Taking a network perspective : friendship and advice network tie strength and in-role and extra-role performance

Taking a network perspective

Friendship and Advice Network Tie Strength and In-Role and Extra-Role Performance

Master Thesis

Fieke Ligthart – 6124674

University of Amsterdam – Amsterdam Business School

Master Business Administration – Strategy Track

Supervisor: Nathan Betancourt

Date of submission: 29

of June, 2015 – final version.

Statement of originality

This document is written by Fieke Ligthart 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.

2

Table of Content

Abstract ... 3

I Introduction ... 4

II Literature Review ... 7

Individual work performance ... 7

Knowledge Sharing ... 10

Social Network Analysis – Relationships. ... 11

Social Networks core concept – Tie Strength ... 12

Type of Network – Friendship and Advice ... 14

Research goal and hypotheses... 17

Friendship Network hypotheses ... 18

Advice Network hypotheses. ... 19

III Data and Method ... 20

Research Design... 20 Sample ... 22 Level of Analysis ... 23 Independent Variable ... 24 Dependent Variables ... 26 Control variables ... 28 IV Results ... 29 Descriptive statistics ... 29 Correlations ... 31 Normality Analysis ... 33 Regression ... 36 Friendship network ... 36 Advice network ... 38 V Discussion ... 41 VI Conclusion ... 46 References ... 47 Appendix ... 51

3

Abstract

Correct understanding of the factors that might enhance or diminish individual performance is essential to monitor and manage performance. This study considers how tie strength, perceived by the ego in his or her friendship and advice network, can be associated with ego’s in-role and extra-role performance. Hereby it is assumed that knowledge sharing is the underlying variance that explains differences among individual performance. This study performed a social network analysis, using data gained from 26 employees of a young professionals organization in Amsterdam. First, the separate concepts ‘individual performance’, ‘knowledge sharing’, ‘tie strength’ and ‘friendship and advice networks’ are integrated leading to 4 hypotheses suggesting that for both the friendship and the advice network tie strength is positively associated with the in-role and extra-role performance of the employee. The in-role and extra-role performance were measured using supervisor ratings. Ultimately, this study was able to show that tie strength in a friendship network is positively associated to the extra-role performance of an employee in the eyes of the supervisor. The last section of this research will discuss the main findings and the proposed managerial implications.

Acknowledgement:

Totally in line with the social network approach I fully agree that relationships are worth everything. Surround yourself with good people, especially in case of a ‘thesis take two’. A special thanks to my supervisor Nathan, my always supporting parents and my dear (library) friends. Enjoy reading!

4 “The richest people in the world look for and build networks.

Everyone else looks for work”

- Robert Kiyosaki, 2013.

I Introduction

‘Why are some people better performers than others?’; one of the enduring questions in today’s society. The last 40 years, tons of scholars addressed the performance domain and tried to understand the antecedents that drive differences in performance. In any organization, the performance of individuals is critical for the success of both the individual and the organization (Chung, 2009) . Therefore, correct understanding of the factors that might enhance or diminish individual performance is essential to monitor and manage performance. Historically, individual performance has mainly been conceptualized as simple ‘task – performance. Recently, a growing body of management research argues that in order to understand individual performance the domain should be decomposed to task-level performance and contextual-performance. Motowidlo and Scotter (1994) showed empirically that both concepts should be distinguished. Besides, the same research showed that supervisors take both task-performance and contextual performance into account when making overall judgments about the performance of their subordinates. Still, splitting up the performance domain doesn’t explain why some individuals perform better than others.

Chung (2009) mentioned that individual job performance is affected by a variety of factors; ‘experience, education, keeping abreast of work-related and technological changes, and so on’ (p. 35) However, leaving this factors and individual properties out of context, job performance is, to some extent, the outcome of obtaining the right knowledge for the task at hand (Cross and Cummings, 2004) . Knowledge sharing represents knowledge exchange

5 among for instance individuals or organizations (Bukowitz and Ruth, 1999). Therefore the, by scholars accepted, underlying variance of differences in in-role and extra-role performance is knowledge sharing. The ability to obtain resources such as information is directly related to individual performance (Mehra, 2001).

One way of understanding the drivers behind employee knowledge sharing is by performing a social networks analysis (Reinholt et all., 2011). According to social network theory, ‘networks spanning social divides are associated with performance-related outcomes’ (Burt, 1992 in Cross and Cummings, 2004, p. 928). Knowledge sharing among individuals in organizations is critically affected by social networks (Borgatti and Cross, 2003). At the individual level, researched showed that people can for instance benefit from their position in a network. If the person can connect others who would otherwise not be connected, they bridge so called, ‘structural holes’. Underlying idea is that these persons have the opportunity to obtain more information and have more resource control (Burt, 1992). Bridging structural holes is associated with earlier promotions, greater career mobility and a more successful adaptation to a changing environment. (Gargiulo & Benassi, 2000; Podolny & Baron, 1997). Social network theorists use also other mechanisms to understand what drives the differences in employees’ knowledge sharing. Levin and Cross (2004) related knowledge sharing to the quality of a dyadic relationship between two actors in a network, similar to the focus of this research: tie strength.

Tie strength determines the quality of a dyadic relationship. Dyadic ties can make a large number of important sources, such as information, available. In that sense, they can contribute to important outcomes as in-role and extra-role job performance. Two streams can be identified in the current studies around tie strength. Granovetter’s (1973) work, ‘the Strength of Weak Ties’ argues that individuals obtain new and novel information, rather from

6 weak ties than from strong ties. Subsequent to this study are scholars who promote the strength of strong ties, arguing that strong ties, more than weak ties, are useful for obtaining the right knowledge.

Obtaining the right knowledge depends not only on the quality of ties, also the type of network an actor is involved in affects knowledge sharing. According to Gibbons (2004) two types of relations commonly occur in organizations; the relation between friends, together forming an individual’s friendship network and the relation between advisor and advisee, forming the advice network.

Previous research have linked certain concepts around tie strength, in-role and extra-role performance both for friendship networks and advice networks. However, an integrating concept considering the strength of a relationship and its association with in-role and extra-role in both a friendship networks and an advice networks has been neglected. This research integrates previous studies around these concepts in order to test whether tie strength in a friendship or advice network can be associated by an ego’s in-role or extra-role performance, as measured by supervisors.

