Does team diversity predict Leader Member Exchange
Differentiation?
Msc. In Business Administration, track: Leadership and Management
Master Thesis – Master’s Thesis Leadership and Management (6314M0253Y) Graduate School of Business (Amsterdam Business School)
University of Amsterdam
Supervisor: Mw. dr. C.K. (Claudia) Buengeler
Author (student number): Mick Spruijt (10872639)
2
Statement of originality
This document is written by Student Mick Spruijt 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
3
Content
Statement of originality ... 2
Abstract ... 5
Introduction ... 6
Theoretical background and Hypotheses ... 10
LMX differentiation in a team context ... 10
Team diversity and LMX differentiation ... 13
The mediation role of similarity attraction ... 16
The mediation role of constraints for the leader ... 16
The mediation role of personal needs ... 17
Method ... 18 Sample ... 18 Resulting sample ... 19 Measures ... 20 Dependent variables ... 21 Independent variables ... 22 Mediator variables ... 22 Control variables ... 24 Statistical procedure ... 24 Basic preparations ... 25
Justifying for aggregation ... 25
Calculating diversity ... 27
Correlation analysis ... 28
Direct regressions ... 28
Control variables ... 28
Testing for mediation ... 31
Results ... 31
Correlation analysis ... 31
Direct effects ... 35
Mediation effects ... 39
Mediation by the procedures of Baron and Kenny (1986) ... 39
Mediation by PROCESS ... 41
Conclusion regarding the mediation effect ... 44
4
Theoretical and practical implications ... 44
Future research ... 45
Limitations ... 45
Conclusion ... 47
References ... 48
Appendix 1 – Remaining variables ... 56
Appendix 2a – Leader questionnaire ... 60
5
Abstract
This research seeks to illuminate the general relationship between ‘team diversity’ and ‘differentiated leadership’, by examining if diversity within a team leads to differentiated leadership by the leader, and in particular if diversity in personality traits among team members lead to LMX differentiation. And potential underlying mechanisms, in the form of mediators are investigated, to know: (1) diversity in personal needs among team members, (2) constraints for the leader in building equal LMX among its team members, and (3) diversity in perceived similarity between the leader and its team members.
According to the current latest theoretical and empirical work on LMX and LMX differentiation there is argued that many characteristics, of followers and leaders, could serve as antecedents of LMX (differentiation) at the dyadic level. But no study has specific examined the antecedents of LMX differentiation at the group level.
Data is gathered through a survey (questionnaire). The final data set consist of in total 68 teams, which consist for at least of one leader and two team members. Data is gathered on a divers set of variables, but the main variables are the five personality traits: openness, neuroticism, extraversion, agreeableness, and conscientiousness among team members. The diversity in on each of those variables is used as independent variables for predicting LMX differentiation, besides that mediator variables have been included to find the potential underlying mechanisms.
The final results does find support that diversity in openness- and diversity in neuroticism among team members does have predicting power in LMX differentiation (at significant level .10). Additionally, there is found that ‘constraints for the leader in building equal LMX among its team members’ mediates the relationship between diversity in openness among team members and LMX differentiation (at significant level .10). But this mediation effect is does not have a significant value at the normality test (Sobel-test), and therefore ultimately there is not found support for any mediation effect.
6
Introduction
In this research I seek to illuminate the general relationship between ‘team diversity’ and
‘differentiated leadership’, by examining if diversity within a team leads to differentiated
leadership by the leader, and in particular if diversity in personality traits among team
members lead to differentiated leadership. In other words, does diversity in personality traits
among team members results in that the leader will differentiate among his team members?
Here, differentiated leadership denotes ‘Leader Member Exchange differentiation’.
LMX differentiation is the extent to which a leader of a team builds different, high or low
quality, exchange relationships with the members of the team he or she leads (e.g.,
Henderson, Liden, Glibkowski, & Chaudhry, 2009; Liden, Erdogan, Wayne & Sparrowe,
2006). Research has shown that LMX differentiation has mixed effects on relevant outcomes
at various levels within an organization (e.g. Henderson et al., 2009; Sui, Wang, Kirkman, & Li, 2015). Whereas some research has shown that differentiation can benefit performance, others studies have revealed detrimental effects of LMX differentiation.
Antecedents can be found at several levels within an organization. Henderson et al.
(2009) proposed the following levels for antecedents of LMX differentiation: individual,
team, and organizational-level antecedents. I have chosen to investigate the antecedents
originating from the team level because modern organizations’ strong reliance on teams has
provoked increased interest in team- and related mechanisms and team characteristics have
been found to impact on important organizational phenomena, including leadership (e.g.
Boies & Howell, 2006; Henderson et al., 2009; Li & Liao, 2014). The notion of the increased
interest in teams is consistent with Kozlowski and Bell (2003) who have argued that there is a
remarkable transformation of organizational structures worldwide and highlighted that there is
an ongoing shift from individual work to team-based work structures (Ledford, Lawler, &
7
Teams are increasingly characterized by high levels of diversity within teams (e.g.
