Das & Teng, 2001a; Mayer, Davis & Schoorman, 1995). In this scenario, the parties control each other through altruism, such as the use of social norms and values.
However, these finding are mainly being applied on an inter-personal or organizational level,
leaving room for speculation on the workings of trust on an inter-organizational level. Also,
when these finding are being compared with TCE with the findings from social sciences,
different arguments on how to reduce opportunism risk can be found. A recent exploratory
study performed by Woolthuis et al. (2005) studied the differences between trust and control
in an inter-organizational setting. After conducting four case studies, Woolthuis et al. (2005)
2
found that contracts and trust can both complement and substitute each other. Moreover, their results suggests that trust plays generally precedes contracts and the outcome of the
collaboration, while no direct evidence has been found for the use of contracts in predicting the success of a collaboration.
These differences with previous literature may be explained by the dependency on the participating parties in a collaboration. In the case of low dependency, an organization may perceive a collaboration as insignificant and thus adjust its effort to its importance, whereas a high level of dependency may increase the motivation to bring the collaboration to a good end (Nooteboom, 1996; 2002; 2006; Woolthuis et al., 2005). Due to specific investments
organizations can become dependent on each other, as the costs of breaking the collaboration may be high (Nooteboom, 2002; 2006) However, due to the limited amount of cases, the generalizability and external validity of these propositions are limited.
While the relation between trust and contracts has been studied, a gap still exists concerning the impact of trust and contracting on collaboration outcomes. By extending the works of Woolthuis et al. (2005), this paper therefore aims to provide a deeper insight in the role of trust, contracts, and dependency on the role of relational risk and outcomes of inter- organizational collaborations. These relations will therefore be studied in this paper to test the propositions made by Woolthuis et al. (2005).
The remainder of the paper is structured as following: First, the research questions will be elaborated in section 2. Next, previous research in the field of trust, contracts, relational risk, collaborations and dependency and the corresponding hypotheses will be presented in section 3. After the literature section, the methodology will be explained in section 4, to be followed by the results of the analyses in section 5. Thereafter, the findings will be discussed in section 6. Finally, the conclusion and its implications will be discussed in section 7.
2. Research question
While the effects of trust, contracts, relational risk and dependency on performance have been analyzed at multiple levels, more research is needed to validate recent findings in the case of collaboration between organizations. Existing literature has thus far offered a partial view on this subject.
Taking trust, contracts and collaboration performance as main themes, the focus of the study will be put on the role of trust and the use of contracts on the outcome of collaborations.
Therefore, the main research question becomes: “What is the influence of trust and contracts
on the outcome of a collaboration between organizations?”.
3
It is important to note that this relation may possibly be mediated by the perceived relational risk. As argued in the introduction, trust and contracts are mainly used reduce risk between parties. As a result, the following two questions arise: “what is the influence of trust and contracts on the perceived relational risk?” and “What is the influence of relational risk on the outcome of the collaboration?”
Besides the previously mentioned research questions, the influence of trust on contracts on the use of contracts will be further investigated. The underlying reason for this investigation is the contradicting literature on the relation between trust and contracts.
Previous research indicates several streams of propositions on this subject. Firstly, it is suggested that trust is a pre-condition for the use of contracts (Woolthuis et al., 2005).
Secondly, it is being argued that trust conflicts with the use of contracts and thus are
substitutes of each other (Lyons & Metha, 1997). Due to the different views on the workings of trust and the use of contracts, the impact of contracts on the outcome of the collaboration will be further investigated, which leads to the second research question: “How do trust and contracts influence each other?”
Finally, it is expected that the dependency of the organizations on the collaboration will moderate the relations. According to Nooteboom (1996; 2002) dependence is being created through investments that are specific to the collaboration, thus creating an incentive not to behave opportunistic and abiding the rules of trust and contracts. Consequently, the following research question will be proposed: “To what extend does dependence on the collaboration moderate the relation of trust and contracts on perceives relational risk?”
3. Literature review
As explained in the previous sections, the paper will study multiple aspects that are expected to be relevant to this research. The key literature will involve collaborations, relational risk, collaboration outcomes, trust, contracts and dependence. Still, since these terms can be interpreted in several ways, an elaboration on defining is required. As a results, this section will review the relevant literature and its implication for this study, followed by their corresponding hypotheses.
3.1 Collaborations
The term collaboration is a subject that can be interpreted in several ways, as there are
many different forms of collaboration. This mean that the exact level must be defined, since
4
inter-organizational relations can range from single transaction to the creation of a strategic alliance. In this paper, the definition of collaboration will be adapted from Coletti et al. (2005) and Dyer & Singh (1998), which will be defined as “two (or more) organizations involved in some type of joint production, where relation specific investments are required to reach a durable competitive advantage”. While the original definition has been used at the
organizational level, it is believed that by adjusting the unit of analysis the most important aspects of collaborations will be captured. By using this definition, according to the resource based view, collaborations will be seen as an instrument to gain a durable competitive advantage which will be hard to imitate (Barney, 1991). Following this approach,
collaborations that are established at arm’s-length in the market will be excluded. The reason behind this definition is to see the collaboration as an asset to be used to gain benefits that otherwise would not be achievable. Therefore, the focus will be put on collaborations instead of a single transaction between organizations. As a result, collaborations ranging from long term buyer-supplier relations to strategic alliances will be considered in this paper.