Hereby this research contributes to the understanding of the factors that might enhance or diminish individual performance. This research assumes that the underlying variance that drives the differences in in-role and extra-role performance is knowledge sharing. In order to explain this variances, this research uses a social network analysis focused on the quality of dyadic ties: tie strength. First the current literature will be outlined, resulting in 4 hypothesis. After, the data and method section will discuss the methodology of this research. Thereafter the analytical strategy will be described in the results section. Finally, the results will be discussed in the discussion, ending by the conclusion of this research.

7

II Literature Review

The last 40 years tons of scholars addressed the performance domain and changed the general meaning of workplace performance excessively. Nowadays, the view of work roles in changing environments have a broader understanding and seem to more dynamic (Griffin et all, 2007). Still, job performance is in its essential ‘the degree to which an individual helps the organization reach its goals’ (Campbell, 1983). Individual job performance is characterized by the extent to which the individual contributes, in the eyes of the management, to the organizational goals (Mehra, Kilduff and Brass, 2001). In any organization the performance of individuals is critical for the success of both the individual as for the organization (Chung, 2009). Therefore, correct understanding of the factors that might enhance or diminish individual performance is essential to monitor and manage performance.

Individual work performance

Work performance is multidimensional concept (Motowidlo et al, 2009), an abstract, latent construct which is not directly measurable (Koopmans et all, 2011). Individual workplace performance is an issue grasping companies all over the world. Scholars agreed that individual workplace performance should be defined in terms of behavior rather than results and that only behaviors relevant for the organizational goals should be seen as individual performance (Koopmans et all, 2011). Where historically individual performance mainly has been conceptualized as simple ‘task-performance’, more recently scholars suggested more behavioral concepts additional to simple-task performance.

Almost all frameworks in the area of task performance mention ‘task-performance’ as the first dimension of individual work performance (Koopmans et all, 2011). Task-performance is defined as ‘the effectiveness with which job incumbents perform activities that contribute to the organization’s technical core either directly by implementing a part of

8 its technological process, or indirectly by providing it with needed materials or services’. (Borman and Motowildo, 1997, p. 99). Almost similar to the concept of ‘task-performance’ is the from Katz (1964) originating concept of ‘in-role behavior’, viewed as the required or expected behavior whereby the formal job-description is met. (Dyne and LePine, 1998). Combining indicators derived from several scholars, the dimension task-performance is associated with for example completing job tasks, work quantity, work quality, administration and solving problems. (Koopmans et all, 2011) .

Researchers came to believe that performance is more than succeeding in prescribed roles. (Koopmans et all, 2011).Two scholars who broaden the understanding of performance are Borman & Motowildo (1997). By blending three previous concepts; role-theory from Katz (1964); Organizational Citizenship Behavior (OCB) categorized by Organ (1988); and the prosocial organizational behavior (POB) from Brief and Motowildo (1986) and taking into account ideas about ‘sportsmanship’ (Organ, 1988), ‘whistleblowing’ (Near and Miceli, 1987) and ‘organizational courtesy’ (Organ, 1988) they brought the term ‘contextual performance’ to the field. Contextual performance is defined as ‘activities as volunteering to carry out task activities that are not formally part of the job and helping and coopering with others in the organization to get tasks accomplished’(Borman and Motowildo, 1997, p. 100) Dimensions, frequently named by scholars in describing contextual-performance are helping others, proactivity, extra tasks and enthusiasm (Koopmans et all, 2011). Lee (et all, 2010) supports this by mentioning indicators as helping behavior, volunteering for other activities that are not formally part of the job and the involvement towards other employees in the organization. These indicators correspondent with the term ‘extra-role behavior’, what represents behaviors that are not formally required and goes beyond the formal job requirements (Williams and Anderson, 1991). According to Motowildo (2000), the label (extra-role performance, contextual performance or OCB) attached is not particularly

9 important, all three refer to the behavioral part of individual performance. Motowildo, Borban and Schmit (2009) showed that performance is the evaluative form of employee behavior –‘what people do while at work’-. in organizations Based on aesthetic preference for one word over another, as suggested by Motowildo (2000) this research will use the terms ‘in-role performance’ and ‘extra-role performance’.

Motowidlo (et all 1997) showed that in-role and extra-role performance are different in three ways; first, task performance activities are different across jobs, in contrast, extra-role performance activities tend to be similar for different jobs. Second, in-extra-role performance activities are more prescribed than extra-role performance activities and third, in-role performance is associated more with cognitive ability and the antecedents of extra-role performance more with personality variables.

Waldman (1994) argued that both types of employee performance are critical factors for organizational success in the current changing business environment. Both types of individual performance contribute to organizational goal accomplishment, but through two quite different means (Borman and Motowidlo, 1997) . The reason that in-role performance is desirable is either that ‘they help the transformation of raw materials into goods or services, or they directly service the organization’s technical core and improve its capability to produce accordingly’. Reversed, the consequences for the absence of individual in-role performance are reprimands and negative financial outcomes. The positive contribution of extra-role performance is that it ‘maintains or improves the organizational, social or psychological environment necessary for the technical core to function effectively and efficiently’. (Motwoldo et all, 1997, p. 76) Extra-role performance represent behaviors that are not formally required or recognized by rewards systems. Scholars showed that extra-role performance is associated with lower turnover intentions (MacKenzie et all, 2009). On the

10 individual level, extra-role performance can turn into systemic rewards. (Motowildo, 2000) Additionally, according to MacKenzi (et all.,1991) also managers define performance broader than simply in-role performance. Their analysis showed that managers take into account additional factors as extra performance to rate their employees.

Summarizing, both in-role performance and extra-role performance are critical for organizational success. In-role performance is an outcome of behavior that is expected and whereby the formal job-description is met. Extra-role performance is the outcome of behaviors that are not formally required, like helping others, and go beyond the formal job description. But, why are some individuals better performers in both dimensions than others?

Knowledge Sharing

The knowledge-based theory of the firm, originated by Penrose in 1959, argues that the ability to transfer knowledge is crucial in reaching an advantage over competitors (Reagans and McEvily, 2003). The ability to share knowledge within the organization, but also between organization represents a specific source of competitive advantage (Arrow, 1974). Chung (2009) mentioned that individual job performance is affected by a variety of factors; ‘experience, education, keeping abreast of work-related and technological changes, and so on’ (p. 35). However, leaving these factors and individual properties out of context, job performance is, to some extent, the outcome of obtaining the right knowledge for the task at hand (Cross and Cummings, 2004) .

Knowledge sharing represents the knowledge exchange among individuals or for example organizations (Bukowitz and Ruth, 1999). Knowledge sharing is defined as the process of making knowledge accessible to others in the organization (Ipe, 2003). Knowledge transfer also identifies a cost to the origin of knowledge, in terms of the effort and time that need to be spent in order to share the knowledge with others (Hansen, 1999).

