Interfirm Absorptive Capacity:
The tacitness of knowledge between buyers and suppliersInterfirm Absorptive Capacity:
The tacitness of knowledge between buyers and suppliersMaster Thesis, Marketing Management
University of Groningen, Faculty of Economics and Business The influence of different levels of tacitness
Abstract
Management Summary
Surviving in a competitive or turbulent environment is difficult, and organizations that are capable of achieving a competitive advantage might be able to cope better with turbulent environments (e.g. Jesus Saenz, Revilla & Knoppen, 2014; Kogut & Zander, 1992; Lewin & Voblerda, 1999; Volberda, Foss & Lyles, 2010;). Over the past decades, literature suggests that the key to achieving such an advantage lies in the knowledge-based view (e.g. Kogut & Zander, 1992; Teece, Pisano & Shuen, 1997; Powel, Koput & Smit-Doer, 1996; Volberda et al., 2010). This knowledge-based view suggests that knowledge absorption is one of the organizations most valuable resource, and it is a necessary asset to survive in a competitive and turbulent environment (Revilla, Jesus Saenz & Knoppen, 2013). Thus, organizations highly depend on external knowledge to be able to innovate and enhance organizational performance (Zollo, Reuer & Singh, 2002; Ireland, Hitt & Vaidyanath, 2002; Laursen & Salter, 2006).
This study investigates the relationship between contracting and PACAP, with the influence of different levels of tacitness. Two dimensions related to PACAP can be identified: assimilation and acquisition (Zahra & George, 2002). In this study PACAP is analyzed with both dimensions included. Although, the relationship between contracting and PACAP already is researched by several studies (e.g., Berger, 2005; Jansen, Van den Bosch & Volberda, 2005; 2006), this study is the first to suggest that different levels of tacitness might influence this relationship. Tacitness is measured with a higher level of tacitness (managerial techniques and marketing expertise) and a lower level of tacitness (manufacturing & production expertise and the new development of product expertise). The managerial relevance of including tacitness in this study is that managers will now know if tacitness influences the before mentioned relationships. Management can then build and integrate several procedures and systems to stimulate and improve the encoding process of tacit knowledge in their organization.
difference for organizations. This is important for management, since responding to a turbulent environment in the right way may create a sustainable competitive advantage.
Lastly, this study investigates the relationship between explorative learning performance and exploitative learning performance. In the literature, conflicting research can be found with regards to this topic. March (1991) suggested that both can’t occur at the same time and Tamayo-Torres, Ruiz-Moreno and Llorens-Montes (2011) suggested that exploration interacted with exploitation. Here, it is taken into consideration that exploration influences the exploitation of potential commercial opportunities (Benson & Ziedonis, 2009). Eventually the goal of this paper is to answer the following research question: Do different levels of tacitness influence the relationship between contracting and PACAP on buyer-supplier relationships? In addition, the following sub-questions ought to be answered; Do different levels of tacitness influence the relationship between PACAP and explorative learning performance? Does explorative learning performance influence exploitative learning performance? Does environmental turbulence moderate the relationship between PACAP and explorative learning performance? These relationships are analyzed based on a questionnaire send to buyers and suppliers. In total 332 participants completed the questionnaire (166 buyers and 166 suppliers). This study investigates the before mentioned relationships from two different perspectives: the buyer perspective and the supplier perspective. The managerial relevance is that when differences occur between the two perspectives management can now understand these differences and develop more effective and successful relationships.
Preface
While attending the course Contemporary Theories my interest for absorptive capacity grew. After reading literature about this topic, I chose absorptive capacity as my first choice for my master thesis. This thesis is the result of the last semester of my Master of Science, Marketing Management. After studying for four and a half years at the University of Applied Sciences in Groningen, I wanted to take on and accomplish a new challenge. In the past two years I have worked incredibly hard to complete the pre-master and master at the University of Groningen, Faculty of Economics and Business successfully. This thesis is the end-result of a period of hard work and stress, but it was also really interesting. I would not have accomplished this thesis without the help of several people.
First I would like to thank supervisor Dr. Hans (J). Berger. He always gave me helpful feedback, provided me with interesting insights and thoughts for my thesis. Besides that Dr. Berger is always there for students to help with any kind of problem. According to me this makes him a good supervisor.
Secondly I would like to thank my second supervisor Dr. J.C. Hoekstra. She already gave me useful feedback during my pre master thesis and now she was willing to evaluate my master thesis as a second supervisor.
Table of contents 1. Introduction 09 2. Theoretical Background 12 2.1 Contracting 12 2.2 Tacitness 12 2.3 Absorptive Capacity 14
2.3.1 Potential Absorptive Capacity 14
2.4 Organizational learning Performances 15
2.4.1 Explorative Learning Performances 15
2.4.2 Exploitative Learning Performances 16
2.5 Environmental Turbulence 17
3 Conceptual model and hypotheses 18
3.1 Conceptual model 18
3.2 Contracting 19
3.3 Potential Absorptive Capacity 20
3.4 Exploration and Exploitation 21
3.5 Environmental Turbulence 22 4 Method 23 4.1 Data Collection 23 4.2 Measurement Scales 23 4.3 Mathematical Model 25 4.4 Statistical Procedure 26
5 Analysis and results 27
5.1 Outer Measurement Model 27
5.2 Inner Structural Model (without Tacitness) 32
5.2.1 Buyer Data 32
5.2.2 Supplier Data 33
5.3 Inner Structural Model (with Tacitness) 34
5.3.1 Buyer Data 35
5.3.2 Supplier Data 36
6. Discussion 38
6.1 Discussion and future research implications 38
6.1.1 Contracting 38
6.1.2 Potential Absorptive Capacity 40
6.1.3 Exploration and Exploitation 40
6.1.4 Moderating effect of environmental turbulence 40
6.2 Managerial and scientific implications 41
6.3 Research limitations 43
6.4 Conclusion 44
References 45
Appendices 49
Appendix 1: Measurements 49
Appendix 2: Robustness check I buyer data 50
Appendix 3: Robustness check I supplier data 51
Appendix 4: Robustness check II buyer data 51
Appendix 5: Robustness check II supplier data 51
1. Introduction
Tacit knowledge has been applied in many different ways and adopted to different definitions, suggesting that it is difficult to grasp the true meaning of tacitness. Indirectly this suggests that the complete process of knowledge transfer is difficult to understand fully and comprehensively. Research suggests that the main problem of tacit knowledge lies in the difficulty of sharing it with others (Leppälä, 2012). To underline this problem, the different definitions of tacitness will be described. Kogut and Zander (1992, p.389) identify tacitness as “noncodifiability, complexity and noteachability”. A second definition from, Spender (1996), suggesting that tacit knowledge is mostly intuitive and the knowledge is not yet abstracted from practice. Thirdly, the definition of tacitness described by Szulanski (1996) suggests that tacit knowledge is sticky and difficult to transfer. The fourth definition of tacitness from Balconi (2002, p.360), who suggested that tacitness refers to “know-how, skills, or habits employed in different tasks”. Fifthly, Nielsen and Nielsen (2009, p.1032) adopted tacitness in their research “as a key attribute that influenced both the access to and application of relevant external knowledge”. Lastly, Simonin (1999, p.599) identified tacit knowledge as “a measure of knowledge of partners” and calculated this with the “degree of codifiability and transferability”.