3.2 Collaboration results
Due to the wide usage of the term results, the need arises to come to a definition that is narrow and fits the context of inter-organizational collaborations. Taking a stockholder
approach, it is to be expected that the actions taken by managers are focused on increasing shareholder value (Blyth, Friskey & Rappaport, 1986). This would mean that the goal of working in a collaboration is to increase shareholder value in the long run, putting the focus on monetary gains. However, when taking financial outcomes as a result, a problem is expected to arise. Due to the nature of many inter-organizational collaborations, the initial focus will be put on different aspects, such as crisis or problem solving, innovation or the exchange of knowledge (Inkpen & Tsang, 2005; Gray, 1985; Powell, Koput & Smith-Doerr, 1996). The following step would be to use these outcomes in creating value for the
organization, shifting the focus from collaboration to the individual organization (Cohen &
Eliashberg, 1996). As a result, it would make more sense to measure the direct output of the collaboration in order to measure the results of the inter-organizational collaboration.
Financial outcomes therefore would not be a proper way to estimate the success of a inter-
organizational collaboration (Lyles & Salk, 1996). According to Yan & Gray (1994),
performance standards exist in a wide range of measures, such as lowering overhead costs,
improved quality, better forecasting, achieving planned goals and making profits. As a result,
in this paper, the collaboration results will be measured on a broad scale by focusing on
5
tangible and intangible outcomes. Therefore, a broad view will be taken on the many different forms of performance measures that have been created as a result of the collaboration. In this paper, the focus will be put on the creation of new expertise across different fields and the improvement of existing operations to measure the results of the collaboration (Selnes &
Sallis, 2003).
3.3 Trust
Trust has been widely studied across many fields of research. In this case, trust in the environment of social studies will used. As trust can be used in explanations across different fields, the concepts used within the social studies offer several perspectives that are relevant at different levels, namely individuals, groups, organizations and institutions (Anderson &
Narus, 1990; Hardin, 2002; Lewis & Weigert, 1985; Rotter, 1971). As Hardin (2002) argues, there are two aspects to trust: trust itself and being trustworthy. Coletti et al (2005) define trustworthiness as “an innate personal characteristic reflecting one’s preference for upholding some social norm of behavior, regardless of economic incentives”, whereas trust is defined as
“the perceived likelihood that another person will cooperate, absent any economic incentives to do so”. When looking at the definitions, the key characteristic of trust and trustworthiness is that a party acts in an expected way. The difference between the two definitions is that trustworthiness involves the act of acting in expected way, while trust concerns the perception of the other party’s trustworthiness. Since this study looks at the how organizations perceive the other party in an organization, the concept of trust in the previous mentioned setting will be used.
However, this leaves different forms of trust. Past studies have mainly distinguished between two forms of trust between organizations, namely competence trust and intentional trust (Nooteboom, Berger & Noorderhaven, 1997). The former is the trust that one has in the abilities and competence of the other party, while the latter refers to the trust in one’s
intentions of the other party. In this paper, the focus will be put on intentional trust, as the problem of this study looks at reducing the risk of opportunistic behavior of the other party.
This form of trust is also closely related to definition used by Coletti et al. (2005), as
described earlier. Finally, Mayer et al. (1995) define trust as “the willingness of a party to be
vulnerable to the action of another party based on the expectation that the other will perform a
particular action important to the trustor.” The mentioned definitions have in common that
trust reduces the risk of opportunism, when the other party treats the other fairly, even when
there is an incentive to behave opportunistically.
6
Using trust to reduce the relational risk between the parties, it is proposed that it will increase the likelihood of favorable outcomes of the collaboration, since both parties are assumed to fulfill their tasks within the relation (Gulati, 1995; Mayer et. Al, 1995;
Nooteboom, 1996). In other words, trust will be treated in this paper as a control mechanism, i.e. a way to mitigate relational risk. Through the use of this mechanism, the parties will depend on the benevolence of the other party to act in the best interest of the collaboration (Nooteboom, 2002; Ring & van de Ven, 1992). Yet, this approach may be conflicting with transaction cost economy theory (TCE), which perceives trust is an intangible asset which is not able to significantly reduce the relational risk in a collaboration (Williamson, 2005). It is argued that since trust is intangible, it may be hard to enforce in situations where the other party does not act as intended. First steps towards solving this difference in literature has been made by Woolthuis et al (2005), whose findings suggest that trust may be a viable way to reduce the risk of opportunism. This way, via trust relational risk within the collaboration will be reduced, leaving the collaboration mainly subject to contingencies from the external environment and thus increasing the likelihood that the collaboration will succeed in achieving its goals. Therefore, the first hypothesis to be tested will be:
H
1: The degree of intentional trust between parties in a collaboration has a negative influence on the degree of perceived relational risk.
3.4 Contracts
As stated earlier, TCE argues that trust is not a viable way to control opportunism.
Several studies have stated that an enforceable, complete contract is required to maintain a
suitable level of control (Das & Teng, 1998; Dekker, 2004; Williamson, 1991). According to
Nooteboom (1996), a person or organization can reduce opportunistic behavior through the
control of opportunity or incentive, or by making use of benevolence. Whereas opportunity
and incentive control mainly base their control on (formal) power, the use of benevolence
limits the risk of opportunism through ‘goodwill’, such as social norms, moral obligations,
empathy and friendship (Nooteboom, 2002; Woolthuis et al., 2005). The focus will be put on
the control of opportunity through the use of contracts, as many collaborations are nonequity
alliances and thus have no overarching hierarchy to control the relation, thus leaving the use
of contracts as the only explicit form of control (Das & Teng, 1998). However, there is also a
difference between contracts themselves. While some contracts are being referred to as
written or verbal and implicit or explicit, this paper will align itself with the study of
7
Woolthuis et al. (2005) and use the definition of contracts as a formal, written and legally binding agreement between two or more parties (Lyons & Metha, 1997). Variations in these formal contracts exist due to completeness (Chen, 2000). A contract that is simple, contains general and few clauses and is hard to enforce in court can be considered as an incomplete contract, while a complex contract with many specific clauses will be considered as a very complete contract. Following TCE, this means that the more complete a contract, the smaller the relational risk will be and thus creating a higher chance of achieving the desired outcome.