11 Motowildo, Borman and Schmit (2009) showed that individual differences in task and contextual performance are partly determined by the knowledge of the individual. This is in line with Cross and Cummings (2004) who argue that job performance is, to some extent, the outcome of obtaining the right knowledge to solve problems. The ability to obtain resources such as information is directly related to individual performance. Intuitively this makes sense as well, in order to perform an employee needs the right information. In order to improve performance, an organization should offer opportunities for mutual learning to facilitate employee knowledge sharing (Reinholt, 2011).

It is widely assumed by scholars that knowledge sharing is the underlying mechanism driving performance outcomes (Reinholt, 2011). High individual performance is associated with successfully adapting to a changing environment, obtaining the right information enhances this adapting process (Cummings, 2004). In knowledge sharing research, scholars mainly mention ‘non-redundant information’ as a key word. This type of information represents new knowledge or project specific knowledge and is seen as valuable information that employees need. (Hansen, 2002) Summarizing, knowledge sharing is crucial for successful individual performance. But what facilitates knowledge sharing among employees?

Social Network Analysis – Relationships.

According to Argote, McEvily and Reagans (2003), social relationships give individuals the opportunity to create, retain and transfer knowledge. Besides, relationships provide also the incentives to share knowledge with others. Supporting this statement, scholars found evidence that knowledge creation and sharing in organizations is critically affected by social networks (Borgatti and Cross, 2003). For social network scholars the raison d’être is that an individual’s social network determines the opportunities that he or she will

12 encounter (Chung, 2009). By analyzing networks, scholars focus on the relationships among the ego and alters (e.g., others, except the ego, in a network) in the system.

Tsai and Ghoshal, (1998) revealed that social networks and performance outcomes are positively associated. Previous scholars showed that in general people prefer to obtain information via other people rather than via documents (Allen, 1977). Other studies showed that relationships in organizations are important for acquiring information (Burt, 1992), learning how to perform a task (Lave and Wenger, 1991), and problem solving (Hutchins, 1991).

Summarizing, the underlying variance for differences between individual’s in-role and extra-role performance is the employees’ possession of the right information. Previous research have inferred an association between a social networks analysis, e.g., analysing the network of for example an individual and the corresponding performance outcomes; whereby knowledge transfer is presumed to be the causal mechanism (Reagans and McEvily, 2003). In order words, in terms of social network theory, relationships are a source of information. But what is the underlying mechanism that explains the differences in knowledge sharing between relationships?

Social Networks core concept – Tie Strength

One way of understanding the mechanism that drives these differences is to focus on the dyadic relationship between actors (Argote et all, 2003). Previous research has shown that dyadic relationships are related to the employees’ access to organizational information (Feeney and Bozeman, 2009). This relationship can be different in several dimensions, including intensity and contact frequency. Social network scholars have focused on these dyadic relational components of an individual’s networks under the name ‘Tie Strength’. Successful knowledge transfer (promoting an individual’s in-role and extra-role performance)

13 can via this concept be predicted by some quality of the dyadic relationship (Chung, 2009). Next to that, the strength of an interpersonal connection can affect the ease of sharing knowledge (Reagans and McEvily, 2003).

In general, scholars make a distinction between ‘strong’ and ‘weak’ ties, both with a significant effect on individual performance, but through different means (Chung, 2009). The strength of a tie is a mixture of three factors; the frequency of contact, reciprocity and friendship (Granovetter, 1973). Strong ties are defined as ‘frequent contacts that have affective and friendly overtones and could possibly include reciprocal actions.’. In contrast, weak ties are less frequent contacts, occasionally there and without affective contents.

When studying tie strength, researchers often start with pointing to Granovetter’s (1973) ‘Strength of Weak Ties’. In its essential, Granovetter (1973) argued that individuals gain new and novel information (non-redundant information) rather from weak ties than from strong ties. Underlying concept in his reasoning is based on the assumption that strong ties are often present between similar people. Similar people tend to cluster and then become all connected. New and novel, non-redundant information, that circulate in these clusters tends to become redundant in a short period of time. Therefore, strong ties, and in particular clusters are not highly sensitive for new information. However, Burt (1992) argued that the information benefits for weak ties are present only in case of structural holes; who provide a relationship between actors divided in a network.

Contradicting the theory around the ‘strength of weak ties’, other scholars noted the importance of strong ties. Krackhardt (1992) concluded that strong ties are important for organizational change, especially in generating trust for the process. In terms of knowledge sharing, Hansen (1999) showed that weak ties enhance project performance when the task simple and strong ties enhance knowledge searching for complex matters. Similar to these

14 findings, McEvily (2003) argued that in performing knowledge-intensive activities there is a positive association between the strength of a tie and knowledge sharing.

Other concepts in terms of tie strength and knowledge sharing that are widely discussed in the literature involve emotional attachment and trust. Studies around the cost of knowledge sharing conclude that the more emotionally involved an ego is with an alter, (e.g., the stronger the relationship) the more time and effort the ego is willing to put into the process of knowledge sharing (Reagan and McEvily, 2003). An additional concept around the cost of knowledge sharing is trust, the more trust is present between the ego and alter (e.g., the stronger the relationship), the more confidence there will be that the shared knowledge is correct and will not be misused. (Krackhardt, 1990). Reagan and McEvily (2003) tested both concepts and concluded that tie strength does ease the knowledge transfer.

Summarizing all the previous findings, Chung (2009) concluded that when knowledge sharing is an important factor for performance, strong ties are important. Although some scholars argue different, ‘the strength of strong ties’ seems to have the overhand in the literature. But, does only the quality of the relationship affects knowledge sharing among individuals?

Type of Network – Friendship and Advice

Knowledge sharing not only varies with the type of tie and it’s characteristics, it also varies with the type of relationship. Gibbon’s (2004) identified two types of relations that are normally present in organizations: ‘the relation between friends and the relation between advisor and advisee’ Together these relationships form an individuals’ friendship network and an individual’s advice network. Both can overlap in organizations (Ibarra, 1992), but the friendship and advice network distinct different functions (Gibbons, 2004). A friendship network is characterized by familiarity and trust enabling open communication and the

15 possibility to change people’s believes. On the other hand the advice network is defined by work-related information transfer and is able to correspond to current activities and the build-up of new organizational norms.

