DISTRIBUTION OF INTERFIRM LEARNING PERFORMANCES:

(1)

DISTRIBUTION OF INTERFIRM LEARNING

PERFORMANCES:

INFLUENCE OF POTENTIAL DRIVERS AND EFFECT ON ACTUAL PERFORMANCE

July, 2015

Koen Jansen

University of Groningen

Faculty of Economics and Business

Master Thesis

Winschoterdiep 35a, 9724 GJ Groningen, The Netherlands

+ 31 (0)6 15 50 14 81

k.w.jansen@student.rug.nl

S2583739

(2)

2

Summary

The main purpose of this study is to research the distribution of interfirm learning performances. On the one hand is it possible that the performance distribution is equally balanced (meaning that both parties are benefiting equally from the partnership), on the other hand is it possible that one of the parties (buyer or supplier) is able to benefit more than the other one. These learning performances are distinguished in both explorative and exploitative performances. Since this topic is not empirically researched in existing literature before, this is an interesting research gap. The first goal was to research the effect of a certain level of performance distribution on the actual interfirm learning performances. In other words, when are the actual learning performances (so the really achieved performances) higher? Is this the case when both parties are benefiting equally from the partnership or when of the parties (buyer or supplier) is able to benefit more than the other? Another research goal was to investigate the influence of several potential drivers of differences in performance distribution. Five different potential drivers, all based on existing literature, are researched: connectedness, market turbulence, idiosyncratic resources, relational norms, and customer diversity.

Regarding the first research goal, there can be concluded that the results are quite different compared to existing literature related to this topic and a few findings are quite surprising. The actual

exploitative performances are higher when the buyer benefits more, where it was expected that the performance score is higher when both parties are benefiting equally. The actual explorative learning performances are higher when both parties are able to benefit equally from the partnership, which was in line with the expectations.

The second goal was to research the influence of potential drivers on the performance distribution. There can be concluded that most levels of these potential drivers are leading to an equally balanced situation, so when both parties are benefiting equally from the partnership. The first researched potential driver is connectedness and all levels are leading to a situation whereby both parties are able to benefit equally from the partnership, for both the distribution of explorative and exploitative learning performances. Next, the data did not support the relationship between market turbulence and performance distribution. The third and fourth researched potential drivers are idiosyncratic resources and relational norms. Regarding the distribution of exploitative performances, there can be concluded that the highest levels of both drivers are leading to a situation whereby the buyer is able to benefit more. The data did not support the relationship between relational norms and the distribution of explorative learning performances. Next to that, all levels of idiosyncratic resources are leading to a situation whereby both parties are able to benefit equally, regarding the distribution of explorative learning performances. For customer diversity, all levels are leading to a situation whereby both parties are benefiting equally, regarding exploitative learning performances. The data did not support the relationship for the distribution of exploitative learning performances.

(3)

3

Table of content

Summary ... 2 1. Theoretical background ... 4 1.1 Introduction ... 4 1.2 Theoretical framework ... 7 1.3 Conceptual model ... 12 2. Methodology ... 14

2.1 Data collection procedure ... 14

2.2 Measures ... 14 2.3 Statistical procedure ... 15 3. Data analysis ... 17 3.1 Convergent validity ... 17 3.2 Descriptive statistics ... 19 3.3 Hypotheses testing ... 20 3.4 Robustness check ... 31 4. Discussion ... 33 4.1 Summary of findings ... 33 4.2 Interpretation of findings ... 34 4.3 Scientific implications ... 35 4.4 Managerial implications ... 36 4.5 Limitations... 36 4.6 Future research ... 36 5. References ... 38

Appendix 1. Scales and measures ... 41

Appendix 2. Numerical results per hypothesis ... 45

Appendix 3. Robustness check ... 56

(4)

4

1. Theoretical background

1.1 Introduction

According to Oosterhuis et al. (2013), most researchers in the field of buyer-supplier relationships agree that perceptions from both buyers and suppliers should be studied in order to gain insights into their relationships. They stated that it seems reasonable to question whether buyers and suppliers do indeed share the same perceptions of their relationships and they concluded that the few studies which researched buyer-supplier relationships are not offering conclusive answers. They investigated

perceptions from both perspectives on different relationship attributes such as communication frequency and media, technology and demand uncertainty, and found that perceptions from both parties of the relationship are different among these attributes.

This is confirmed by the study of Ambrose et al. (2010), which concluded that buyers and suppliers have significantly different perceptions of their relationship. Suppliers typically rate the relationship satisfaction higher than buyers, for example. Furthermore, buyers have greater expectations and less commitment than suppliers. They proposed that buyers and suppliers have significantly different perceptions of commitment, adaption communication, resource dependence, trust, uncertainty, power, and relationship success. So, this means that buyers and suppliers could have different perceptions from the relationship.

Both parties can also have different perceptions from the benefits which are created by the partnership, according to a conceptual paper of Lyons et al. (1990). They found that buyers, for example, are particularly benefiting from alliances in the field of reduction of manufacturing and labour costs, improved quality, reduction of complexity, supply assurance etc. Suppliers, on the other hand, are particularly benefiting from these alliances in the field of increased R&D effectiveness, information about competition, inside information on buying decisions etc. This means that buyers and suppliers are perceiving different benefits from interfirm relationships.

This is confirmed by the study of Sweeney and Webb (2002). The relational benefits to be gained by supplier organizations are for example cost efficiencies and enhanced revenues. They also stated that supplier benefits can include buyer loyalty and associated profitability, greater business stability and marketing efficiency, together with increased flow of information between buyers and suppliers. However, they also argued that buyers have the “upper hand” in the relationship, meaning that long-term relationships are usually initiated by the buyer. From the buyer’s viewpoint, relationships are mostly associated with increased performance, gaining a stronger competitive position, reliability of supply, improved delivery schedules, lower production costs, shortened product development time, improved product and service quality, and the ability to resolve conflicts satisfactorily. According to this research, the importance of relationship marketing in an increasing competitive environment is relatively unknown. The motivation for maintaining a relationship, however, depends on benefits gained by relationship partners. The authors concluded that it is not only logical but also necessary to address questions regarding the extent to which both parties are benefiting from interfirm

relationships.

They identified seven benefit categories: symbiotic (derived from a sense of sharing, togetherness, and common understanding), psychological (feelings of trust and confidence in the other party),

(5)

5 advantage, for example reputation, expertise, opportunities), and customization (customizing or tailoring the nature of the product).

