Interfirm Absorptive Capacity: Antecedents in the Institutional & Relational Context and Innovative Learning Performance Outcomes

(1)

Interfirm Absorptive Capacity:

Antecedents in the Institutional & Relational Context and

Innovative Learning Performance Outcomes

Joint Master Thesis

MSc Marketing & MSc Strategic Innovation Management

Bas van Herwaarden

University of Groningen - Faculty of Economics & Business

(2)

2 Joint Master Thesis

University of Groningen – Faculty of Economics & Business

MSc Marketing

MSc Business Administration - Strategic Innovation Management

Interfirm Absorptive Capacity:

Antecedents in the Institutional & Relational Context and Innovative Learning Performance Outcomes

by

Bas van Herwaarden January 15th, 2016

Jan Pieter Heijestraat 118-2 1054MH Amsterdam

The Netherlands +31 6 43 97 37 19

s.t.m.m.van.herwaarden@student.rug.nl S1895737

(3)

3

Abstract

Outside sources of knowledge have become critical to the innovation process of firms. The degree to which a firm is able to accumulate and effectively use this knowledge is called absorptive capacity (ACAP). Previous research has identified multiple antecedents of strong interfirm ACAP, originating in mainly three contexts: knowledge, relational and institutional. In a response to fill the gap in literature of missing empirical evidence concerning effects of relational and institutional antecedents, this paper analyses data from a database covering 166 dyadic relations between buyers and suppliers. We empirically test the effects of relationship duration and product importance on theoretically and related concepts PACAP and RACAP. Additionally, the link between interfirm ACAP and innovative learning performance is empirically researched. The analysis proves that contrary to literature, relationship duration has an indirect negative effect on buyer ACAP, while product importance in general positively influences ACAP. The relationship between ACAP and innovative learning performance is indisputably proven. This study draws conclusions regarding these findings and makes implications for managers.

(4)

4

Preface

“Absorptive Capacity”: the concept that has been circling in my head since early June 2015.

Nearly eight months later I find myself finalizing this thesis with pride. This work marks the finalization of six years at the University of Groningen and is the closure to my two master degrees. Over the last two years I have grown fond of the combination of Marketing Management and Strategic Innovation Management and as cliché as it might sound, I would not have been able to finish it alone. Therefore, there are people to thank.

Firstly, I would like to thank my first supervisor Hans Berger for his early support, for guiding me through this joint thesis and for his honest feedback on my work. Secondly, I want to thank my second supervisor Isabel Estrada Vaquero for ensuring that I would not lose sight of the innovation perspective as well as for her constructive notes. Thirdly, thank you to all of my fellow thesis group mates. I am sure we’ve been able to greatly help each other out along the way. Lastly, a big thank you to my family and friends for their support throughout the process of writing this paper and moreover so, throughout the past six years.

This thesis has been a challenge at times. Nonetheless, I believe I have provided the reader with interesting new insights concerning interfirm absorptive capacity, its antecedents and its effects. Hereby I hope to add to the literature of both strategic innovation as well as marketing.

(5)

5

1. Introduction

In the last twenty years, the view that knowledge can be the basis of competitive advantage for organizations has gained widespread acceptance (Argote & Ingram, 2000). Knowledge can be considered a resource that is tacit, sticky, complex and difficult to codify, making it an inimitable resource that has potential for becoming a competitive advantage (Dyer & Singh, 1998; Nelson & Winter, 1982). From a firm perspective, knowledge can traditionally be created by internal R&D activities of the firm but firms are increasingly relying on knowledge acquired from other firms to facilitate the development of their own capabilities and competitive advantages (Lane & Lubatkin, 1998).

Outside sources of knowledge are often of critical importance to the innovation process (Cohen & Levinthal, 1990; Lane & Lubatkin, 1998). Firms that are more effective in acquiring and developing these resources are more likely than their competitors to achieve competitive advantages, successfully innovate and operate in turbulent markets (Barney, 1991). The degree to which a firm is successful in gaining and effectively using outside knowledge resources is called absorptive capacity (also abbreviated to: ACAP) and is according to literature defined as: “a set of organizational routines and processes by which

firms acquire, assimilate, transform, and exploit knowledge to produce a dynamic organizational capability” (Zahra & George, 2002). Absorptive capacity of an interfirm

relationship is defined as interfirm absorptive capacity. This concept will be one of the main concepts of the following research.

The topic of interfirm ACAP has been widely researched and defined. Cohen & Levinthal (1990), in their early research labeled ACAP as the ability that stems from prior related knowledge, to recognize the value of new information, assimilate it, and apply it to commercial ends. Zahra & George (2002), reconceptualized ACAP, as mentioned before, as a bundle of complementary knowledge-based capabilities consisting of acquisition, assimilation, transformation and exploitation. Within this definition of the concept, acquisition and assimilation together form the potential absorptive capacity (PACAP) while transformation and exploitation shape the realized absorptive capacity (RACAP).

(7)

7 concept of interfirm absorptive capacity and their contexts. Concepts as compatibility, connectedness, idiosyncratic resources, contracting and relational norms have empirically proven to mediate the impact of complementarity on PACAP and RACAP and with that ambidexterity in innovative learning performance (Berger, 2015). These concepts will be discussed in more detail later. Within this field of literature, a vast amount of research on primarily organizational and environmental antecedents of absorptive capacity and (with it) organizational performance has been conducted (e.g. Berger, 2015; Fosfuri & Tribo, 2008; Jansen, van den Bosch & Volberda, 2005; Jansen, van den Bosch & Volberda 2006; Van den Bosch, van Wijk & Volberda; 2003).

Referring to the initial pioneering work of Cohen & Levinthal (1990) on the topic of ACAP, the authors mention that a firm’s ACAP will depend on the ACAP of their independent employees however, “it is not merely an accumulation of the absorptive

capacities of the employees”. Concluding from the extensive academic research available,

Cohen & Levinthal’s statement (1990) can be assumed to be steering in the right direction however, is inconclusive about what does accumulate to ACAP. It is therefore of practical importance to investigate how different antecedents influence ACAP. As mentioned, antecedents that are organizational have extensively been researched. Coordinating and socializing capabilities that positively influence PACAP and RACAP are exemplary of findings concerning antecedents in the organizational context (Jansen et al., 2005). Similarly, knowledge context antecedents like complementarity of resources have shown to benefit joint alliance success (Lambe, Spekman & Hunt, 2002). Evidently, within the innovation literature, antecedents are recognized to originate from different contexts. For this research we consider the complete context to be most clearly subdivided in the knowledge, institutional and relational context (Dougherty & Hardy, 1996; Yeoh 2009). As we will see, there are prior indications that antecedents in the institutional (Yeoh, 2009) and relational contexts (Jansen et al., 2006) have on ACAP. Nonetheless there appears to be a lack of empirical research and evidence into these antecedents and their effects on the dynamic capability of interfirm ACAP. This research will attempt to fill this gap.