Friends are defined as people with whom you like to spend your free time, people you have been with most often for informal social activities, such as visiting each other’s houses and going to the movies. Friendship between two people emerge if and when their paths cross. Yang and Tang (2003) describe this as ‘meet’ before you can ‘mate’. If people have the same interest and share their social networks it would be more likely that they meet. Friendship is voluntary, democratic, trusting and enduring (Bell, 1981). Lee (et all, 2010) mentioned that informal, friendship networks are often a way for employees to find information. Friends might be willing to provide more information or help to each other (Kilduff, 1992). Moreover, people are more likely to see out for the people they trust when searching for information (Levin and Cross, 2004) . According to Lee (et all, 2010), the friendship network is an important measure of the informal relationships of employees, and the associated hidden power it has in finding work-related information. Friendship networks can be informal, and in organizations employees tend to have informal relationships with people who are close to them (Brass, 1985). This suggests, that ties in friendship networks are per se close, friendship and tie strength are in most cases positively related, however Reagans and McEvily (2003) showed that not all strong ties grasp friends. Not only informal information is shared via friendship networks, also important (task) information can be transported (Burt, 1992) . Altogether , supported by Yang and Tang (2003), a relationship between two friends can provide access to information and promote knowledge transfer.

An advice networks consists of relationships through which employees’ share their resources such as information, assistance and guidance in order to complete a task at hand

16 (Sparrowe, 2001). The focus of an advice network is more instrumental than the more informal friendship network. Advice networks develop over time, and can be fostered by closeness, similarity between tasks or by formal lines of communication. (Sparrowe, 2001). Research towards the characteristics of advice networks mainly showed that centrality, e.g., the position of an individual is in an advice network is positively related to in-role and extra-role performance.

As described, the movement of information between employees depends on the strength of relationships. On the other hand, with respect to the type of network, the focus is more on the type of information , depending on the type of relationship (friendship/advice) . Within the friendship network, information tends to be more informal, but research proved that also important work-flow information can be shared (Lee et all, 2010). Advice networks consist of more formal, work-flow information sharing (Sparrowe, 2011). Cross (2000) showed that in an advice network people mainly provide solutions, meta-knowledge, problem reformulation, validation and legitimation for other people. In the research of Cross et all (2001), finding solutions and the strength of ties was positively associated.

Summarizing, scholars distinguished workplace performance in in-role and extra-role performance. Both, although through different means, contribute to organizational performance. Individual performance is defined as the extent to which the individual contributes, in the eyes of the management, to the organizational goals (Mehra, Kilduff and Brass, 2001). Motowildo (et all, 2009) showed that individual differences in in-role and extra-performance are partly determined by the knowledge of the individual. Additionally, job performance, is to some degree, obtaining the right knowledge in order to execute the task at hand (Cross, 2004). Therefore the current study makes the assumption that the underlying variance that determines the differences in knowledge among individuals is

17 knowledge sharing. Scholars found that social networks are related to knowledge creating and sharing in organizations (Borgatti and Cross, 2003). Social network analyses can help to understand how a given network of people create and share knowledge (Cross et all, 2001). An important aspect of a social network, is the dyadic relationship between the ego and the alter. (Argote et all, 2003). The quality of a relationship, in other words, tie strength, affects the movement of knowledge. Obtaining the right knowledge is not only depends on the quality of ties, also the type of network determines knowledge sharing among individuals. This research distinguishes, in line with Gibbons (2004), an individual’s friendship network and an individual’s advice network.

The current study contributes to the understanding of the underlying variances and mechanisms that drive performance differences among employees. By integrating friendship network and advice network concepts, tie strength and in-role and extra-role performance this thesis tries to explain a small piece of the puzzle towards understanding of performance differences among individuals.

Research goal and hypotheses

The purpose of this study is to consider an association between strength of relationship in a friendship or advice network, on the dyadic level, and in-role performance and extra-performance of an ego. Underlying mechanism in these associations is knowledge sharing, assuming that knowledge sharing is positively related to in-role and extra-role performance.

18 Friendship Network hypotheses

Research that investigated friendship networks and in-role performance provided mixed results. Some scholars showed that friendship increased the social focus, distracting employees from the task (Bramel and Friend, 1987). But for example Shah and Jehn (1993) concluded that groups of friends spend time on discussing non-task related issues (socialization, for example gossip) but also on task-related activities as the planning of tasks and task-discussion. They suggest that friendship networks facilitate an open way of communication, what can enhance in-role performance. Therefore, following Ibarra (1992), in order to get things done in an organization, friendship network are particularly important. Besides, Hansen (1999) revealed that strong friendship ties are good channels for tacit information flow, frequently associated with in-role performance. Therefore this research argues that an ego’s perception of tie strength in a friendship network is positively associated with an ego’s in-role performance.

H1: Tie strength in a friendship network is positively associated with in-role performance

In terms of extra-role performance, Bowler and Brass (2005) found friendship as a key predictor for interpersonal citizen behavior. Additionally, Brass and Burkhardt (1992) found that, since friends are more willing to help each other strong friendship ties positively influence this interpersonal citizen behavior as well. In this study, interpersonal citizen behavior, when employees help others and go beyond their job prescription, is similar to the description of extra-role performance. Assuming that stronger friendship relationships contain lots of informal information; this might enhance an individual’s helping behavior. Therefore this study argues that an ego’s perception of tie strength in a friendship network is positively associated with an ego’s extra-role performance.

19 Advice Network hypotheses.

The information that flows within an ego’s advice network is mostly work-related information. This research assumes that ego’s seek out for advice if their task is somehow complex and the can’t figure it out themselves. Strong ties facilitate the sharing of complex knowledge (Hansen, 1999). Additionally, trust is important indicator considering knowledge sharing and advice network. Stronger ties include more trust. Reagans and McEvily argued that trust and lower cost of knowledge transfer result in easier knowledge sharing. Therefore this research made the suggesting that tie strength in an advice network, as perceived by the ego is positively associated with in-role performance.

H3: Tie strength in an advice network is positively associated with in-role performance

This research expect stronger relations between the advisor and advisee will enhance extra-role performance. The information flow in an advice network is mainly formal. It’s reasonable to assume that strong ties in an advice network are an outcome of finding right solutions in previous interactions. After finding a right solution, the advisee might be willing to help others; e.g.: extra-role performance. There a positive association between the tie strength, as perceived by the ego (advisee) and ego’s extra-role performance is proposed. H4: Tie strength in an advice network is positively associated with extra-role performance

20

III Data and Method

This study considers the association between the strength of a relationship, on the dyadic level, and in-role performance and extra-performance for both a friendship network and an advice network. The following section will describe the methodology and corresponding data used to test the hypotheses.