This research of Sweeney and Webb (2002), provided a basis for understanding the balances of benefits gained by both suppliers and buyers in a dyadic relationship. Given that both parties are interested in maximizing the potential of relationship marketing in order to achieve organizational outcomes, it is important for each party to recognize the benefits of the relationship to the other party. In total, they concluded that buyers perceived more benefits than suppliers but this difference was not significant. Nevertheless, buyers and suppliers clearly perceived different types of relationship benefits. Firstly, suppliers perceived significantly more strategic benefits than buyers. Similarly, suppliers reported more economic benefits than buyers, although these differences were not significant. Buyers, in contrast, perceived significantly more psychological and symbiotic benefits. Customization benefits were almost exclusively perceived by buyers, although this category was minor in comparison to all others with the exception of social benefits.

While operational benefits were the main benefits derived for both buyers and suppliers, buyers viewed the ‘process’ benefits of psychological and symbiotic benefits as particularly relevant. This contrasted with suppliers who were more ‘outcome’ orientated, viewing economic and strategic benefits as more salient than process benefits. Economic benefits were of greater salience to suppliers. The authors suggested that from the buyer point-of-view, the relationship does not have a strong effect on the economic performance of the organization directly. Rather, the attainment of the process benefits ultimately leads to the achievement of organizational objectives which are related to financial outcomes.

So, they found that the two parties in the dyad benefited in different ways. Hence, while both parties desire to gain from a relationship, the specific received benefits are different between the two parties. Furthermore, they concluded that this topic is not studied directly in existing literature before and that additional research is needed.

That the performance distribution between buyers and suppliers in interfirm relationships can be different is confirmed by the study of Yao et al. (2007), in which the extent to which the benefits in a supply chain are equally distributed between buyers and suppliers is researched. This study

investigated which party is able to benefit more from different types of inventory management

systems. For example, they found that the benefits from ‘just-in-time practices’ are most likely to flow to buyers rather than to suppliers.

(6)

6 the buyer receives 60% and the supplier only 40% of the benefits from the partnership, illustrates a situation whereby the performance distribution is unequally balanced, because one of the parties is able to benefit considerably more than the other one.

Next to that, it could also be interesting to research when interfirm relationship performances are higher, for example when both parties are benefiting equally from the partnership or when one of the parties is able to benefit more than the other one. This is not empirically researched before. However, Bucklin and Sengupta (1993) argued that the effectiveness and success of interfirm relationships will reduce in case of imbalances in the relationship, implying that equally balanced relationships, so when both parties are able to benefit equally, are more successful.

Therefore, the main research questions of this study are:

- Are interfirm performances better when both companies are benefiting equally from the partnership or when one company is able to benefit more than the other one?

- Which potential drivers are influencing the distribution of the interfirm relationship success?

The potential drivers which will be used for the purpose of this research are concepts which could influence the distribution of interfirm relationship success. As aforementioned, according to Jap (2001), additional research on the distribution of interfirm relationship success should focus on when and how differences can occur, so when are both parties benefiting equally from the partnership (equally balanced distribution) and when is one of the parties (buyer or supplier) benefiting

considerably more than the other one? The second research question will address this research gap. The theoretical contribution from this study to existing literature is twofold. First, the effect of the distribution of interfirm relationship success on the actual interfirm relationship will be researched. Second, there will be investigated when the distribution of performance is equally or unequally balanced. In order to be able to determine this, different potential drivers of differences in the distribution of interfirm performances will be researched.

Existing literature (Jap, 2001, and Sweeney and Webb, 2002) provided some interesting statements on this topic and concluded that future research is needed. This study will focus on these gaps (see aforementioned research questions), since there is a lack of empirical research to this topic. Although some existing studies have noticed that it could be interesting to investigate the distribution of interfirm performances between buyers and suppliers, this is the first study which will empirically research this topic and the above mentioned research questions are not studied before.

Regarding the contribution of this research for managers in interfirm relationships, the findings will help to better understand when on the one hand buyers and on the other hand suppliers are able to benefit more. Next to that, there will be determined how to achieve higher interfirm performances. Finally, the scales and measures which will be used in this research can be used as practical guidelines for improving interfirm partnerships.

(7)

7

1.2 Theoretical framework

Interfirm relationship success

According to Selnes and Sallis (2003), the primary purpose of an interfirm relationship is to connect the buying activities of the customer with the selling activities of the supplier. They also state that companies can cooperate in other activities as well, such as R&D, marketing, quality control, and so forth. Furthermore, they concluded that the purpose of interfirm relationships is to enhance the effectiveness and efficiency of the relationship. So, a well-performing relationship exists if both parties are satisfied with the effectiveness and efficiency of the cooperation.

They defined the concept interfirm relationship performance based on indicators such as decreasing logistics costs, flexibility to handle unforeseen fluctuations in demand, better product quality,

synergies in marketing efforts, ability to develop successful new products, extent to which investments in the relationships paid off very well, and ability to detect changes in end-user needs before

competitors do.

Berger (2015) also mentioned these performance indicators in order to describe the exploitative success of interfirm relationships. This research distinguished interfirm relationship success in both explorative and exploitative learning performances. The performance indicators which are used for determining the explorative learning performance were about the extent to which the cooperation between both companies resulted in the joint creation of new expertise in the field of manufacturing and production, product development, technology, marketing, and managerial.

March (1991) explained the differences between explorative and exploitative learning and underlines the importance of both concepts for companies. The author concluded that both exploration and exploitation are essential for organizations and that organizational learning is about the exploration of new possibilities and exploitation of old certainties. Exploration includes things captured by terms such as search, variation, risk taking, experimentation, play, flexibility, discovery, and innovation. Exploitation includes such things as refinement, choice, production, efficiency, selection,

implementation, and execution. Effective selection among forms, routines, or practices is essential to survive, but so also is the generation of new alternative practices, particularly in a changing

environment. Maintaining an appropriate balance between exploration and exploitation is a primary factor in system survival, but they compete for scarce resources. As a result, organizations are making explicit and implicit choices between the two. The search for new ideas, markets, or relations

(explorative) has less certain outcomes, longer time horizons, and more diffuse effects than development of existing ones (exploitative). Because of these differences, adaptive processes characteristically improve exploitation more rapidly than exploration.

(8)

8 Bucklin and Sengupta (1993) argued that the effectiveness and success of interfirm relationships will reduce in case of imbalances in the relationship, implying that equally balanced relationships are more successful. They concluded that a possible explanation for this could be that these imbalances are leading to relational conflicts and unsatisfactory alliance performances. Furthermore, based on the study of Anderson and Narus (1990), there can be concluded that there is a negative relationship between unequally balanced interfirm relationships and alliance satisfaction, implying that these unequally balanced relationships are not desired.

Although above mentioned studies argued that balanced relationship are more effective and desired than unequally balanced relationships, there is no research which found empirical evidence for the statement that balanced relationships are leading to better actual interfirm learning performances. Therefore, it is interesting to determine if the actual interfirm relationship success (so the really achieved relationship success) is higher when the supplier or buyer benefits more from the cooperation or that the distribution of relationship success should indeed be equally distributed in order to achieve the best results, as expected in existing literature. Thus, the first hypothesis is:

H1a: An equally balanced distribution in the interfirm relationship performances, leads to better interfirm explorative learning performances, compared to situations whereby one of the parties is benefiting more than the other one.