Van Knippenberg & Schippers, 2007; Kearney & Gebert, 2009). This fact drives the
considerable research that has focused on team composition (Kozlowski & Bell, 2003). As
Kozlowski and Bell (2003) argue, team composition deserves research interest because
combinations of member attributes can have powerful effects on relevant outcomes (e.g. team
processes and outcomes). A better understanding of such effects could thus help practitioners
to select and construct more effective teams. By understanding the diversity effects better, it is
more likely that leaders could anticipate on the effects of a divers team and can apply certain
levels of LMX differentiation, which should result in effectively and efficiently managing a
team. This is underlined by Harrison, Price, Gavin, and Florey (2002), who argue that
managing diverse teams is difficult and a pressing challenge in modern organizations.
Harrison and Klein (2007, p. 2000) define diversity as ‘the distribution of differences
among the members of a unit with respect to a common attribute’. Van Knippenberg, De
Dreu, and Homan (2004) mentioned that “diversity refers to differences between individuals
on any attribute that may lead to the perception that another person is different from self (e.g.
Jackson,1992; Triandis et al., 1994; Williams & O’Reilly, 1998)”. There are two main
distinctions in forms of diversity, namely ‘surface-level’ diversity and ‘deep-level’ diversity
(Harrison, Price, & Bell, 1998). Surface-level diversity refers to diversity which could be
observed easily, like demographics (i.e. age, sex, etc.). Deep-level diversity refers to diversity
which is not (easily) observable such as a person’s attitudes, beliefs and values.
Harrison et al. (1998) argue that deep level diversity is not readily detectable, but
becomes visible over time through member interactions. In their study they found evidence
that deep-level differences between people, in comparison to surface-level differences,
becomes more influential over time. This is because when people interact they get to know
8
effect and people starts to judge people on their real being (deep-level) (Amir, 1976).
Personality traits (deep level) influence people’s behaviour. In teams, differences in
personality can mean differences in how people behave and interact, with potentially powerful
consequences on outcomes. Not surprisingly, it is suggested that the personality composition
within a team has important implications for team performance (e.g. Lievens, Chasteen, Day,
& Christiansen, 2006; Kozlowski & Bell, 2003).
I argue that it also influences how leaders interact with the team, for example when a
team is divers in extraversion (e.g. one extravert person and one introvert person) it would be
logic that the leader should manage each one (slightly) different and do build up a relationship
differently (thus LMX differentiation). Within this line of reasoning, it is shown that both
individual personality traits of the leader and of the followers (team members) does have its
influence on the perceptions of the LMX (Bernerth, Armenakis, Feild, Giles, & Walker,
2007). Within this context, according to Schyns (2015), does some leaders are better able to
develop good quality LMX relationships among its followers and does others have more
difficulties building good quality LMX if there are more diversities in personality traits
among his or her followers. Additionally, Kozlowski and Ilgen (2006) argue that leaders are
key players in enhancing team effectiveness. This argument is understandable, because the
leader is one of the two players in the interaction between the leader and his or her team
members, and being most of the time responsible for achieving their common (organizational)
goal.
Besides of the direct effects of team personality diversity to LMX differentiation, I am
interested in potential underlying mediations mechanisms between this relationship. Three
potential mediators are investigated.
The first mediation effect that is investigated is regarding the diversity in perceived
9
people do perceive similarities among each other and do share beliefs, values, and have more
interaction with people who are perceived as similar. Therefore diversity in personality traits
among team leaders could be a reason why leader do perceive different similarities, and
therefore interact and build (LMX) relationships unequally.
The second mediation effect is expected by constraints for the leader to build up and
sustain equal LMX with all his or her team members. Graen and Uhl-Bien (1995) argue that a
leader does not have the time, effort and/or resources to create and sustain equal relationships
with all team member. As every personality trait does have its own characteristics within a
group context (Bernerth et al., 2007), it would be logical that when a team is high on diversity
in personality traits among its team members, that a leader should manage its resources
critically to give every as equal as possible their resources.
The third and latest investigated potential mediation effect is diversity in (work
related) personal needs among the team members. Le Blance & Gonzales-Roma (2012) does
underline this in their research that dissimilarities among team members regarding work
values is a potential antecedent of LMX differentiation. When there is a team which contains
of members who are different in personality trait to each other, it is plausible that these
members also do have different personal needs in their work. This potential mediator is
therefore investigated in this research.
By conducting this research on team diversity as antecedent of LMX differentiation, I
aim to increase insights into team personality diversity as an important reason for ‘why’
leaders use LMX differentiation. If there is more insight into which team diversity variables
have influences on LMX differentiation, it is more likely that we can more adequately act
on/control certain situations where LMX differentiation is desired or not. Think for example
10
but also on their personality traits. Or a leader who through insights in the diversity among the
personality traits of team member can more adequately manage his or her team.
By examining team personality diversity in relation to LMX differentiation, I seek to
illuminate the relevance of this potential antecedent of LMX differentiation. Furthermore I try
to show potential underlying mechanisms through which team personality diversity leads to
LMX differentiation. This research should help the academic field move forward in
discovering the underlying mechanism of our daily lives. This research is also a direct
contribution to the practical field of leadership, in particular regarding LMX differentiation,
and is therefore a minor step in creating guidelines for leading teams effectively and
efficiently. This should ultimately serves the interest of all (i.e., leaders, followers,
stakeholders, and ultimately whole society), by reducing the negative consequences of LMX
differentiation through more and grounded understanding in the effects of team diversity.
Research question: ‘Does team personality diversity predict LMX differentiation?’
Theoretical background and Hypotheses
Before elaborating on why team diversity should lead to LMX differentiation, and giving the
hypotheses, a theoretical background of the main concepts is given.