Research Design

As described in the literature review, analyzing an individual’s social network can make substantial contributions in order to understand variances in, in this case, ego’s in-role and extra-role performance. The network analysis used in this study is called a basic research approach (Borgatti et all, 2013) whereby the variances in in-role and extra-role performance are described as a function of perceived tie strength. The level of analysis is on the dyadic level, from an ego’s perspective. This thesis is deductive in nature, using a top-down approach, by starting with the existing theory (Saunders et all., 2009). A young professional organization, WidgetCo was chosen to collect the data.

The research technique used in this study consisted of two traditional survey questionnaires to collect the required data. The employee survey questionnaire, was used to gather network data and additional information about the employees (ego’s). The supervisor questionnaire measured the supervisors ratings for both the in-role and extra-role performance of the employees. This survey method is chosen in order to gather a large amount of data with standardized variables across the sample. Also, this instrument ensures that the researcher can analyze and interpret the data easily (Saunders et al., 2009).

In order to indicate habits, cultural characteristics and the use of nicknames, an informational interview was held with one supervisor and one employee of WidgetCo before drafting the surveys. After, a trial version of both composed surveys was send to the same

21 employee and supervisor in order to control the clearness of the survey. Some spelling mistakes and system errors were removed where after the final versions of the surveys were made. Both final versions of the surveys can be found in Appendix A (employee survey) and Appendix B (Supervisor Survey).

The employee survey used the egocentric approach for collecting network data. This approach is, according to Chung (2009) practical and feasible. The ego in this approach is the actor of interest; the employee of WidgetCo. The ego’s alters – the other nodes in his/her network- got identified by using the name generator technique. The purpose of this technique is to create a list of distinct names (Chung, 2009). After, the ego can systematically answer questions about the names on the list. An electronic survey instrument named Qualtrics, was used to make and send out the surveys. After the raw was obtained from Qualtrics the researcher replaced all the names by random numbers. The list with the name–random number classification was deleted after the supervisor surveys were processed.

After the employee data was gathered, the supervisor survey was send out to the 6 supervisors of WidgetCo. Supervisors rated only the subordinates they closely worked with. Since some supervisors preferred filling out the survey on hard copy copies of the surveys were hand out to the supervisors. The name of the employees were presented on a list and the supervisor received the same amount of survey copies as names on the list. Other supervisors received no information on the other employees of WidgetCo. After the supervisors completed their rating, the ratings were converted to an excel document and both the employee and the supervisor names got randomized by numbers corresponding with the already mentioned name-random number list.

Since this research tests hypotheses both for a friendship network and an advice network, 2 separate excel documents got prepared. This documents included random

22 numbers, the corresponding network characteristic scores and supervisor ratings per employee. Both anonymized data sets can be requested from the researcher.

Sample

This research is conducted in a young professionals organization in Amsterdam, for privacy matters called ‘WidgetCo’. Inspired by the research of Chung (2009) this company was chosen because of the importance of knowledge sharing. Next to that, WidgetCo uses formal job-prescription and the researcher had easy access to both the employees and supervisors. WigdetCo is a medium large organization in Amsterdam, with 100 active young professionals and 600 alumni, ex active young professionals. To work for WidgetCo, candidates must have a university background. They can sign up where after they got selected on basis of their skills, knowledge and motivation. In small teams, of approximately six persons the young professionals work together for a relatively short period of nine months. Staying in touch with the organization is however consuetudinary and signing up for another project is a possibility. Within the teams, members perform a separated function described by a formal job-prescription. Supervisors coordinate the process of the several teams and have in-depth information on the behavior, knowledge and skills of the young professionals.

The initial target sample of the study was a total of 70 respondents. Due to holidays and study circumstances 59 respondents, using a semi-random sampling technique were approached by electronic email and were presented an online survey in Dutch. 26 completed the 40 minute survey, a response rate of 44%.

A cover letter opened both the survey, explaining the goals of this research (Bryman, 2008). Network research is different from conventional person-based research (Borgatti et all., 2013). A network design requires that the respondents identify themselves. Therefore in order to offer confidentiality the cover letter also made clear that only the researcher had

23 access to the raw data and that all the names in the raw data would get anonymized by random numbers.

It must be acknowledged the chosen sample includes some limitations. First, the sample is derived from a single, young professional organization which result in reduced generalizability. The relatively low response rate also influences the generalizability of the study. Due to the duration time that was needed in order to complete this survey there were a lot employees of WidgetCo that didn’t participate. Young professionals sign up for WigdetCo on voluntary basis so an obligated note from above wasn’t possible, to make sure the research process didn’t delay too much the decision was made to continue with the 26 networks and performance ratings that were available. Furthermore, the data included a big amount of non-participants. These non- participants were named by the respondents but didn’t fill out the survey themselves, for example because they were alumni. In order to overcome this limitation, this research measures tie strength on a dyadic level, from an ego perspective. The level of analysis will elaborate on this.

Level of Analysis

As mentioned before, the level of analysis in this study is on the ‘dyadic level’, from the ego’s perspective. This level of analysis demands some more explanation. Respondents were asked for 10 names representing their friendship network and for 6 names representing their advice network, 26 employees filled out the survey. This resulted in 260 dyadic relations in the friendship network, and 156 dyadic relations in the advice network. With respect to the independent variable; from an ego perspective means that it measures who the ego perceive his relation with the alter. The dependent variable, measured using supervisor ratings, is per respondent in both networks the same. As example: for the 10 relationships in the friendship network of respondent 1, the dependent variable stayed for the 10 relationship the same; the

24 interdependent variable changed per relationship. Graph 1 – Ego #2: Friendship Network & performance outcomes shows the underlying idea. The bold number indicate the tie strength perceived by respondent number 2 in his friendship network for WidgetCo.

Graph 1 – Ego #2: Friendship Network & performance outcomes

Independent Variable

The independent variable is this research is tie strength, e.g., the quality of a dyadic relationship between the ego and the actor in a network. Tie strength is measured on the dyadic level from an ego’s perspective. Inspired by studies as Marsden and Campbell (1984) and Hansen (1992), this study used conventional network measures to measure tie strength as the average of frequency and closeness from the ego’s perspective as reported on a 4-point Likert scale in the employee survey. Table 1 – Questionnaire Items Tie Strength shows the questionnaire items used for the calculation of the means for both the friendship and advice network.