However, several findings contradict this belief and argue that the use of contracts may not always lead to the desired results (Bradach & Eccles, 1989; Goshal & Moran, 1996;
Lyons & Metha, 1997). Woolthuis et al. (2005) also found other uses of contracts that may be relevant in determining the success of a collaboration. They argue that setting up a complete contract does not only limit relational risk, but also help to coordinate, safeguard against contingencies from the environment and act as a sign of commitment (Ring & van de Ven, 1994; Woolthuis, 2005). Following this statement, the assumption that using a complete contract diminishes opportunistic behavior will be made in this paper (Das & Teng, 2001b).
For this reason, it is expected that the a more complete contract will lead to better results from the collaboration, leading to the following hypothesis:
H
2: Contract completeness has a negative influence on the perceived relational risk within a collaboration.
3.5 Perceived relational risk
As mentioned earlier, the expectation is that the perceived amount of relational risk is influenced by trust and contracts in an inter-organizational setting. In this section the role of perceived relational risk will be further explained.
When an organization starts a collaboration with another firm, the organization will be subject to risk that comes with this collaboration. Following Das & Teng (1996), a distinction between two forms of risks can be distinguished in collaborations: Relational risk and
performance risk. While the former refers to the risk involving the possibility that the partner
firm does not behave in a desired way and thus behaves opportunistically, the latter refers to
the risk involving the possibility that the collaboration may fail due to bad performance, even
though the parties may be fully cooperating (Das & Teng, 2001b). Since the paper looks at
reducing opportunism, only relational risk will be used in this study. A problem with
relational risk on the other hand is its subjectivity (Nooteboom et al., 1997). Since the
8
subjectivity stems from experiences and believes, relational risk should be measured through how an individual or organization perceives this phenomenon (Nooteboom et al., 1997).
When following this approach regarding relational risk, the risk translates itself in variation in outcomes of the collaboration, meaning that increased risk dictates a sub-optimal
collaboration outcome (Das & Teng, 1996; Das & Teng, 1998). Consequently, the hypothesis regarding perceived relational risk and collaboration outcome is as following:
H
3: Perceived relational risk has a negative influence on collaboration outcomes.
3.6 Relation between trust and contract
While the main focus of the paper remains on the effect of trust and contracts and their impact on the perceived relational risk within an inter-organizational collaboration, this paper also seeks to contribute to the ongoing discussion of the relation between trust and contracts.
According to TCE, a complete contract is needed before trust can be developed between two parties (Williamson, 1985). As stated earlier, several researchers disagree with this view and propose that trust is able to reduce opportunistic risk and through this way decrease the need of contracts (Bradach & Eccles, 1989). However, empirical studies have shown mixed results, where trust and control could be either substitutes or complements of each other (Costa &
Bijisma-Frankema, 2007; Lyons & Metha, 1997; Zaheer & Venkatraman, 1995). When looking at the research done by Woolthuis et al. (2005) and Ring & van de Ven (1992,1994), it is noticed that the answer may not as black and white as proposed. It is argued that
intentional trust and formal control can both be substitutes and complementing and it can also precede the introduction of formal control. When trust in one’s intentions has developed between the parties, the possibilities to set up complete contracts are improved (Larson, 1992;
Woolthuis et al, 2005). Due to the conflicting arguments of TCE, the relation between trust and contract will be re-examined. However, due to the conflicting views on this relationship, it is unclear which view will be dominant. Studying the amount of empirical evidence that argues that trust acts as a substitute in respect to contract completeness, it is expected that similar results will be found in this study. Consequently, when looking at this relationship, the following is hypothesized:
H
4: Intentional trust has a negative influence on contract completeness.
9
3.7 Dependency
Finally, it is expected that the influence of trust and contracts on the perceived relational risk of the collaboration will be moderated by the dependency of the organization on the said collaboration. Through large relation specific investments or the need of a positive result organizations are more likely to behave as expected without resorting to opportunism for fear of high costs, thus reducing relational risk (Noorderhaven, Nooteboom & Berger, 1998; Woolthuis et al., 2005). Nooteboom (2002) refers to these specific investments as incentive control, where higher relation specific investments lead to higher switching costs, thus being more dependent on the relationship. This would mean in the case of a high
dependency, an organization is more inclined to work towards a positive outcome, increasing the effects of trust and contracts due to the increase in predictability of one’s actions. This leads to the final hypotheses, namely:
H
5a: The dependency of an organization on a collaboration positively moderates the relation between trust and perceived relational risk.
H
5b: The dependency of an organization on a collaboration positively moderates the relation between contract completeness and perceived relational risk.
3.8 Control variables
Finally, several control variables will be taken into that account that may influence the perceived relational risk and the results of a collaboration. These variables include firm size, market growth and technological change.
Firstly, the respondents need to be controlled for trust propensity. As Mayer et al.
