The effect of affective state on knowledge transfer in an organizational social network

(1)

A.H. (Anneleen) Overbeek – 11130083

MSc. In Business Administration – Strategy track

Amsterdam Business School - University of Amsterdam

Supervisor: dr. N.E. (Nathan) Betancourt

Date of submission final version: 24th_{of June 2016}

(2)

2 Statement of Originality

This document is written by Anneleen Overbeek who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Abstract

The present research determines when the affective states, positive and negative, influence the relationship between the strength of ties and tacitness of knowledge received. Previous

research showed that affective state influences the way people interact, but it was not

implemented in research concerning knowledge transfer before. It was expected to have most effect on the dimension of tacitness of knowledge, as this determines if the knowledge is transferred in face-to-face interaction or not. A survey was send out to employees in four software-departments of a company. In contrast to what was expected the relationship between strength of ties and tacitness of knowledge received was negative. Average to high levels of positive affect and low to average levels of negative affect moderated the

relationship between strength of ties so that this relationship was strengthened and became more negative. However, the relationships that were found explained only half of the results as the scale of tacitness of knowledge received was not internally consistent (Cronbach’s Alpha=0.52). Suggestions for further research are given.

(6)

6

1.Introduction

Theory and research in strategic management have evolved dramatically over the years (Hoskisson, Hitt, Wan, & Yiu, 1999). Hoskisson et al. (1999) argue that the field has swung like a pendulum when looking for ways to explain competitive advantage; starting with theory and research focusing on the inside of the organisation in the 60s, later looking at the outside of the organisation and ending up looking at the inside again from the 90s onward. In recent years looking at the inside of the organisation has focused on the micro-level, the level of the actor, for explanations of competitive advantage (Foss, 2011; Ployhart & Hale, 2014). How actors interact and share knowledge can give a starting point for determining the

microfoundations of competitive advantage.

As strategic management is a practical field of research in search of explanations for firm performance, it is likely to benefit from using a wide variety of theoretical perspectives and methodologies (Hoskisson et al., 1999). One theoretical perspective that has gained increasing importance is that of the social network (S. P. Borgatti, Mehra, Brass, & Labianca, 2009; S. Borgatti, 2003; Burt, Kilduff, & Tasselli, 2013; Wasserman & Faust, 1994). The view of the organisation as a social network has shown to be able to answer standard social and

behavioural questions (Wasserman & Faust, 1994). In the field of Economics it has shown to be able to contribute to decision-making theory and economic behaviour in general (Jackson, 2009). For example, it can shed light on the influence of others on an individual that is making a decision.

The social network perspective sees employees as actors (or nodes). The actors are connected to each other with ties, which can be used for the transfer or flow of resources (Wasserman & Faust, 1994). The network is the set of actors connected through ties (S. Borgatti, 2003). A lot

(7)

7

of factors influence these ties and the resources, knowledge in specific, which flow through them.

One of the most influential studies on knowledge flowing through ties was conducted by Granovetter who made a distinction between strong and weak ties and the information passing through the ties (S. P. Borgatti et al., 2009; Granovetter, 1973). Since then research has tried to specify the conditions and influencers on this knowledge transfer, looking at interpersonal trust, willingness to incur costs for transfer and types of knowledge that were transferred (Phelps, Heidl, & Wadhwa, 2012). It was consistently found that strong ties are better at facilitating knowledge transfer than weak ties, especially when this knowledge is tacit, but results for knowledge creation have been contradictory (Phelps et al., 2012). This means that strong ties may facilitate knowledge transfer better, but that does not mean the receiver will use this knowledge to create more knowledge for himself. While Phelps and colleagues (2012) call for more research into the benefits and costs of strong and weak ties, the field of psychology has already found a possible explanation on the level of the individual actor. The field of psychology has developed a lot of research concerning the individual actor and his willingness to share or receive information. Fredrickson (2001) found that the affective state of the actor has an influence on the actor’s ability and will to transfer knowledge. When an actor receiving knowledge is in a positive affective state, he will be open to exploration and he will build personal (knowledge) resources. Another study confirmed that actors in a positive affective state are more likely to absorb and act on new knowledge (Levin,

Kurtzberg, Phillips, & Lount, 2010). Both studies focused on face-to-face conversations, which is what tacit knowledge requires. To make a distinction between the sharing of tacit

(8)

8

and codified knowledge when researching the influence of affective state on knowledge transfer will therefore be especially interesting.

In the present research the influence of affective state on knowledge transfer is measured. This will be done using a social network perspective. Both strength of ties and the affective state have shown to have an impact on the transfer of knowledge, tacit or codified. However, the two have not been combined in one research up until this point. The present research will combine the findings of the effect of the strength of ties on knowledge transfer, tacit or codified, with the findings of the effect of the affective state of the actor on knowledge transfer and answer the following the research question by reviewing data that was collected via a survey.

Research question: When does affective state influence knowledge-sharing via ties?

By answering the research question the present research will contribute to existing literature by giving more insight on the microfoundations of competitive advantage. We will know if the relationship between strength of ties and sharing tacit versus codified knowledge is influenced by the actor’s individual affective state, and if the affective state offers an

explanation for the contradictory results of knowledge creating. Additionally we will know if it is useful to look at the individual actors’ dispositions when researching knowledge transfer. To answer the research question a literature review will be given first, which generates three hypotheses. Then the study that was conducted will be elaborated upon in the research design chapter. The results from this study will be presented and discussed and lead to a conclusion regarding theoretical and practical implications.

(9)

9

2.Literature review

This literature review will explicate the theory of the strength of ties of social networks, after which the knowledge transfer via these ties will be discussed. Then the literature on affective state of people and its influence on knowledge transfer and interaction in general will be elaborated upon and implemented in the theoretical model.

2.1.Strength of Ties

According to Granovetter (1973) ties between actors can be strong, weak or absent. He defines the strength of a tie as “a combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the

tie”(Granovetter, 1973, p.1361). With this definition he identifies four dimensions of the strength of a tie. He argues that when these dimensions become higher (simultaneously, as is usually the case), the tie becomes stronger. In order to determine the strength of a tie

respondents are asked to tell how much time they spend with a certain actor, how much emotion goes into the bond with this actor, how intimate the bond is and how reciprocal the bond is in terms of actual services. These dimensions have been criticized to be subjective, excluding the dimension concerning time (Krackhardt, 1992), because time is the only dimension that can be accurately measured while the rest is an interpretation of the actor. Therefore a slightly different set of dimensions has been introduced by Krackhardt consisting of interaction, affection, and time (1992). These dimensions consider the frequency of

interaction between two people, the affection they have for one another, and how long they know each other. Affection, the only seemingly subjective dimension, can be determined by both actors, leaving this to the interpretations of the two actors involved. Together these dimensions determine the strength of a tie.