25 Table 1 – Questionnaire Items Tie Strength

Concept Dimensions Question

Tie Strength in Friendship network

Frequency Hoe vaak praat je met mensen over niet MAA- gerelateerde zaken (1x keer per maand of minder – wekelijks – 1 of meer keer per week – dagelijks) Closeness ‘Geef aan hoe close je bent met elk van de

volgende personen’ (Helemaal niet close – niet close – close – erg close)

Concept Dimensions Question

Tie Strength in Advice network

Frequency - Hoe vaak praat je met de volgende mensen over MAA-zaken (1x keer per maand of minder – wekelijks – 1 of meer keer per week – dagelijks) - Hoe vaak ga je voor advies naar de volgende personen (Bijna nooit – soms – regelmatig – vaak) - Hoe vaak komen de volgende personen naar jou toe voor advies (Bijna nooit – soms – regelmatig – vaak)

Closeness Close (Helemaal niet close – niet close – close – erg close)

For the Friendship network, Cronbach’s alpha coefficient for ‘frequency’ and ‘closeness’ indicated 0.770, > 0.700, suggesting acceptable reliability (Pallant, 2007). The correlation coefficient in the comparable study of Hansen (1999) between ‘frequency’ and ‘closeness’ was 0.83, in this study the coefficient showed 0.700 (p = 0.000). Both ‘frequency’ and ‘closeness’ were computed into the variable ‘tie strength friendship network’.

The advice network tie strength measurements for the ‘frequency’ dimension resulted in a Cronbach’s alpha coefficient of 0.714, >0.700, suggesting good internal consistency. The reliability analysis for ‘frequency’ and ‘closeness’ in the advice network reported a Cronbach’s alpha coefficient of 0.770. The correlation coefficient for ‘frequency’ and ‘closeness’ was 0.542 (p = 0.000). Both items were computed into the variable ‘tie strength advice network’ Table 2 provides an outline of the correlation and reliability statistics.

26 Table 2 – Reliability and Correlation statistics for Tie Strength.

Tie Strength = average (Frequency + Closeness)

Network Variable Measurement Cronbach Alpha

Advice Network Frequency N = 3

See table 1 dimensions

0.714 Closeness N = 1

Tie Strength (Correlation = 0.543) 0.703

Friendship Network Frequency N = 1

Closeness N = 1

Tie Strength (correlation = 0.700) 0.770

Dependent Variables

The dependent variables, ego’s in-role performance and extra-role performance were based on supervisors ratings. Supervisors in WigdetCo are relative close involved with the small employee teams, and they perform certain similar tasks on a different level. Therefore, supervisor ratings are a sufficient way to indicate employees’ in-role and extra-role performance. Using a 5-point Likert scale (1 = totally disagree, 5 = totally agree) and items developed by Willliams and Anderson (1991) the survey included 17 items measuring several aspects of in-role performance, extra-role performance. In random order the 17 items pointed to either in-role performance or extra-role performance.

The dependent variable is measured based on 7 items. (Tijd, Taak, Klaag, Kwaliteit, Aanwezig, Op tijd en Rotklus). The first 5 items were ‘positive’ predictors of in-role performance, ‘Klaag (complaining)’ and ‘Rotklus (unpleasant task) were ‘negative’ predictors for in-role performance and were recoded , (1=5, 2 = 4 ,3 = 3, 4=2 and 5 -1) into rKlaag and rRotklus. After, the reliability analysis, Cronbach alpha for the various items showed 0.738 (N = 7) < 0.700, suggesting acceptable reliability. (Pallant, 2007) . Deleting ‘tijd’ and ‘klaag’ would have raised the Cronbach’s alpha if the item got deleted (table 3 –

27 Reliability statistics in-role performance). However, WidgetCo expects, as part of the formal job description that employees are doing their tasks on time (‘Tijd’) and without complaining (‘Klaag’). For that reason, both items were not excluded from the new computed variable: ‘In-Role performance’

Table 3 - Reliability Statistics In-Role Performance Friendship & Advice Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted P_Tijd 23.7308 7.885 0.373 0.733 P_taak 23.8077 7.762 0.664 0.661 rP_klaag 23.7308 9.725 0.153 0.762 P_kwaliteit 23.6538 8.955 0.425 0.715 P_aanwezig 23.6538 8.555 0.365 0.726 P_optijd 23.8462 6.775 0.626 0.658 rP_rotklus 23.5 7.86 0.634 0.668

Extra-role performance is based on 6 items, (betrokken, interesse, hulp, helpen, vertrouwen and carrier). All items were affecting extra-role performance in a positive manner so no items needed to be recoded. Cronbach alpha’s for extra-role performance items

reported 0.7888 ( N= 6), indicating good internal reliablitiy. An overview of the reliability statistics for extra-role performance can be found in table 4 – reliability statistics Extra-Role performance. Together the items for the dependent variable ‘extra-role performance’.

Table 4 - Reliability Statistics Extra-Role Performance Friendship & Advice Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted P_betrokken 19.31 7.342 0.575 0.748 P_interesse 19.27 7.405 0.634 0.731 P_Hulp 19.46 7.698 0.54 0.756 P_helpen 19.54 9.138 0.337 0.798 P_vertrouwen 19.54 7.378 0.636 0.731 P_Carriere 19.62 8.966 0.559 0.762

28 From the original supervisor survey, 4 items (bron, oplossing, overall and actief) were deleted. These items don’t influence the in-role or extra-role performance of the employees from WidgetCo and were mainly included for an eventual creative performance measurement and to build up a logical survey.

Control variables

This study controls for age (age) and for the time that an employee is active at WidgetCo (time). The age-performance relationship varies across samples. For example, Jang, Borenstein, Chiriboga & Mortimer (2005) found that the relationships of age to task performance (in-role performance) and to OCB (extra-role performance) were more positive. For WidgetCo the difference among the age of the employees are not substantial big, however it remains helpful to control for the extra variance that might be explained by the age of an employee. Second, the study controlled for ‘time’, which derived from the informational interview. Young professionals sign up voluntary for WidgetCo, naturally if an employee is performing well there is a bigger chance that they stay at WidgetCo. In the informational interview it got repeatedly mentioned that the longer the employee was active for WidgetCo, the better they performed.

The control variables were measured at the end of the survey. For age, respondents fulfilled the box ‘age’ and for time the scores were measured using a 4point scale. (0 – 6 months, 6 – 12 months, 1 – 2 years, 2 years or longer).

29

IV Results

The following section reports the results of this research. First, the descriptive data for the independent, dependent and control variables are presented . After, a correlation analysis is reported to point out the correlations among the variables. Next, the results of a normality analysis are reported to ensure that the dependent variables, in-role and extra-role performance were normally distributed. Finally, a hierarchical multiple regression analysis was performed in order to test the earlier formulated hypotheses.