H1b: An equally balanced distribution in the interfirm relationship performances, leads to better interfirm exploitative learning performances, compared to situations whereby one of the parties is benefiting more than the other one.

Drivers from differences in performance distribution

Although this topic is not empirically researched before, existing literature is used in order to find potential drivers of differences in the distribution of interfirm performances. Finally, two concepts are found: connectedness and market turbulence. These concepts are directly related to differences between buyers and suppliers in the distribution of interfirm performances in existing literature. Below, these drivers are described and hypotheses are developed based on related literature to this topic.

Connectedness and distribution of performances

According to Johnson and Sohi (2001) connectedness is about the strength of the interfirm relationship and the extent to which the two parties communicate and coordinate. Jansen et al. (2006) concluded that the higher is the level of connectedness, the higher are the interfirm explorative and exploitative learning performances. Next to that, interaction characterizes high levels of connectedness between firms. High levels of connectedness can be visualized as thick interfirm boundary spanning structures with strong healthy communication patterns. Importantly, along with frequent and intense

communication, connectedness entails high quality and open communication. It suggests easy, ready, and substantive communication in boundary spanning between firms. It also includes the extent to which the interface between interfirm relationships spans multiple levels. High levels of

connectedness means extensive, high quality communication via multiple means between multiple managers at multiple points and multiple levels of the firm’s managerial hierarchy.

(9)

9 contact among employees across departments. These kind of activities are highly related to

communicative capabilities. Furthermore, according to Ambrose et al. (2010) effective communication is desirable for both parties but particularly important for the supplier, because they concluded that knowing the buyer is the key challenge for suppliers and this requires effective communication. Therefore, it is expected that lower levels of connectedness are particularly harming the success of the relationship for the supplier and creates an unequally balanced performance distribution, whereby the buyer benefits more. So, the second hypothesis is:

H2a: Lower levels of connectedness are leading to an unequally balanced distribution of interfirm explorative relationship performances, whereby the buyer is able to benefit more. H2b: Lower levels of connectedness are leading to an unequally balanced distribution of interfirm exploitative relationship performances, whereby the buyer is able to benefit more. Market turbulence and distribution of performances

According to Santos-Vijande and Alvarez-Gonzalez (2007), market turbulence is about the changes in the composition of customers and their preferences, ongoing buyer entries and exits from the market place, characteristics of competitors, and the extent to which it is possible to predict the future of the market accurately. A basic requirement for relationship performance is the reduction of uncertainty for both parties (Morris and Carter, 2005). According to Chan et al. (2009), flexibility of suppliers is a tool to cope with environmental uncertainty.

Furthermore, according to Noorderwier et al. (1990) suppliers often are called upon to react to unforeseen and unforeseeable changes. This element defines the flexibility displayed by suppliers towards buyer-requested adjustments. Buyer requests for adjustments constitute opportunities for a supplier to display flexibility. Buyers are expecting flexibility in response to request for changes. According to Ambrose et al. (2010) uncertainty is significantly and negatively related to interfirm performances for both buyers and suppliers, however, this negative effect is slightly stronger in case of buyers. This means that suppliers are slightly better in dealing with uncertainty.

Nevertheless, the study of Buchanan (1992) argued that under conditions of market uncertainty, buyers have a stronger position than suppliers and that suppliers are a liability under conditions of high uncertainty, implying that buyers are better able to deal with situations of market uncertainty.

So, existing research is not clear about which company experiences the most negative effect from higher levels of uncertainty, but the above mentioned studies stated that market uncertainty/turbulence is related to differences in the distribution of interfirm performances.

(10)

10 Therefore, the third hypothesis is:

H3a: Higher levels of market turbulence are leading to an unequally balanced distribution of interfirm explorative relationship performances, whereby the buyer benefits more.

H3b: Higher levels of market turbulence are leading to an unequally balanced distribution of interfirm exploitative relationship performances, whereby the buyer benefits more.

As aforementioned, both connectedness and market turbulence are potential drivers of differences in the distribution of interfirm performances, according to existing literature. However, since this topic is not empirically researched before which means that there is not a lot of literature available, it is possible that there are also other concepts which could create differences in the distribution of interfirm performances, next to connectedness and market turbulence. Therefore, concepts which are, according to literature, related to these concepts will also be taken into account. This means that the following potential drivers are not directly related to the distribution of learning performances in existing literature, but could still be very interesting and are, especially regarding the exploratory nature of this study, worth to investigate.

So, the following potential drivers are selected because there are some overlaps between these concepts and the concepts of connectedness and market turbulence, which are mentioned as potential drivers of differences in distribution of interfirm performances in existing research, found in literature. Below, these drivers are described and hypotheses are developed based on related literature.

Idiosyncratic resources and distribution of performances

According to Jap (1999), idiosyncratic investments are non-fungible investments that uniquely support the buyer-supplier relationship. These investments can be tangible (e.g. a manufacturing facility) or intangible (e.g. tacit knowledge). Their non-fungible nature means that these investments are not easily transferable to other relationships. So, they lose their value when the relationship is terminated. Four different dimensions of idiosyncratic resources are mentioned in the paper of Berger (2015): site specificity, physical asset specificity, human asset specificity and dedicated assets, but primarily investments in human assets, which is one of the four asset specificity dimensions of idiosyncratic resources, seems to play an important role in interfirm relationships. According to Dyer (1996) human assets will increase when the partners are developing experience by working together and accumulate specialized information, language, and know-how that allows them to communicate efficiently and effectively. These advanced communicative capabilities are important since information sharing is necessary in relational learning (Selnes and Salis, 2003).

So, these human-related assets are highly related to an efficient and effective communication between both parties. This means that, although the concept idiosyncratic resources depends on more

dimensions than the human-related assets (see above mentioned, Berger, 2015), it could be possible that the level of idiosyncratic resources is also related to the distribution of interfirm performances. Therefore, it is interesting to research the effect of idiosyncratic resources as a potential driver of these differences in distribution. Ambrose et al. (2010) concluded that effective communication is desirable for both parties but particularly important for the supplier, as aforementioned (see hypotheses about connectedness). Therefore, it is expected that lower levels of idiosyncratic resources are particularly harming the success of the relationship for the suppliers and creates an unequally balanced

(11)

11 Thus, the fourth hypothesis is:

H4a: Lower levels of idiosyncratic resources are leading to an unequally balanced distribution of interfirm explorative relationship performances, whereby the buyer benefits more.

H4b: Lower levels of idiosyncratic resources are leading to an unequally balanced distribution of interfirm exploitative relationship performances, whereby the buyer benefits more.