The aim of this study is to understand the role of tacitness. This will be the first study to empirically test if different levels of tacitness influence: (1) the relationship between contracting and PACAP, (2) the relationship between PACAP and explorative learning performance, and (3) takes into account two different perspectives ‘the buyers perspective and the suppliers perspective’. The role of tacitness is ambiguous and unclear, however according to Lane, Salk and Lyles (2001, p.1158) there are different levels of tacitness, “a higher level (managerial techniques) and a lower level (manufacturing and production expertise)”. This study will empirically investigate higher and lower level of tacitness on the before mentioned relationships. Lane et al. (2001) suggested that a higher level of tacitness is more difficult to transfer and to assimilate, because of difficulties with encoding managerial techniques into manuals and procedures (ibid.).
This study takes into consideration PACAP and the two underlying dimensions: acquisition- and assimilation (Zahra & George, 2002). The first refers to organizations being able to identify and acquire new knowledge, while the latter refers to being able to analyze and interpret the new acquired knowledge (ibid.). These capabilities are closely related to explorative learning (ibid.). When a buyer and supplier want to share knowledge in the form of relationship, a contract is often developed (Jansen et al., 2005). However, formalization tends to decrease the assimilation of newly acquired knowledge (ibid.). Furthermore, in this study, exploration and exploitation is taken into account. Exploration refers to learning through new possibilities, and exploitation refers to incrementally refining old certainties (March, 1991). Knowledge is important in all of the constructs mentioned before, especially in turbulent environments, because with more knowledge a competitive advantage can be achieved (Teece et al., 1997; Powell et al., 1996).
competitive, uncertain and unstable times, knowledge about the role of environmental turbulence is extremely important for management, since responding to a turbulent environment in the right way may create a competitive advantage. Third, this study aims to investigate the before mentioned relationships from two different perspectives: the buyer perspective and the supplier perspective. The managerial relevance is that when differences occur between the two perspectives management can now understand these differences and develop more effective and successful relationships.
To conclude, this study investigates the relationship between contracting and PACAP while being moderated with different levels of tacitness. Secondly, it is investigated whether or not the relationship between PACAP and explorative learning performances changed, when being moderated with different levels of tacitness. Another moderator is added to the relationship between PACAP and explorative learning performance, namely environmental turbulence. Lastly, the relationship between explorative learning performance and exploitative learning performance is researched. The goal of this paper is to answer the following research question: Do different levels of tacitness influence the relationship between contracting and PACAP on buyer-supplier relationships?
2 Theoretical background
In this chapter, the theoretical background of the variables that are used in this study will be discussed. Over the past decades, the interest for the topic ‘absorptive capacity’ has increased, and numerous research has been done with regards to this topic (e.g., Berger, 2015; Cohen & Levinthal, 1990; Jansen et al., 2005; Jaworski & Kohli, 1993; Lane et al., 2001; Lane & Pathak, 2006; March, 1991; Teece et al., 1997; Volberda et al., 2010; Zahra & George, 2002). The literature above is the basis for the theoretical background and conceptual model in the current study.
2.1 Contracting
Research identifies two important elements that explain contracting: formalization and governance. According to Khandwalla (1977) and Jansen et al. (2005, p.1002) formalization could be identified as “the degree to which rules, procedures, instructions and communications are formally written in documents or formal systems”. When an organization relies on these rules and procedures this can “reduce the likelihood that individuals will deviate from established behavior” (ibid.). However, “formalization prevents the knowledge interaction and hinders the assimilation of new knowledge between individuals” (ibid.). In this case formalization “acts as a frame of reference that constrains exploration efforts and directs attention toward restricted aspects of an external environment” (ibid.). This results into “a limited scope of efforts in knowledge acquisition”.
According to Gosh and John (2005, p.346) governance can be defined as “the explicit and implicit rules of exchange between economic parties”. Which follows is that, “these rules allow each stakeholder to make a claim” (Gosh & John 2005, p.347). This is based on “the expected share of value generated with the alliance” (ibid.). However, literature suggests that the contract should include “a specification of assets and investments made by both parties” (Ambrose, Marshall & Lynch 2010, p.1271). When these are formally written down this will result into a “reduction of uncertainty and increases trust” (ibid.).
2.2 Tacitness
reflection” (Faulkner & Runde 2004, p.473). To get a better understanding of ‘tacitness’ it is important to understand the underlying elements. Research suggests that tacitness is a feature of knowledge (Martin & Salomon, 2003). Ambrosini and Bowman (2001, p.811) identified several elements related to tacitness: “difficult to write down or formalize, personal knowledge is needed, it’s difficult to communicate to others, it’s practical, describes processes and is context specific”. Tacit knowledge is extremely hard to codify and is difficult for competitors to imitate, because know-how is needed to use tacitness in practice (Kogut and Zander, 1993). The process of codification will ensure that organizations protect their own information instead of losing it to their main competitors (ibid.). However, acquiring knowledge is a costly process and therefore organizations might not engage in the process of encoding knowledge (Schulz & Jobe, 2001). However, when an organization is successful in acquiring tacit knowledge this might results in a sustainable competitive advantage (Bierly III et al., 2009). Organizations that stimulate the acquisition of tacit knowledge they are viewed as open for creativity, innovation and coordination (Schulz & Jobe, 2001).
Research suggests that there is a moderating effect of tacitness on the relationship between external learning capabilities and knowledge application (Bierly III et al., 2009). Other literature suggests that tacitness mediated the relationship between partner characteristics and alliance outcomes (Nielsen & Nielsen, 2009). They concluded that tacitness enhanced organizational learning and performance. Moreover, numerous research suggests that tacitness stimulates the process of exploration and strengthens the effect of exploration on new product innovation and product development (e.g., Zhang et al., 2015; De Luca et al., 2007; McEvily & Chakravarthy, 2002). Research suggests the differences in the degree of tacitness with opportunity identification (Smith, Matthews & Schenkel, 2009). They suggest that “the degree of tacitness may result in a contingent relationship” and “the utility of knowledge is contingent upon its degree of tacitness” (2009, p.43). Moreover, research shows that depending on the type of knowledge there is a higher level of tacitness and a lower level of tacitness (Lane et al. 2001, p.1158). They identify five separate types of tacit knowledge: “managerial techniques, technological expertise, marketing expertise, manufacturing & production processes and product development expertise” (ibid.).
2.3 Absorptive Capacity (AC)
The reconceptualization and extension of Zahra & George (2002) clarified the ambiguity and diversity of the definitions, components, antecedents and outcomes of AC. Zahra and George (2002, p.186) identified absorptive capacity as “a set of organizational routines and processes by which firms acquire, assimilate, transform, and exploit knowledge to produce a dynamic organizational capability”. Whilst other research, a more process-based perspective refers to AC as “a firm’s ability to utilize external knowledge through the three sequential processes: (1) recognizing and understanding potentially valuable new knowledge outside the firm through exploratory learning, (2) assimilating valuable new knowledge through transformative learning, and (3) using the assimilated knowledge to create new knowledge and commercial outputs through exploitative learning” (Lane & Pathak 2006, p.856). In line with this Cohen and Levinthal (1990, p.128) suggested that AC could be “recognized, valued, assimilated and applied to commercial ends”.