(10)

10

The stronger the tie, the more similar the actors (Granovetter, 1973). Other researchers call this homophily, which literally means ‘love of the same’. More figuratively speaking it is the assumption that actors tend to bond with actors that are similar to themselves and have the same characteristics (Jackson, 2009). This works the other way around as well. Actors with weak ties to one another are expected to differ more. When actors differ more from each other, it can be assumed that their knowledge differs more as well. Knowledge flowing through weak ties is therefore more novel than knowledge flowing through strong ties (Granovetter, 1973; Levin & Cross, 2004; Perry-Smith, 2014).

As two out of the three dimensions do not entail affection of some sort, a tie can be rather strong without the actors actually liking each other. Take for example Jenny and Ben who work in the same department for a couple of years now. Due to their specified tasks they have to interact a lot. Ben and Jenny have known each other for a long time and interact a lot. Assuming that they do not feel much affection for each other, their tie will still be considered rather strong due to their high scores on the other two dimensions. Therefore a relatively strong tie does not necessarily mean that two actors like each other.

The assumption that the strength of a tie can determine how the knowledge transfer takes place has led numerous researchers to investigate the specifics of this relationship and what it means for organizations. The strength of ties has been linked to the type of knowledge

transferred (Hansen, 1999; Levin & Cross, 2004), creativity (Baer, 2010; Perry-Smith, 2006; Zhou, Shin, Brass, Choi, & Zhang, 2009), individual and group performance (Sparrowe, Liden, Wayne, & Kraimer, 2001) and more. The present research focuses on knowledge transfer via ties.

(11)

11 2.2.Knowledge Transfer via Ties

In the last two decades research on knowledge flowing through ties has gained attention. A fundamental finding is that the flow of knowledge through strong ties is easier than through weak ties (Krackhardt, 1992; Perry-Smith, 2014; Reagans & McEvily, 2003). Actors who spend more time together, have more affection for one another and know each other for a longer time, are expected to feel more comfortable sharing knowledge with each other. In addition to that they are more likely to share certain knowledge which allows them to communicate easily. However, the relationship between the strength of ties and knowledge transfer is moderated by certain factors, the type of knowledge in specific.

Zander and Kogut were one of the first to give knowledge dimensions. They identified five dimensions, namely codifiability, teachability, complexity, system dependence and product observability (Zander & Kogut, 1995). The level of codifiability determines if the knowledge can be encoded or written down, even if one does not understand the knowledge. The level of teachability determines how much of the knowledge can be trained. The level of complexity determines how difficult it is to understand the knowledge. The level of system dependence determines the extent to which knowledge in the form of a capability depends on a system or group of people. The level of product observability determines how easy it is for competitors to imitate the knowledge.

Hansen has combined two dimensions of Zander and Kogut (1995) in his research into the relationship between strength of ties and complex knowledge (Hansen, 1999). Hansen (1999) studied the effects of tie strength on completion time of projects of units (a set of actors) linked to each other by ties for different levels of complexity of knowledge. The complexity of the knowledge is determined in Hansen’s research by its codifiability and its system

(12)

12

dependency which show significant influence separately and combined. He finds that the more codified and independent the knowledge, the shorter the completion time of projects in the case of weak ties (Hansen, 1999). However, as knowledge becomes more system

dependent and tacit, strong ties are required to ensure knowledge transfer.

Levin and Cross continued on the proposed distinction between codified and tacit knowledge from Hansen in their research (Levin & Cross, 2004). In their research they look at the mediating effect of competence- and benevolence-based trust on the relationship between the strength of ties and the receipt of useful knowledge, either tacit or codified. This research finds that there is a significant effect of strong ties on the receipt of useful knowledge, more so compared to weak ties. This effect is largely mediated by competence- and benevolence-based trust, leading to the conclusion that weak ties have a bigger effect on the receipt of useful knowledge. The results show that this is especially so for tacit knowledge in

combination with competence-based trust. The authors argue for competence-based trust even for weak ties, but acknowledge that it might be misplaced since a weak tie does not provide enough information for competence-based trust (Levin & Cross, 2004). Even so, the type of knowledge does influence the ease of the knowledge transfer; tacit knowledge is better transferred via ties with competence-based trust that are most likely strong ties.

In another study tacitness of knowledge was found to modify the negative relation between the strength of ties and project completion time (Hansen, Mors, & Lovas, 2005). This study argues that for units (a set of actors) novel knowledge can be found via weak ties, as opposed to via strong ties, and therefore leads to a shorter project completion time. However, when the knowledge being sought is tacit, as opposed to codified, a strong tie will be more useful.

(13)

13

Reagans and McEvily (2003) investigated multiple features of social networks affecting knowledge transfer, but also looked at the tacitness of the knowledge transferred. They found that the relation between the strength of ties and transferred knowledge was moderated by the tacitness of the knowledge transferred, so that this relationship was more positive when knowledge was tacit and less positive when knowledge was codified (Reagans & McEvily, 2003). Later on they find that this effect disappears once controlled for the two social network characteristics. Even so they found that tacit knowledge is easier to transfer via strong ties than weak ties, meaning it is more useful to transfer tacit knowledge via strong ties and codified knowledge via weak ties (Reagans & McEvily, 2003).

The research discussed above had different dependent variables which makes it harder to compare. In any way, apart from the results of Levin and Cross (2004), the results so far indicate that strong ties ease the transfer of tacit knowledge (Hansen et al., 2005; Hansen, 1999; Reagans & McEvily, 2003). So it is expected that strong ties will be used more often for the transfer of tacit knowledge than weak ties. Therefore the following hypothesis is formulated.

Hypothesis 1: The strength of ties will be positively related to the tacitness of knowledge received, so that strong ties lead to receiving tacit knowledge and weak ties lead to receiving codified knowledge.