Relational norms and distribution of performances

Norms can be defined as expectations about behavior that are at least partially shared by a group of decision makers (Heide and John, 1992). The study of Berger (2015), distinguished three types of relational norms: flexibility, information exchange, and solidarity. This study defined the norm type flexibility as the expectation that the alliance partners are willing to make adaptations to deal with changing circumstances, taking into account partner’s interests. The norm type information exchange reflects the expectation that the alliance partners will proactively provide information that is helpful for the partner. The norm type solidarity reflects the expectation that the alliance partners value the relationship and are committed to the relationship (Heide and John, 1992).

According to Berger (2015) relational norms foster trust and when trusts increases, communication between the parties will increase and are the parties more prepared to share information (Inkpen and Tsang, 2005). Furthermore, relational norms reduce the barriers to information sharing and encourage the interaction between alliance members (Berger, 2015). Thus, the higher the level of relational norms, the higher the level of communication between the two parties. Since Ambrose et al. (2010) concluded that communication is particularly important for the supplier, as earlier mentioned (see hypotheses about connectedness), it is expected that lower levels of relational norms are particularly harming the success of the relationship for the supplier and therefore creates an unequally balanced distribution, whereby the buyer benefits more. Thus, the fifth hypothesis is:

- H5a: Lower levels of relational norms are leading to an unequally balanced distribution of interfirm explorative relationship performances, whereby the buyer benefits more.

- H5b: Lower levels of relational norms are leading to an unequally balanced distribution of interfirm exploitative relationship performances, whereby the buyer benefits more. Customer diversity and distribution of performances

According to Basset-Jones (2005) the concept customer diversity encompasses a range of differences in ethnicity/nationality, gender, function, ability, language, religion and lifestyle.

Furthermore, according to Chowdhury and Miles (2006), organizational uncertainty depends on the ambiguity that arises from customer diversity, since a highly diverse customer base requires

organizational agents to continually adjust their approach to completing transactions. Organizations with highly diverse customers will therefore have difficulties in predicting activities to successfully complete their transactions, raising the level of organizational uncertainty.

(12)

12 So, because organizational uncertainty depends on customer diversity (Chowdhury and Miles, 2006) and Buchanan (1992) argued that suppliers are a liability under conditions of high uncertainty, it is expected that higher levels of customer diversity are leading to a situation whereby the buyer benefits more. Thus, the sixth hypothesis is1:

- H6a: Higher levels of customer diversity are leading to an unequally balanced distribution of interfirm explorative relationship performance, whereby the buyer benefits more.

- H6b: Higher levels of customer diversity are leading to an unequally balanced distribution of interfirm exploitative relationship performance, whereby the buyer benefits more.

1.3 Conceptual model

The conceptual model is shown in below figure. The five potential drivers of differences in the distribution relationship performances are shown on the left side. The variables in the middle are representing the distribution of relationship performances. These constructs are measuring the extent to which both parties are able to benefit equally from the partnership. A balanced distribution of relationship performances means that both parties are able to benefit equally from the partnership. So, when both buyers and suppliers are receiving approximately 50% of the benefits. An imbalanced distribution means that one of the parties is benefiting considerably more than the other, so when buyers are considerably benefiting more than suppliers, or the other way around. In this case, the performance distribution is unequally balanced. The variables on the right side are measuring the actual interfirm performances (the extent to which the partnership resulted in the creation/development of explorative and exploitative learning performances). So, the difference between the constructs in the middle and the constructs on the right side is that the constructs in the middle are measuring the extent to which the performance distribution is equally balanced and the constructs on the right side are measuring the really achieved learning performances.

1

(13)

13 The constructs about the distribution of learning performances will be used twice. Once as

independent variables (for H1a and H1b) for determining when the actual learning performances are higher, for example when the distribution is equally balanced or when one of the parties is able to benefit more than the other. These constructs will also be used as dependent variables (for H2a-H6b), to determine how the performances are distributed among the two parties under different market (market turbulence, and customer diversity) or relational (connectedness, idiosyncratic resources, and relational norms) conditions (the five drivers on the left side of the conceptual model).

(14)

14

2. Methodology

2.1 Data collection procedure

All data are obtained from the research of Berger (2015), whereby heads of purchasing or high-ranking technology managers were invited to participate and asked to supply contact data of four people within their companies who are central to their customer relationships. In order to analyze data from both perspectives of the relationship, data were obtained from informants of both sides of the dyad.

Since manufacturing firms are showing a higher probability to systematically acquire the external knowledge necessary to conceive new or improved products and or processes (Arbussà and Coenders, 2007), service firms were excluded from the study. To support variance, the choice of industries (viz. automotive, machinery, chemicals, pharmaceuticals, semiconductors and electronics) is based on prior research describing different knowledge strategies in these industries (Lichtenthaler and Ernst, 2007). In order to prevent selection bias, and following Johnson et al. (2004), four relationships were selected following a 2x2 design, with a relationship duration greater or less than two years on one dimension, and average or crucial importance of products or components on the other dimension. After the buyer informants were recruited, they were asked to supply the names of their contacts in the supplier organization. The researchers then contacted the identified informant in the supplier organization and asked them to participate in the study. Customer informants, all located in the Netherlands, were interviewed following a standardized questionnaire. Supplier informants located all over of the world, received the standardized questionnaire by (e-)mail. All responses were treated with absolute confidentiality; i.e. only company codes were included into the database in order to be able to match both sides of the dyad.

In order to check the involvement and knowledge of the respondents, their current working position, the number of years they have been involved in the relationship, the percentage of time they spend on the relationship, and the total sales volume in the previous year were asked. In total, 166 matched-pair relationships were obtained.

2.2 Measures

As aforementioned, the data are obtained from the research of Berger (2015), so all measures and scales are also based on this research. In this study, existing scales from previous research were used. Some were previously discussed in buyer-supplier relationship literature, while others were developed in the context of firm learning, intrafirm learning, or interfirm learning not related to buyer-supplier relationships (e.g., international joint ventures, mergers and acquisitions). These measures were extended to vertical interfirm relationships.

This research will use several variables and constructs from the study of Berger (2015): distribution of both interfirm explorative and exploitative performances, actual interfirm explorative and exploitative performances, connectedness, market turbulence, idiosyncratic resources, relational norms, and

(15)

15

2.3 Statistical procedure

The associated variables are combined into constructs, based on the convergent validity (see next chapter). Since all constructs are based on continuous data, a regression analysis is most useful. With help of a regression analysis there can be determined to which extent the relationship between the independent and dependent variables is significant, as well as the direction of this relationship (either positive, which means that the buyer benefits more, or negative, which means that the supplier benefits more), and the size of the effect. Therefore, this analysis will determine under which levels of

performance distribution, the actual performances are higher (hypotheses 1a and 1b) and how the performances are distributed over different levels of the aforementioned potential drivers (hypotheses 2a-6b), which is necessary in order to be able to answer the hypotheses.