One can define four capabilities/dimension relevant to AC: “acquisition, assimilation, transformation and exploitation” (Zahra & Geoge 2002, p.189). These four capabilities are divided into two different, but complementary dimensions: Potential Absorptive Capacity (PACAP) and Realized Absorptive Capacity (RACAP). When an organization is efficient in applying PACAP/RACAP this could result in an organizational dynamic capability. This is, according to Teece et al. (1997, p.516), “a firm’s ability to integrate, build, and reconfigure internal and external competences to address rapidly changing environments”. Based on a study conducted by Volberda et al. (2010), there are multiple possible outcomes when applying the AC construct; a competitive advantage could be obtained (Cohen & Levinthal, 1990), it can lead to innovation (Stock, Greis & Fischer, 2001), performance can increase (Lane et al., 2001) and exploration/ exploitation performance can increase (Lewin & Volberda, 1999).
2.3.1 Potential Absorptive Capacity
allow it to analyze, process, interpret, and understand the information obtained from external sources with routines and processes” (ibid.). This dimension often contains “sticky” tacit knowledge that needs to be codified and understood (Dyer & Singh 1998, p.665; Kogut & Zander, 1992; Szulanski, 1996; Teece et al., 1997). Often, this is a difficult, costly and complex process (ibid.). After successfully acquiring and assimilating external knowledge, these knowledge-based capabilities are often impossible for competitors to imitate and thus create a competitive advantage (Cohen & Levinthal, 1990; McEvily & Chakravarthy, 2002; Zahra & George, 2002).
2.4 Organizational Learning Performances
March (1991) is a pioneer in the field of exploration and exploitation in organizational learning. He suggested a trade-off between explorative- and exploitative learning performance (ibid.). Tamayo-Torres et al. (2011, p.6175) disagreed and hypothesized “that organizations develop exploration of new knowledge at the same time as they exploit their abilities”, which suggests that a combination of exploration and exploitation at the same time is a feasible strategy. However, both strategies are competing for the same scarce resources (ibid.). March (1991) proposed that an organization should always make a strategic decision. This decisions is based on acquiring new knowledge that will improve future performance, or instead use the current available knowledge to improve performance directly (ibid.). Another important implication is that, when an organization aims to survive in a rapidly changing environment, the organization should effectively select organizational forms, routines and practices to be able to adapt to the environment (ibid.). A flexible organizational design of a firm encourages the development of exploration and exploitation (Tamayo-Torres et al., 2011). Moreover, this relationship becomes stronger when an organization is performing in a turbulent and dynamic environment (ibid.).
2.4.1 Explorative Learning Performance
kinds of tests” (2000, p.768; Tamayo-Torres et al., 2011). Beckman (2006, p.744) suggested that exploration is a form of behavior of organizations that “seek to win the technological race in new niches and achieve competitive advantages in order to develop new products and technologies”.
An important objective for exploration is “experimentation with new alternatives” (March 1990, p.85). Other research suggests that “flexibility and openness for innovation is important” (Tamayo-Torres et al. 2011, p.6180). Which results in structuring the organization in such a way that it allows organizations to acquire new knowledge and solve problems in different ways (March, 1996). Organizational forms that allow firms to be flexible and adaptable to the exploration of new possibilities are decentralized and organic structures (Brown & Eisenhardt, 1997; Nickerson & Zenger, 2002; Siggelkow & Levinthal, 2003). The adoption of such a system allows organizations to increase experimentation (March 1991, p.71). However, there are also high costs involved and there will be no guarantee that the outcome is beneficial (ibid.). Uncertainty and the development of new ideas play a vital, but risky role in the exploration approach. Elaborating further on this risk, an organization could develop numerous new ideas, but not one of them is developed completely or properly (ibid.). This might have a negative influence on organizational performance (ibid.). Zahra and George (2002) suggested that there is a link between exploratory learning performance and the process of acquiring external knowledge. Moreover, it is suggested that this would have an impact on PACAP (ibid.).
2.4.2 Exploitative Learning Performance
According to March (1991), the adoption of exploitation can be best explained through the immediate rewards organizations use. These old certainties can be identified as “refinement, choice, production, efficiency, selection, implementation and execution” (March 1991, p.71). In addition, these organizations adapt incremental extension to old certainties (ibid.). Baum et al. (2000, p.768), have a slightly different view on exploitation, namely “learning in the firm through search, experimental refining, selection and reuse of existing routines in the firm”.
(Borwn & Eisenhardt, 1997; Nickerson & Zenger, 2002; Siggelkow & Levinthal, 2003). Organizational forms that allow firms to exploit old certainties and perform in a stable environment are centralized and mechanistic structures (ibid.). In addition, these centralized and mechanistic structures allow an organization to continue with standardized processes (ibid.). A negative outcome of this could be that organizations are reluctant to change which can cause organizational inertia (March 1996; Leonard-Barton 1992). Zahra and George’s study (2002) proposed an important link between exploitative learning performance and the process of applying acquired knowledge.
2.5 Environmental Turbulence
According to the literature, environmental turbulence refers to “the degree of instability and uncertainty within a firm’s market” (Helfat, Mitchell, Peteraf, Singh, Teece, & Winter, 2007; Jaworski & Kohli 1993, p.57). There are two important distinctions observable: between ‘market- and technological turbulence’ and between ‘stable- and turbulent environments’. Market turbulence refers to “the rate of change in composition of customers and customer preferences” (1993, p.57) and technological turbulence refers to “the rate of change in technological change” (ibid.). When this distinction is made, the measure environmental turbulence includes both elements. The second distinction is based on stable environments (for example a medium-tech sector like a chemical industry or a mature single industry) and turbulent environments (for example high-tech sector like a high tech industry or emerging industrial complex (Cloodt, Hagedoorn & Van Kranenburg, 2006; Glazer & Weiss, 1993; Van den Bosch, Volberda & de Boer 1999). In a stable environment, “organizations tend do develop AC aimed at exploitation, with high efficiency, a narrow scope, and little flexibility” (Van den Bosch et al. 1999, p.553). However, in a turbulent environment “organizations are likely to dedicate efforts exclusively to increasing their absorptive capacity” (ibid.). Thus, in a turbulent environment the aim will be on exploration with a high flexibility for acquiring new knowledge (ibid.). A turbulent environment results for organizations in: less loyal consumers, an intense competitive environment, increase of ambiguity and a rapid product innovation cycle (ibid.). Moreover, the ability to imitate firm’s capabilities decreases and the need for systems that are open for innovation increases (Helfat et al., 2007; Ravi & Stern, 1988; Song, Droge, Hanvanich & Calantone 2005).
sustainable competitive advantage (Kogut & Zander, 1992; Lewin & Voblerda, 1999; Volberda et al., 2010; Jesus Saenz et al., 2014). The key to achieve such an advantage appears to be knowledge-based (e.g., Teece et al., 1997; Powell et al., 1996). Vega-Jurado, Guttierrez-Gracia & Fernandez-de-Lucio (2008) suggested that different mechanisms and forms could be adopted to increase knowledge within a firm. However, these mechanisms should analyze the environment, because differences in environments require different evaluations of dynamic capabilities on AC (Eisenhardt & Martin, 2000).