2.3.Affective state

The affective state of an actor can be positive or negative. Positive affect is defined as an individual’s disposition to experience positive mood states (Watson, Clark, & Tellegen, 1988). Negative affect is the exact opposite. When having a high positive affective state,

(14)

14

people tend to see things that happen to them in a positive way. When having a high negative affective state, people tend to see things that happen to them in a negative way. Both ways reinforce their individual disposition, either positive or negative, and determine the affective state of an individual. In most research the affective state is operationalised as a temporal state (Fredrickson, 2001; Levin & Cross, 2004; Levin et al., 2010; Lount, 2010; Lyubomirsky, King, & Diener, 2005; Lyubomirsky, 2001; Russell, 1980). This temporal state of a positive or negative disposition has shown to have implications on behaviour in relation to the transfer of knowledge between people.

In general, the affective state of an actor influences the way he interacts. Berry and Hansen (1996) found that actors in a high positive affective state interact more frequently in general and consider their interactions to be of a higher quality and more pleasant, when compared to actors in a low positive affective state. Actors in a high negative affective state also interact more frequently, but only with actors from the same sex, and they consider their own interactions as involving higher levels of disclosure, when compared to actors in a low negative affective state (Berry & Hansen, 1996). These results show that a high level of positive and negative affect leads to actors who interact more compared to actors with a low level of positive and negative affect. However, the interactions from actors with a high level of negative affect are only with actors from the same sex and therefore suggest a moderate level of openness to interaction.

The discussion of the affective state’s influence on knowledge transfer starts with the

broaden-and-build theory (Fredrickson, 2001). This theory states that the receiver’s temporary positive emotions can have long-lasting effects. Because positive affect broadens the

(15)

15

physical, social, psychological and, most important for the present research, intellectual resources (Fredrickson, 2001). This implies that a temporal state, such as joy, can lead actors to increase their intellectual resource, such as knowledge, for the long term. This openness suggests a willingness to talk to other actors to exchange knowledge, leading to positive actors who receive more tacit knowledge than actors that are not positive.

This implication was confirmed by a study which found that the positive affect has a positive impact on the knowledge-transfer process (Levin et al., 2010). This was mainly due to the impact of positive affect on the actors in need of knowledge, as opposed to the actors in possession of knowledge, where there was no effect. Actors in need of knowledge with a positive affective state were more likely to absorb and act on new information because they were more open minded. Actors in need of knowledge with a negative affective state were less likely to absorb and act on new information because they were less open minded. Another study that elaborated on the actor in need of knowledge found that the actor in need of knowledge, the receiver, will likely be more critical and less willing to comply if in a negative affective state, while receivers in a positive affective state are more generous and willing to comply (Forgas & George, 2001). This insinuates that actors in a positive affective state are more likely to cooperate and are more willing to transfer knowledge, as stated by Fredrickson (2001) and Levin, Kurtzberg, Phillips and Lount (2010).

The affective state of actors has already been implemented in the research on creativity. Research on creativity resembles research on knowledge transfer, as knowledge transfer is a central construct in research on creativity. Research on creativity has shown that actors in need of knowledge will receive more novel knowledge via their weak ties as opposed to their strong ties (Perry-Smith & Shalley, 2003), which leads to more creativity for the individual

(16)

16

actor (Baer, 2010; Perry-Smith, 2006; Zhou et al., 2009). For strong ties the results have been inconclusive; on the one hand showing that strong ties are not related to less creativity (Zhou et al., 2009), and on the other hand showing that strong ties are not related to creativity at all (Perry-Smith, 2006). In any way, to be able to make use of their ties in order to be creative, actors should take initiative and display their ideas to their colleagues (George & Zhou, 2002). In other words: they should engage in interaction with colleagues. Researchers looking at the arising of creativity in interaction turned to affective state to explain the differences in outcome. They found that positive affect leads to higher creativity (Amabile, Barsade, Mueller, & Staw, 2005; Binnewies & Wörnlein, 2011). Actors in a positive affective state displayed their ideas more to attribute to the group performance, and so appearing more creative. This means that actors in a positive affective state engage more in interaction, or the exchange of tacit knowledge.

The information about the transfer of knowledge and the affective state of the receiver shows that if the receiver is happy, he should be able to absorb more knowledge and be more willing to comply with this knowledge(Forgas & George, 2001; Levin et al., 2010). In general it can be stated that an actor with a positive affective state welcomes the interaction with other people. As the transfer of tacit knowledge requires face-to-face interaction because it is not codified, it is expected that actors with a high level of positive affect will receive more tacit knowledge than actors with a low level of positive affect. The following hypothesis is formulated:

Hypothesis 2: Positive affect moderates the relationship between the strength of ties and the tacitness of knowledge received, so that this relationship is stronger for a high level of positive affect and weaker for a low level of positive affect.

(17)

17

The information about the transfer of knowledge and negative affect shows that a high level of negative affect leads to more interaction, but only with people from the same sex (Berry & Hansen, 1996). Additionally it was shown that receivers with a high level of negative affect were less open minded and more critical in interaction (Forgas & George, 2001; Levin et al., 2010). Taking this information into account it can be stated that receivers of knowledge with a high negative affective state are more negative about interaction and thus less likely to engage in face-to-face interaction, which is required for sharing tacit knowledge. The following hypothesis is formulated:

Hypothesis 3: Negative affect moderates the relationship between the strength of ties and the tacitness of knowledge received, so that this relationship is weaker for a high level of negative affect and stronger for a low level of negative affect.

(18)

18

3.Research Design

The present research is quantitative and was conducted using a cross-sectional survey design. This facilitates the needs for testing the research design shown in figure 1 and was feasible within the time limits of this study. This research takes on a network-perspective, which means that the focus will lie on relationships between actors instead of actors as sole individuals, on dyadic ties.

In this chapter the sample and its characteristics will be outlined, after which the measurement of the different variables in the survey is explicated. Lastly, the analytical strategy is

explained. 3.1.Sample

The population for this research is employees working in organizations that have internal knowledge that can be categorized as tacit or codified. The sampling technique being used is non-probability purposive heterogeneous sampling. This technique was chosen to be able to test the hypotheses in a real-life setting with all the different variables present, within the time constraints of the present research. Non-probability purposive heterogeneous sampling may however lead to bias due to the small number of respondents that are not necessarily

representative for the population. To make sure the sample was as representative as possible the following measures were taken. Firstly an introductory conversation with the CEO and the Human Resources Manager took place that confirmed the company had specialized

knowledge internally which had to be shared within and across departments to achieve the company’s goals. This conversation indicated that the software-department of the company would be most interesting as this department has specialized knowledge that was shared

(19)

19

among employees in a tacit or codified way and had to be combined to produce an offer for each client. It was then concluded that the sample met the requirements of the research’s population. An interview with the manager of the software-department in the Netherlands confirmed that the survey questions were applicable to the sample.