(8)

8 relationship between cooperation and knowledge transfer yet, evidence was found of a direct effect of relationship duration on knowledge transfer in a buyer-supplier context (Squire, Brown & Cousins, 2009). Presumably relationship duration has more than a singular effect as no conclusive evidence exists on its direct effect on ACAP. Potentially it also adds to the development of a shared understanding of values and expectations (Heide & John, 1992). As so, shared norms might evolve out of a longer lasting relationship (Ring & van de Ven, 1992) and create a beneficial institutional base that can enhance absorptive capacity between partner firms. Consequentially, relationship duration therefore might not only directly influence absorptive capacity, but also strengthen one of the mediating variables distinguished by Berger (2015): the relational norm that exists between two partners. Adding to the institutions governing absorptive capacity and knowledge transfer, both the relational norm as well as relationship duration are therefore considered to belong to the institutional context. Besides the implicit governance structure that is the relational norm, a second governance structure exists within the institutional context: explicit governance (Berger, 2015; Ghosh & John, 2005). Explicit governance mechanisms include explicit rules in for example contracts (Ghost & John, 2005). Nonetheless, to limit the scope of this research, explicit governance will be left out of this paper, thereby from this point narrowing the institutional context down to the implicit governance mechanism.

(9)

9 understanding of who knows what and where which knowledge resides (Jaworski & Kohli, 1993). This degree of contact (formal and informal) between personnel or departments is called connectedness (Jaworski & Kohli, 1993) and as it is determined by (individual) relationship specific aspects, together with product importance is assumed to belong in the relational context.

Where previous research proved the mediating effect of antecedents in the three main contexts on ACAP and ambidexterity, this research will extend the work of Berger (2015) by empirically testing whether relationship duration and product importance are antecedents of ACAP and whether their influence is mediated by the relational norm and connectedness. Moreover so, this study will test whether these antecedents matter in the light of the continuous strive for innovative learning performance. In doing so, the paper aims on providing empirical proof of the before mentioned relationships within the scope of buyer-supplier relationships as well as to buttress the scales developed by Berger (2015). Empirical evidence is needed to confirm the inconclusive findings in literature and can potentially provide managers with interesting insights in their buyer-supplier relationships. Based on the identified research gap, the following research question has been formed:

To what extent do antecedents in the relational and institutional context influence the interfirm absorptive capacity and innovative learning performance of buyer-supplier relationships?

(10)

10

2. Literature Review & Hypotheses Development

2.1 Knowledge-based resources

According to the knowledge-based view (KBV), knowledge is the most valuable resource of an organization and can be a form of sustained competitive advantage (Grant, 1996). Knowledge forms the basis of interfirm learning, which is more effectively enabled by interfirm absorptive capacity. As mentioned in the introduction, valuable knowledge based resources may lie outside the focal organization and when internalized can become a source of competitive advantage (Argote & Ingram, 2010). Accurately summarized, Galunic & Rodan (1997) describe knowledge-based resources as the way that firms acts upon and transforms tangible input resources to provide value for the firm. “These knowledge

resources may be distinguished according to their basic form (information, know-how, understanding), their degree of tacitness, their dispersion (concentrated or dispersed) and their context specificity.” (Galunic & Rodan, 1997, p.7). Knowledge-based resources are

tacit, sticky and difficult to codify (Dyer & Singh, 1998).

2.2 Absorptive Capacity (ACAP)

Absorptive capacity is defined as the collective abilities of a firm that prior knowledge provides ‘to recognize the value of new information, assimilate it, and apply it to commercial ends (Cohen & Levinthal, 1990). Research has claimed that absorptive capacity entails a broad array of skills reflecting the need to deal with the tacit components of transferred technology, as well as the need to modify a foreign-sourced technology for domestic application (Mowery & Oxley, 1995). Building on this research, Zahra & George (2002), reconceptualized absorptive capacity to be a dynamic capability of the firm that is embedded in the firm’s routines and processes. In line with Winter (2000), a capability is defined as “a

high level routine that, together with its implementing input flows, confers upon an organization's management a set of decision options for producing significant outputs of a particular type" (p.938). Absorptive capacity as a capability therefore allows a firm to

(11)

11 capabilities (Zahra & George, 2002). Within absorptive capacity a subdivision between potential and realized absorptive capacity can be made.

2.2.1 Potential absorptive capacity (PACAP)

Potential absorptive capacity (PACAP) contains the two capabilities of acquisition and assimilation and potentially leads to explorative learning. PACAP mostly reflects Cohen & Levinthal’s (1990) description of the ability to value and assimilate external knowledge.

“Acquisition refers to a firm's capability to identify and acquire externally generated knowledge that is critical to its operations, while assimilation refers to the firm's routines and processes that allow it to analyze, process, interpret, and understand the information obtained from external sources” (Zahra & George, 2002, p.189).

2.2.2 Realized absorptive capacity (RACAP)

More related to exploitative learning, realized absorptive capacity includes the two later capabilities proposed by Zahra & George (2002): transformation & exploitation. RACAP allows a firm to leverage the knowledge that has been acquired and assimilated by PACAP capabilities (Zahra & George, 2002). Transformation includes a firm's capability to develop and refine the routines that make the combination of existing knowledge and the newly acquired and assimilated knowledge possible. Lastly, Zahra & George (2002) confirm the definition of exploitation of Cohen & Levinthal (1990) to be the application of knowledge.

PACAP and RACAP have been proven to be interlinked several times and research has proposed that the two dimensions belong to the same construct (ACAP) that might lead to ambidexterity (Berger, 2015; Yeoh 2009; Zahra & George, 2000; 2002). In this research we will therefore not hypothesize and test the relationship between PACAP and RACAP.

(12)

12 2.3 Institutional context

The institutional context involves the setting within which knowledge transfer is enabled resulting from factors, variables and antecedents shaping the organizational form of a buyer-supplier relationship, reliance and trust on the partner (Morgan & Hunt, 1994; Yeoh, 2009). Generally, within the institutional context, a distinction can be made between implicit and explicit governance mechanisms (i.e. contracting). As mentioned, this research will solely focus on the implicit governance side of this distinction.

2.3.1 Relational norm

The interrelation between organizations has frequently been found to be norm driven, or

"held together and coordinated by market driven focal organizations" by means of "norms of sharing and commitment based on trust" (Morgan & Hunt, 1994, p.20). More clearly

elaborated, relational norms have been defined as the partially shared values of exchange partners about what constitutes appropriate behavior and what is the expected behavior in a relationship (Heide & John, 1992). Naturally, over time relational norms evolve depending on the characteristics of the relationship (Dwyer, Schurr & Oh, 1987; Ring & Van de Ven, 1992). In the research of Berger (2015) a relational norm between two parties to an exchange has found to be directly positively related to PACAP as well as RACAP.

2.3.2 Relationship duration

As we have seen in the section above, relational norms are shared values that constitute appropriate behavior (Heide & John, 1992). Over time (with growing relationship durations) this relational norms, and with it the relation’s specifications and characteristics will evolve:

“Reliance on trust will emerge only as a consequence of repeated market transactions between the parties affirming the observance of norms of equity by both parties” (Ring & van

de Ven, 1992, p.489). Hence, from repeated interaction between two partners originates trust (Gulati, 1995). Trust, which has been extensively researched, will here be defined as the confidence that a partner will not exploit the vulnerabilities of the other (Barney and Hansen, 1994). Firms with prior connections and longer lasting relations are likely to have a greater awareness of the rules, routines, and procedures that each needs to follow (Gulati, Nohria & Zaheer, 2000).