3. Conceptual model and hypotheses
This chapter discusses the conceptual model of this study and the hypotheses. The conceptual model is displayed in figure 1.
3.1 Conceptual model
This study is aimed at answering the following research question: Do different levels of tacitness influence the relationship between contracting and PACAP on buyer-supplier relationships? First, this study hypothesizes that contracting influences PACAP, and this relationship is moderated with tacitness. Secondly, PACAP influences explorative learning performance and this relationship is also moderated with tacitness. Thirdly, PACAP influences explorative learning performance and this relationship is moderated with environmental turbulence. Lastly, exploitative learning performance is influenced by explorative learning performance. The hypotheses are explained in the coming paragraphs.
3.2 Contracting
According to Jansen et al. (2005, p.1002) a contract consists of rules that are formally written down by both parties involved. When an organization relies on these rules and procedures this can “reduce the likelihood that individuals will deviate from established behavior” (ibid.). However, “formalization prevents the knowledge interaction and hinders the assimilation of new knowledge between individuals” (ibid.). Moreover, based on the arguments mentioned before they hypothesized that, “formalization negatively influences acquisition and assimilation of the new external knowledge underlying PACAP” (ibid.).
In this study contracting is expected to negatively influence PACAP, because contracting is based on routines, rules and procedures (ibid.). These routines and rules have a restrictive effect the flexibility of an organization, and therefore there is a higher likelihood that an organization becomes standardized and inflexible. Which follows is that there will be less new knowledge acquired, since the organization is not open for new information due to these routines, rules and procedures that are familiar. Change is less likely to occur, because there are certain standardized, inflexible procedures that the organization already follows. However, PACAP represents the exact opposite, because it searches for new external knowledge, is open for change, is driven by flexibility and stimulates the transfer of (new) knowledge (Lane & Pathak, 2006; March, 1991; Zahra & George, 2002). As mentioned before, contracting avoids change, reduces knowledge interaction and hinders the assimilation of new knowledge (Jansen et al., 2005). Therefore, contracting is negatively related to PACAP. This results in the following hypothesis:
H1: Contracting will be negatively related to PACAP.
Tacit knowledge is identified as “difficult to pin down or formalize” (Ambrosini & Bowman 2001, p.811). In contrast, contracting aims to write down specific rules, procedures and instructions (Khandwalla, 1977; Jansen et al. 2005). In this case, specifying tacit knowledge in a contract accompanies numerous problems. Moreover, the process of encoding tacitness is difficult, because know-how is needed. This suggests that when new knowledge is acquired it needs to be encoded first before the knowledge can be assimilated and understood. Therefore, in the current study, contracting and PACAP will be positively moderated by tacitness (strengthen the relationship and becomes more negative), because of the complexity associated with tacitness and the difficulties with transferring, encoding, communicating and writing down tacitness. This results in the following hypothesis:
H2: The relationship between contracting and PACAP will be positively moderated by
tacitness (strengthens the relations and becomes more negative).
3.3 Potential Absorptive Capacity
In this study, PACAP refers to “a firm’s capability to value and acquire external knowledge but does not guarantee the exploitation of this knowledge” (Zahra & George 2002, p.190). Literature suggests that PACAP can create a competitive advantage, which might result in the creation of new possibilities (Fosfuri & Tribó, 2008). These new possibilities can help an organization to improve, and are part of organizational learning. March (1991) identified these new possibilities as result of the explorative learning performance of an organization. March (1991) suggested that exploration referred to “search, variation, risk taking, experimentation, flexibility, play, discovery and innovation” (1991,p.71). In the current study PACAP will be positively related to explorative learning performance, because of the similarities between both constructs. Both constructs search for new knowledge, new possibilities, improvements and are externally focused. In addition, PACAP and exploration aim to adopt systems that are flexible, adaptive and open for innovation (Tamayo-Torres et al. 2011). When organizations acquire and assimilate new knowledge, the exploration of new possibilities will be stimulated (March, 1991; Zahra & George, 2002). This results in the following hypothesis:
The assimilation capability often contains tacit knowledge that needs to be codified and understood (Dyer & Singh, 1998; Kogut & Zander, 1992; Szulanski, 1996; Teece et al., 1997). This is often a difficult, costly and complex process (ibid.). However, after successfully acquiring and assimilating external knowledge, these knowledge-based capabilities are often impossible for competitors to imitate and therefore create a competitive advantage (e.g., Cohen & Levinthal, 1990; McEvily & Chakravarthy, 2002; Zahra & George, 2002). In this study PACAP and explorative learning performance are positively moderated by tacitness. This is because encoding tacit knowledge is a complex process, but when organizations succeed in encoding, transferring and assimilating tacit knowledge this can create new alternative knowledge perspectives. This results in the following hypothesis:
H4: The effect between PACAP and explorative learning performance is negatively1 moderated by tacitness.
3.4 Exploration and exploitation
The distinction between exploration and exploitation is clear. Tamayo-Torres et al. (2011, p.6175), views of exploitation are preferred over Marchs’ views since his study suggests that organizations can “develop exploration of new knowledge at the same time as they exploit their abilities”. In line with this, a different study suggests that exploration influences the exploitation of potential commercial opportunities (Benson & Ziedonis, 2009). Moreover, He and Wong (2004) showed that the interaction between explorative and exploitative learning performance was beneficial and positively related to performance (ibid.). This suggests that explorative learning performance positively relates to exploitative learning performance, because the search for and experimentation of new possibilities can improve both explorative and exploitativelearning performances of an organization at the same time. This results in the following hypothesis:
H5: Explorative learning performance will be positively related to exploitative learning
performance.
3.5 Environmental Turbulence
Environmental turbulence refers to “the degree of instability and uncertainty in a market” (Jaworski & Kohli 1993, p.57). In a turbulent environment, the influence of AC on new product development was positive, and this relationship was stronger than when there was less turbulence (McMillan, Mauri & Halmilton, 2003). In a turbulent environment the “focus will be on exploration with a high flexibility for acquiring new knowledge” (Van den Bosch et al. 1999, p.553). In line with this, research suggests that the greater the exposure to external environmental sources, the more influence a turbulent environment has on PACAP (Zahra & George, 2002; Sun & Anderson, 2010). This study suggests that PACAP and explorative learning performance are positively moderated by environmental turbulence, because the acquisition and assimilation of new knowledge is extremely important to be able to survive in a turbulent environment. In addition, exploration enables an organization to react flexible, and experiment with new possibilities in uncertain times (March, 1991). This results in the following hypothesis:
H6: The effect between PACAP and explorative learning performance is positively
4Method2
This chapter discusses the data collection, the measurement scales, mathematical model and statistical procedure of this research paper.