The respondents come from the software-departments of an international firm that specializes in system integration for logistic issues. This department is present in all four locations of the firm, namely in the Netherlands, Moldavia, the United States of America (USA) and China. The employees at the different locations are supposed to share knowledge and work together via e-mail and via Skype.

In total 32 employees from the four software-departments received the survey, of which 28 employees filled out the survey (response rate 87.5%). In total this lead to 168 dyadic ties. Table 1 shows the frequencies in percentages of all the control variables, except tenure. When looking at these characteristics of the sample in table 1 it is important to note the following. First, one out of 32 employees works at the software-department in China, and this employee did not fill out the survey. Second, only 10.7% of the dyadic ties was indicated by a woman which is a very low amount. Third, only 4.2% of the dyadic ties is from employees of the software-department in the USA which is a very low amount. Fourth, only 9.5% of the dyadic ties has an Associate Degree which is a low amount. The second, third and fourth notion are important because the distribution is so unequal that significant results with these variables are likely to be biased by the sample.

Tenure of the respondents was between zero and ten years, with the average on three years and a standard deviation of three years; ten employees were working at the firm for a year or

(20)

20

less (42.9% of all dyadic ties), twelve employees were working at the firm more than one to five years (36.9% of all dyadic ties) and six employees were working for more than five years at the firm (20.2% of all dyadic ties). This shows that most employees have not worked for the company for a long time, which could be expected due to their age and relatively high level of education.

(21)

21 3.2.Measures

One questionnaire in English was distributed to all participants, see appendix 1. This

questionnaire consisted of questions that determined the level of each variable; the strength of ties, the tacitness of the knowledge received, the level of positive affect and the level of negative affect. All items were measured using a 5-point scale.

Before sending out the questionnaire an interview took place with the manager of the software-department in the Netherlands. This interview confirmed that there were no exceptional circumstances during the last month and that it was a month as usual for the employees at the software-departments of the firm. This information was needed as the strength of ties was determined using an advice-network that was geared to the past month. In addition this interview provided additional information regarding the codified knowledge within the software-departments, which was added to one of the questions regarding the tacitness of the knowledge received, to make this question more understandable for the respondents.

For each scale that was composed of multiple items the Cronbach’s alpha was measured to ensure the internal consistency, see appendix 2.

3.2.1.Independent variable: Strength of ties

The strength of ties was measured using an advice-network because it was important to link this to the tacitness of the knowledge that was received. Respondents were asked to identify their sources for advice in the last month from a list of the employees working at the software-departments. Asking respondents to identify their own sources of advice leads to a lower risk of measurement error and a more accurate response (Wong, 2008; Zhou et al., 2009). For the

(22)

22

colleagues the respondents selected, they received three additional questions to determine the strength of the ties.

Theory indicated that the items should reflect the frequency of the interaction, the duration of the relationship and the affection they felt for the other person (Krackhardt, 1992). Since most employees worked for the company for a short time, the duration of the tie was not expected to give information that adequately reflected the advice tie strength. Instead the importance of the other person’s advice was questioned, which was previously used as a sole measure of advice tie strength (Zhou et al., 2009).

First, the respondents were asked to indicate the closeness of their working relationship with the selected colleagues, with ‘working’ in bold. Questioning the closeness of a relationship reflects the affection-part of a tie. It was decided to focus this on a working relationship instead of a social relationship as the content flowing through the ties concerned advice. The question was asked to assess the impression of the strength of the advice-tie that the receiver of knowledge has (Levin & Cross, 2004; Reagans & McEvily, 2003). The closeness of the working relationship was not confirmed by the other actor in order to not reduce the number of dyadic ties.

Second, the respondents were asked to indicate the frequency of their interaction with the selected colleagues. This question was asked as actors who have a strong tie are expected to interact often (Krackhardt, 1992; Reagans & McEvily, 2003). Information gained during the interview indicated that it was normal to talk about work-related issues almost once per day, when both actors work in the software-department of one country. This frequency was taken as the average. This question was reverse coded for analysis.

(23)

23

Third, the respondents were asked to assess the importance of the selected colleague as a source of advice for work-related problems. This question was chosen to see if the actor actually valued the advice from their selected colleague and had been proven to be of value in previous studies (Zhou et al., 2009). This question was reverse coded for analysis.

The strength of ties-scale showed to have an acceptable reliability, with Cronbach’s Alpha=.74. The corrected item-total correlations indicate that all the items have a good correlation with the total score of the scale (all above .30). Also, deleting one of the items would either bring Cronbach’s Alpha down or not affect reliability substantially. It was decided to use all three items of strength of ties.

A high score on the scale of strength of ties indicates a strong tie and a low score on the scale of strength of ties indicates a weak tie.

3.2.2.Dependent variable: Tacitness of knowledge received

To measure the tacitness of the knowledge that was received via the advice ties that were indicated, three questions were asked based on Hansen (1999) and Levin and Cross (2004). The first questions let the respondents identify the type of knowledge they received from their selected colleague. The answers were formulated by bringing the information from the

interview and the information from previous studies together.

The second questions asked if the knowledge was sufficiently explained in writing to the actor by the selected colleague. This question was reversed coded for analysis.

The third question asked the respondent to assess how well documented the knowledge was that he received from the selected colleague.

(24)

24

The tacitness of knowledge received-scale showed to have a poor reliability, with Cronbach’s Alpha=.52. The corrected item-total correlations indicate that only one item, the recoded item, has good correlation with the total score of the scale (above .30), the other two do not have good correlation with the total score. However, deleting one of the items would bring Cronbach’s Alpha down, and this difference becomes substantial for the second question. This means that this scale does not comply with the rule of thumb for requirements for Cronbach’s Alpha. It was decided to use this scale nonetheless for the following reasons. First, the scale is based on only three items that measure different aspects of the tacitness of knowledge received and leaving one or two out would mean the construct does not measure what it is supposed to measure. Second, even though the scale was developed in a specific setting (Hansen, 1999), it has showed to be a reliable scale in other settings before (Hansen et al., 2005; Levin & Cross, 2004). Therefore it was decided to use all three items for the

tacitness of knowledge received-scale and take the poor reliability into account when discussing the results and their implications.

A high score on the scale of the tacitness of knowledge received indicates the receiving of tacit knowledge and a low score on the scale of the tacitness of knowledge received indicates the receiving of codified knowledge.