(13)

13 Cousins and Brown (2008) confirmed that relationship duration has a direct negative effect on the transfer of knowledge. Reasoning leads to the conclusion that firms share more knowledge in the beginning of the relationship compared to later phases when the relationship matures (Squire, Cousins & Brown, 2008). These findings suggest that somewhere in the process of knowledge transfer the duration of the relationship has an influence. Nevertheless, this does not mean that the negativity of the above mentioned effect will always hold. Therefore, we follow the reasoning that resulting from a longer relationship duration, trust and familiarity between two partners to a relationship will grow and that this will enable the firm to acquire, assimilate, transform and exploit external resources the longer the relationship exists. This results in an expected positive relationship between relationship duration and ACAP. Consecutively, reasoning from the finding that firms share more knowledge in the beginning of the relationship (Squire et al., 2008), we expect the benefits from this knowledge sharing for PACAP and RACAP to be highest in the beginning of relationships and to level off, the longer the relationship lasts (when the relational norm has been formed). Accordingly, literature suggests that in the beginning of a relationship, focus lies on dedicating resources to effectively educating the partner (Ahuja & Katila, 2001) while, when a substantial level of interfirm ACAP has been attained, resources previously dedicated to education can be reconfigured to other organizational goals. Alternatively, such diminishing returns originating from a longer relationship duration can be considered to exists as a result of the “dark-side” of trust (Grayson & Ambler, 1999). The theory of this negative side to the concept of trust as we have identified it, proposes dampened positive results can originate from various situations (Moorman, Zaltman & Deshpande, 1992). Reasons and situations relevant to this research (and therefore the effect of a longer relationship duration on PACAP and RACAP) are multiple: for example the buyer’s belief that resulting from experience, a partner has lost its objectivity or has become too similar to the focal company and therefore has less value to add (Moorman et al. 1992). Alternatively, increased involvement, experience or trust may breed boredom that invokes a desire for something new or increases expectations that lead to dissatisfaction with the situation at hand (Moorman et al. 1992). As we assume that trust has a positive effect on both dimensions of ACAP however, it also has an inherent “dark-side” and drawbacks (Grayson & Ambler, 1999). We therefore hypothesize:

(14)

14

H2: The relational norm will mediate the relationship between relationship duration and potential absorptive capacity (PACAP)

H3: There will be a positive non-linear relationship between relationship duration and realized absorptive capacity (RACAP).

H4: The relational norm will mediate the relationship between relationship duration and realized absorptive capacity (RACAP)

2.4 Relational context

The relational context involves the setting within which knowledge transfer is enabled resulting from factors, variables and antecedents that belong to the relationship itself (Dougherty & Hardy, 1996; Yeoh, 2009)

2.4.1 Connectedness

Connectedness is the degree of formal and informal direct contact among employees across organizations that can enable the preservation of relational rents between two firms. It will occur in cumulative increments on an existing stock of assets held by a firm and will enable individual members to develop an understanding of who knows what and where which knowledge resides within and outside the organization (Dyer & Singh, 1998; Jaworski and Kohli, 1993). A higher degree of connectedness implies shorter network tie and relationship lengths that, together with relatedness of the knowledge, enable effective inter-unit knowledge sharing (Hansen, 2002).

2.4.2 Product importance

Within a relationship dyad, the importance of the product or service to either of the two parties might differ. This research will focus on relationships including tangible products. Findings from organizational buyer-supplier research indicates that “perceived product

importance leads to higher levels of cooperation, information exchange and social exchange, in addition to extensive inter-organizational networks involving many functional areas and hierarchical levels” (Metcalf & Frear, 1993, p.63). The variety of exchanges that follows

(15)

15 knowledge and its locus outside the company. Continuing this reasoning, we assume that when product importance is high, the firm takes a more central position in the network as it is more dependent on the product exchanged. The central position in a network provides a firm with better access to the desired strategic knowledge resources that can enhance the unit's ability to create new value and to achieve economic goals (Coleman, 1990; Tsai, 2001). Hence, high product importance and thereby high dependence positively influence openness to draw in external sources of knowledge (Laursen & Salter, 2006). This may be in the form of product related and non-product related information search that potentially increases the firm’s own knowledge base (Bloch & Richins, 1983). In other cases, dependency on a partner can also induce a reaction of opportunism towards this partner. This can result in search for alternative knowledge sources or even more direct threats (i.e. switching partner) and conflict (Joshi & Arnold, 1997). The hypothesis that a central network position, moderated by absorptive capacity, is positively related to a firm’s innovation has been confirmed by Tsai (2001) in his research on network position, absorptive capacity and business unit innovation performance. Hence we hypothesize that:

H5: Product importance is positively related to potential absorptive capacity (PACAP)

H6: Connectedness will mediate the relationship between product importance and potential absorptive capacity (PACAP)

H7: Product importance is positively related to realized absorptive capacity (RACAP) H8: Connectedness will mediate the relationship between product importance and realized absorptive capacity (RACAP).

2.5 Learning performance

(16)

16 to as relationship learning, exploitative learning can enhance the efficiency and effectiveness of a relationship when the partners share knowledge (Selnes & Sallis, 2003). Notably, besides from direct benefits of innovation for the firm, from a buyer-supplier relationship, “product

and process innovations developed together with a customer may improve the value of the supplier’s offerings to this customer in the future as well as to other customers” (Walter,

Ritter & Gemünden, 2001, p.368). The innovation function of a buyer-supplier relationship therefore can have a positive impact on knowledge exchanges in other relationships (Walter et al., 2001).

Literature confirms that relationship learning positively influences the innovation performance of the relationship and that, in the process of absorption and transformation of external flows of knowledge into innovation outcomes, the role played by absorptive capacity changes continuously. Moreover so, ACAP affects different capabilities and routines as well as the innovation performance outcomes (Chen, Lin & Chang, 2009; Fosfuri & Tribo, 2008; Selnes & Sallis, 2003). Academic research suggests that firms with high PACAP are more adept at updating their stock of knowledge, thereby attaining higher levels of explorative learning. On the other side of the coin, RACAP includes knowledge transformation and exploitation and will typically enhance exploitative performance (Liebeskind, 1996; Zahra & George, 2002). Lastly, pioneering work in this field of research has assumed that performance differences may arise when units follow different developmental paths of PACAP and RACAP (Jansen et al. 2005) and that ACAP positively influences business unit performance (Tsai, 2001). Even though the relationship between PACAP and RACAP and respectively, explorative and exploitative learning performance therefore does not appear to need additional testing, this research chooses to do so as previously research on the relationships mentioned (Jansen et al. 2005; Tsai, 2001) analyzed data from an intra-firm setting. This study will focus on an inter-firm setting and as so the results can be of importance for any implications made. From this reasoning we therefore hypothesize that:

H9: Potential absorptive capacity (PACAP) is positively related to explorative learning performance.