A curve estimation analysis will be executed in order to conclude which specific type of regression analysis is most suitable for a specific hypothesis (e.g. linear, logarithmic, inverse, quadratic, cubic, power, compound, S-curve, logistic, growth, or exponential). With help of this estimation analysis, it is possible to determine for which regression type a significant relationship between the independent and dependent variables can be found. After that, the suitable type of regression analysis will be executed. When none of the regression types is appropriate (so when the significance levels of all types of regression analysis are too high), there is no significant relationship between the variables, which means that the hypothesis could not be supported and, therefore, will not be investigated. Based on the values which are found (e.g. significance level, constant, regression coefficient), it is possible to draw graphs which are showing the direction and size of the effect of the independent variable on the dependent variable. With help of these graphs, it is possible to determine under which levels of performance distribution, the actual performances are higher (hypotheses 1a and 1b) and how the performances are distributed over different levels of the aforementioned potential drivers

(hypotheses 2a-6b), which is necessary in order to be able to answer the hypotheses.

Cut-off points distribution of performances

In order to be able to answer the hypotheses, it is necessary to determine a range which is representing the equally balanced distribution situation (so when both parties are able to benefit equally from the partnership) and the situations whereby the buyer or supplier is able to benefit more. There is no study which empirically researched this topic before, so it is not possible to compare different statistical procedures. Therefore, a combination between existing literature and statistical analysis is used in order to determine the most reliable range.

As aforementioned, there is no study which researched this topic before, so it is not possible to compare different statistical procedures. But this is not the first research in which it is necessary to determine a range in order to represent a situation around the middle point of the scale (e.g. point 4 on a 7-point scale). From an extensive search to literature in which a certain range around the middle point is used, can be concluded that a range based on ± 1 standard deviation from the middle point is most commonly used (Laskowski, 1993, Pearson and Lipman, 1988, Wyse et al., 2002, Guinhouya et al., 2006, Eyberg and Ross, 1978, Waterlow et al., 1977, Amato et al., 2011, and Hänninen et al., 1996).

(16)

16 Therefore, a range of ± 1 standard deviation will be used for this research. Although the range is based on extensive analysis of literature in which a certain range around the middle point is determined, this choice is a bit arbitrary since it is also possible to select other ranges (e.g. ± 2 and ± 1,5 standard deviations, as aforementioned). Hence, a robustness check will be executed in order to check whether or not the results are comparable when other ranges would have been used.

The standard deviation for the construct about the distribution of explorative learning performances is 4. As aforementioned, a 7-point scale is used which means that the absolute middle point is exactly 4. There are five different variables used for the questions about explorative learning performances, so when the scores for all of these variables are summed up, the middle point is 20. This means that the range for the explorative learning performances is between 20 (middle point) ± 4 (one standard deviation), so between 16 and 24. Dividing these scores by five, since there are five different variables related to the explorative learning performances, leads to a range between 3,2 and 4,8 (on a 7-point scale) is representing the situation whereby both parties are able to benefit equally from the

partnership. A final score lower than 3,2, so a score below the earlier mentioned range, means that the supplier is benefiting more, and a score higher than 4,8, so a score above the earlier mentioned range, means that the buyer is benefiting more. This is because the questions in the questionnaire were based on a 7-point scale, whereby respondents had to select one of the boxes. The boxes on the left side (so the lowest scores) represented the situation whereby the supplier benefits more and the boxes on the right side (so the highest scores) the situation whereby the buyer benefits more, as shown in Appendix 1.

The standard deviation for the construct about the distribution of exploitative learning performances is 4,7. As aforementioned, a 7-point scale is used which means that the absolute middle point is exactly 4. There are seven different variables used for the questions about exploitative learning performances, so when the scores for all of these variables are summed up, the middle point is 28. This means that the range for the exploitative learning performances is 28 (middle point) ± 4,7 (one standard

(17)

17

3. Data analysis

3.1 Convergent validity

The study of Coltman et al. (2007), concluded that it is important to distinguish formative measures on the one hand and reflective measures on the other hand. Therefore, testing the convergent validity for both types of measures requires a different approach.

According to Diamantopoulos and Winklhofer (2001), the choice of a formative versus a reflective specification depends on the causal priority between the indicator and the latent variable. More specifically, reflective constructs are typically viewed as underlying factors that are ‘giving rise’ to something that is observed. On the other hand, when constructs are conceived as explanatory combinations of indicators that are determined by a combination of variables, their indicators are formative. Formative indicators are expected to measure different aspects of the same construct, therefore formative indicators of the same latent construct are not related. Next to that, in case of reflective scales the items are sharing a common theme and are interchangeable. On the other hand, in case of formative constructs, the indicators are not necessarily related to each other and adding or removing an indicator can change the conceptual domain of the construct (Coltman et al., 2007). All constructs and associated scales which are used for this research are shown in Appendix 1. There can be concluded that the constructs about customer diversity, market turbulence, and (distribution of) explorative and exploitative learning performances are based on variables which can be seen as quite divergent, which is a characteristic of formative scales, as aforementioned. For example, the constructs about customer diversity and market turbulence are based on variables which are not necessarily related to each other (e.g. on the one hand aggressiveness of competitors and the extent to which it is difficult to forecast where technology within the market will be in the next 2-3 years are part of the construct about market turbulence and on the other hand nationality and preferred product related features are part of the construct about customer diversity). Furthermore, the constructs about the (distribution of) learning performances, both explorative and exploitative, are based on variables which are quite divergent (e.g. on the one hand reduced logistic costs and flexibility to handle unforeseen fluctuations are variables which are describing exploitative learning performances and on the other hand manufacturing/production expertise and marketing expertise are part of the construct about explorative learning performances). Therefore, these constructs will be used as formative scales. The variables which will be used for the constructs about connectedness, idiosyncratic resources, and relational norms are highly related to each other, which is a characteristic of reflective scales, as aforementioned. For example, all variables about connectedness are about the extent to which employees of both companies are able to connect with each other (e.g. extent to which the people are accessible for each other, and the extent to which there are opportunities for informal ‘hall talk’ among employees). All variables about idiosyncratic resources are about the extent to which both companies jointly contributed to and invested in the relationship (e.g. jointly created knowledge, capabilities, and investments). Finally, all variables about relational norms are describing the relationship’s atmosphere (e.g. how the companies are dealing with certain situations, sharing information, etc.). Therefore, these constructs will be used as reflective scales.

(18)

18 Factor) in order to determine if the formative measures are too highly correlated. Traditionally, general statistics theory suggests that multicollinearity is a concern if the VIF-values are higher than 10; however, with formative measures multicollinearity poses more of a problem. Therefore, the threshold for formative measures is 3,3 (Petter et al., 2007). Table 1 shows that, for all formative constructs used in this research, the significance levels are lower than the threshold of 0,05, and all VIF-scores are lower than the threshold of 3,3.