(1995) argue, people differ in approach on how to trust others. More precisely, it shows the likelihood that a person will trust someone else. Following this argument, it can be stated that people have different attitudes towards trusting another party, also called trust propensity. As the values and believes held by the individual may affect the perceived trust, the need arises to control for this relation and will therefore be measured as well.
It is argued that the firm size may give larger organizations an advantage over smaller
organizations (Veugelers, 1997). Since large organizations usually have a larger pool of
financial assets and are in possession of a better diversified portfolio, lowering the overall
risks for organizations which are considered large. Besides being able to spread risks, Teece
(1986) argues that large organizations are more able to acquire assets that are needed to make
10
the collaboration work, thus improving the chance of getting positive results. As a result, there is need to control this possible influence of firm size on collaboration results.
When looking at the changes within the market of the organizations, Park, Chen and Gallagher (2002) find that the market demands play an important role in the establishment of collaborations. They argue that a volatile market offers more opportunities for collaborations and alliances, and possibly influences the outcome of the collaboration. Therefore, the volatility of the market will be accounted for in this research.
Finally, the technological change within the industry will be measured. An analysis performed by Dodgson (1993) showed that in a technical turbulent industry the need for successful collaborations was higher when compared to more stable industries. In the case of a turbulent environment, it is argued that collaborations are needed to overcome the turbulence.
Therefore, by increasing the need for inter-organizational collaborations, it is possible that the turbulent environments causes an incentive to make the collaboration succeed. Due to this possibility, technological change will be addressed as a possible control variable.
When taking the hypotheses together, this leads to the following conceptual model as shown in figure 1.
Figure 1: Conceptual model
Intentional trust
Contract completeness
Dependence on collaboration
Perceived
relational risk Collaboration outcome
H
1: -
H
2: -
H
3: - H
4: -
H
5a: +
H
5b: +
Control variables:
Trust propensity, market growth, technological
change, organization size.
11
4. Research methodology
In this section, the procedures involving the measurement of the proposed model will be discussed. First the data sample will described, follow by the measurements used to
measure the constructs, to conclude with a description of the analysis that have been executed to come to the results.
4.1 Data
Due to the empirical nature of the paper, large scale data on inter-organizational collaborations need to be collected. To design an appropriate research tool, different research tools to test the conceptual model have been considered. In this study, data will be collected through the use of self-administered surveys. Advantages of this approach allows for large, quantitative data collection, which will be easy to use in statistical analyzes (Saunders, Thornhill & Lewis, 2009: 144). Taking this approach also brings disadvantages, such as the low amount of control once the survey has been send or contamination of the respondents’
answers (Saunders et al., 2009: 363).
Industries and sectors where collaborations are common and/or important will be defined and targeted. Therefore, the target group mainly exists from organizations from the industrial and technological sector. According to Bougrain & Haudeville (2002), increased emphasis is being put on inter-firm collaboration in these sectors to facilitate innovation projects. Consequently, members of the Dutch FME association have been targeted, an association that has been specializing in supporting their members in innovation, legal issues and developments within the technological industry. Organizations within this sector have been approached through several steps. First, the questionnaires and letters have been printed, to be send through postal services to the addressed respondents. In the envelope, the cover letter, two copies of the questionnaire and a return envelope have been enclosed. Second, after one week a follow-up letter has been sent. Third, a second follow-up letter with new
questionnaires, cover letter and return envelope have been send three weeks after the first follow-up. In total, 1617 organizations have been approached through email and mail, of which 83 organizations filled in the survey, leading to a 5.1% response rate. In a follow up call or email with several contacts it was found that the main reasons for the low response rate were (1) the length of the survey, (2) the disinterest in the subject of the survey and (3)
company policies. While the amount of responses is not optimal by multiple standards for
sample size, due to practical limitations the decision has been made to use these responses for
analysis. Taking the standard of Tabachniek & Fidell (2001: 588), the sample size is small
12
and not optimal, it is useable for analysis. The data itself has been collected through the use of a survey, which can be found in the appendix. The content of the survey will be discussed in the next section.
4.2 Measures.
As argued in the previous section, to measure the outcome of collaborations, a wide perspective must be maintained. Since the primary focus is not on financial gain, but on the sharing of knowledge to develop tangible and intangible assets, there is a need to incorporate the possibilities that can act as output of a collaboration. However, this leaves several
possibilities, as the outcome of the relation may be expertise that has been developed or the improvement of concrete operations (Lyles & Salk, 1996; Yan & Gray, 1994). Therefore, the survey exists of two parts that will measure collaboration outcomes: the development of new knowledge and expertise will be measured through an existing, tested construct developed by Lyles & Salk (1996). The second part exists of more concrete measures of relationship results.
In this case, the construct used to measure a more concrete form of collaboration results will be used. A similar construct has been developed by Selnes and Sallis (2003) to study the relationship performance. Due to the similar characteristics between this study and the one performed by Selnes and Sallis (2003), it is believed that an adoption of the existing will provide the most reliable representation. Both constructs will be measured on a seven point Likert scale. By moving away from a five point scale it is believed to reduce the central tendency error (Matell & Jacoby, 1972).
In this paper it is important, when measuring inter-organizational trust, that the trust in the intention of the other party is being captured. The underlying reason being that
opportunism plays a role in the perceived trust of the other party. Therefore, an existing construct developed by Mayer, Davis & Schoorman (1995) will be adapted. This construct exists of three aspects regarding inter-organizational trust, namely the measurement of trust in the ability, benevolence and integrity of the partner. Sharing the same characteristics of inter- organizational trust as the definitions used in this paper, the existing construct will be adopted and used as a part in the survey. This way, it is possible to analyze the overall trust that an organization has in its partner, or a more specific aspect such as the perceived integrity and benevolence of the partner. Similar to the construct for the outcomes of the collaboration, the construct is based on a seven point Likert scale.