H10: Realized absorptive capacity (RACAP) is positively related to exploitative learning performance.

(17)

17

3. Methodology

3.1 Research design

The concepts and relations identified in the literature review have not been fully empirically addressed in literature and no conclusive evidence concerning the effects of the variables in these relations has been found. In a situation that phenomena faced by many companies have not been fully addressed by academic literature and where empirical evidence is still lacking or inconclusive, a theory testing approach is deemed to be most suitable (Van Aken, Van der Bij & Berends, 2012). In this study, relationship duration and product importance are hypothesized to affect the two dimensions of ACAP and to potentially be mediated by constructs in the relational and institutional context. PACAP and RACAP are in turn, hypothesized to positively influence explorative as well as exploitative learning performance. Originating from literature we have generated hypotheses that will be tested by means of the collected data as described below.

3.2 Data collection

(18)

18 3.3 Measures

Most scales used in the questionnaire originate from previous research. Technically, to test our conceptual model we will make use of first order and second order constructs. “A

first-order construct is a latent construct that has observed variables as indicators, whereas second-order constructs have other latent (first-order) constructs as their indicators”

(Berger, 2015, p.91). Secondly, the constructs are measured by scales that are reflective or formative. The difference between these types of constructs is most clearly explained by Jarvis, MacKenzie & Podsakoff (2003). The authors explain that with reflective latent constructs, covariation among the indicators is caused by, and therefore reflects, variation in the underlying latent factor. Alternatively, in formative (Fornell & Bookstein, 1982) latent constructs, changes in the indicators cause changes to the underlying construct. Naturally, the indicators to such a constructs are not caused by a single latent construct but together cause the latent construct (Jarvis et al. 2003). Resulting from this distinction, according to literature, high correlations and internal consistency are expected for reflective constructs (Bollen & Lennox, 1991). However, as this cannot be assumed for formative latent constructs, literature suggests that to assess the validity of formative latent constructs (Jarvis et al. 2003) researchers must assess validity by means of to nomological and/or criterion-related reasoning (Bollen & Lennox, 1991).

3.3.1 Potential and Realized Absorptive Capacity (second-order constructs)

We make use of the reconceptualization of ACAP into PACAP and RACAP (Zahra & George, 2002). As we have seen, PACAP exists of the capabilities of knowledge acquisition and knowledge assimilation while RACAP comprises transformation and exploitation capabilities. Before the data collection was conducted, no appropriate scale of ACAP existed, neither of PACAP and RACAP. Scales for all four ACAP dimensions were generated by Berger (2015) based on a definitions from Zahra & George (2002), Lane et al. (2006) and the multilevel perspective of Sun and Anderson (2010). In the questionnaire thus the capability of a firm to recognize, acquire and assimilate knowledge is measured by questions 1.1 – 1.5 (Acquisition, formative scale) and questions 2.1 – 2.5 (Assimilation, reflective scale). RACAP was measured through questions 3.1 – 4.4 (Transformation, formative scale) and 5.1 – 5.5 (Exploitation, reflective scale). Participants were asked to answer these questions on a 1 – 7 Likert Scale with answers ranging from (1) “strongly disagree” to (7) “strongly agree”.

(19)

19 constructs for inner model testing (Jarvis et al. 2003). The same validity and reliability tests will be applied to the second-order constructs (Bollen & Lennox, 1991).

3.3.2 Explorative and Exploitative Learning Performance

As seen in the literature review, explorative learning refers to the application of external knowledge and the integration of the knowledge with the organizations activities to generate new knowledge from it (Lane et al. 2001). For the measurement of this construct, questions 8.1 – 8.5 of the questionnaire are used (formative scale). This scale was taken from Lane et al. (2001) after which the wording was slightly adjusted to relate to interfirm buyer-supplier relations rather than relations between joint ventures and their international partners (Berger, 2015). Five questions on a 1 -7 Likert scale were filled out by participants with answers ranging from (1) “to no extent” to (7) “to a great extent”.

Secondly, exploitative learning performance was referred to as the ability to enhance the efficiency and effectiveness of a relationship when the partners share knowledge (Selnes & Sallis, 2003). The scale is use originates from the same study from which we lend our definition of exploitative learning performance and consists of 7 questions to be answered on a 1 – 7 Likert scale (questions 9.1 – 9.7, formative scale) with answers ranging from (1) “strongly disagree” to (7) “strongly agree” (Selnes & Sallis, 2003).

3.3.3 Relational Norm (second-order construct) and Connectedness

To measure the two hypothesized mediators, scales from existing research were used. To measure the relational norm, statements from Heide and John (1992) representing flexibility (15.1 – 15.3), information exchange (15.4 – 15.7) and solidarity norm (15.8 – 15.10) were used (Berger, 2015). All three first-order constructs are reflective. Similar to PACAP and RACAP, the first-order constructs will be tested separately for validity and reliability before they can be merged in to a second-order latent variable for inner model testing.

(20)

20

3.3.4 Relationship Duration and Product Importance

The relationship duration is measured by the first question from the general questions section in the questionnaire (0.1): “How long have the two companies been involved in the

relationship?” (Berger, 2015). The response is expressed in years.

Product importance is not represented in the questionnaire however has been derived through the 2x2 research design. The research design was used for selection of respondent companies for the research based on their relationship length and the importance of the product (Berger, 2015). The product or component importance can be either “average” (1) or “crucial” (2) for the responding buyers and suppliers.

3.3.5 Control variables

(21)

21 3.4 Statistical Analysis

As hypothesized and shown in the conceptual model (Appendix 1), four hypotheses state to expect a mediating effect of Connectedness and the Relational Norm. Baron and Kenny (1986) explain that “a given variable may be said to function as a mediator to the extent that

it accounts for the relation between the predictor and the criterion” (p.1176). For mediation

to hold, three conditions must be met: (1) there must be an effect of the independent variable on the dependent variable, (2) the independent variable must affect the mediator and (3) the mediator must affect the dependent variable. When the effect of the independent on the dependent variable strongly decreases, when a mediator is included in the model, we speak of partial mediation. In the more extreme situation that the relationship between dependent and independent variables is reduced to zero when introducing a mediating variable, we speak of full mediation (Baron & Kenny, 1986).

(22)

22 The statistical analysis of the model and its inner relationships will be done with the SmartPLS software tool. SmartPLS is one of the leading PLS-SEM tools that is highly suitable for latent variable modeling. In doing so the default settings were used.

3.5 Feasibility (Controllability, Reliability and Validity)

To account for the feasibility and quality of this research, according to Van Aken et al. (2012) a study must be controllable, reliable and valid. Referring to the controllability of this research the previous sections have already slightly elaborated the process of the study. To sum up, a standardized questionnaire was held among 332 participants that were key informants of either side to 166 buyer supplier relationships. The dyads were selected based on the Dutch buyers and matched with the suppliers from around the world. All dyads trade in tangible products. The questionnaire was used to interview the Dutch buyers in person. Afterwards, the questionnaire was translated in English and extensively compared to the original version before it was sent to the supplier counterparts by mail or e-mail. The questions and statements used for this research can be found in Appendix 3.