Descriptive statistics

For the dependent variable in-role performance, supervisors rated their subordinates with values ranging from 2.62 to 4.75 with a mean of 3.97 and a standard deviation of 0.47 (N = 26). The second dependent variable, extra-role performance, consisted of values ranging from 2.50 to 4.67 with a mean of 3.89 and a standard deviation of 0.54 (N = 26) . The higher the employee’s value, the better his or her in-role or extra-role performance judged by their supervisor.

This study used the age of an employee (age) and the time the employee is employed in WidgetCo (time) as control variables. The participants in this research were between 21 and 25 years old, with a mean of 23,5 years and a standard deviation of 1,09 years. 1 participant was 21 years old, and 5 were 25 years old. The majority of the respondents (10) were 24 years old. With respect to the time the participants are employed in WidgetCo the values range between 1 and 4, with a mean of 1.96, meaning that most participants were active in WidgetCo between 6 – 12 months. Noteworthy, this means that the date that the employee started at WidgetCo differences between 6 and 12 months before the survey was filled out. 4 respondents were active for WidgetCo for 2 years or longer and 6 just started less than 6 months before they filled out the survey. The survey was completed by 10 male

30 respondents (39%) and 16 respondents were female (61%). Gender is not used as control variable.

The independent variable of this research, tie strength was measured on the dyadic level for both the friendship and the advice network of the participating employees of WidgetCo. For the friendship network, 10 relationships per ego were obtained, with respect to the advice network, 6 relationships per ego were obtained. In total, 260 relations were found for the friendship network, and 156 for the advice network.

For the friendship network, the strength of tie values ranged between 1 and 4 with a mean of 2.44 and a standard deviation of 0.77 (N = 260). The higher this value, the stronger the ego perceives his or her relationship with the alter. The tie strength in the advice network, as perceived by the participating ego differed between 1 and 3.83, with a mean of 2.35 and a stand deviation of 0.59. Overall, tie strength as perceived by the ego towards the alter was higher for the friendship network compared to the advice network. Table 5- Descriptive statistics provides an overview of all the descriptive variables.

Table 5 - Descriptive Variables

Variables Friendship Network N Min Max Mean SD

Dependent Variables In-Role Performance 26 2.63 4.75 3.97 0.47

Extra-Role Performance 26 2.5 4.67 3.89 0.54

Control Variables Age 26 21 25 23.54 1.09

Time 26 1 4 1.96

Independent Variables Tie Strenght 260 1 4 2.44 0.77

Variables Advice Network N Min Max Mean SD

Dependent Variables In-Role Performance 26 2.63 4.75 3.97 0.47

Extra-Role Performance 26 2.5 4.67 3.89 0.54

Control Variables Age 26 21 25 23.54 1.09

Time 26 1 4 1.96

31 Correlations

The following part will report the significant results of the bivariate correlation analysis performed to indicate if the variables in this study were correlated. The correlations were made using a Pearson product-moment correlation coefficient. The results of the bivariate correlation analysis for the friendship network are reported in table 6 – Correlations Friendship Network, for the advice network the correlations are outlined in table 7 – Advice Network Correlations.

The only significant correlation for independent variable tie strength in the friendship network is the correlation between friendship network tie strength and the dependent variable extra role performance. (Pearson correlation coefficient, r = 0.171, N = 260, p = 0.006) This slightly positive correlation indicates that an ego’s perception of tie strength in the friendship network is somehow correlated with ego’s extra-role performance in the eyes of the supervisor.

Furthermore, the analysis showed that there was a strong, positive correlation between the dependent variable in-role performance and the dependent variable extra-role performance. (r = 0.850, N = 26, p = 0.000). The dependent variable in-role performance was medium, negatively correlated to the control variable age. (r = -0.315, N = 260, p = 0.000), and the correlation analysis showed a medium, negative correlation between the dependent variable extra-role performance and the control variable age as well. (r = -0.448, N = 260, p = 0.000). In contrast, the dependent variable in-role performance was positively correlated with the control variable time ( r = 0.363, N = 260, p = 0.000). Also the dependent variable extra-role performance showed a positive correlation with the control variable time ( r = 0.305, N = 260, p = 0.000). Due to, perhaps a type-error the Pearson correlation coefficient, between the control variables in the friendship network and advice

32 network slightly differ. The error couldn’t be found in the data, since the difference is really small and it doesn’t affect the significant scores enormously the researcher continued the analysis. But, this medium difference should be acknowledged. For the friendship network, the control variable age was weakly negative correlated with the control variable time (r = -0.217, N = 260, p = 0.000), de values differed slightly for the advice network. ( r= -0.225, N = 156, p = 0.005). Also the correlation between the control variable time with the dependent variable in-role performance slightly differed in the advice network but showed still a medium, positive correlation. ( r = 0.371, N = 156, p = 0.000) The same goes for the dependent variable extra-role performance and age (r = -0.446, N = 156, p = 0.000) and extra-role performance and the control variable time ( r = 0.307, N = 156, p = 0.000).

Table 6 - Correlations - Friendship Network

1 2 3 4 5 Friendship Network Tie Strength - 1 Pearson Correlation 1 Sig. (2-tailed) In Role Performance - 2 Pearson Correlation .093 1 Sig. (2-tailed) .133 Extra Role Performance - 3 Pearson Correlation .171** .850** 1 Sig. (2-tailed) .006 .000 Age - 4 Pearson Correlation .024 -.315** -.448** 1 Sig. (2-tailed) .703 .000 .000

Hoe lang actief - 5

Pearson Correlation

-.001 .363** .305** -.217** 1 Sig. (2-tailed) .984 .000 .000 .000

33

Table 7 – Correlation – Friendship Network

1 2 3 4 5 Advice Network Tie Strenght - 1 Pearson Correlation 1 Sig. (2-tailed) In Role performance - 2 Pearson Correlation .122 1 Sig. (2-tailed) .130 Extra Role performance - 3 Pearson Correlation .111 .850** 1 Sig. (2-tailed) .166 .000 Age - 4 Pearson Correlation .087 -.315** -.446** 1 Sig. (2-tailed) .280 .000 .000

Hoe lang actief - 5

Pearson Correlation

.130 .371** .307** -.225** 1 Sig. (2-tailed) .107 .000 .000 .005

**. Correlation is significant at the 0.01 level (2-tailed).

Normality Analysis

In order to test whether the dependent variables, in-role performance, extra-role performance were normally distributed this study conducted a normality test. The normality descriptive statistics are outlined in table 8 – Normality Analysis statistics.