Sig. VIF Customer diversity Variable 1 Variable 2 Variable 3 Variable 4 Variable 5 <0,001 <0,001 <0,001 <0,001 <0,001 2,453 2,345 1,508 1,523 1,049

Distribution exploitative performance

Variable 1 Variable 2 Variable 3 Variable 4 Variable 5 Variable 6 Variable 7 <0,001 <0,001 <0,001 <0,001 <0,001 <0,001 <0,001 1,424 1,416 1,279 1,156 1,479 1,448 1,296

Distribution explorative performance

Variable 1 Variable 2 Variable 3 Variable 4 Variable 5 <0,001 <0,001 <0,001 <0,001 <0,001 1,462 1,609 1,558 1,518 1,543

Exploitative learning performance

Variable 1 Variable 2 Variable 3 Variable 4 Variable 5 Variable 6 Variable 7 <0,001 <0,001 <0,001 <0,001 <0,001 <0,001 <0,001 1,430 1,679 1,382 1,439 1,698 1,691 1,456

Explorative learning performance

Variable 1 Variable 2 Variable 3 Variable 4 Variable 5 <0,001 <0,001 <0,001 <0,001 <0,001 2,095 1,941 1,996 1,882 1,963 Market turbulence Variable 1 Variable 2 Variable 3 Variable 4 Variable 5 <0,001 <0,001 <0,001 <0,001 <0,001 1,365 1,225 1,109 1,199 1,260 Table 1. Convergent validity formative scales

According to Henseler, Ringl, and Sinkovics (2009), reflective measurement models should be assessed with regard to their reliability and validity. Usually, the first measure which is checked is internal consistency reliability. The traditional measure for internal consistency is Cronbach’s Alpha, which provides an estimate for the reliability based on the indicator intercorrelations. While

(19)

19 composite reliability takes into account that indicators have different loadings, and can be interpreted in the same way as Cronbach’s Alpha. For both reliability coefficients, an internal consistency reliability value above 0,7 is satisfactory. As shown in table 2, the values for both Cronbach’s Alpha and Composite Reliability are higher than the threshold of 0,7.

Cronbach’s Alpha (PLS) Composite Reliability (PLS)

Connectedness 0,71 0,77

Idiosyncratic resources 0,9 0,91

Relational norms 0,87 0,88

Table 2. Convergent validity reflective scales

3.2 Descriptive statistics

In order to start with the data analysis, some descriptive statistics are presented. As shown in table 3, the number of observations (N) for each construct is higher than 300. Next to that, the minimum, maximum and mean score, the standard deviations, the number of items associated to the construct and the mean score on a 7-point scale are shown.

N Min Max Mean Standard

deviation

Number of variables

Mean 7-point scale

Exploitative learning performances 322 7 49 30,6 6,9 7 4,4

Explorative learning performances 329 5 32 17,5 6,5 5 3,5

Exploitative learning performances distribution 308 12 46 30,5 4,7 7 4,4

Explorative learning performances distribution 308 7 35 20,9 4 5 4,2

Relational norms 329 17 70 54,3 8,2 10 5,4

Market turbulence 323 10 34 22,4 4,7 5 4,5

Idiosyncratic resources 330 6 42 28,3 7,5 6 4,7

Customer diversity 330 5 35 21,8 6,2 5 4,4

Connectedness 326 7 35 25,3 5,5 5 5,1

Table 3. Descriptive statistics

Furthermore, it is interesting to determine to which extent the data between on the one hand the supplier and on the other hand the buyer are different, so an Independent Sample T-Test is executed. The results are shown in table 4.

Levene’s Test Sig. Mean difference

(supplier – buyer)

Exploitative learning performances 0,761 0,357 0,71

Explorative learning performances 0,160 0,052 1,38

Exploitative learning performances distribution 0,255 0,204 0,49

Explorative learning performances distribution 0,230 0,291 -0,69

Relational norms 0,101 0,009 2,36

Market turbulence 0,858 0,428 0,42

Idiosyncratic resources 0,003 0,000 3,13

Customer diversity 0,049 0,769 -0,2

Connectedness 0,024 0,024 1,37

Table 4. Results Independent Sample T-Test

The Levene’s Test determines whether or not the variances between the two groups (buyers and suppliers) are equal. If the result of the Levene’s Test is lower than 0,05, it is assumed that the variances between the two groups are not equal. If the significance level is lower than 0,05, the differences between the two groups are significant.

(20)

20 significantly higher than the buyer, because the mean differences are higher than zero which means that the supplier scores are higher. For the other variables, the differences between suppliers and buyers are not significant.

3.3 Hypotheses testing

First, for all hypotheses is estimated, with help of the Curve Estimation analysis, which type of regression analysis is suitable in order to describe the effect between the independent and dependent variable. Finally, a linear relationship is found for six hypotheses, and a quadratic relationship is found for two hypotheses. The other hypotheses could not be supported, meaning that there is no statistical significant relationship between the variables. In tables 5 and 6 is shown for which hypotheses which type of relationship is found, the linear relationships are shown in table 5 and the quadratic

relationships in table 6.

Hypothesis Significance level

H1b: Performance distribution x actual performance score (exploitative) 0,000 H2a: Connectedness x Performance distribution (explorative) 0,005 H2b: Connectedness x Performance distribution (exploitative) 0,008 H4a: Idiosyncratic resources x Performance distribution (explorative) 0,004 H4b: Idiosyncratic resources x Performance distribution (exploitative) 0,000 H5b: Relational norms x Performance distribution (exploitative) 0,007

Table 5. Significance levels linear relationships

Hypothesis Significance level

H1a: Performance distribution x actual performance score (explorative) 0,000 H6a: Customer diversity x Performance distribution (explorative) 0,026

Table 6. Significance level quadratic relationships

The basic formula for linear relationships is Y = Constant + β1x and the basic formula for quadratic relationships is Y = Constant + β1x + β2 x*x.

The other hypotheses could not be supported, because the significance levels (max. = 0,05) for all of the regression types are too high. All significance levels for the hypotheses which could not be supported are shown in Appendix 2.7. Since the significance levels for each type of regression analysis are too high, none of the types is suitable. As aforementioned, for the eight other hypotheses the significance level was sufficient for a regression analysis. The results of all of these hypotheses are shown below.