To ensure reliable research, potential sources of bias have been accounted for. Firstly, with regard to respondents bias, the relatively large sample size of 332 respondents increases the reliability of the research. The fact, however, that only one observation was made for every organization obliges us to assume a certain degree of respondents bias. Secondly, the research did not incur researcher bias as the questionnaire was developed and the data collected by a different researcher than the one conducting the study. We can therefore assume questionnaire questions or results not to be steered into a certain direction and researcher reliability to be high. Next, the consistency of the survey and the research on the scales done beforehand allows us to assume that there is no instrument reliability bias involved in this research. Lastly, as respondents answered the questionnaire in a self-determined time and place we assume there is little bias stemming from the situation in which the research took place. Nonetheless, as we did not control for this situation we cannot conclude it to be bias free (Van Aken et al. 2012).

(23)

23 high proportion of variance and whether we can discriminate between the constructs that should not be related (Hair et al. 2014).

(24)

24

4. Analysis and results

In general PLS-SEM exists of two phases. Firstly, the quality of the measurement (outer) model is evaluated based on the reliability and validity measures of the constructs (Hair et al. 2014). In this phase, construct reliability, convergent and discriminant validity as explained in the methodology will be estimated to verify the quality of the outer model. Within the process of outer model testing, the researcher must distinguish between formative and reflective constructs. The two measurement approaches require different evaluative measures (Hair et al. 2014). The measurement model is tested using the full (332 observations) database. Secondly, if the inter-correlations between the indicators and the difference between constructs appear to be sufficiently strong, the structural (inner) model is used to assess the relationships between the latent constructs (Hair, Ringle & Sarstedt, 2011). Additionally, perceptional differences of their relationships might exist between buyers and suppliers (Ambrose, Marshall & Lynch, 2010). Therefore, besides the full database, the structural model will also be tested using separate buyer-only and supplier-only data. The second phase allows us to make inferences about the hypotheses and draw conclusions. 4.1 Measurement model

We will begin the assessment of the measurement model by testing the reliability and validity of the reflective constructs. Traditionally, Cronbach’s alpha is used to measure construct reliability however, composite reliability provides a more adequate measure of internal consistency within the working principles of PLS-SEM (Hair et al. 2014). We will however provide both, to assess the fit and usefulness of all reflective indicators.

(25)

25

TABLE 1:

Construct reliability and convergent validity of reflective scales

Constructs/indicators Outer Loading p-value Composite Reliability Cronbach's Alpha Assimilation 0.853 0.786 VAR2.1 0.618 < 0.01 VAR2.2 0.752 < 0.01 VAR2.3 0.778 < 0.01 VAR2.4 0.774 < 0.01 VAR2.5 0.734 < 0.01 Exploitation 0.842 0.765 VAR5.1 0.664 < 0.01 VAR5.2 0.663 < 0.01 VAR5.3 0.786 < 0.01 VAR5.4 0.862 < 0.01 VAR5.5 0.600 < 0.01 Connectedness 0.886 0.827 VAR12.1 0.748 < 0.01 VAR12.2 0.883 < 0.01 VAR12.3 0.823 < 0.01 VAR12.5 0.792 < 0.01 Flexibility 0.845 0.728 VAR15.1 0.836 < 0.01 VAR15.2 0.853 < 0.01 VAR15.3 0.716 < 0.01 Information Exchange 0.852 0.773 VAR15.4 0.695 < 0.01 VAR15.5 0.813 < 0.01 VAR15.6 0.754 < 0.01 VAR15.7 0.808 < 0.01 Solidarity 0.899 0.832 VAR15.8 0.876 < 0.01 VAR15.9 0.907 < 0.01 VAR15.10 0.810 < 0.01 Second-Order Constructs Relational Norm 0.875 0.786 Flexibility 0.748 < 0.01 Info Exchange 0.882 < 0.01 Solidarity 0.874 < 0.01

(26)

26 Formative scales require a different measurement approach to assess the reliability and validity of the construct since formative indicators measure different aspects of the same latent variables (Berger, 2015). According to Hulland (1999) reliability in an internal consistency sense is not meaningful for formative indicators. Furthermore, to assess reliability of a formative construct Bollen and Lennox (1991) explicitly advise researchers not to rely on correlation matrices to select or delete indicators from a construct. This might lead to the loss of valid measures and data. In fact, Hair et al. (2014) state that eliminating any formative indicator should be an exception as “omitting an indicator is equivalent to

omitting a part of the construct” (p.113). Comparing this literature with the accounted for

reliability bias as described in a previous section we therefore assume the formative constructs to be reliable. A next step is to determine the validity of formative constructs (Diamantopoulos, Riefler & Roth, 2008). Validity measurement of formative constructs is a controversial issue. Opinions of authors differentiate as to whether quantitative measures to assess formative validity exist at all or are limited (Diamantopoulos et al. 2008). Like Edwards and Bagozzi (2000) however, we believe it is bad practice to claim a formative measure and thereby assume validity. Therefore we will use several measures to assess formative construct validity while keeping the advice of Hair et al. (2014) not to delete indicators in mind.

(27)

27 not exceed the threshold of 3.3 (Petter, Straub & Rai, 2007). The above mentioned measures are depicted for all formative constructs in Table 2.

TABLE 2:

Convergent validity of formative scales

Constructs/indicators _Outer

Weight p-value VIF

Acquisition VAR1.1 0.328 0.044 1.281 VAR1.3 0.049 0.750 1.194 VAR1.4 0.719 < 0.001 1.415 VAR1.5 0.157 0.262 1.443 Transformation VAR3.1 0.254 < 0.001 1.417 VAR3.3 0.178 0.005 1.069 VAR3.4 0.366 < 0.001 1.234 VAR4.2 0.239 0.007 1.728 VAR4.3 0.270 0.001 1.747 VAR4.4 0.196 0.046 2.066 Explorative LP VAR8.1 -0.256 0.147 2.095 VAR8.2 0.300 0.076 1.937 VAR8.3 0.612 < 0.001 2.002 VAR8.4 0.120 0.553 1.873 VAR8.5 0.432 0.033 1.958 Exploitative LP VAR9.1 0.097 0.371 1.461 VAR9.2 0.155 0.256 1.692 VAR9.3 0.053 0.586 1.372 VAR9.4 0.121 0.210 1.401 VAR9.5 0.088 0.410 1.697 VAR9.6 0.205 0.100 1.875 VAR9.7 0.678 < 0.001 1.441

Second Order Constructs

PACAP Acquisition 0.335 0.003 1.502 Assimilation 0.768 < 0.001 1.502 RACAP Transformation 0.842 < 0.001 1.848 Exploitation 0.215 0.016 1.848

(28)

28 VAR4.3 & VAR4.4 relate to how knowledge is shared between and transferred to different parts of the organization. Secondly, looking at the formative indicator loadings, VAR1.2 showed the last insignificant p-value (p=0.245) and was deleted. Like VAR1.2, VAR1.1 and VAR1.3 question “awareness of relevant external technology” therefore allowing for deletion of the former indicator. The last variable that was considered for deletion was VAR3.2 (Transformation). Unlike VAR1.3, VAR9.3 and all indicators belonging to Explorative and Exploitative Learning Performance that were also considered based on their outer weight significance, deletion of VAR3.2 (p=0.245) was allowed resulting from the compensation of other construct indicators; in this case VAR3.1, VAR3.3 and VAR3.4 relating to “matching new to what is already known” (Berger, 2015). Prioritizing the completeness of the constructs (Hair et al., 2014), it was decided to leave the remaining indicators untouched despite the remaining low outer weight p-values. The model was reassessed to come to the final scores in Table 2. Based on the assumed validity of the first-order formative constructs, creation of the second-order constructs was permitted and validity assessment shows adequate p-values for its outer weights and loadings together with strong conceptual support as seen in previous sections (e.g. Zahra & George, 2002).