3.2.3.Moderating variables: Positive and negative affect

The moderating variables are positive affect and negative affect. They are measured with the Positive Affect and Negative Affect Schedule (PANAS)-Scales. These scales use ten items per construct to assess the level of positive and negative affective state of a person (Watson et al., 1988). The scale was developed in 1988 and confirmed as a reliable and valid

(25)

25

indicate to what extent they experienced the emotions. The items of the two different constructs were ordered so that there was no logic sequence in answering the question. The positive affect-scale was found to have a good reliability, with Cronbach’s Alpha=.90. The corrected item-total correlations indicate that all but one item, “Alert”, have a good correlation with the total score of the scale (all above .30, the one .02). After deleting “Alert” from the positive affect scale it becomes excellent: Cronbach’s Alpha=.92. It was thus decided to use only nine variables for the positive affect-scale.

A high score on the scale of positive affect indicates a high level of positive affect and a low score on the scale of positive affect indicates a low level of positive affect.

The negative affect-scale was found to have a good reliability, with Cronbach’s Alpha =.79. The corrected item-total correlations indicate that all but one item, “Hostile”, have a good correlation with the total score of the scale (all above .30, the one -.05). After deleting “Hostile” from the negative affect scale it becomes excellent Cronbach’s Alpha=.84. It was thus decided to use only nine variables for the negative affect-scale.

A high score on the scale of negative affect indicates a high level of negative affect and a low score on the scale of negative affect indicates a low level of negative affect.

3.2.4.Control variables

Control variables that were measured are gender, country of workplace, age, education level and tenure. Gender was measured with two categories, namely (1) male, and (2) female, resulting in a nominal variable. The country of workplace was measured with four categories, namely (1) China, (2) Moldavia, (3) the Netherlands, and (4) the USA, resulting in a nominal variable. Age was measured with six categories, namely (1) 24 or younger, (2) 25 to 34, (3)

(26)

26

35 to 44, (4) 45 to 54, (5) 55 to 64, and (6) 65 or older, resulting in an ordinal variable. Education level was measured using five categories, namely (1) no schooling completed, (2) high school graduate, diploma or the equivalent, (3) associate degree, (4) bachelor’s degree, (5) master’s degree or higher, resulting in an ordinal variable. Tenure was measured with an open question that asked respondents to write down how many years they had been working for the company, resulting in a scale.

3.3.Analytical strategy

The questionnaire was made using an online service called Qualtrics Survey Software. The link to this questionnaire was accompanied by an explanatory e-mail and distributed to all respondents via the manager of the Software-department in the Netherlands. After two

reminders the survey was closed on the 12th of May, after which the data was retrieved. As the present research has dyadic ties instead of actors as the object, the data first had to be

transformed in Excel to make it suitable for analyses. It was transformed into a vertical edge list, where every row represents a tie between the receiver and the sender of knowledge as indicated by the receiver. The results were then statistically analysed using Statistical Package for the Social Science (SPSS) 23 in four steps.

First, a correlation matrix (table 2) was developed to find significant correlations between the variables. The variable ratings of each scale were calculated by computing scale means of the total number of items that were used to measure each variable. Additionally, the standard deviation was calculated for all variables. For the department variable dummy variables were created as this concerned a nominal variable with more than two categories. Gender was recoded so that (0) is female and (1) is male, to make interpretation easier.

(27)

27

For the following steps the variables of strength of ties, positive affect and negative affect were mean-centered by subtracting the mean from the scores. Mean-centering decreases the risk of multicollinearity, which is especially high when testing moderation. Multicollinearity occurs when the correlation between the interaction of variables and the variables themselves is high, which leads to inaccurate coefficient estimates. Thus mean-centering was done to make coefficients more meaningful.

Second, to find out if positive affect and negative affect independently moderate the relationship between strength of ties and tacitness of knowledge received, hierarchical multiple regressions with two-way interactions were executed for both positive affect (table 3A) and negative affect (table 3B). Figure 2 and figure 3 show what was measured.

Third, to find out if positive affect and negative affect jointly moderate the relationship between strength of ties and tacitness of knowledge received, a hierarchical multiple

regression with both two-way interactions was executed (table 4). Figure 4 shows what was measured.

(28)

28

Fourth, to find out what levels of the moderating variables significantly influence the relationship between strength of ties and tacitness of knowledge received, Hayes’

computational tool called PROCESS was used. This tool allows the user to do a path analysis-based moderation and mediation analysis as well as a combination of the two. Based on the SPSS PROCESS manual model 2 was chosen. This model displays one independent variable, one dependent variable, and two moderators. The output gave the significant levels of positive affect and negative affect (table 5).

(29)

(30)

30

4.Results

This chapter will discuss the results of the statistical analyses. It will start off with discussing the significant correlations that can be found in table 2. It then continues with explaining the results of the two-way interactions for positive affect and negative affect independently. After that the results of the two-way interaction including positive affect and negative affect as a moderator will be discussed. Finally, the significant levels of positive affect and negative affect will be elaborated upon.

4.1.Significant correlations

Table 2 shows the correlations for all variables. The significant correlations for the control variables will be discussed first, after which the correlations with the independent variable, dependent variable and moderator of the research design will be discussed.

4.1.1.Control variables

Table 2 shows that gender correlates positively (r =0.30, p<.01) with the department in Moldavia and correlates negatively (r=-0.33, p<.01) with the department in the Netherlands This means that the department in Moldavia mostly employs males, while the department in the Netherlands also employs females. Gender also correlates negatively with age (r=-0.52, p<.01), and correlates positively with level of education (r=0.38, p<.01). This means that female respondents tend to be younger than male respondents, and that they have a lower level of education when compared to the male respondents. Lastly, gender correlates positively with both positive affect (r=0.17, p<.05) and the tacitness of knowledge received (r=0.31, p<.01). This means that the male respondents are happier than the female

(31)

31

at least partly, be explained by the unequal distribution of gender, taking into account that the number of male respondents was much higher than the number of female respondents.

When we look at the correlations of the different departments the following can be found. The department of Moldavia correlates negatively with age (r=-0.57, p<.01), meaning that

respondents who work in Moldavia are rather young. Additionally the department in Moldavia correlates positively (r=0.27, p<.01) with the level of education, meaning that respondents who work in Moldavia usually have finished higher levels of education. The department of Moldavia correlates negatively (r=-0.17, p<.05) with tenure, meaning that respondents who work in Moldavia have been working for the company for a short time. Lastly, the department in Moldavia correlates positively (r=0.39, p<.01) with positive affect, meaning that the respondents who work in Moldavia are happier.