First, Appendix C – graph 1 shows the imposed normal curve of the in-role performance dependent variables in a histogram. The skewness score for in-role performance showed -.603. A negative skewness score has scores clustered to the right, with the tail extending to the left.. Since the Kurtosis score is positive (1.366) the peak can be explained. For this small sample, the Kolmogorov (K-S) test is added to the study. A significant K-S score (K-S = 0.230, p = 0.000) indicates that the in-role performance data are non-normal. However, looking at the normal-probability plot; there was an acceptable straight line for in-role performance; signaling normal distribution of the in-in-role performance scores. The boxplot Appendix C – graph 1a was used to indicate certain outliers for in role performance. Due to the multiplication of the performance scores more outliers were found than actual

34 present. Since the 5% trimmed mean value (3.9923) was relatively equal to the mean value for in-role performance (3.9692) the outliers were not excluded from the data set. Appendix C – graph 1c shows the normal probability plot of in-role performance.

Second, extra-role performance scores are reported in Appendix C – graph 2. The same procedure was conducted as for the in-role performance. The negative skewness score (-0.846) indicates scores clustered to the right, with the tail extending to the left. A small Kurtosis score (0.071) showed that the data is close to normally distributed. Since the significant K-S (K-S = 0.194, p = 0.000) score is under 0.05 the data cannot be seen as normal distributed according the Kolmogorov test. Although, the normal-probability plot (Appendix C – graph 2c) showed an relatively straight line, indicating normal distribution of the extra-role performance scores. To indicate outliers a boxplot, Appendix C – graph 2a was

35

Table 8 - Normality Analysis statistics

Variable Descriptives Statistic Std. Error

In-Role Performance Mean 3.9692 0.02828

95% Confidence interval for Mean - Lower Bound 3.9135

95% Confidence interval for Mean - Upper Bound 4.0249

5% trimmed mean 3.9923 Median 3.94 Variance 0.208 Std Deviation 0.45604 Minimum 2.63 Maximum 4.75 Range 2.12 Interquartile Range 0.25 Skewness -0.603 0.151 Kurtosis 1.366 0.301

Variable Descriptives Statistic Std. Error

Extra Role Performance Mean 3.8915 0.03379

95% Confidence interval for Mean - Lower Bound 3.825

95% Confidence interval for Mean - Upper Bound 3.9581

5% trimmed mean 3.9235 Median 4 Variance 0.297 Std Deviation 0.54485 Minimum 2.5 Maximum 4.67 Range 2.17 Interquartile Range 0.67 Skewness -0.843 0.151 Kurtosis 0.055 0.301

36 Regression

In order to test the four hypotheses drawn in the first chapter of this study, a hierarchical multiple regression analysis was used. With this analysis this research test whether the independent variable (strength of ties) is associated with the dependent variables (in-role performance and extra-role performance). The independent variable tie-strength was tested for both the friendship and advice network via a linear regression model while controlling for age and time. The presence of these control variables determined a step-wise regression, were the variables were entered in a predetermined order. First, the control variables were included, corresponding to model 1 in the regression tables. After, the independent variable tie strength was added, corresponding to model 2. The results per hypothesis are reported below.

Friendship network

H1: Tie strength in a friendship network is positively associated with in-role performance

The control variables, age and time explained a statistical significant portion of the variance in in-role performance. (F = 30.29, p = 0.000). The two control variables had a significant result. In random order; age (B = -0.248, p = 0.000) and time (B = 0.309, p = 0.000) contributed to in-role performance. This would suggest that age is negatively related to in-role performance and the time employed for WidgetCo is positively affecting the in-role performance.

By adding the tie strength in friendship network, the model explained 20% of the variance for in-role performance, an increase of 1%. However this addition was not significant (F change = 3.18, p = 0.076). The regression coefficient for friendship network tie strength reported no significant result (B = 0.100, p = 0.076). Therefore hypothesis 1 is not supported, table 9 – regression analysis H1 outlines the statistical results.

37

Table 9 - Regression Analysis H1

Dependent variable: In-Role Performance

Friendship network Model 1 Model 2

Control B St Error B St Error

Variable Age -0.248* 0.024 -0.25* 0.024 Time 0.309* 0.037 0.309* 0.037 p = IV Tie Strength 0.1 0.076 Rsquare 0.19 0.2 F change 30.29 3.181 Rsquare change 0.19 0.01 Sign (p) 0 0.076 N 260 260

H2: Tie strength in a friendship network is positively associated with extra-role performance

The control variables age and time explained 24,6% of the variances of extra-role performance. (F = 41.868, p = 0.000). Age was negatively related to the dependent variable (B = -0.400, p = 0.000). The control variable time reported a score that was positively related to extra role performance (0.218, p = 0.000).

After the addition of friendship network tie strength, the model explain an extra 3,3%; resulting in a Rsquare of 27,9%. This result was significant (F Change = 11.621, p = 0.001). Likewise, the standardized regression coefficient demonstrated a significant result (B = 0.181, p = 0.001) . Therefore hypothesis 2 is supported, suggesting that ego’s perceived tie strength in a friendship network is positively associated to the ego’s extra-role performance as judged by the supervisors. Table 10 – Regression analysis H2 shows the statistical results.

38

Table 10 - Regression Analysis H2

Dependent variable: Extra-Role Performance

Friendship network Model 1 Model 2

Control B St Error B St Error

Variable Age -0.400* 0.028 -0.405* 0.027 Time 0.218* 0.043 0.218* 0.042 P = IV Tie Strength 0.181 0.001 Rsquare 0.246 0.279 F change 41.868 11.621 Rsquare change 0.246 0.033 Sign (p) 0.000 0.001 N 260 260 Advice network

H3: Tie strength in an advice network is positively associated with in-role performance

For the advice network, the control variables age and time explained 19,4% of the variances in the dependent variable in-role performance (F = 18.447, p = 0.000). A brief analysis resulted in age being negatively related to in-role performance (B = -.0244, p = 0.001) and the regression coefficient suggested time as being positively related to in-role performance (B = 0.316, p = 0.000).

After this first step, independent variable, tie strength in the advice network, was added to the model. Together the control variables and the independent variable explained 20,5 % of the variances of in-role performance, however this result was not statistically significant (F change = 2.058, p = 0.153). Moreover, the regression coefficient report a small positive, but non-significant score as well (B = 0.105, p = 0.153). Therefore hypothesis 3 is not supported. An overview can be found in table 11 – regression analysis H3