The final quality check of the measurement model consisted of an assessment of the discriminant validity of all constructs (first- and second-order) included by means of the Fornell-Larcker criterion. This criterion suggests that the squared root of average variance extracted (AVE) for a construct must be greater than the correlations with the other latent variables (Fornell & Larcker, 1981). The table below provides an overview of all first-order constructs.

TABLE 3:

Discriminant validity first-order constructs

(29)

29

TABLE 4:

Discriminant validity second-order constructs

1 2 3

1. PACAP (Acquisition & Assimilation) 0.887

2. RACAP (Transformation & Exploitation) 0.730 0.916

3. Relational Norm (Flexibility, Information Exchange, Solidarity) 0.469 0.565 0.837

Leaving aside Transformation, all first-order constructs show discriminant validity based on their AVE coefficients. The lesser ability to discriminate between Transformation and Exploitation can be explained as both capabilities conceptually belong to RACAP (Zahra & George, 2002). We ascribe Transformation and Assimilation’s lack of discriminant validity to an overlap in questions regarding “matching and understanding new knowledge to a firm’s existing knowledge base”. Nonetheless, as this research and the hypotheses focus on second-order constructs rather than separate ACAP capabilities, a more relevant assessment of discriminant validity is one that includes the second-order constructs. The squared roots of the AVE per constructs are higher than the correlation coefficients of other second-order construct therefore we can conclude the measurement model by confirming discriminant validity (Table 4).

4.2 Structural model

The structural model consists of first-order and second-order constructs and intends to control for three variables: Firms Size, R&D Intensity and Industry. Before testing the direct and indirect relations among the constructs, we will assess whether the proposed control variables are relevant to the model. This analysis is done for each individual database (Buyer & Supplier, Buyer-only and Supplier only). An appropriate measure of fit between variables and a model is R squared (R2). R2 represents the variance of the dependent variable (DV) explained by independent variables (IV). When adding a variable to a model, R2 automatically increases therefore making it an unsuitable coefficient to assess the fit of the control variables. Selection of control variables based on R2 when adding variables to the model would therefore lead to biased selection. Adjusted R2 however, accounts for this phenomenon and instead is a better measure to compare the suitability of control variables (Gefen, Straub & Boudreau, 2000).Table 5 provides the adjusted R2 values of the dependent variable constructs and the significance level of the direct effect of the variables on the constructs when separately adding variables.

(30)

30 Performance implies that larger firms in addition to explorative learning performance might also lose the flexibility (Jansen et al., 2006) to become more effective and efficient. The more distinguishable positive effect of R&D Intensity on PACAP confirms theory that states that internal R&D enhances a firm’s ability to identify and assimilate external technology and makes a firm more receptive to external knowledge (Cohen & Levinthal, 1990; Veugelers, 1997; Coombs & Bierly, 2006). Moreover, buyers in particular appear to benefit from internal higher R&D expenditures, allowing them to more easily exploit the knowledge they internalize and generate new knowledge. Industry differences did not show to have significant effects on the DV’s. For the full as well as the supplier database, adjusted R2

values increase when all control variables are included. Despite the little decrease of adjusted R2 in the buyer database (PACAP & Explorative LP) it was decided to include the control variables in all models in for comparability.

TABLE 5:

Adjusted R-squared values, path coefficients and p-values per database Buyer & Supplier PACAP

Explor. LP RACAP Exploit. LP Buyer-only PACAP Explor. LP RACAP Exploit. LP Adjusted R2 0.310 0.150 0.453 0.326 Adjusted R2 0.357 0.180 0.528 0.446

with Firm Size only 0.312 0.154 0.453 0.333 with Firm Size only 0.357 0.188 0.525 0.453

Path coefficient -0.066 -0.081 -0.049 -0.099 Path coefficient -0.068 -0.130 0.007 -0.111

P-value 0.198 0.258 0.322 0.088† P-value 0.356 0.330 0.926 0.216

with R&D Intensity only 0.315 0.152 0.457 0.323 with R&D Intensity only 0.383 0.209 0.538 0.442

Path coefficient 0.101 0.044 0.065 0.005 Path coefficient 0.183 0.233 0.101 0.064

P-value 0.018* 0.534 0.135 0.936 P-value 0.002** 0.076† 0.063† 0.440

with Industry only 0.309 0.150 0.453 0.326 with Industry only 0.355 0.180 0.525 0.449

Path coefficient 0.044 -0.036 0.031 -0.051 Path coefficient 0.067 -0.065 -0.019 -0.082

P-value 0.367 0.525 0.476 0.390 P-value 0.320 0.488 0.735 0.331

all Control Variables 0.316 0.154 0.456 0.330 all Control Variables 0.381 0.220 0.532 0.454

Supplier-only PACAP Explor. LP RACAP Exploit. LP Adjusted R2 0.308 0.181 0.402 0.227

with Firm Size only 0.308 0.179 0.413 0.235

Path coefficient -0.067 -0.048 -0.118 -0.118

P-value 0.287 0.541 0.085† 0.204

with R&D Intensity only 0.306 0.182 0.400 0.226

Path coefficient 0.008 -0.100 0.047 -0.076

P-value 0.915 0.376 0.487 0.648

with Industry only 0.304 0.179 0.402 0.222

Path coefficient 0.052 -0.022 0.056 0.005

P-value 0.474 0.829 0.377 0.959

all Control Variables 0.303 0.178 0.411 0.229

(31)

31 Assessing whether mediation holds, requires analysis of the path coefficients of direct as well as indirect effects. In this analysis, direct effects represent the effect of an IV on a DV without a mediator included. Indirect effects assume the effect of an IV on a DV when a mediator is included in the model. Coefficients and significance levels of these effects determine whether mediation holds. All p-values of the effects were generated by using the bootstrapping algorithm in SmartPLS.

Hypotheses 1 and 3 hypothesize a positive non-linear effect. Therefore, a natural logarithm of Relationship Duration has been added to the model as an IV affecting PACAP and RACAP (Rel. Dur. LOG). Results for the model tested with the full database, buyer database and supplier database are depicted in Table 6, Table 7 & Table 8 below. Based on the result we can confirm or reject the hypotheses. Hypotheses will primarily be accepted or rejected based on the full data model however, discrepancies in results between the different datasets compel us to nuance our conclusions (Ambrose et al., 2010). Within the full database the possibility of buyer and supplier observations neutralizing each other exists thereby leading to nonsignificant results.