Effect performance distribution on actual performance

Explorative learning performances (hypothesis 1a)

As shown in table 6 (see above), a significant quadratic relationship is found (sig = 0,000). Based on the constant and the regression coefficients, the following quadratic formula is determined: Y = -2,480 + 1,914 * x - 0,043* x*x. The adjusted R² value is 0,967, which means that 96,7% of the variation in the actual performance level can be explained by the performance distribution level, which is quite high. This quadratic function is shown in below graph. In case of quadratic functions the average levels, rather than the lowest or highest ones, of the independent variable are leading to higher levels of the dependent variable. So, this means that in case of the average levels of performance distribution (so when both parties are able to benefit equally from the partnership), the actual learning

(21)

21 As explained earlier, for explorative learning performances a range between a score of 3,2 and 4,8 (+ and – one standard deviation) on a 7-point scale is most reliable in order to represent the situation where both parties are able to benefit equally and will be used for this research. There are five different variables related to the explorative learning performances, so one can multiply 3,2 and 4,8 with five in order to determine for which levels the performance distribution is equally balanced (so when both parties are able to benefit equally from the partnership). The results are shown below, the area between the orange lines is representing the situation whereby both parties are able to benefit equally from the cooperation, the area on the left side the situation whereby the supplier benefits more, and the area on the right side the situation whereby the buyer benefits more.

Figure 2. Result hypothesis 1a

One can interpret the quadratic function in the graph in order to draw conclusions. There can be concluded that in case of the situation whereby both parties are able to benefit equally from the cooperation, the actual performance scores are higher. The levels of performance distribution between the orange lines are clearly leading to higher actual performance scores than the levels on the left and right side. Therefore, hypothesis 1a is supported. The numerical results, whereby the actual

performance scores are calculated for each level of performance distribution, are shown in Appendix 2.1a.

Exploitative learning performances (hypothesis 1b)

(22)

22 As explained earlier, for exploitative learning performances a range between a score of 3,4 and 4,6 (+ and – one standard deviation) on a 7-point scale is most reliable in order to represent the situation where both parties are able to benefit equally and will be used for this research. There are seven different variables related to the exploitative learning performances, so one can multiply 3,4 and 4,6 with seven in order to determine for which levels the performance distribution is equally balanced (so when both parties are able to benefit equally from the partnership). The results are shown below, the area between the orange lines is representing the situation whereby both parties are able to benefit equally from the cooperation, the area on the left side the situation whereby the supplier benefits more, and the area on the right side the situation whereby the buyer benefits more.

Figure 3. Result hypothesis 1b

(23)

23

Effect potential drivers on performance distribution (H2a-H6b) Effect connectedness on performance distribution

As shown in table 5 (see above), a significant linear relationship is found (sig = 0,005). This means that the level of connectedness does significantly influence the distribution of explorative learning performances. Based on the constant and the regression coefficient, the following linear formula is determined: Y = 17,231 + 0,145 * x. The adjusted R² value is 0,022, which means that 2,2% of the variation in the performance distribution level can be explained by the level of connectedness, which is quite low. This linear function is shown in below graph. The regression coefficient is positive (0,145), so the higher the levels of connectedness, the more the line will shift towards a situation whereby the buyer benefits more.

As explained earlier, for explorative learning performances a range between a score of 3,2 and 4,8 (+ and – one standard deviation) on a 7-point scale is most reliable in order to represent the situation where both parties are able to benefit equally and will be used for this research. There are five different variables related to the explorative learning performances, so one can multiply 3,2 and 4,8 with five in order to determine for which levels the performance distribution is equally balanced (so when both parties are able to benefit equally from the partnership). The results are shown below, the area between the orange lines is representing the situation whereby both parties are able to benefit equally from the cooperation, the area on the left side the situation whereby the supplier benefits more, and the area on the right side the situation whereby the buyer benefits more.

(24)

24 regression coefficient is very low (0,145) which means that the line is not steep enough to create an unequally balanced distribution. Therefore, all levels of connectedness are leading to a situation whereby both parties are able to benefit equally from the partnership. It was expected that lower levels of connectedness will create an imbalance in performance distribution, but this cannot be confirmed with help of this research. Therefore, hypothesis 2a is rejected. The numerical results, whereby the distribution levels are calculated for each level of connectedness, are shown in Appendix 2.2a.

As shown in table 5 (see above), a significant linear relationship is found (sig = 0,008). This means that the level of connectedness does significantly influence the distribution of exploitative learning performances. Based on the constant and the regression coefficient, the following linear formula is determined: Y = 27,248 + 0,133 * x. The adjusted R² value is 0,015, which means that 1,5% of the variation in the performance distribution level can be explained by the level of connectedness, which is quite low. This linear function is shown in below graph. The regression coefficient is positive (0,133), so the higher the levels of connectedness, the more the line will shift towards a situation whereby the buyer benefits more.

As explained earlier, for exploitative learning performances a range between a score of 3,4 and 4,6 (+ and – one standard deviation) on a 7-point scale is most reliable in order to represent the situation where both parties are able to benefit equally and will be used for this research. There are seven different variables related to the exploitative learning performances, so one can multiply 3,4 and 4,6 with seven in order to determine for which levels the performance distribution is equally balanced (so when both parties are able to benefit equally from the partnership). The results are shown below, the area between the orange lines is representing the situation whereby both parties are able to benefit equally from the cooperation, the area on the left side the situation whereby the supplier benefits more, and the area on the right side the situation whereby the buyer benefits more.

(25)

25 One can interpret the linear function in the graph in order to draw conclusions. There can be concluded that all levels of connectedness are leading to a situation whereby both parties are able to benefit equally from the partnership. Higher levels are leading to higher levels of performance distribution (and the higher the level of distribution, the more the distribution is beneficial for the buyer), but the regression coefficient is very low (0,133) which means that the line is not steep enough to create an unequally balanced distribution. Therefore, all levels of connectedness are leading to a situation whereby both parties are able to benefit equally from the partnership. It was expected that lower levels of connectedness will create an imbalance in performance distribution, but this cannot be confirmed with help of this research. Therefore, hypothesis 2b is rejected. The numerical results, whereby the distribution levels are calculated for each level of connectedness, are shown in Appendix 2.2b.

Effect market turbulence on performance distribution

As shown in Appendix 2.7, none of the possible regression types (both linear and non-linear) found a significant relationship for this hypothesis. The significance levels for all regression types are higher than the threshold of 0,05, so none of the types is suitable for answering this hypothesis. This means that the level of market turbulence does not significantly influence the distribution of explorative learning performances. Therefore, hypothesis 3a could not be supported, where it was expected that higher levels of market turbulence are leading to a situation whereby the buyer benefits more.

As shown in Appendix 2.7, none of the possible regression types (both linear and non-linear) found a significant relationship for this hypothesis. The significance levels for all regression types are higher than the threshold of 0,05, so none of the types is suitable for answering this hypothesis. This means that the level of market turbulence does not significantly influence the distribution of exploitative learning performances. Therefore, hypothesis 3b could not be supported, where it was expected that higher levels of market turbulence are leading to a situation whereby the buyer benefits more.