4.1 Data collection
The data is collected from key informants form both sides of the dyad (buyers, n=166 and suppliers, n=166), because “relationships are not always observable and this might cause measurement errors” (Berger 2015, p.90; Selness & Sallis, 2003). The data of the buyers was collected from organizations in the Netherlands, and the data of the suppliers was obtained from all around the world. At the time of the study, the focus was on a variance of industries: automotive, chemicals, machinery, pharmaceuticals, semiconductors and electronics. The aim of the study was “to get insights in the vertical relationship between independent firms operating at successive stages in the production chain” (Berger 2015, p.90). In this research data was obtained from 166 matched-pair relationships between June 2011 and April 2013 (ibid.).
4.2 Measurement scales
Garson (2016) suggested that, depending on the causal links between the indicators and the constructs in the path model, these models could be reflective, formative or a combination of both (very rare). In the reflective model “the causal arrows go from the latent variable (factor) to the indicators, these indicators represent a reflection of the latent variable that has been measured” (Garson, 2016, p.17; Hair, Ringle & Sarstedt, 2011). The reflective model assumes high correlations among the reflective measures (ibid.). In the formative model “the causal arrows go from the measures to the latent variables. In this case each indicator represents a dimension of the latent variable” (ibid.). In addition, it is suggested that there are no correlations between the formative indicators of the same latent variable. In appendix 1 the reflective and formative measurement scales are indicated. In addition, some of these measurement scales were recoded. This is also indicated in appendix 1.
Existing measurement scales of previous research are used in this study. In appendix 1, the summary of the measures is presented. Contracting refers to “the degree to which rules, procedures, instructions and communications are formally written in documents or formal systems” (Khandwalla, 1977; Jansen et al., 2005). The measures of contracting were obtained
from Desphande and Zaltman (1982), Cannon and Perreault, (1999), Buvik and Reve (2002) and Jansen et al. (2006). The questionnaire proposed by Berger (2015), included first- and second-order constructs. A first order construct is identified as “a latent construct that has observed variables as indicators, while the latter has other latent constructs as indicators” (2015, p.91). “PACAP is a second-order latent factor shown by two first order dimensions: acquisition and assimilation” (Camisón & Forés 2010, p.710; Zahra & George, 2002). The first order dimension ‘acquisition’ refers to “a firm’s capability to identify and acquire externally generated knowledge” (Zahra & George, 2002). The measures of acquiring were obtained from Szulanski (1996) and Camison and Fores (2010). The other first order dimension ‘assimilation’ refers to “a firm’s capability to analyze, process, interpret and understand information assimilated with routines and processes” (Zahra & George, 2002). The measures of assimilation were obtained from Jansen et al. (2005) and Camisón and Forés (2010).
Explorative learning performance refers to “acquiring knowledge through changes in established firm processes, planned experiments, or different kinds of tests” (Baum et al. 2000, p.768). The measures of explorative learning performance were obtained from Lane et al. (2001). In this study, the measure for explorative learning performance described by Lane et al. (2001) will be used for measuring tacitness as well, however in a different way. Several studies have described Tacitness as “a feature of knowledge, valuable, know-how and skills are needed, complexity, unarticulated knowledge, abstract, sticky, difficult to transfer, describes processes and is context depended” (Ambrosini & Bowman 2001, p.811; Martin & Salomon, 2003; Kogut & Zander, 1993). However, the different levels of tacitness will be measured separately, because literature suggests “the higher the level of tacitness, the more difficult it is to transfer” (Lane et al. 2001, p.1158). This suggests that different levels of tacitness have different impacts on performance. For example, managerial and marketing skills are highly tacit and difficult to transfer, since they depend on the social aspect and experiences of organizations (ibid.). As explained before, these managerial and marketing skills are difficult to encode (ibid.). In contrast, manufacturing and production expertise are more explicit, because they are written down in procedures, rules and routines (ibid.). Therefore, a higher level of tacitness refers to: 8.5 (appendix 1)3 the creation of new managerial expertise and 8.4 creation of new marketing expertise (ibid.). However, a lower
level of tacitness refers to 8.2 (appendix 1) the creation of new product development expertise and 8.1 the creation of new manufacturing and production expertise (ibid.). Multi-Group-Analysis in PLS allows measuring the proposed differential effect in these separate tacitness-groups. The conceptual model will be estimated four times with a decreasing level of tacitness (Group A: highest level of tacitness versus group B: lowest level of tacitness, this is done from both buyer and supplier perspectives)4. These estimations will investigate the influence of tacitness on the relationships that are conceptualized before. Exploitative learning performance refers to “learning in the firm through search, experimental refining, selection and reuse of existing routines in the firm” (Baum et al. 2000, p.768). The measures of exploitative learning performance were obtained from Selnes and Sallis (2003). Environmental turbulence refers to “the degree of instability and uncertainty within a firm’s market” (Helfa et al., 2007; Jaworski & Kohli 1993, p.57). The measures of environmental turbulence were obtained from Berger (2015) and Jaworski and Kohli (1993).
4.3 Mathematical model
The hypotheses will be tested with three regression analyses. This study uses multigroup analysis to test for different levels of tacitness, because of this tacitness is not included in the mathematical model (explained in paragraph 4.2). The following equations will be estimated:
1. PACAPi = α0 + β1CTRi + ε
2. EXPLOREi = α0 + β2PACAPi + β3ETi + β4ETi × PACAPi+ ε 3. EXLPOITi = α0 + β5 EXPLOREi + ε
PACAPi = Potential Absorptive Capacity (i = 1…250) CTRi = Contracting (i = 1…250)
CONN = Connectedness (i = 1…250)
EXPLORE = Explorative Learning Performance (i = 1…250) ET = Environmental Turbulence (i = 1…250)
EXLPOIT = Exploitative Learning Performance (i = 1…250) ε = Standard error term
4.4 Statistical procedure
In this study, a software tool for Partial Least Squares (PLS) path modeling is used (Hair et al., 2011). The program used for running the software is called Smart PLS. This way of path modeling has become increasingly important for analyzing and estimating relationships between complex, latent variables (Hair et al., 2011). “The path modeling uses least squares algorithms that are based on the principal component analysis (PCA) and canonical correlations analysis to identify relationships between latent variables” (Garson 2016, p.8; Hair et al., 2011). Compared to regression analysis, PLS-SEM (Partial Least Squares – Regression & Structural Equation Models) allows analysis for “complex models, non-normal data and formative measures” (Berger 2015, p.94; Hair et al., 2011; Ringle, Sarstedt & Straub, 2012). Since the conceptual model (figure 1) includes second-order construct, formative constructs and multiple equations5 the choice for PLS is therefore logic.
Multigroup analysis is one of the statistical procedures in PLS path modeling, this Multi Group Analysis (MGA) 6, allows for comparison of group-specific effects. “A result is significant at 5% probability of error, if the p-value is smaller than 0.05 or higher than 0.95 for a certain difference of group-specific path coefficients” (Garson 2016, p.180; Henseler, Ringle & Sinkovics, 2009). Moreover, “MGA allows to first test if pre-defined data groups have significant differences in their group-specific parameter estimates (Garson, 2016, p.180; Hair, Hult, Ringle & Sarstedt, 2014; Henseler et al., 2009; Sarstedt et al., 2011)”. These estimates are “e.g., outer loadings, outer weights and path coefficients” (ibid.). Thus, “PLS-MGA tests if the PLS model significantly differs between groups” (Garson 2016, p.166).