TABLE 6:

Direct and indirect effects buyer & supplier data combined

PACAP p-value RACAP p- value Explor. LP p-value Exploit. LP p-value Rel. Norm p-value Connect. p-value Direct Effects Rel. Duration 0.075 0.457 0.010 0.911 -0.046 0.499

Rel. Dur. LOG -0.036 0.573 -0.038 0.495

Prod. Importance 0.188 0.000** 0.135 0.003** 0.061 0.264 PACAP 0.383 0.000** RACAP 0.566 0.000** Relational Norm 0.325 0.000** 0.338 0.000** Connectedness 0.251 0.000** 0.401 0.000** Ìndirect Effects Rel. Duration -0.015 0.526 -0.015 0.520 Prod. Importance 0.015 0.283 0.024 0.283 Note: ** p ≤ 0.01 ; p ≤ .05 *; p ≤ .10 †

4.2.1 Full database (buyer & supplier)

(32)

32 PACAP (x=0.118, p<0.001) as well as RACAP (x=0.135, p=0.003) which indicates support for H5 & H7. Nonetheless, the indirect effects of product importance are not significant, therefore renouncing mediation and rejecting H6 (x=0.015, p=0.283) and H8 (0.024, p=0.283). As expected, PACAP and RACAP show significant effects on the learning performances they were hypothesized to be linked with, thereby supporting H9 (x=0.383, p<0.001) and H10 (x=566, p<0.001).

TABLE 7:

Direct and indirect effects of buyer-only data

PACAP p-value RACAP p- value Explor. LP p-value Exploit. LP p-value Rel. Norm p-value Connect. p-value Direct Effects Rel. Duration 0.012 0.928 -0.004 0.974 -0.157 0.016*

Rel. Dur. LOG 0.093 0.257 -0.002 0.976

Prod. Importance 0.146 0.023* 0.087 0.115 0.104 0.184 PACAP 0.349 0.012* RACAP 0.659 0.000** Relational Norm 0.416 0.000** 0.435 0.000** Connectedness 0.216 0.004** 0.387 0.000** Ìndirect Effects Rel. Duration -0.065 0.033* _-0.068 _0.025* Prod. Importance 0.023 0.235 0.040 0.219 Note: ** p ≤ 0.01 ; p ≤ .05 *; p ≤ .10 † 4.2.2 Buyer-only database

(33)

33

TABLE 8:

Direct and indirect effects of supplier-only data

PACAP p-value RACAP p- value Explor. LP p-value Exploit. LP p-value Rel. Norm p-value Connect. p-value Direct Effects Rel. Duration 0.201 0.217 0.078 0.632 0.042 0.642

Rel. Dur. LOG -0.179 0.073† -0.088 0.384

Prod. Importance 0.247 0.000** 0.145 0.024* 0.029 0.703 PACAP 0.429 0.000** RACAP 0.468 0.000** Relational Norm 0.145 0.216 0.253 0.020* Connectedness 0.405 0.000** _0.441 _0.000** Ìndirect Effects Rel. Duration 0.006 0.738 0.011 0.687 Prod. Importance 0.012 0.695 0.013 0.699 Note: ** p ≤ 0.01 ; p ≤ .05 *; p ≤ .10 † 4.2.3 Supplier database

In Table 8, we find that for the supplier database, conversely to what we hypothesized, a negative nonlinear effect between relationship duration and PACAP exists (x=-0.179, p=0.073). The same effect on RACAP is insignificant therefore rejecting H1 and H3. For the supplier database the direct effect of product importance on PACAP (x=0.247, p<0.001) and RACAP (x=0.145, p=0.024) are both significant indicating support for H5 and H7. None of the hypothesized mediated indirect effects are significant as can be seen on the bottom of Table 8. We therefore reject H2, H4, H6 and H8 for the supplier data. Lastly, also this data sample shows strong proof for H9 and H10 (x=0.429, p<0.001 and x=0.468, p<0.001).

4.2.4 Additional analyses

(34)

34 as RACAP therefore in fact diminishes when the duration of the relationship increases for buyers.

To further investigate the hypothesized direct non-linear effects of duration on PACAP and RACAP, quadratic effects of the IV were included in the model. None of these effects proved to be significant. Neither short (< 10 years) nor long (> 10 years) duration subsamples showed any significant effects. These additional analyses give us no proof of a non-linear relation between the two variables. Results can be found in Table 9 however, other statistical applications might contradict our findings and might be able to assess the relationship more effectively.

TABLE 9:

Subsampled and squared effects on PACAP and RACAP

PACAP p-value RACAP p-value

Buyer-only Database

Rel. Duration (< 10 years) 0.053 0.931 0.032 0.935 Rel. Duration (> 10 years) -0.029 0.871 -0.032 0.863 Rel. Duration squared (quadratic) -0.024 0.347 -0.013 0.421 Relational Norm (< 10 years) 0.512 < 0.01** 0.501 0.016* Relational Norm (> 10 years) 0.441 < 0.01** 0.421 < 0.01** Connectedness (Average Prod. Importance) 0.273 0.014* 0.469 < 0.01** Connectedness (Crucial Prod. Importance) 0.185 0.167 0.299 < 0.01**

Supplier-only Database

Rel. Duration (< 10 years) -0.036 0.947 0.018 0.970 Rel. Duration (> 10 years) 0.245 0.242 -0.126 0.570 Rel. Duration squared (quadratic) -0.056 0.195 -0.037 0.339 Connectedness (Average Prod. Importance) 0.457 < 0.01** 0.456 0.025* Connectedness (Crucial Prod. Importance) 0.268 0.099† _{0.372 0.021}*

Buyer & Supplier Database

Rel. Duration (< 10 years) -0.019 0.924 0.060 0.651 Rel. Duration (> 10 years) 0.110 0.421 -0.060 0.623 Rel. Duration squared (quadratic) -0.024 0.347 -0.013 0.421 Connectedness (Average Prod. Importance) 0.347 < 0.01** 0.496 < 0.01** Connectedness (Crucial Prod. Importance) 0.185 0.048* 0.314 < 0.01**

Note: ** p ≤ 0.01 ; p ≤ .05 *; p ≤ .10 †

(35)

35 we will assume the following: differences in perceived product importance are incorporated in the degree of connectedness for each subsample. Therefore we consult the effects of connectedness on PACAP and RACAP within each subsample and compare the results. From Table 9 we see that, compared to average importance, perceived crucial products cause a weaker positive effect of connectedness on PACAP and RACAP for all databases.

As a final additional analysis we have taken a step back and analyzed to what extent the first-order constructs of the relational norm were influenced by the relationship duration for all three databases. The results are depicted in Table 10 and will be discussed onward in the next section.