LMX differentiation in a team context
Differentiated leadership occurs when a leader uses different degrees of a certain leadership
behaviour when leading the members of his or her team, or develop relationships of differing
quality with the various members of his or her team (Wu, Tsui, & Kinicki, 2010). The most
commonly studied form of leadership differentiation is Leader Member Exchange (LMX)
11
LMX refers to the unique interpersonal (dyadic) exchange relationship between a
manager (as leader) and his or her follower (as member) (Graen & Uhl-Bien, 1995). High
LMX differentiation results when the leader of a team establishes different relation qualities
with their team members, contrary when leaders establish equal relationship qualities with
their team members, LMX differentiation is characterized as low(e.g. Henderson et al., 2009;
Liden et al., 2006). According to Epitropaki, Kapoutsis, Ferris, Drivas, and Ntorsi (2014) it
showed that LMX differentiation does have influence on several employee outcomes, such as
performance, satisfaction and interpersonal deviance. Therefore LMX, and LMX
differentiation, are important determinant in managing a team effective and efficient.
Managing teams, especially those that are diverse in their composition, is one of the
most difficult and pressing challenges in modern organizations (Harrison et al., 2002). But
what is a team within this context? According to Kozlowski and Ilgen (2006, p. 79), teams
can be defined as ‘(a) two or more individuals who (b) socially interact (face-to-face or,
increasingly, virtually); (c) possess one or more common goals; (d) are brought together to
perform organizationally relevant tasks; (e) exhibit interdependencies with respect to
workflow, goals, and outcomes; (f) have different roles and responsibilities; and (g) are
together embedded in an encompassing organizational system, with boundaries and linkages
to the broader system context and task environment (Alderfer, 1977; Argote & McGrath,
1993; Hackman, 1992; Hollenbeck et al., 1995; Kozlowski & Bell, 2003; Kozlowski, Gully,
McHugh, Salas, & Cannon-Bowers, 1996; Kozlowski et al., 1999; Salas, Dickinson,
Converse, & Tannenbaum, 1992)’. Teams are viewed as complex, adaptive, dynamic systems
within organizations (Ilgen, 1999). Kozlowski and Ilgen (2006) argue that leaders are the key
players in enhancing team effectiveness. LMX has been argued to be an effective form of
leadership not only of individuals, but also of teams (Boies & Howell, 2006). But what
12
According to Erdogan and Bauer (2010), LMX differentiation has the potential to
influence employee attitudes, interactions among coworkers, and the level of attachment of an
individual to one’s group (Martin, Epitropaki, Thomas, & Topakas, 2010). Empirical research
shows that the effects of LMX differentiation on outcomes such as Organizational Citizenship
Behaviour (OCB) and turnover intentions are mixed (Harris, Li, & Kirkman, 2014). Hooper
and Martin (2008) mentioned that individuals in teams can be very sensitive to social
comparison information (Festinger, 1954), which can influence employee reactions by
affecting the self-concept and perceptions of fairness and several authors have suggested that
inequalities in LMX distribution may negatively affect relations among team members (e.g.,
Graen & Uhl-Bien, 1995; Liden, Sparrowe & Wayne, 1997; Sias & Jablin, 1995). However,
as Erdogan and Bauer (2010) have argued, team members do not necessarily find
differentiation unacceptable and mentioned that it could be acceptable and even expected,
given that member contributions, loyalty, and interest in developing the relationship further
will vary (e.g., Erdogan & Liden, 2002; Scandura, 1999; Scandura & Lankau, 1996; Stewart
& Johnson, 2009). In conclusion, LMX differentiation can have both positive and negative
effects on team performance depending on the contextual conditions present (Sui et al., 2015).
What is less understood however is where LMX differentiation comes from, thus what
are the antecedents of LMX differentiation (Chen, He, & Weng, 2015). This is worthwhile to
know because, as described here above, LMX differentiation can have positive and negative
effects, and therefore it is relevant to predict its emergence. This to be able to adequately deal
with the negative consequences of LMX differentiation, or even be able to avoid negative
consequences. One of the potential antecedents for LMX differentiation lies at the team level
(Henderson et al., 2009), in the upcoming section there is explained how/why team
13
Team diversity and LMX differentiation
According to Chen, He and Weng (2015), there is little research that has examined the
antecedents of LMX differentiation. Referring to Chen, Yu, and Son (2014), these authors
indicate that it is necessary to tackle this limitation because employees pay attention to the
grounds on which leaders treat employees differently, which could influence the effects
differentiation will have. Additionally, Henderson et al. (2009) argue that current attempts to
identify the antecedents of LMX differentiation is lacking. In a recent study, Chen et al.
(2015) mentioned that theoretical and empirical work on LMX (Dienesch & Liden, 1986;
Graen & Scandura, 1987; Dulebohn, Bommer, Liden, Brouer, & Ferris, 2012; Gerstner &
Day, 1997) has indicated that many characteristics of followers and leaders could serve as
antecedents of LMX at the dyadic level. But no study has examined the antecedents of LMX
differentiation at the group level.
Following this call for research on the antecedents of LMX differentiation, I aim to
examining diversity as group-level antecedent of LMX differentiation. Particularly, the
diversity in personality traits among team member as antecedent for LMX differentiation.