5 Multiple equations: three equations are estimated between four different groups to infer differences among a higher versus lower level of tacitness and between buyer versus supplier.
5. Analysis and Results
This chapter discusses the analyses and results of this study. First off, the outer measurement model is assessed. Secondly, the inner structural model is assessed for both buyer and supplier data. Lastly, a robustness check will be performed. The inner structural model consists of “the relationship between the constructs” (Berger, 2015; Garson 2016, p.21). The outer measurement model consists of “the relationships between the indicators and each latent variable” (Berger, 2015; Garson 2016, p.21). The outer measurement model tests if the outer measurement model is sufficiently strong enough, based on the reliability and validity of the latent constructs (Garson, 2016). When the model is sufficiently strong the inner structural model will be assessed.
5.1 Outer measurement model
The outer measurement model is based on the combined dataset of buyer and supplier (n=332), because research on the same dataset suggests, “that the combined dataset shows the exact same statically significant indicators and constructs compared to the separate dataset of buyer and supplier” (Berger 2015, p.103). The convergent validity and discriminant validity tests if the measurement model is adequate enough. The convergent validity tests if “indicators of the same construct score highly on common variance” (Berger 2015, p.95; Hair, Black, Babin & Anderson 2010). The discriminant validity tests if a reflective construct has the strongest relationship with its own indicators (Hair et al., 2014). In addition, discriminant validity assessment tests if the construct is truly different compared with any other construct in the measurement model (ibid.).
Reflective and formative scales.
are reliable. Lastly, the composite reliability is assessed. This alternative method of Cronbach’s Alpha is preferred, because “the latter may under- or overestimate scale reliability and can create higher values when more indicators are included” (Garson 2016, p.64). Thus, the composite reliability seems to create a more realistic measure of reliability. Research suggests, “values above .8 are considered good for confirmatory research” (Daskalakis & Mantas 2008, p.288). In this study, all composite reliability indicates values higher than .8 and therefore are considered reliable. Table 1 represent the construct reliability and convergent validity of reflective scales and the before mentioned measurements.
Table 1: Construct reliability and convergent validity of reflective scales.
Constructs/Indicators: Loadings Outer Cronbach's alpha Composite reliability p-value
PACAP .841 .875 Acquire 1.1 0,586 .000 Acquire 1.3 0,509 .000 Acquire 1.4 0,686 .000 Acquire 1.5 0,671 .000 Assimilate 2.1 0,587 .000 Assimilate 2.2 0,742 .000 Assimilate 2.3 0,76 .000 Assimilate 2.4 0,758 .000 Assimilate 2.5 0,643 .000 CTR .827 .863 Contracting 14.1 0,686 .010 Contracting 14.2 0,798 .002 Contracting 14.3 0,798 .005 Contracting 14.4 0,729 .017 Contracting 14.5 0,595 .041 Contracting 14.6 0,674 .016
Kohli 1993, p.55). However, an additional test will indicate the consequences of deleting the two items from the original scale of environmental turbulence. These additional tests 7 indicate different results (appendix 6 and 7). Therefore, in this study although ET6.2 and ET6.4 show insignificant values they are necessary to capture the complete meaning of a turbulent environment. The second step in assessing the convergent validity of formative scales is the variance inflation factor (VIF). Research suggests that VIF values should be below 3.3 (Cenfetelli & Bassellier, 2009; Petter, Straub & Rai, 2007). In this study all VIF values are below the threshold of 3.3 and therefore suggest that there is no multicollinearity among the indicators. Table 2 represents the convergent validity of formative scales and the aforementioned measures.
Table 2: Convergent validity of formative scales.
Constructs/Indicators: Weights VIF value
p-EXPLORE EXLP8.1 .240 2.197 .000 EXLP8.2 .250 2.030 .000 EXLP8.3 .281 2.100 .000 EXLP8.4 .258 2.022 .000 EXLP8.5 .244 2.120 .000 ET ET 6.1 .541 1.388 .000 ET 6.2 .182 1.208 .123 ET 6.3 .245 1.089 .050 ET 6.4 .159 1.177 .228 ET 6.5 .412 1.257 .000 EXPLOIT EXPLOIT 9.1 .130 1.426 .000 EXPLOIT 9.2 .116 1.635 .000 EXPLOIT 9.3 .225 1.385 .000 EXPLOIT 9.4 .281 1.439 .000 EXPLOIT 9.5 .293 1.759 .000 EXPLOIT 9.6 .281 1.923 .000 EXPLOIT 9.7 .192 1.437 .000 Discriminant validity
The objective of the discriminant validity test is determining whether or not both a reflective and a formative construct have the strongest relationships with its own indicators (in comparison with any other construct) in the PLS path model (Hair et al., 2014). To assess discriminant validity, literature suggests using Fornell-lacker Criterion and a test for cross
loadings. However, recent research shows that these traditional approaches “do not reliably identify discriminant validity issues and lack the ability to detect discriminant validity in common research situations” (Henseler, Ringle & Sarstedt 2015, p.121). Heterotrait-Monotrait Ratio of correlations (HTMT) method is “superior to these two traditional approaches”(ibid.). Accordingly, “HTMT achieves high specificity and sensitivity rates across all simulations conditions”(ibid.). In addition, there are two different criteria to assess the discriminant validity with HTMT: the first is the ‘conservative’ HTMT.85 and the second is the ‘liberal’ HTMT.90 (ibid). To be as complete and robust as possible, this study performs all three measures of discriminant validity.
In table 3 the results of the first discriminant validity test are presented based on the Fornell-Lacker Criterion (first-order construct). The results indicate that the square root of the AVE is always higher than the correlations between the constructs (requirement), therefore it can be concluded that based on these results discriminant validity is established (Hair et al., 2014; Henseler et al., 2015).
Table 3: Fornell-Lacker Criterion
Square roots of the average variance extracted (AVE) and correlation matrix (first-order constructs).
1 2 3 4 5 6
1. Acquisition 0,726
2. Assimilation 0,645 0,739
3. Contracting 0,109 0,086 0,717
4. Environmental Turbulence 0,095 0,194 -0,065 0,682 5. Exploitative Learning Performance 0,354 0,428 0,055 0,288 0,656
6. Explorative Learning Performance 0,287 0,331 0,071 0,296 0,621 0,786
Table 4: Cross loadings.