To measure the completeness from a contract, it is needed to have a look at how
important formal contracts are within the collaboration. Following the definition by Woolthuis
13
et al. (2005), it can be stated that the more important formal, written agreements are, the more complete the contract is. Therefore, the survey questions will be measuring to what extent formally written agreements are playing a role within the collaboration. By looking at past research, two studies have been found that investigated the role of formal agreements between organizations (Buvik & Reve, 2002; Cannon & Perreault, 1999). However, small changes have been made to the original measures to match this research, since either the question was based on a single form of collaboration, such as a buyer-supplier relationship, or only
regarded one aspect of a collaboration, such as price setting. It is not expected to have
negative influences for the results, due to the same unit of analysis, namely inter-organization collaborations.
Measuring the perceived relational risk will be done with the use of an existing construct. Das & Teng (2001b) developed a construct used in surveys where perceived
relational risk has been successfully measured in the setting of strategic alliances. The authors developed the construct by involving different aspects of perceived relational risk, such as the possibility of unfair dealings, hidden agendas and manipulation. The construct is almost completely adopted from the original, save for a few small changes in the addressing of the target sample. Also, one variable has been omitted from the construct, as it was found not to be fitting within the construct (Das & Teng, 2001b). Similar to the previous constructs, perceived relation risk will also be measured on a seven point Likert scale.
The moderator dependency on the relationship will measure two aspects of
dependency: the dependency on the relation with the other party, and the dependency on the other party. While the latter may seem out of place since the paper focuses on collaborations, however, the decision to involve both aspects is based on the fact that while both are closely connected to each other, there is a possibility that an organization may be dependent on the relation, but not on the partner or vice versa. Therefore, both forms of dependency have been adapted from Handfield & Bechtel (2002), Heide & John (1988) and Noorderhaven et al.
(1998) to form the construct of dependency. Again, small changes have been made to fit the unit of analysis, as the original studies focused exclusively on buyer-supplier relationships.
Similar to the previous constructs, the construct will be measured on a seven point Likert scale.
Lastly, there is a need to set up measurements for the control variables. As argued in
the literature section, the model will be controlled for trust propensity, market growth,
technological change and firm size. While these specific control variables can be found in
many organization studies, with exception of trust propensity, the choice must be made
14
whether these variables will be measured in relative or absolute data. While the absolute data may involve less respondent bias, these may be difficult to compare across branches.
Therefore, market growth, technological change and firm size will be measured on a relative base by asking the respondents how they perceive the variables in their environment. To accomplish this, questions that measure these variables have been adopted from Song, van der Bij & Weggeman (2006). For trust propensity a separated construct will be used due to fact that the concept is not as popular as the other control variables. To measure trust propensity, an existing construct will be adapted from Mayer & Davis (1999), measuring several items indicating the likelihood that the respondent puts trust in the other party. Like previous constructs, the same seven point Likert scale will be applied to the control variables.
4.3 Analysis.
After collecting the data, several analyses will be conducted. To analyze the results the Structural Equations Method will be used (SEM). This test will be executed with the help of software computer program SPSS and SPSS Amos. Due to the nature of the conceptual model, it is believed single regression analyses will not suffice. Therefore, it is expected that SEM will offer a possible solution, as there is a need to estimate multiple and interrelated relationships, a need to account for measurement error and a need to define the model to describe the relations (Hair, Black, Babin & Anderson, 2009: 610). The relationship between constructs described earlier will be analyzed by the constructs shown in the appendix.
The following steps will be taken in order to execute the analysis. First, a confirmatory factor analysis, goodness of fit test (GOF) and Cronbach’s alpha test will be executed to check the validity of the model. Via these tests it is possible to determine if a construct can be used, and if not, which variables from the construct should be removed to make the construct ready for analysis. Secondly, a test for multicollinearity will be performed to determine whether the constructs are truly different from each other. The underlying reason to prevent
multicollinearity is to make sure that the variables are not overlapping in a significant way and thus act independently from each other. Thirdly, a path analysis will be executed,
resulting in an overview of the strength of relationships. These results will be shown through
the covariance of the respective constructs to show the strength of their influence on the other
variable. Fourthly, the role of the mediators will be analyzed for possible confounding or
suppression (MacKinnon, Krull, Lockwood, 2000). Fifthly, the influence of the moderator
construct will be tested for significance on the relations towards the dependent variable by
15
performing a moderator analysis. Finally, the model will be tested for other unobserved relations that may have not been considered so far.
5. Results
In this section the outcomes of the proposed analyses will be reported and explained.
First, the factor analysis and Cronbach’s alpha will be reported. Secondly, descriptive
statistics, including multicollinearity, will be discussed. Thirdly, the proposed model, and thus the hypotheses, have been tested and will be explained. Fourthly, the influence of control variables will be reported. Finally, the findings of unobserved relations will be reported.