Naturally, not every person is identical and therefore it is most likely that there will be,
at least some, diversity among members of a team. Focusing on teams in particular, Van
Knippenberg and Schippers (2007) define team diversity as the varied perspectives and
approaches to work on the part of individuals from different identity groups.
As mentioned by Kozlowski and Bell (2003) it is relevant for research and practice to
understand the effects of a team composition, because the composition of teams in form of the
combinations of member attributes can have powerful effects on team outcomes. Deep-level
diversity such as of personality has the characteristics of being not readily detectable, but
14
arguable that deep level diversity over time will become visible and therefore becomes a form
of surface level diversity (Harrison et al., 1998).
In the context of research on personality composition of teams, the Five-factor model
of personality (Big five personality factors) figures prominently (Bernerth et al., 2007). The
Five-factor model illustrates that personality (deep level) consists of five relatively
independent dimensions which provide a meaningful taxonomy for studying individual
differences (Barrick & Mount, 1991).
It is important to pay attention to personality traits, because they determine the way in
which people behave (Lievens et al., 2006). The Big-Five personality traits are relatively
stable among working age adults, in other words, the personality traits of an adult individual
do typically not change a lot over time (Cobb-Clark & Schurer 2012). Kozlowski and Bell
(2003) suggest that having members with different personalities (and thus high personality
diversity) in a team has important implications for team performance.
The Five-factor model consists of the following traits: Extraversion, Emotional
Stability, Agreeableness, Conscientiousness, and Openness to Experience (Judge & Bono,
2000). Extraversion is associated with traits as socially oriented (outgoing), active
(adventurous and assertive), dominant and ambitious (Judge, Higgins, Thoresen & Barrick,
1999). Neuroticism (as “opposite” of Emotional stability) refers to a lack of positive
psychological adjustment and emotional stability (Judge et al., 1999). People who score high
on agreeableness are cooperative and likeable. They are more likely to trust other people and
care for them, and they are seen as cheerful and gentle (Judge et al., 1999). Conscientiousness
is related to self-control. People who are very conscientious are hardworking,
15
openness to experience is related to people’s tendency to be original, imaginative, to have
broad interests and to be daring (Costa & McCrae’s, 1992).
Based the finding in role theory it is suggested that personal characteristics are likely
to influence the dyadic (leader – follower) interaction (Kahn, Wolfe, Quinn, Snoek, &
Rosenthal, 1964). With this foundation, LMX has examined a variety of personal
characteristics as antecedents for the development of LMX (Nahrgang & Seo, 2015). These
researchers use the term personal characteristics as a sort of general/broader term for
personality. But researchers as Schyns (2015), Bernerth et al. (2007), Kamdar and Van Dyne
(2007) does literally investigated to what s certain personality traits does have for
consequences in the creation of LMX. All these research do conclude that specific personality
traits does have specific outcomes on the development of LMX, for example Kamdar and Van
Dyne (2007) have found evidence that conscientiousness (on an individual level) is related to
the LMX quality.
Based on these previous researches it shown that personality traits does have influence
on the development in LMX, and therefore I argue that this would also be the case for LMX
differentiation. Therefore I argue that diversity in personality traits among team members
leads to LMX differentiation, and therefore I hypothesize the following:
H1: Team personality diversity positively relates to LMX differentiation.
16 The mediation role of similarity attraction
According to Stewart and Johnson (2009), similarity-attraction is about similarity with regard
to salient or contextually relevant demographic group memberships has been shown to foster
assumptions of competence and of shared values and goals (Byrne, 1971; Riordan, 2000).
Ford and Seers (2006) argue, based on the organizational demography literature (e.g.,
Kanter, 1977; Pfeffer, 1983), that similarity in demographic characteristics is likely to mean
that individuals share beliefs and values, which results in increased attraction and interaction.
Individuals tend to compare themselves with others around them to determine whether or not
they share similar characteristics (Tsui & O'Reilly, 1989), and that similarity or dissimilarity
is thought to have a significant impact on the extent to which the individual identifies with the
group (Ashforth & Mael, 1989). Similarity in attitudes also lead to interpersonal interaction
when it is seen within the exchange theory framework (Thibaut & Kelley, 1959). Furthermore
attitudinal similarity seems to facilitate communication, and communication on the job
increases (Tsui & O'Reilly, 1989).
Within diverse teams it is most likely that there are members who have more in
similarity with the leader than others. Consequently it would be conceivable that a leader
“voluntarily” shares more with members who have more similar personality characteristics,
and therefore building different LMX relationship among team members. Therefore I
hypothesized the following:
H2: The positive relationship between team personality diversity and LMX differentiation is
mediated by diversity in similarity attraction.
The mediation role of constraints for the leader
It could be that leaders are “forced”, due their resources (time, effort, and resources)
constraints to sustain and maintain the same LMX relationship among team members (Graen
17
Conlon (2010) argue that organizational factors could act as constraints for exchanges
between the leader and his or her team members. Henderson et al. (2009) underline these form
of constraints and argue that the amount of resources available for a leader does have its
influences on LMX differentiation.
As every personality trait does have its own characteristics within a group context
(Bernerth et al., 2007), it would be logical that when a team is high on diversity in personality
traits among its team members, that a leader should manage an every team member in another
way to get the best out an individual team member and that a specific resource could be
“spend” only ones and therefore leads to LMX differentiation. Therefore I hypothesize the
following:
H3: The positive relationship between team personality diversity and LMX differentiation is
mediated by constraints for the leader.