Effect idiosyncratic resources on performance distribution

As shown in table 5 (see above), a significant linear relationship is found (sig = 0,004). This means that the level of idiosyncratic resources does significantly influence the distribution of explorative learning performances. Based on the constant and the regression coefficient, the following linear formula is determined: Y = 18,358 + 0,09 * x. The adjusted R² value is 0,024, which means that 2,4% of the variation in the performance distribution level can be explained by the level of idiosyncratic resources, which is quite low. This linear function is shown in below graph. The regression coefficient is positive (0,09), so the higher the levels of idiosyncratic resources, the more the line will shift towards a situation whereby the buyer benefits more.

(26)

26 equally from the cooperation, the area on the left side the situation whereby the supplier benefits more, and the area on the right side the situation whereby the buyer benefits more.

One can interpret the linear function in the graph in order to draw conclusions. There can be concluded that all levels of idiosyncratic resources are leading to a situation whereby both parties are able to benefit equally from the partnership. Higher levels of idiosyncratic resources are leading to higher levels of performance distribution (and the higher the level of distribution, the more the distribution is beneficial for the buyer), but the regression coefficient is very low (0,09) which means that the line is not steep enough to create an unequally balanced distribution. Therefore, all levels of idiosyncratic resources are leading to a situation whereby both parties are able to benefit equally from the partnership. It was expected that lower levels of idiosyncratic resources will create an imbalance in performance distribution, but this cannot be confirmed with help of this research. So, hypothesis 4a is rejected. The numerical results, whereby the distribution levels are calculated for each level of idiosyncratic resources, are shown in Appendix 2.4a.

(27)

27 As explained earlier, for exploitative learning performances a range between a score of 3,4 and 4,6 (+ and – one standard deviation) on a 7-point scale is most reliable in order to represent the situation where both parties are able to benefit equally and will be used for this research. There are seven different variables related to the exploitative learning performances, so one can multiply 3,4 and 4,6 with seven in order to determine for which levels the performance distribution is equally balanced (so when both parties are able to benefit equally from the partnership). The results are shown below, the area between the orange lines is representing the situation whereby both parties are able to benefit equally from the cooperation, the area on the left side the situation whereby the supplier benefits more, and the area on the right side the situation whereby the buyer benefits more.

Figure 7. Result hypothesis 4b

One can interpret the linear function in the graph in order to draw conclusions. There can be concluded that higher levels of idiosyncratic resources are leading to a situation whereby the buyer is able to benefit more from the partnership, in contrast to the previous hypothesis, where all levels of idiosyncratic resources are leading to a situation whereby the performance distribution is equally balanced. This is because the regression coefficient of hypothesis 4b is higher (0,156) than the coefficient of hypothesis 4a (0,09). Due to this, in case of the higher levels of idiosyncratic resources, the distribution level is higher than the threshold of 4,6 which means that the buyer is able to benefit more. It was expected that lower levels of idiosyncratic resources will create an imbalance in performance distribution, but this cannot be confirmed with help of this research. Therefore,

(28)

28

Effect relational norms on performance distribution

As shown in Appendix 2.7, none of the possible regression types (both linear and non-linear) found a significant relationship for this hypothesis. The significance levels for all regression types are higher than the threshold of 0,05, so none of the types is suitable for answering this hypothesis. This means that the level of relational norms does not significantly influence the distribution of explorative learning performances. Therefore, hypothesis 5a could not be supported, where it was expected that lower levels of relational norms are leading to a situation whereby the buyer benefits more.

As shown in table 5 (see above), a significant linear relationship is found (sig = 0,007). This means that the level of relational norms does significantly influence the distribution of exploitative learning performances. Based on the constant and the regression coefficient, the following linear formula is determined: Y = 25,704 + 0,089 * x. The adjusted R² value is 0,021, which means that 2,1% of the variation in the performance distribution level can be explained by the level of relational norms, which is quite low. This linear function is shown in below graph. The regression coefficient is positive (0,089), so the higher the levels of relational norms, the more the line will shift towards a situation whereby the buyer benefits more.

As explained earlier, for exploitative learning performances a range between a score of 3,4 and 4,6 (+ and – one standard deviation) on a 7-point scale is most reliable in order to represent the situation where both parties are able to benefit equally and will be used for this research. There are seven different variables related to the exploitative learning performances, so one can multiply 3,4 and 4,6 with seven in order to determine for which levels the performance distribution is equally balanced (so when both parties are able to benefit equally from the partnership). The results are shown below, the area between the orange lines is representing the situation whereby both parties are able to benefit equally from the cooperation, the area on the left side the situation whereby the supplier benefits more, and the area on the right side the situation whereby the buyer benefits more.

(29)

29 One can interpret the linear function in the graph in order to draw conclusions. There can be concluded that almost all levels of relational norms are leading to a situation whereby both parties are able to benefit from the partnership. Nevertheless, the highest levels are leading to a situation whereby the buyer benefits more, since in case of the highest levels the performance distribution level is slightly above the threshold of 4,6. It was expected that lower levels of relational norms will create an imbalance in performance distribution, but this cannot be confirmed with help of this research. Therefore, hypothesis 5b is rejected. The numerical results, whereby the distribution levels are calculated for each level of relational norms, are shown in Appendix 2.5b.

Effect customer diversity on performance distribution

As shown in table 6 (see above), a significant quadratic relationship is found (sig = 0,026). This means that the level of customer diversity does significantly influence the distribution of explorative learning performances. Based on the constant and the regression coefficients, the following quadratic formula is determined: Y = 18,930 + 0,3 * x - 0,009 * x * x. The adjusted R² value is 0,018, which means that 1,8% of the variation in the performance distribution level can be explained by the level of customer diversity, which is quite low. This function is shown in below graph. In case of quadratic functions the average levels, rather than the lowest or highest ones, of the independent variable are leading to higher levels of the dependent variable. So, this means that in case of the average levels of customer

diversity, the level of performance distribution will be higher compared to the lowest and highest levels (and the higher levels of performance distribution are representing the situation whereby the buyer benefits more). Whether or not these average levels of customer diversity are indeed leading to situation whereby the buyer benefits more depends on the size of the effect. The results can be determined with help of below graph.

(30)

30

One can interpret the quadratic function in the graph in order to draw conclusions. There can be concluded that all levels of customer diversity are leading to a situation whereby both parties are able to benefit equally from the partnership. Average levels are leading to higher levels of performance distribution than the lowest and highest levels (and the higher the level of distribution, the more the distribution is beneficial for the buyer), but this increase is not strong enough to create an unequally balanced distribution. Therefore, all levels of customer diversity are leading to a situation whereby both parties are able to benefit equally from the partnership. It was expected that higher levels of customer diversity will create an imbalance in performance distribution whereby the supplier benefits more, but this cannot be confirmed with help of this research. Therefore, hypothesis 6a is rejected. The numerical results, whereby the distribution levels are calculated for each level of customer diversity, are shown in Appendix 2.6a.