5.1 Factor analysis.
After several items had been recoded, the items got labeled respective to the items in the survey that can be found in the appendix. Intentional trust has been labeled as Trust7- Trust17 (Trust1-Trust6 represents ability trust and has thus been removed from the analysis), contract completeness ass Contr1-Contr5, perceived relational risk as RelRisk1-RelRisk13, collaboration outcomes as Outcome1-Outcome12, dependency as Dep1-Dep9 and trust
propensity as Prop1-Prop8. The confirmatory factor analysis has then been performed to make sure that the items match the proposed constructs and therefore do not have cross loading across constructs. The results of this analysis can be found below in table 1. To make sure the factors are appropriate, a KMO and Bartlett’s test of sphericity have been performed, resulting in a KMO of 0
.624 (p=<0.001).The Bartlett’s test of sphericity resulted in a significant chi- square of 3670 (df=2016, p=<0.001). The factors have been rotated using varimax rotation in a six factor setting. The remaining items have been tested for validity in a Cronbach’s alpha test to ensure the items measure the same construct. Afterwards the constructs have been tested for multicollineairity. The measured levels of variance inflation factors (VIF) and other descriptive statistics can be found in table 2.
The VIFs in table 2 suggest the constructs hardly overlap each other, meaning
multicollinearity does not play a significant role in the following analysis (Farrar & Glauber, 1967; Kutner, Nachtsheim & Neter, 2004). The factor analysis and alphas give further evidence of sufficient discriminant validity. However, the construct for collaboration
outcomes, adapted from Lyles & Salk (1996), turned out to be incompatible in this analysis as
no item from this construct remained after the factor analysis. It is not expected to have
significant consequences as the remaining items stay close to the aspects of this research.
16
5.2 Model testing
The next step was testing the proposed hypotheses by using structural equation
modeling, which will be discussed one by one in this section. For an overview of the relations found, please see figure 2.
Model statistics
To assure a decent fit a goodness of fit (GOF) test has been performed. The GOF test suggested a decent model fit with a significant Chi-square of 766.3 (df=429, p=<.001).
Moreover, when looking at the root mean square error of approximation (RMSEA), a decent fit was suggest, as the RMSEA had a level of 0.07 (p=<.001). While this does not suggest a perfect fit, it suggests it is adequate for further analysis (Hair et. al, 2009). This is also
confirmed by the comparative fit index (CFI), which reports a fit of 0.810. As a result, there is a possibility the model may not be accepted as whole, as not all indications of a perfect model fit are satisfactory. Nevertheless, there is also evidence for an imperfect but adequate fit. As a result, the relations of the model have been studied which will be explained below.
Intentional trust – perceived relational risk
When testing the influence of intentional trust on the level of perceived relational risk, a significant negative effect was found, with a significant estimation of -0.76 (p=0.05). As a result of this finding it can be stated that hypothesis 1 has been supported. This suggests that an increase in intentional trust reduces the relational risk that is being perceived by the organization. The same relation has been found when looking at the influence of perceived relational risk on intentional trust. With a significant estimate of -0.273 (p=<0.01), it can be stated that lower relational risk also improves trust in return. This would suggest a reinforcing cycle between intentional trust and the relational risk that is perceived in the collaboration.
Contract completeness – perceived relational risk
After performing the analysis between contract completeness and the relational risk
that the organization perceives, it can be concluded that there is a significant and positive path
from contract completeness to risk at a ten percent significance level. The strength of this link
is estimated at 0.193 (p=0.063), which contradicts hypothesis two. While the expectation was
to find a negative relation, the opposite has been found, meaning contracts do not reduce the
perceived relational risk, but rather increase the distrust between parties. Also did the analysis
show that the perceived relational risk is significant on contracts (estimate of 0.291, p=0.088),
17 0.291 (p=.088)
0.193 (p=.063) N.S.
Intentional trust
Contract completeness
Dependence on collaboration
Perceived
relational risk Collaboration outcome
N.S.
-0.439 (p=.118)
-0.098 (p=.106)
-0.76 (p=.05)
-0.273 (p=<.001)
N.S Trust
Propensity 0.284 (p=.06)
Organization Size 0.158 (p=.075) -0.891 ( p=.004)
thus suggesting another reinforcing cycle. Following this finding it can be concluded that the need for contracts increase perceived relational risk, which in turn increase the need for contracts.
Perceived relational risk – collaboration outcomes
The structural equation modeling did not provide support for hypothesis three, as the relation of perceived relational risk was negatively insignificant on the outcomes of the collaboration (estimate of -0.248, p=0.193). This would suggest that other factors are more important for explaining the performance of the collaboration.
Intentional trust – contract completeness
For this relation, a negative insignificant influence has been found from intentional trust on contract completeness (estimation of -0.439, p=0.118). However, given the fact the significance level borders on the ten percent significance border, it may be useful to consider this relation. In that case, this result suggests that trust can act as a substitute for the use of contracts, which indicates a relation as suggested in hypothesis four. A similar relation can be found from contract completeness on intentional trust (estimate of -0.098, p=0.106),
suggesting there is a small change in the intentional trust when the role of contracts become more important. This supports the idea that trust and contracts act as substitutes, in contrary of the believes that the relation is either complementary independent of each other.
Figure 2: Representation of results. First number represents estimates, second represents significance levels.