The mediation role of personal needs
Diverse teams have different personal needs among team members, it is most likely that not
every team member personal needs exactly the same amount of empowerment, mentoring,
and emotional support to deliver performance and to develop him- or herself (Henderson et
al., 2009). Differences in personality traits will most probably lead to different personal needs
among people. For example a team consisting of members who all score differently on
extraversion having a training on ‘cold calling’, a telephone call soliciting business made directly to a potential customer without prior contact. It will be most likely that a member high on extraversion does not have any issues with ‘cold calling’, but someone low on extraversion will most likely loathe ‘cold calling’. Consequently, leaders will need to provide different levels of support, in
this case more support, to the member low on extraversion. Thereby a different LMX
relationship takes place among the leader and the various members of his or her team,
Gonzales-18
Roma (2012) does underline that dissimilarities among team members regarding work values
is a potential antecedent of LMX differentiation.
Based on the abovementioned line of reasoning I hypnotized the following:
H4: The positive relationship between team personality diversity and LMX differentiation is
mediated by diversity in personal needs.
Method
The data for this thesis research is gathered through a survey study in the context of a larger
data collection project. Two questionnaires were composed based on available measures, one
for team members and one for team leaders. As entry requirement, at least two team members
and their leaders needed to participate. Both questionnaires measuring other variables, this to
see certain constructs from the perspective of the leader and / or the perspective of the team
members. Only the questions regarding the personality traits and (most) demographic
questions are in both questionnaires.
Sample
Data were collected by means of an online survey, through the online software
program of Qualtrics (freely available for UvA students). Survey administration started on
April 25th 2016 and was closed six weeks later on June 6th 2016. All participants did fill in
the digital questionnaire and could only move on in answering the survey if they completed
all questions (by the Qualtrics software) to avoid incomplete data.
Together with my data collection partner we aimed to collect data from 80 teams.
Every team in the world had the possibility to be included in this thesis, but it is was on
forehand already most likely that we collect our data by convenience and volunteer sampling
Zuid-19
Holland, and Utrecht. These provinces are part of the ‘de Randstad’, the so called economic
engine of the Netherlands where enough potential respondent companies and associated work
teams are located (Rijksoverheid, 2011).
During the acquisition of teams, who were necessary for generating our data-set, it
seemed more difficult than expected to find participants who are suitable and willing to
participate. There were many reasons why potential teams were not suitable to participate, for
example within my search for teams one team did not get approval of higher management
(they did not respond to the request) to participate, and there where for around five potential
respondents who were a ‘self managed team’ and therefore had no “true leader” who gives
direction and facilitates all team members.
Resulting sample
The resulting sample consists of Dutch employees who working in a team within their
organization. The final sample consist of 68 teams clustered in 61 companies. Of those teams
42 consist of a leader with two members (61.8%), 11 teams consist of a leader and three
members (16.2%), seven teams consist of a leader and four members (10.3%), five teams
consist of a leader and five members (7.4%), and three teams consist of a leader and six
members (4.4%). Which ultimately results in a total of 256 respondents, 68 team leaders and
188 team members.
The majority of leaders is male (66.2%) and also at the team members level males are
more presented (53.7%). The average age of our team member sample is 37.4 years and of our
team leader sample is 43.8 years, resulting in an overall average of 39.1 years. This is below
the Dutch national age level of working people, 42.2 years (CBS Statline, 2016). Our sample
consist of employees who are, in comparison with Dutch average (Bierings, 2013), highly
20
‘university of applied science’ (HBO) or at a ‘university of science’ (WO). Of the leaders 58
are highly educated (85.3%) and of the members 130 are highly educated (69.5%).
Most of the 68 leaders are working on a fulltime basis (86.8%) and the majority of the
188 members work on a fulltime basis (56.4%). Our data set consists of a workforce which
has a high average tenure of almost nine years (8.8 years) within their organization in
comparison with Dutch trends (Smits & De Vries, 2013). But the most frequent length is
around one year of employment (45 times of all 256 respondents). The average length of
being part of their particular team is 4.7 years, but again the most frequent duration within the
team is around one year (52 times of all respondents).
Measures
Most of the items, used in both questionnaire (team leader, and team member survey), were
derived from previous studies in English-speaking settings. To make sure that we could gather
as much data as possible, we translated the items for which no translation was available, into
Dutch. This was done by the back-translation method, as described by Brislin (1970). To
make sure that the content of the items remains the same, the translated Dutch items were
back translated into English by Anne Diederen (master student colleague) and items which
were “content wise” changed were discussed and adapted together, and this under supervision
of dr. C.K. Buengeler. Therefore we corrected the small number of discrepancies between the
back-translated and the original items, to make sure we have a reliable, valid and
understandable questionnaire. There are a lot of variables who are characterized as control
variables, this is because we (Anne Diederen) both have another perspective on the overall
topic of this research (differentiated leadership). And therefore we both want to control for
different variables in testing our hypotheses, which lead to the relatively high amount of
21
The two questionnaires mainly measure other variables, only personality traits and
most demographic questions are in both questionnaires. In the following paragraphs the
measurements of the variables are explained, beginning with the dependent variables, then the
independent variables, thereafter the mediator variables, and ending with the control
variables.