TABLE 10:

Effects of Relationship Duration on Flexibility, Information Exchange and Solidarity

Flexibility p-value Info Exchange p-value Solidarity p-value Relationship Duration Buyer-only -0.198 0.004** -0.110 0.087† _{-0.111 0.069}† (<10 years) -0.029 0.845 0.082 0.460 -0.022 0.861 (>10 years) -0.156 0.121 -0.051 0.571 -0.089 0.269 Supplier-only 0.021 0.834 -0.005 0.962 0.106 0.082† (<10 years) -0.050 0.677 0.004 0.974 0.071 0.506 (>10 years) 0.172 0.187 0.205 0.201 0.251 0.022* Buyer & Supplier -0.074 0.290 -0.054 0.376 0.008 0.885

(36)

36

5. Discussion, Implications and Conclusion

5.1 Discussion

Do the duration of a relationship and the importance of the product that is exchanged increase or decrease interfirm absorptive capacity and do these antecedents matter in the continuous strive for innovative learning performance? As organizations attempt to improve their ability to acquire, assimilate, transform and exploit new knowledge to benefit their innovative performance, these questions are crucial for managing, creating and terminating relationships between buyers and suppliers. As the quality of complementarity between buyer and supplier enables the realization of explorative as well as exploitative learning (Berger, 2015), in an attempt to contribute to literature we empirically tested the effects of two under researched antecedents in the relational and institutional context on interfirm absorptive capacity. Our results showed no support for any hypothesized direct linear effects of relationship duration (Table 6 & Table 9) on ACAP although significant differences exist between the indirect influence of the relationship’s duration for buyer firms and that for supplier firms separated. Within the buyer-only analysis we encountered full mediation.

(37)

37 relationship duration by the relational norm on PACAP and RACAP we see in Table 7 (PACAP; x= -0.065, p=0.033 and PACAP; x= -0.068, p=0.025).

Nevertheless, why do we see proof for this reasoning for the buyer sample but not for our supplier sample? And moreover, why is this relationship a negative one contrary to our expectations and previous research? We appropriate this first difference to the higher value buyers accredit to credibility and more importantly trust, relative to suppliers when assessing relationship success (Ambrose et al., 2010). As we have seen, trust is strongly related to the creation of a relational norm (Gulati, 1995; Gulati et al. 2000) and interfirm ACAP can result from successful relationships. Therefore, buyers typically will focus on the building of trust and such understanding, thereby increasing their potential for explorative and exploitative innovation performance. Suppliers on the other hand more highly value communication and commitment in their assessment of relationship performance (Ambrose et al., 2010) and as such will focus on the strengthening of these concepts. This perceptional difference might for example explain why in the strive for relationship success buyer’s PACAP is strongly affected by the relational norm (Table 7: x=0.416, p<0.01 vs. x=0.216, p=0.004) and for suppliers PACAP is only strongly affected (Table 8: x=0.405, p<0.01 vs. x=0.145, p=0.216) by connectedness that inherently links to communication (Ambrose et al., 2010; Jansen et al., 2005). The results lead us to argue that a supplier’s PACAP does not benefit from the existence of a relational norm which, in line with Sethi, Smith & Park (2001), beyond a moderate level therefore might even have a negative effect on the explorative innovativeness of the focal company. Referring to RACAP, supplier and buyer firms both benefit from the relational norm, however comparable to PACAP, buyers seem to benefit much more from the existence of a strong relational norm.

(38)

38 willingness to make adaptions within the relationship (x=-0.198, p=0.004). The proactive sharing of information and commitment to the relationship in the forms of information exchange and solidarity also appear to suffer when relationship lengths increase (x=-0.110, p=0.087; x=-0.111, p=0.069). These findings imply that a longer lasting relationship leads to a perception among buyers that not only the focal company, but also its partners are becoming lethargic or uninterested. Suppliers on the other hand, seem to differ in perception when the relationship is aging. The more years pass, the stronger the suppliers feel that a greater degree of commitment to the relationship starts to exist (x=0.106, p=0.082). When subsamples were created, high and low duration subsamples even proved that in earlier stages solidarity does not seem benefit along the duration of the relationship, however, after ten years it benefits from the shared experience and multiple interactions (Rel. Dur. < 10 years: x= 0.071 p=0.506, Rel. Dur. > 10 years x=0.251, p=0.022).

The significant diminishing negative effect of duration on PACAP (x=-0.179, p=0.073) as seen in Table 8 can be attributed to more knowledge transferred from buyer to supplier in early stages with the intention to educate the supplier in the buyer’s needs (Ahuja & Katila, 2001). The hereafter increased familiarity and trust with buyers generates a sense of boredom for the supplier which thereby becomes subject to the dark side of trust (Grayson & Ambler, 1999; Moorman et al., 1992). The supplier loses its edge and thereby lowers its ability to acquire and assimilate crucial external knowledge. However, we refrain from drawing final conclusions concerning nonlinear effects of duration from this particular finding as it is isolated from the other results, and while we have empirically proven that relationship duration is mediated by the relational norm.

(39)

39 external sources of knowledge. We explain this direct positive link between product importance and PACAP and RACAP as the result of wider knowledge search to lower dependence on an existing partner. Namely, these firms seem to pay less attention and resources to formal and informal contact with their known partners in search of useful knowledge. Instead these firms become more receptive to broad sources of (product as well as non-product) related knowledge that can enhance a firm’s own knowledge base (Bloch & Richins, 1983). These knowledge sources potentially allow the firm to equally strengthen their PACAP and RACAP capabilities.

This presumed choice for more broad knowledge search and absence of a relation between importance and connectedness implies less communication with partners and therefore less trust building (Gulati, Nohria & Zaheer, 2000). As discussed before, the only inferences this research can make concerning the effect of average or crucial product importance on ACAP are based on connectedness levels. Two subsamples were created based on the product importance. Referring to Table 9, for firms dealing with products of average importance, connectedness has a distinguished stronger effect on PACAP and RACAP compared with connectedness of firms trading in perceived crucially important products. The weakening effect of a perceived crucially important product on connectedness can be explained through the Not-Invented-Here syndrome (Katz & Allen, 1982). Firms trading crucially important products acquire a more central position in the network. The perception that all required knowledge is already owned by the firm then results in rejection of knowledge from outsiders (Katz & Allen, 1982). Nevertheless, empirical research is needed to test whether crucial product importance leads to a lower degree of interfirm contact. Our research has not been able to empirically confirm these assumptions due to a lack of variance in the importance variable.

(40)

40 effects of PACAP and RACAP on innovation of the firm might matter, moreover the innovative learning performance might increase future attractiveness for partners (Walter et al. 2001). With regard to a more theoretical function, these findings allow us to extend similar conclusions of previous research beyond the intra-firm level to the interfirm level (Jansen et al. 2005; Tsai, 2001). Lastly, the composition of the formative scales included in PACAP and RACAP (Acquisition and Transformation) was different than the composition of those used by Berger (2015). A comparison of constructs and results between Berger’s work (2015) and this study shows both researchers yielded similar results. Similarity in results despite the deletion of different statements within several constructs reconfirms validity and instrument reliability for the newly developed measurement scales of ACAP’s dimensions.

5.2 Implications

5.2.1 Managerial Implications

Interfirm Absorptive Capacity: Antecedents in the Institutional & Relational Context and Innovative Learning Performance Outcomes