The department of the Netherlands correlates positively (r=0.56, p<.01) with age, meaning that the respondents working in the Netherlands are older on average. Additionally the

department of the Netherlands correlates negatively (r=-0.30, p<.01) with education, meaning that respondents who work in the Netherlands have finished lower levels of education on average. The department of the Netherlands correlates positively with tenure (r=0.17, p<.05), meaning that the respondents working in the Netherlands work for the company for a longer time. This is expected since the company was founded in the Netherlands and has been there the longest, when compared to the other departments which show no significant correlations or a negative correlation with tenure. Lastly, the department of the Netherlands correlates negatively (r=-0.23, p<.01) with positive affect, meaning that respondents who work in the Netherlands are less happy.

(32)

32

meaning that respondents who work in the USA are less happy, even more so than the respondents who work in the Netherlands.

When we look at the remaining correlations for age the following can be found. Age correlates negatively (r=-0.19, p<.05) with education, meaning that the respondents with a higher age have a lower level of education on average. Additionally age correlates positively with tenure (r=0.37, p<.01), meaning that respondents with a higher age have worked for the company for a longer time, which can be expected. Lastly, age correlates negatively (r=-0.44, p<.01) with positive affect, meaning that the older respondents are less happy than the

younger respondents.

When we look at remaining correlations for the level of education we first find that education is positively correlated (r=0.38, p<.01) with tenure, meaning that respondents who have a higher level of education work for the company for a longer time. Second, education level is positively correlated (r=0.21, p<.01) with positive affect, meaning that the respondents with a higher level of education are happier than the respondents with a lower level of education. Third, education level is negatively correlated (r=-0.20, p<.05) with the tacitness of

knowledge received, meaning that respondents with a higher level of education receive more codified knowledge than respondents with a low level of education.

Finally, tenure is negatively correlated (r=-0.28, p<.01) with positive affect and positively correlated (r=0.24, p<.01) with negative affect, meaning that respondents who have worked for the company for a longer time have a lower level of positive affect and a higher level of negative affect.

(33)

33

4.1.2.Independent, moderating and dependent variables

The strength of ties is negatively correlated (r=-0.32, p<.01) with tacitness of knowledge received. This means that respondents who received knowledge via strong ties have received this in a more codified form. This is the opposite of what was expected and further analyses will shed more light on this correlation.

Positive affect correlates negatively (r=-0.20, p<.01) with negative affect, meaning that respondents with a higher level of positive affect have a lower level of negative affect, which can be expected.

Positive affect also correlates negatively (r=-0.30, p<.01) with tacitness of knowledge received, meaning that respondents with a higher level of positive affect received more codified knowledge than respondents with a lower level of positive affect. This is also the opposite of what was expected and further analyses will shed more light on this correlation. Negative affect correlates positively (r=0.31, p<.01) with tacitness of knowledge received, meaning that respondents with a higher level of negative affect received more tacit knowledge than respondents with a lower level of negative affect. This is also the opposite of what was expected and further analyses will shed more light on this correlation.

4.2.Independent two-way interactions

4.2.1.Direct effects

Table 3A and table 3B show results of the hierarchical multiple regressions and the direct effects of multiple variables on the tacitness of knowledge received. Table 3A shows the results with positive affect as a moderator and table 3B shows the results with negative affect as a moderator. In both tables the department of Moldavia is not displayed as this is

(34)

(35)

(36)

36

a reference category, meaning that scoring zero on working in the department of the Netherlands and the USA means working in the department of Moldavia. The first step in both tables tests the effects of control variables gender, department, age, education and tenure on the tacitness of knowledge received. This shows that gender (β=.48, p<.01), and education (β=-.42, p<.01) significantly affect the tacitness of knowledge received. In the second and third step the department of the USA also significantly affects (β=-.33 and β=-.30, p<.01) the tacitness of knowledge received. As the regression coefficient is negative it means that the department of Moldavia positively affects the tacitness of knowledge received.

To test hypothesis 1, tie strength is added in the second step of the regression for both tables. Results show that tie strength in table 3A is negatively associated with the tacitness of

knowledge received (β=-.25, p<.01). A significant amount of variance is explained by adding tie strength and positive affect to the regression (ΔR²=.19, p<.01). Results of table 3B confirm that tie strength is negatively associated with the tacitness of knowledge received (β=-.30, p<.01). A significant amount of variance is explained by adding tie strength and negative affect to the regression (ΔR²=.15, p<.01). The results show that tie strength in both instances has a negative relationship with the tacitness of knowledge received, which means that hypothesis 1 is not supported. The opposite is true, meaning that actors receive more codified knowledge from actors they have strong ties with.

4.2.2.Moderation effects

To test hypothesis 2 and 3, interaction variables were constructed to test the moderating effect of positive affect and negative affect on the relationship between tie strength and the tacitness of knowledge received. Hypothesis 2 predicted that positive affect would moderate the relationship between the strength of ties and the tacitness of knowledge received, so that it

(37)

37

becomes stronger for a high level of positive affect and weaker for a low level of positive affect. The results show that the interaction between tie strength and positive affect was negative and significant (β=-.18, p<.01). The interaction added a small but significant amount of variance explained of the tacitness of knowledge received (ΔR²=.03, p<.01). This means that hypothesis 2 is partially supported, but that the opposite of what was expected is true; positive affect moderates the relationship so that a higher level of positive affect strengthens the relationship between tie strength and the tacitness of knowledge received, however, this relationship was expected to be positive and turned out to be negative. In wording the hypothesis is confirmed, but in meaning it is not.

Dawson’s two-way interaction plot sheets allow the depiction of slopes that show the

(38)

38

a low level of positive affect (-1 SD) and a high level of positive affect (+1 SD). Dawson states on his website that when control variables are present in the regression, the values of the dependent variable displayed on the plot will be inaccurate, but the pattern and its interpretation will be correct. The slopes are depicted in graph 1. The figure shows that positive affect moderates the relationship between the strength of ties and the tacitness of knowledge received so that higher levels of positive affect result in receiving less tacit knowledge, while lower levels of positive affect do not seem to have an effect on the relationship. The fourth section of this chapter will conclude on that.