CTR ET: Moderator ET EXPLOIT EXPLORE PACAP CTR14.1 0,686 -0,045 -0,138 0,009 0,017 0,100 CTR14.2 0,798 -0,078 -0,140 0,055 0,085 0,087 CTR14.3 0,798 -0,061 -0,002 0,039 0,090 0,082 CTR14.4 0,729 0,014 0,151 0,048 0,119 0,011 CTR14.5 0,595 -0,047 0,050 0,130 0,094 0,052 CTR14.6 0,674 0,009 0,152 0,108 0,175 0,046 ET6.1 -0,063 0,168 0,869 0,163 0,251 0,151 ET6.2 0,016 0,121 0,290 0,029 0,084 0,079 ET6.3 0,102 0,039 0,394 0,172 0,114 0,201 ET6.4 -0,058 0,083 0,255 0,147 0,074 0,046 ET6.5 -0,111 0,203 0,662 0,193 0,191 0,041 EXPLOIT9.1 0,164 0,068 0,014 0,400 0,223 0,197 EXPLOIT9.2 0,089 0,060 -0,041 0,537 0,199 0,340 EXPLOIT9.3 -0,020 0,063 0,193 0,624 0,387 0,288 EXPLOIT9.4 0,082 0,169 0,201 0,649 0,482 0,239 EXPLOIT9.5 0,002 0,093 0,172 0,781 0,503 0,347 EXPLOIT9.6 0,040 0,054 0,264 0,775 0,483 0,369 EXPLOIT9.7 0,080 0,011 0,122 0,609 0,329 0,506 EXPLORE8.1 0,125 0,111 0,248 0,467 0,811 0,214 EXPLORE8.2 0,080 0,186 0,206 0,482 0,790 0,257 EXPLORE8.3 0,030 0,100 0,257 0,529 0,817 0,312 EXPLORE8.4 0,045 0,157 0,253 0,484 0,748 0,257 EXPLORE8.5 0,175 0,162 0,164 0,470 0,761 0,260 PACAP 0,107 0,073 0,183 0,499 0,333 1,000 PACAP * ET -0,067 1,000 0,201 0,122 0,181 0,073
In table 5 the results of the third discriminant validity test with the HTMT method are presented. The results show that the upper confidence interval limit is below 0.85. The HTMT inference criteria indicate that all HTMT values are significantly different from one (Henseler et al., 2015). Therefore, discriminant validity is established.
Table 5: Discriminant Validity: HTMT.
To conclude, based on the before mentioned analyses the outer measurement model is adequately strong to test the hypotheses, and therefore the hypothesis will be tested based on the inner structural model.
5.2 Inner structural model (without tacitness)
The inner structural model is assessed for both supplier and buyer separately, because research suggest that there are differences among the ‘buyer, n=166” and ‘supplier, n=166’ dataset on most of the dimensions (Ambrose et al., 2010; Berger 2015, p.103). In table 6, the results of the paired sample T-test are presented. For contracting, exploitative learning performance and environmental turbulence the t-value indicate insignificant values. This suggests that equality is established for both buyer and supplier data on the aforementioned variables. However, explorative learning performance indicates a significant higher value. This suggests inequality and presumes that for the supplier data a higher value is established (resulting in a negative t-value). In addition the Levene test is included, this test shows that for PACAP unequal variances are also established. A higher variance is established for the buyer perspective. To conclude, unequal variances are established for two variables: PACAP and explorative learning performance. This indicates that buyers and suppliers have different perceptions on PACAP and explorative learning performance.
Table 6: Paired sample T-test, independent sample t-test (buyer and supplier).
Buyer Supplier t-value: p-value: Levene: p-value
Mean: Std. Dev: Mean: Std. Dev: (of differences) Sig. (2-tailed) Test for Equality of variances Sig. (2-tailed) Contracting 3,70 1,31 3,86 1,26 -1,07 0,29 0,04 0,84 PACAP 4,63 0,94 4,91 0,86 -2,80 ,005*** 2,72 0,10* Explorative 3,37 1,23 3,65 1,35 -1,95 ,052* 1,99 0,60 Exploitative 4,32 0,99 4,42 0,98 -0,92 0,36 0,09 0,76 ET 4,45 0,99 4,53 0,90 -0,79 0,43 0,03 0,86 Note: ***p ≤.01; ** p ≤. 05; * p≤.10. 5.2.1 Buyer data
significant (β =.296, p <.01) and therefore it can be concluded that hypothesis 3 is supported. Thirdly, The hypothesized positive relationship between explorative learning performance and exploitative learning performance is significant (β =.610, p <.01) and therefore it can be concluded that hypothesis 5 is supported. Lastly, The hypothesized positively moderated role of environmental turbulence on the relationship between PACAP and explorative learning performance is significant (β =.180, p <.10). Therefore it can be concluded that hypothesis 6 is supported as well.
Table 7: Hypotheses without considering ‘tacitness’ based on buyer data only.
Relation β (Beta) p-value Rejected/Supported
H1 Contracting - PACAP .167 .043** Rejected
H3 PACAP - Explorative Learning Performance .296 .000*** Supported
H5 Explorative Learning Performance - Exploitative Learning Performance .610 .000*** Supported
H6 PACAP - Explorative, moderated by Environmental Turbulence .180 .086* Supported Note: ***p ≤.01; ** p ≤. 05; * p≤.10.
5.2.2. Supplier data
In table 8 the hypotheses are tested without the role of tacitness taken into account (based on the supplier data only). Firstly, the hypothesized negative relationship between contracting and PACAP is not significant (β = -.169, p >.10) and therefore it can be concluded that hypothesis 1 is rejected. Second, the hypothesized positive relationship between PACAP and explorative learning performance is significant (β =.240, p= <.01) and therefore it can be concluded that hypothesis 3 is supported. Thirdly, the hypothesized positive relationship between explorative learning performance and exploitative learning performance is significant (β =.630, p= <.01) and therefore it can be concluded that hypothesis 5 is supported as well. Lastly, the hypothesized positively moderated role of environmental turbulence on the relationship between PACAP and explorative learning performance is not significant (β =.084, p >.10). Therefore it can be concluded that hypothesis 6 is rejected.
Table 8: Hypotheses without considering ‘tacitness’ based on supplier data.
Relation β (Beta) p-value Rejected/Supported
H1 Contracting - PACAP -.168 .433 Rejected
H3 PACAP - Explorative Learning Performance .240 .002*** Supported
H5 Explorative Learning Performance - Exploitative Learning Performance .631 .000*** Supported
5.3 Inner structural model (with tacitness)
In this study, as mentioned before, the PLS-MGA analysis allows to test if pre-defined data groups are significantly different (Garson, 2016). In addition, to be able to analyze different effects of tacitness, four data groups are made in Smart PLS 8. Four data groups are made, because firstly table 6 indicates that differences occur among buyer and supplier perceptions. In addition, other research also suggests that there are differences between buyer and supplier (Ambrose et al., 2010; Berger, 2015). Secondly there is a higher and lower level of tacitness according to the literature (Lane et al., 2001). The highest level of tacitness refers to: 8.4 (new marketing expertise) and 8.5 (new managerial expertise). This item was tested for on a 7-point likert scale; only the values between 4 and 7 will indicate a truly high level of tacitness, because higher scores infer “the relatively ease that organizations have with transferring and assimilating tacit knowledge, although these high levels are the most difficult to assimilate, transform and encode” (Lane et al. 2001, p.1158). These high values indicate that the relationship is capable of transforming highly tacit information into actual knowledge, indicating success at the highest and most difficult levels of tacitness (ibid.). The lowest level of tacitness refers to: 8.1 (new manufacturing and production expertise) and 8.2 (new product development expertise). This item was also tested for on a 7-point likert scale; only the values between 1 and 3 will indicate a truly low level of tacitness, because lower scores infer the difficulty of transferring and assimilating relatively low levels of tacit knowledge (ibid.). These low values indicate that the relationship is not capable of transforming low tacit information into actual knowledge. Research suggests that “the minimum sample should be equal to or larger than; (1) ten times the amount of indicators of the scale with the largest number of formative indicators (in this study: exploitative learning performance is indicated by seven items, 7 x 10 = 70), or ten times the largest number of structural paths directed to a dependent variable (in this study: potential absorptive capacity is indicated by two items, 2 x 10 = 20)” (Berger 2015, p.95; Chin 1998, p.311). Since this study consists of four separate data groups8 there will be ample respondents to perform the analysis.