Dependent variables
LMX. Leader-member exchange was measured by, a Dutch translated scale of Breukelen,
Konst, and Van der Vlist (2002), the LMX-7 scale of Graen and Uhl-Bien (1995). It scores a
Cronbach’s alpha of .796. This scale has seven items which stresses the leader’s contribution
to the exchange relationship and the overall quality of the working relationship, seen from the
perspective of the memver (Liden & Maslyn, 1998). All items were indicative for a “good”
LMX and measured on a 5 point Likert-scale (with “each time” different answer options). An
example item is: “How would you characterize your working relationship with your
supervisor?”.
LMX differentiation, measured by the team members. To measure LMX
differentiation, adapted versions, of the abovementioned (Dutch) LMX items, are created and
used. It has a Cronbach’s alpha’s score of .862. The measurement consists of seven items and
assesses the perception of (directly visual) differentiation used by a leader in building
relationships with other team members. An example item is “My leader is more satisfied for
some team members their delivered work, than the work delivered by another team
members”. All items were indicative for seeing the leader differentiating among team
members. A 5 point Likert-scale ranging from 1 (strongly disagree) to 5 (strongly agree) was
used.
LMX differentiation, measured by the leader. LMX differentiation was measured via
22
versions, of the abovementioned (Dutch) LMX items. It has found a Cronbach’s alpha score
of .745. The measure consists of seven items and assesses the perception of the leader in using
differentiation at building relationships with team members. An example item is “I have with
some team members a better (working)relationship than with other team members”. All items
were indicative for differentiating among team members. A 5 point Likert-scale ranging from
1 (strongly disagree) to 5 (strongly agree) was used.
Independent variables
Personality traits. A measurement scale developed by Costa and McCrae (1985; 1992) was
used to measure the personality traits of an individual person. With their measurement, NEO
Personality Inventory-Reversed (NEO PI-R), the Big Five personality factors were assessed.
Extraversion, Neuroticism (Emotional Stability), Conscientiousness, Openness and
Agreeableness each personality factor were measured using 12 items, thus in total 60 items.
Respondents answered these questions on a 5 point Likert-scale ranging from 1 (strongly
disagree) to 5 (strongly agree), and there were various reversed items (27 spread among the
personality factors). Examples of items included are “I like to have many people around me”
for extraversion; “When things go wrong, I often get discouraged and feel like giving up” for
Neuroticism; “If I promise something, people can count on it that I do what I promised” for
conscientiousness; “I rather work together with others than to compete with them” for
agreeableness, and “When I read a poem or look at a piece of art I sometimes feel a cold
shiver or excitement” for Openness.
Mediator variables
Perceived similarity. There is captured to what extent a leader perceived similarities between
him/her and their team members, this to examine if there is a basis for similarity attraction.
23
Cronbach’s alpha is .875). A 5-point Likert-scale, ranging from 1 (strongly disagree) to 5
(strongly agree) was used.
Constraints for the leader. This variable was included to capture potential constraints
on leaders for building equally high LMX relationships with all team members. To measure
these constraints we combined and adapted items which have been used in three different
studies. Four items were used which have been used by Steffensmeier (2008), with an
example item: “please indicate how often you experience the following in your job: too much
paperwork and reporting requirements”. Three items were found in a study on ‘time urgency’
by Landy, Rastegary, Thayer, and Colvin (1991). An example item is: “please indicate how
often you experience the following in your job: I often feel pressed for time”. And ultimately
four items were created based on the study of Van Breukelen, Konst, and Van der Vlist
(2002). Together this variable measurement scores a Cronbach’s alpha of .825. In their study
they mentioned that leader distribute resources (among others, information, support, feedback,
and attention) differently among their employees. With this as basis, four items were created
(thus no history on internal consistency, Cronbach’s alpha score), an example item is: “please
indicate how often you experience the following in your job: I am not able to provide every
team member the same amount (and/or quality) of feedback”. All items were measured on a 5
point Likert-scale, ranging from 1 (never) to 5 (always).
Personal needs. Personal needs refers to an employee’s emotional feelings and needs
at the work floor (La Guardia, Ryan, Couchman & Deci, 2000). To measure this variable, La
Guardia et al. (2000) scale (8 items) for ‘need satisfaction’ was adapted and used (Cronbach’s
alpha’s ranging from .85 to .94 per item). An example item is “At work, I feel like a
competent person”. Two of the eight items were reverse coded, meaning that a relatively low
score on those items refers to relatively high “needs”. The measurement was conducted by
24 Control variables
Here below are the control variables explained which have been used during this research. In
the ‘statistical procedures’ section is explained why these control variables have been used.
But as already explained, the questionnaire have been measured more variables than only the
used variables. These measurements of those “remaining” variables could by find in,
‘Appendix 1 – Remaining variables’.
Team size. Team size is the amount of members who did participate in this research,
and is calculated through the amount of completed survey’s with the same team code. This
amount of people could therefore differ from the total amount people who are part of the
team, or the amount of people the leader do manage.
Task interdependence. To measure the task interdependence among team members.
An adopted version of Pearce and Gregersen (1991) was used (Cronbach’s alpha = .76). The
measure consists of five items and assesses the leader his perception of the task
interdependence among his or her team members. An example item is: “Their work requires
team members to consult with others fairly frequently”. All items were indicative for seeing
the team member being interdependence to each other. A 5 point Likert-scale ranging from 1
(strongly disagree) to 5 (strongly agree) was used.