18 Table 1: Results of factor analysis. Numbers represent correlations with the corresponding constructs.
Factors
1 2 3 4 5 6
Dep1 .883 .117 -.063 .038 -.008 .109
Dep2 .833 .124 -.064 .049 -.107 .008
Dep3 .784 .042 -.028 -.075 -.051 .066
Dep4 .703 -.028 -.148 .027 -.033 -.055
Dep8 .676 .124 .060 .061 -.052 -.027
Dep6 .648 -.067 -.179 -.080 -.173 -.238
Trust10 .163 .821 -.059 -.012 -.081 -.129
Trust8 -.042 .783 -.187 -.055 -.023 -.105
Trust11 -.085 .721 -.248 -.041 -.060 .059
Trust7 .086 .686 -.004 -.066 .235 .021
Trust9 -.033 .646 -.141 -.190 .083 .097
Trust14 .024 .602 -.208 -.025 .117 .165
Trust16 .213 .576 -.119 .006 .193 .185
Relrisk8 -.002 -.269 .762 .012 -.029 -.068
Relrisk6 .052 -.146 .743 -.106 -.124 .027
Relrisk4 -.151 -.121 .700 .242 -.141 .035
Relrisk13 .066 -.114 .692 .015 -.183 -.215
Relrisk3 -.222 -.132 .671 .061 -.142 -.246
Relrisk2 -.081 -.141 .562 .158 -.187 .071
Contr4 -.003 -.003 .036 .882 -.093 -.005
Contr3 -.051 -.099 .093 .846 -.059 .072
Contr5 -.023 -.046 .019 .774 -.094 .049
Contr2 .130 -.234 .129 .643 .178 .179
Prop6 -.025 .017 -.247 .079 .729 .044
Prop5 -.016 .175 -.204 -.064 .723 -.064
Prop8 -.078 .225 -.027 -.154 .690 -.038
Prop7 -.203 -.213 -.117 -.283 .648 .052
Prop3 .045 .094 -.159 .144 .632 -.150
Outcome8 .022 .181 -.082 -.074 .026 .789
Outcome7 -.011 .083 -.028 .121 -.132 .767
Outcome6 .149 .005 -.176 .076 .037 .699
19 Table 2: Means, standard deviations, alpha's, VIF and correlations of the factors..
Mean(s.d.) Cronbach's Alpha
Highest VIF
Intentional
Trust Contract. Relational
Risk Outcome Depen- dency
Trust Propensity
Market Growth
Technological Change
Organization Size
Intentional
Trust 5.15 (0.73) 0.836 1.38 1
Contract 3.74 (1.31) 0.825 1.48 -.195 1
Relational
Risk 3.02 (1.01) 0.843 1.15 -.399* .198 1
Outcome 4.39 (1.24) 0.73 1.46 .141 .134 -.141 1
Dependency 3.94 (1.36) 0.875 1.48 .134 .011 -.085 .085 1
Trust
Propensity 4.03 (0.85) 0.76 1.30 .195 -.134 -.376* -.042 -.121 1
Market
Growth 4.47 (1.57) - 1.48 -.073 .060 .157 -.042 0.231 .011 1
Technological
Change 4.17 (1.79) - 1.46 .085 -.069 .088 .105 .166 .054 .196 1
Organization
Size 3.77 (1.76) - 1.48 -.032 .011 .109 .183 -.069 .026 .092 .024 1
*. Correlation is significant at the 0.01 level (2-tailed).
20
Dependence on the relation
Finally, the moderating effect of dependency on the relation has been tested by
performing a moderator analysis. However, no significant relations have been found on either the relation between intentional trust and perceived relational risk (p=0.139) or contract completeness and perceived relational risk (p=0.650). Consequently, hypothesis 5a and 5b can be rejected. As a result, it is assumable that the dependency on the relation does not influence the amount of risk that is perceived. Further analysis of the influence of dependence on other constructs did not uncover any other relations with other constructs, suggesting that the dependency of an organization in a collaboration plays an insignificant role in this model.
5.3 Controlling factors
As described in the methods section, the constructs have been controlled for several aspects, namely market growth, technological change, organization size and trust propensity.
The main finding regarding the control factors was the role of trust propensity. Following the results, trust propensity significantly influences intentional trust (estimate of 0.284, p=0.06) and perceived relational risk (estimate of -0.891, p=0.004). Therefore, it is likely to assume that the role of trust and relational risk depend on the values and believes of an individual.
Finally, the relative size of the organization has a significant positive influence on the outcome of the collaboration (estimate of 0.158, p=0.075). This finding suggests that larger organizations will have better outcomes in collaborations in contrast to smaller organizations.
6. Discussion
This study aimed to extend the work of Woolthuis et. al (2005), who suggested that trust might be a viable way to reduce relational risk. Other studies however suggested that the use of formal contracts are more important in reducing relational risk (Das & Teng, 1998, 2001b). There is also the discussion on how trust and contracts relate to each other in a
collaboration, where it is believed that the factors have a complement, substitute each other or that the factors simply do not influence each other (Costa & Bijisma-Frankema, 2007; Larson, 1992; Lyons & Metha, 1997; Ring & van de Ven, 1992, 1994; Woolthuis et al., 2005; Zaheer
& Venkatraman, 1995). These issues form the basis of this study and therefore the results of this research will be compared to the previous literature.
To answer the main research question, we need to look at the influence of intentional
trust and contract completeness on the outcomes of the collaboration, both directly and
21
indirectly through the perceived relational risk mediator. The findings of this paper suggest that trust does indeed lower the perceived relational risk, which matches the findings of Mayer et. al (1995) & Nooteboom (1996). Since trust is developed, the relational risk becomes lower as both parties signal they want to do what is best for the collaboration. Another result of the study suggests that the relation between intentional trust and relational risk is a dynamic one.