To test for differences in segments, three different approaches are suggested: the Multi-Group Analysis, a Parametric Test and a Welch-Satterthwait Test (Henseler et al., 2009). For completeness and robustness purposes this study performs all three tests. The results are presented in table 9 and 10. The MGA analysis can be identified as “a non-parametric significance test where a p-value smaller than .05 or higher than .95 suggests significance” (Garson 2016, p.180; Henseler et al., 2009; Sarstedt, Henseler & Ringle, 2011). According to these researchers this method is often used to test for group differences. The second (=Parametric Test) approach can be compared with the first (= MGA) but differs on two important points: “it’s parametric and it assumes that groups have equal variances” (ibid.). The Welch-Satterthwait Test can be identified as an alternative parametric approach, but it differs from the Parametric Test, because: “it assumes unequal variances between groups” (ibid.). In addition, table 6 indicates unequal variances between buyer and supplier data on two dimensions: PACAP and explorative learning performance. Therefore in this study the Welch-Satterthwait Test will be used to determine the significant differences between the different groups. The Multi-Group Analysis and Parametric Test are used to determine the robustness of the results.
5.3.1 Buyer data
Table 9 indicates that for the buyer data there are no significant differences between Group A (highest level of tacitness) and Group B (lowest level of tacitness) based on the Welch-Satterthwait Test. In line with these findings, the other tests (robustness checks) show similar results. Therefore it can be concluded that hypothesis 2 and 4 are rejected for the buyer data.
Table 9: Standardized path coefficients, PLS-MGA, Parametric Test and Welch-Satterthwait Test.
PLS-MGA Parametric Test Welch-Satterthwait Test
Path Coefficients-of differences T: High – Low9 p-Value T: High vs Low Path Coefficients-of differences T: High - Low) p-Value T: High vs Low Path Coefficients-of differences T: High - Low) p-Value T:High-Low) Contracting - PACAP 0,202 0,247 0,202 0,458 0,202 0,412
PACAP - EXPLORE, mod: ET 0,214 0,785 0,214 0,423 0,214 0,423
EXPLORE - EXPLOIT 0,146 0,067 0,146 0,188 0,146 0,165
PACAP - EXPLORE 0,046 0,405 0,046 0,859 0,046 0,863
Note: *** p ≤.05 or ≥.95
5.3.2 Supplier data
Table 10 indicates that for the supplier data there are no significant differences between Group A (highest level of tacitness) and Group B (lowest level of tacitness) based on the Welch-Satterthwait Test. In line with these findings, the other tests (robustness checks) show similar results. Therefore it can be concluded that hypothesis 2 and 4 are rejected also for the supplier data.
Table 10: Standardized path coefficients, PLS-MGA, Parametric Test and Welch-Satterthwait Test.
PLS-MGA Parametric Test Welch-Satterthwait Test
Path Coefficients-of differences T: High – Low p-Value T: High vs Low Path Coefficients-of differences T: High – Low p-Value T: High vs Low Path Coefficients-of differences T: High – Low p-Value T:High-Low) Contracting - PACAP 0,130 0,608 0,130 0,652 0,130 0,630
PACAP - EXPLORE, mod: ET 0,070 0,352 0,070 0,706 0,070 0,704
EXPLORE - EXPLOIT 0,013 0,432 0,013 0,884 0,013 0,882
PACAP - EXPLORE 0,062 0,355 0,062 0,710 0,062 0,709
Note: *** p ≤.05 or ≥.95
Finally, based on the before mentioned analyses the results of all hypotheses are summarized in table 11. In addition, the insignificant values of the Welch-Satterthwait Test indicate that there are no significant differences between the different tacitness groups. Therefore it can be concluded that tacitness does not moderate; (1) the relationship between contracting and PACAP and (2) also does not moderate the relationship between PACAP and explorative learning performance (for both buyer and supplier).
Table 11 Overview of supported and rejected hypotheses.
Relation Buyer data Supplier data
H1 Contracting - PACAP Rejected* Rejected
H2 Contracting - PACAP, moderated by Tacitness Rejected Rejected
H3 PACAP - Explorative Learning Performance Supported Supported
H4 PACAP - Explorative, moderated by Tacitness Rejected Rejected
H5 Explorative Learning Perfomance - Exploitative Learning Performance Supported Supported
H6 PACAP - Explorative, moderated by Environmental Turbulence Supported Rejected
5.4 Robustness, correctness check
To be able to test the robustness of the outcomes, a Multi-Group Analysis and Parametric Test are included in table 9 and 10. In addition, two different robustness tests are included, based on the representativeness and operationalization of the different data-groups10. The first robustness test checks if the outcomes change if the ‘average’ level of tacitness is added (8.3) to the lowest level of tacitness (8.1, & 8.2). For the lowest level of tacitness the values 1-3 will still indicate a low score. For the highest level of tacitness (8.4 & 8.5) the values 4-7 will still indicate a higher score. The buyer data results are presented in appendix 2 and the supplier data results in appendix 3 (also included Multi-Group Analysis and Parametric Test, because of guaranteeing robustness). The second11 robustness test checks if the outcomes change if the highest level of tacitness (only 8.5) is compared with lowest tacitness level (only 8.1). The buyer data results are presented in appendix 4 and the supplier data results in appendix 5 (also included Multi-Group Analysis and Parametric Test, because of guaranteeing robustness). Based on the before mentioned robustness checks it can be concluded that the results are almost identical.
10 The first data group represents the ‘buyer data’ with the ‘highest level of tacitness’ (n=67), values between 4-7. The second data group represents the ‘buyer data’ with the ‘lowest level of tacitness’ (n=91), values between 1-3. The third data group represents the ‘supplier data’ with the ‘highest level of tacitness’ (n=92), values between 4-7. The last data group represents the ‘supplier data’ with the ‘lowest level of tacitness’ (n=81), values between 1-3. In addition, a multigroup analysis compares different groups and this influences (reduces) the amount of respondents per group, because of the conditions (high and low level of tacitness and the specified scores. The difference between the requirement (n >70) and actual observations in the first data group (n=67) however are minor and therefore are ample to perform the analysis. The difference between this analysis and the analysis in paragraph 5.3 is that in the lowest level of tacitness item 8.3 is included.