Team tenure. This is the duration that a team member is part of the team. This is
questioned by one open question, ‘how long are you be member of this team?’
Age. This is the age of the team member. And is asked by the following open
question, ‘how old are you?’.
Statistical procedure
To perform the statistical analyses, the Statistical software Package for Social Sciences
(SPSS) was used and a data “dump” out of Qualtrics have been imported. Before being able to
25 Basic preparations
First, the counter-indicative items were recoded, to make sure that the items are measuring
with the same “yardstick”. Second, a reliability analysis was done to assess if the items
measured the same construct and to check is they all meet the required standards of having a
Cronbach’s Alpha score of at least > 0.70 (Cronbach’s Alpha scores could be found at the
measurement section). Third, all corresponding items (e.g. all items concerning ‘leader
conscientiousness’) were bundled into one variable, the mean variable. The fourth data
preparation stage was testing for kurtosis and skewness, to check for normality in the
responses on all variables. The main variables, the ones who have been described in the
measurement section, did not have kurtosis and/or skewness issues. Four other variables, of in
total 26 variables (excluding demographics), had kurtosis and/or skewness issues. These
skewness and kurtosis issues could have multiple causes, some of them have been described
in the limitation section. But because the interest of this research is in the diversity among
members which should lead to LMX differentiation, it is almost wishful to have these kind of
issues, because this would indicate that there is a diversity between and within the teams.
Justifying for aggregation
At the abovementioned third stage, of computing mean values of the variables, there is also
made used of aggregation. As this research is at the team level, the individual responses
regarding a variable are aggregated to create one score per team on a certain variable. This is
done for the variables (which have been measured by the team members): openness,
neuroticism, agreeableness, extraversion, conscientiousness, LMX, LMX differentiation and
personal needs. This is done by the SPSS function ‘aggregate’, but first there should be
justified that it is allowed to aggregate the individual responses to a team level. This is done
by a Excel-tool, the ‘IRA (interrater agreement) and IRR (interrater reliability) for Consensus
Composition Constructs test’, developed by Biemann, Cole and Voelpel (2012). With this test
26
and show the degree of agreement between the individuals within the teams. In other words,
the rwg[J] does estimates the variance of every item across the respondents in each team.
When the rwg[J] is higher than 0.7 is statistically justifies aggregation, because it then
suggests a strong agreement among the respondents within each team. The intraclass
correlation coefficients (Bliese, 2000) are calculated to display the ratio of between-group to
total variance (ICC[1]) when corrected for the average team size (Biemann et al., 2012). The
F-tests, and the reliability of team members’ average ratings (ICC[2]) are also given.
The values for openness are, 0.88 (rwg[J],uniform), 0.28 (ICC[1]), F = 2.06, p-value <
.001, and 0.51 (ICC[2]). Neuroticism, 0.91 (rwg[J],uniform), 0.06 (ICC[1]), F = 1.17, p-value =
.229, and 0.14 (ICC[2]). Agreeableness are, 0.96 (rwg[J],uniform), 0.21 (ICC[1]), F = 1.72,
p-value < .005, and 0.42 (ICC[2]). Extraversion, 0.96 (rwg[J],uniform), 0.03 (ICC[1]), F = 1.10,
p-value = .328, and 0.09 (ICC[2]). Conscientiousness 0.94 (rwg[J],uniform), 0.05 (ICC[1]), F =
1.04, p-value = .265, and 0.12 (ICC[2]). LMX, 0.97 (rwg[J],uniform), 0.28 (ICC[1]), F = 2.08,
p-value < .001, and 0.52 (ICC[2]). LMX differentiation, 0.90 (rwg[J],uniform), 0.34 (ICC[1]), F
= 2.42, p-value < .001, and 0.59 (ICC[2]). And personal needs, 0.93 (rwg[J],uniform), 0.01
(ICC[1]), F = 1.02, p-value = .460, and 0.02 (ICC[2]).
As the results indicated, all jwg[J]’s scores are higher than 0.7 and therefore the
aggregation to the team level should be permitted. But not all values were significant, to
know: neuroticism, extraversion, conscientiousness and personal needs. This indicates that
there is not enough degree of agreement between the individuals within a team. This should
be an issue for not continuing with aggregating those variables, but because the interest of this
research is in the diversity among members which should lead to LMX differentiation, this
disagreement among team members is just a reason to continue. Thus although it does not met
all formal requirements for aggregation, there is continued with doing it through the underling
27 Calculating diversity
To be able to tests the diversity related hypotheses, thus all four hypotheses, diversity has to
be calculated and a diversity variables has been created. Diversity is a term which is
mentioned throughout this research, but as Harrison and Klein (2007) explains in their article
diversity, is not one thing/concept but consist of three things. In other words, diversity could
be seen as an umbrella concept which covers three specific concepts. When there is diversity
within an organizational unit there it is called ‘separation’, Harrison and Klein (2007, p. 1200)
provides the following description, ‘differences in position or opinion among unit members.
Such differences reflect disagreement or opposition – horizontal distance along a single
continuum representing dissimilarity in a particular attitude, or value, for example.’
Within-unit diversity could be indicative for variety, ‘differences in kind or category, primarily of
information, knowledge, or experience among unit members.’ (Harrison & Klein, 2007, p.