The finding that low relational risk increases trust suggest that a reinforcing cycle is possible when establishing trust, thus reinforcing the (dis)trust over time, which matches Hirschman’s (1984) view, who suggests that “trusts grows in use”. This finding contradicts the transaction cost economics theory however, which assumes a tangible safeguard is required to hedge against relational risk (Williamson, 1991). According to TCE, a contract is needed to lower the relational risk. However, this proposition does not match the findings of this paper, which suggests that contracts do the exact opposite: the more the relationship depends on contracts, the higher the perceived relational risk will be. A possible explanation for this finding is that while contracts can be used to set authority between the parties, increasing the amount of rules does not take away the perceived risk in the same that trust in one another does (Ghoshal &
Moran, 1996). As Ghoshal & Moran (1996) explain, there is need for a social construct such a trust to make a collaboration work, as using only contracting will be insufficient. Also,
focusing too much on contracting may signal increased risk (Perks & Halliday, 2003). This explains the finding that similar to trust, the relation between contacts and relational risk is a reinforcing cycle, where increased emphasis on contracts lead to increased perceived
relational risk which in turn increases the emphasis on contractual agreements.
The relation between trust and contracts has also been studied, but does not provide for conclusive evidence. While the results do suggest a substituting relation, this cannot be said without caution. At first sight this findings seems to contradict Woolthuis et. al (2005), as trust and contracts have a negative relation and thus would be considered substitutes. In this research, contracts have been defined as an instrument to control the actions of the other party in the collaboration. However, contracts can also have different functions, such as
coordination, safeguarding and acting as a sign of commitment (Ring & van de Ven, 1994;
Woolthuis, 2005). Depending upon these roles, it might be possible that the influence of contracts on trust depends on the function of the contract, instead of its completeness.
Therefore, there is too little evidence to consider contracts and trust as substitutes.
The relation between relational risk and the outcome of the collaboration has been
found insignificant, meaning that a lower relational risk does not always improve the outcome
of the collaboration. A possible explanation is provided by Mayer et. al (1995), who makes
22
the differences between the perception and the attitude towards relational risk on the one hand and the actual risk taking behavior on the other hand. The authors propose that the perception of risk and trust does not influence the outcome directly, but that it is being mediated through the actual risk taking behavior of the organizations.
When looking at the role of dependency, no support for the moderating effect has been found. Analyzing other possible relations did not show any signs of significance of
dependency either. Based on these outcomes, it is likely to assume that the dependency on the relationship is not linked to trust, control, relational risk or the outcomes of the collaboration.
This result contradicts the findings from Nooteboom (1997), who found that mutual
dependence decreased the perceived risk. This difference may originate from several sources.
Firstly, Nooteboom et. al (1997) used a dyadic approach when studying the inter-
organizational relations, while this study took an unilateral approach. Secondly, the definition of risk that has been used by Nooteboom et. al (1997) involves the risk in the form of the probability of a loss in the collaboration, while this paper looks at the probability that one of the parties behaves opportunistically. In other words, dependence may be of influence in collaborations, this would need a focus on different kinds of risk.
The analysis also showed a significant influence of organization size on the outcome of collaborations. A possible explanation for this finding might be that larger organizations usually have more resources at their disposal. Moreover, a large organization may be more experienced due to collaborations in the past and thus being more able to improve their chances for a successful collaboration. These findings match the results of Teece (1986) and Veugelers (1997), confirming that past expertise and increased access to resources may give organizations an advantage in a successful collaboration.
Finally, trust propensity was found to have a large influence on both intentional trust and perceived relational risk. This matches the description of Mayer et. al (1995), who argue that how risk and trust is perceived is largely dependent on the values and believes of the individual. Consequently, it is difficult to treat trust and relational risk as standalone
constructs, as the content of these construct is largely dependent on how likely a person is to
trust someone.
23
7. Conclusion, implications, limitations and future research.
In this section, the conclusion and their implications will be discussed. Furthermore, the limitations of the study will be discussed. Concluding, suggestions for future research will be made.
7.1 Conclusion
The main question of this paper questioned the influence of trust and contracts on relational risk in inter-organizational relations. It was hypothesized that both aspects would lower relational risk which in turn would improve the outcomes of the collaboration. After collecting data from organizations in the technological industry it was found that trust indeed lowers relational risk, while the use of contracts increase relational risk, meaning that risk is more effective at lowering the relational risk in a collaboration. There was no relation found between relational risk and the outcomes of the collaboration however, meaning that the risk does not necessarily improve or worsen the collaboration outcomes. Analysis also suggested that the relation between trust and contracts is a dynamic one and dependable on the context.
Finally, the dependence on the relation did not uncover any relations, suggesting that it is either trivial or more complex than initially thought.
7.2 Implications
One of the main findings of this paper is the finding that trust is a better control to reduce relational risk than use of contracts. This does not necessarily mean that contracts should not be used, but when looking for ways to reduce relational risk, developing trust between the parties appears to be the optimal way. As a result, organizations may want to get to know each other and develop mutual trust before agreeing to a collaboration if they wish to have low relational risk during the partnership.
However, it is important to note that lowering the relational risk may not always be of importance to the parties in the collaboration, as it is unsure if this will help achieving the desired outcomes of the collaboration. While not found explicitly in this paper, other authors suggest the chance of reaching these outcomes being higher in case of lower relational risk (Mayer et. al, 1995; Woolthuis et. al, 2005). Possible ways of achieving this may be through the increased will to perform well out of benevolence, or because the perceived relational risk influences the actual risk taking behavior which influences the outcomes of the collaboration.
These findings are not directly found in this study and should therefore be used with caution.
24