Hypothesis 3 predicted that negative affect would moderate the relationship between strength of ties and tacitness of knowledge received, so that the relationship would become weaker for a high level of negative affect and stronger for a low level of negative affect. The results show that the interaction between tie strength and negative affect was positive and significant (β=.15, p<.05). The interaction added a small but significant amount of variance explained of the tacitness of knowledge received (ΔR²=.02, p<.01). This means that hypothesis 3 is

partially supported; negative affect moderates the relationship so that a higher level of negative affect weakens the relationship between tie strength and the tacitness of knowledge received. This means the hypothesis is partially confirmed, as the wording indicated a weaker relationship for higher levels. However, the hypothesis meant to say that higher levels

negative affect would lead the receiving actor to receive more codified knowledge, while lower levels of negative would lead the receiving actor to receive more tacit knowledge, and this is not the case.

The slopes for the conditional effect of the strength of ties on the tacitness of knowledge received, depending on a low level of negative affect and a high level of negative affect are

(39)

39

depicted in graph 2. The figure shows that negative affect moderates the relationship between the strength of ties and the tacitness of knowledge received so that lower levels of negative affect result in receiving less tacit knowledge, while a higher level of negative affect does not seem to have an effect. The fourth section of this chapter will determine the significant levels of negative affect.

4.3.Combined two-way interactions

In this section both two-way interactions of positive affect and negative affect are presented to see if positive affect and negative affect together influence the relationship between the

(40)

(41)

41

Model 1 presents the control variables and their influence on tacitness of knowledge received. Education and gender were found to significantly influence the tacitness of knowledge

received, as was the case before.

Model 2 added positive affect and negative affect as influencers on tacitness of knowledge received. The results of model 2 show that both positive affect (β=-.42, p<.01) and negative affect (β=.25, p<.01) significantly influence the level of tacitness of knowledge received. Model 2 also showed to explain a significant amount of variance (ΔR²=.25, p<.05).

Model 3 added the interaction effects of positive affect and negative affect with strength of ties on the tacitness of knowledge received. The results of step 3 show a significant effect of the interaction between strength of ties and positive affect (β=-.14, p<.05), but an insignificant effect of the interaction between strength of ties and negative affect (β=.08, p>.05). However, including both interactions explained a small but significant amount of variance (ΔR²=.03, p<.05). It can thus be stated that both positive affect and negative affect moderate the relationship between strength of ties and tacitness of knowledge received. Higher levels of positive affect make this relationship more negative and higher levels of negative affect make this relationship less negative.

4.4.Significant levels of moderation

To find the significant levels of moderation of positive affect and negative affect a Hayes’ computational tool called PROCESS was used. This tool calculates the significant levels based on the Johnson-Neyman technique. The results showed the conditional effect of X on Y at different values of the moderator (see table 5). This means it shows what levels of positive affect and negative affect have a significant influence on the relationship between strength of ties and tacitness of knowledge received. The different values of the moderators are the mean

(42)

42

(average), the mean minus one standard deviation (low), and the mean plus one standard deviation (high).

A low level of negative affect has a significant effect, when the level of positive affect is average (β=-.36, p<.01), or when positive affect is high (β=-.49, p<.01). So when negative feelings are limited to a minimum and positive feelings are present or at a maximum, this has a significant influence on the relationship between strength of ties and tacitness of knowledge received.

An average level of negative affect has a significant effect when the level of positive affect is also average (β=-0.25, p<.01), or when positive affect is high (β=-0.39, p<.01). So when negative feelings are present and positive feelings are present or at a maximum, this influences the relationship between strength of ties and tacitness of knowledge received. A high level of negative affect has a significant effect when the level of positive affect is also high (β=-0.29, p<.05). This means that when negative feelings and positive feelings are at a

(43)

43

maximum, this influences the relationship between strength of ties and tacitness of knowledge received. The presence of both emotions at the same time seems highly unlikely.

The results show that a low level of positive affect never has a significant effect on the relationship between strength of ties and tacitness of knowledge received, while a high level of positive affect always has a significant effect. A low or average level of negative affect has a significant effect, except when positive affect is low. Additionally, a low level of negative affect has a larger effect than an average level negative affect, and a high level of positive affect has a larger effect than an average level of negative affect. Overall it can be stated that a low to overage level of negative affect and an average to high level of positive affect have a significant effect on the relationship between strength of ties and tacitness of knowledge received.

Hypothesis 2 and 3 thus remain partially confirmed: for a high level of positive affect and a low level of negative affect. Hypothesis 2 and 3 are not confirmed for a low level of positive affect and a high level of negative affect. Additionally, the influence of positive affect was expected to be positive and for negative affect to be negative, but it was shown that this was exactly the opposite.

(44)

44

5.Discussion

5.1.Theoretical and practical implications

The present research aims for a deeper understanding of the relationship between the strength of ties and knowledge transfer, the tacitness of knowledge in specific, and the variables that influence this relationship. It contributes to existing research by taking the affective state of actors into account, which was not done before. Affective state is a concept that was

introduced in psychology and was found to have an influence on the openness in interaction of people (Fredrickson, 2001). Since the transfer of tacit knowledge requires face-to-face interaction, while the transfer of codified knowledge does not, it was especially interesting to determine the influence of affective state on the transfer of tacit or codified knowledge. 5.1.1.Control variables

The first section of the results show the direct effects of the control variables, strength of ties, positive affect and negative affect on the tacitness of knowledge received. Gender showed to have a positive effect on the tacitness of knowledge received, meaning that male actors received more tacit knowledge than female actors. Most researchers of social networks do not include gender as a sole variable, but rather if the two actors involved in a relationship share the same gender (Levin & Cross, 2004; Reagans & McEvily, 2003) which was not found to have an effect. However, this finding could be explained by the notion of homophily in combination with the gender segregation of the network. Women talk more with women because it feels more comfortable (Burt, 1998; McPherson & Smith-Lovin, 1987). It can be expected that when advice needs to be asked and someone does not feel comfortable talking to the advice-giver in person, he would ask the advice in a codified way. Considering only

(45)

45

10% of the respondents is female, this would lead women to interact less face-to-face in general and men to interact more face-to-face.

Education showed to have a negative effect on the tacitness of knowledge received, meaning that a higher level of education led to receiving less tacit or more codified knowledge. Previous research on tacitness of knowledge did not include education in their analyses (Hansen, 1999; Levin & Cross, 2004), or included the dyadic dissimilarity in education (Reagans & McEvily, 2003) which did not have an effect. The results could be explained by noting that the company employed actors with a rather low level of education and a high level of education in this department. Considering the employees all work in the same department, actors with a higher level of education most probably find the work less difficult. For teams it was found that using codified knowledge leads to saving time, while using tacit knowledge improves the quality of a teams’ work and their ability to signal competence (Haas & Hansen, 2007). It is likely that this finding also applies to individuals because a team consists of individuals that receive the tacit or codified knowledge. As well-educated actors are more intelligent, improving the quality of their work might not have their focus while saving time does. It can thus be explained that well-educated actors are more likely to receive codified knowledge, instead of tacit knowledge.

In subsequent analyses the department of the USA showed to have a negative influence on the tacitness of knowledge received, compared to the department of Moldavia. This means that actors in the USA shared more codified knowledge, while actors in Moldavia shared more tacit knowledge. This can be explained by the organizational culture of those specific departments, as the organizational culture determines how actors interact with each other (Schein, 1984).

(46)

46 5.1.2.Independent variable

The first section of the results also showed the negative direct effect of strength of ties on tacitness of knowledge received. Subsequent analyses confirmed this finding. Previous research found that tacit knowledge should flow through strong ties in order to save time (Hansen et al., 2005; Hansen, 1999), and that tacit knowledge was easier transferred and considered more useful when transferred via strong ties (Levin & Cross, 2004; Reagans & McEvily, 2003). Hence a positive direct effect of strength of ties was expected. Since this was not found hypothesis 1 was not supported. The simplest explanation is that although these findings prove that transferring tacit knowledge via strong ties is most useful, it might be the case that this is not actually happening in organizations. A more theoretical explanation is that different types of advice may require different levels of tacitness. Previous research has shown that advice is not a concrete concept (Cross, Borgatti, & Parker, 2001). Actors may ask very different kinds of advice, which connects different people, makes different structures, and may even hold different meanings for people (Cross et al., 2001). It may thus be the case that the kind of advice that actors in the company ask determines that actors mainly acquire codified knowledge via their strong ties and tacit knowledge via their weak ties.

5.1.3.Moderating variables

Previous research showed that actors with a high level of positive affect were more open to interaction or open-minded and more generous in the interaction (Forgas & George, 2001; Fredrickson, 2001; Levin et al., 2010). It also showed that actors with a positive affective state engaged more in interaction with colleagues to attain higher levels of creativity (Amabile et al., 2005; Binnewies & Wörnlein, 2011). Hence it was expected that a high level of positive affect would strengthen the (positive) relationship between strength of ties and tacitness of

(47)

47

knowledge received, while a low level would weaken this relationship. The results show that a high level of positive affect strengthens the relationship, while a low level of positive affect does not influence the relationship. Since the relationship between strength of ties and

tacitness of knowledge received was negative, hypothesis 2 is partially confirmed in wording, but not confirmed in meaning. A high positive affective state thus leads to receiving more codified knowledge as the tie between actors becomes stronger. This could be explained by findings of recent research that a too high level of positive affect does not have the same or even contrary effects as a result (Fredrickson, 2013; Lam, Spreitzer, & Fritz, 2013).

Previous research showed that actors with a high level of negative affect were more critical and less open-minded in interactions (Forgas & George, 2001; Levin et al., 2010). It also showed that actors with a high level of negative affect were likely to interact more, but only with actors from the same sex (Berry & Hansen, 1996). Hence it was expected that a high level of negative affect would weaken the (positive) relationship between strength of ties and tacitness of knowledge received, while a low level would strengthen the relationship. The results show that a high level of negative affect does not influence the relationship, while a low to average level of negative affect strengthens the relationship. Since the relationship between strength of ties and tacitness of knowledge received was negative, hypothesis 3 is partially confirmed in wording, but not confirmed in meaning. A low to average negative affect state thus leads to receiving more codified knowledge, as the tie between actors become stronger. However, this influence is limited as the two-way interaction with both positive affect and negative affect as a moderator showed that the interaction effect between negative affect and strength of ties was not significant. The results for negative affect as a moderator can be explained by the dynamic perspective introduced by recent research on creativity

(48)

48

(Bledow, Rosing, & Frese, 2013). The authors of this research find that it is not the affective state that the actors are in, but the change of states, that leads to results for level o f creativity. Another way to explain the results is by arguing that the results of previous research have been wrongly interpreted. This would mean that being more critical and less open-minded due to a high level of negative affect, does not mean that actors avoid interaction, but actually look for it and receive more tacit knowledge.

5.1.4.Overall

The results show that there is a relationship between the strength of ties and the tacitness of knowledge received, and that this relationship is moderated by positive affect and for a small part by negative affect.

5.2.Limitations

The present research has four limitations. First, the present research is limited by its sample, which consists of 28 employees of software-departments of one company. The tacitness of knowledge in general might differ in different companies. Codifiability, the other side of the same coin, was one of five constructs to measure the degree to which a capability can be easily communicated (Zander & Kogut, 1995). The level of codifiability and tacitness thus depends on the company’s capability.

Second, the scale of tacitness of knowledge received was not found to be reliable. Only half of the answers were internally consistent. The reason for this internal inconsistency could be that the scale-items were not consistent, as three different aspects of tacitness of knowledge were queried. However, this seems unlikely as the scale has proven to be internally consistent before (Hansen et al., 2005; Hansen, 1999; Levin & Cross, 2004). It could also be due to a

(49)

49

misunderstanding regarding the questions, meaning that the questions were not rightfully understood and respondents could not give the right answer. To prevent this from happening contact details of the researcher and her supervisor were given in case the respondents had questions or remarks concerning the research or the survey. No respondents contacted either the researcher or the supervisor with questions or remarks, which makes a misunderstanding regarding the questions unlikely. Another explanation, which seems more likely, is that the advice that was recorded concerned multiple types of advice. As was stated earlier, previous research has shown that advice is not a concrete concept, leading to different types of advice with different structures (Cross et al., 2001). As the questions were geared at the month prior to filling out the survey, it could be the case that respondents filled out the questions per person for different kinds of advice that came in different structures.

This limitation has a big impact on the present research as it means that the relationship that was found and the influence of the affective state on this relationship only accounts for half of the results of the dependent variable. This means that only for half of the times when

respondents received advice, the tacitness of the knowledge that they received can be predicted based on the predicting variables.

Third, the survey was administered in departments in three different countries, which might have influenced the interpretation of the questions and their answers. To overcome this issue the questions were made as objective as possible and the survey was administered in English, as advised by the CEO of the company. The questions that were expected to suffer most from different interpretations were the questions about the strength of ties and the moderating variables. The scale of the strength of ties showed to be internally consistent, as well as the moderating variables after deleting one question from both moderating variable-scales. Thus