1200). And within-unit diversity could be indicative for disparity and is explained by Harrison
and Klein (2007, p. 1200) as follows, ‘differences in concentration of valued social assets or
resources such as pay and status among unit members – vertical differences that, at their
extreme, privilege a few over many’.
When looking at the main aim of this research, finding evidence that diversity in
personality traits among team members will result in LMX differentiation, it is clear that there
is not only researched within a unit (e.g. a team) but that it is more about a organizational unit.
Within this research there is interest in, for example, does a team which contains of people
who are different to each other on extraversion (e.g. one members is very extravert, the other
member is more introvert) leads to that a leader will differentiate on/in LMX? Thus within
this context there is searched for dissimilarities among team members on the same continuum
28
The most common indicator for separation is Standard Deviation (SD), it is the square
root of the variance and it measures the dispersion or spread of the data around the mean
(average score). For example, when a team does have a large SD at agreeableness this means
that the values of individual members on agreeableness are at distance from the mean value,
contrary when a team does have a low SD on agreeableness is indicates that the members are
all almost equal to each other on agreeableness and that their values lie around the mean
value.
From this point on, when the term diversity is named, it is known that it is about
separation with as measure the standard deviation (SD).
Correlation analysis
After preparing the data for analysis, the first analysis is the correlation analysis, this analysis
gives an overview about if variables are related to each other and how much. By doing this
analysis it already becomes visible if hypotheses could find support.
Direct regressions
Regression analyses were undertaken to further examining the hypotheses. Therefore
standardized values (Z) of the variables had be created and direct relationships were examined
by the use of hierarchical regression. From this point on there was also controlled for some
variables (control variables), these variables are included into the model given their potential
influence on the dependent variable (LMX differentiation).
Control variables
The first variable where is controlled for is team size, the amount of people (team members)
in a particular team. The amount of team members is exclusive the team leader, because the
interest of this research lies in the effects of teams on the leader his or her LMX
differentiation. In their reviewing article Henderson et al. (2009) states several propositions
29
the LMX differentiation, such that the larger the group, the greater the LMX differentiation.”.
Sui et al. (2015) also endorse the impotence of team size on the outcomes of LMX
differentiation, in their study they argue that social- identity and categorization is affected by
team size, and this social categorization is affecting internal team dynamics which on his turn
is seen as a antecedent for LMX differentiation (Stewart, 2006; Henderson et al., 2009).
Based on the proposition of Henderson et al. (2009) and the espousing arguments of the study
by Sui et al. (2015) it is convincing that team size, on its own, does have influence on LMX
differentiation and should therefore be controlled for finding the clearest/finest effects of the
independent- and mediator variables on LMX differentiation.
The second variable where is controlled for is task interdependence among the team
members. This variable was included to examine how dependent the team members are
among each other to fulfil their own tasks, in other words does team members need to consult
and or collaborate with their colleagues to fulfil tasks or can the tasks be done (more) solely.
Research on task interdependence shows that higher levels of task interdependence leads to
higher levels of team cooperation, than when team members are less interdependent
(McShane & Von Glinow, 2010). Lee, Lin, Huang, Huang and Teng (2015) in their study
underline this phenomenon and also state that higher intensity task interdependence will both
positively affect task conflict and relationship conflict. Thus both studies do underline that
higher levels in task interdependence will lead to a more collective team and team oriented
team, because they are dependent to each other in fulfilling their tasks. In relation to LMX
differentiation Henderson et al. (2009) state the following proposition (p.523), “P4c. The
more a work group's culture is characterized by collectivism and a team orientation, the less
the LMX differentiation.”. Therefore it seems to be that task interdependence does have some
predicting power in LMX differentiation and to find solely the predicting power of the
30
Furthermore there is controlled for team tenure to find the clearest/finest effects of the
independent- and mediator variables. In this context team tenure is the duration in time, of a
member, as being part of the team (Schippers, Den hartog, Koopman & Wienk, 2003) and the
duration in team tenure can also been seen as indicator for work experience (Tesluk and
Jacobs, 1998). According to Stoker (2007) there are various leadership theories (e.g. Hersey
and Blanchard, 1969, 1977) who have identified that team member experience is an important
factor in determining which leadership behavior is effective and or efficient for a team which
is high/low experienced. Additionally to this preceding explanation for the importance of
adding team tenure, is that team tenure also have its influence on for example task
interdependence. It is naturally that proportions in task interdependence between team
members will evolve during the time that an individual is part of a team, just as an
individual’s cognitive base evolves through experiences that have been experienced
(Wiersema & Bantel, 1993).
Age (aggregated team mean, and standard deviation) is the last variable were is
controlled for at the team level. Controlling for age was for a twofold reason, one to show that
surface-level demographics do not have (much) direct predicting power in explaining the
variance in LMX differentiation and second because age on its own does have influences in
the interaction among people and therefore influence the team processes and perceived LMX
differentiation. This is underlined in team composition research which have used, among
other things, age as important (visible) variable for social categorization (e.g. Tajfel, 1981;
Turner, 1987). And according to Timmerman (2000) teams which consist of people who are
more similar to each other should have better cooperation and will have less internal conflict,
and vice versa teams who consist of people who are less similar to each other will have more
lack of cooperation and more conflict among each other. But other researchers does conclude
31
diversity leads to less (emotional) conflict. In conclusion, research findings on age and its
potential effects on various team processes are inconsistent (Timmerman, 2000) and therefore
a reason to control for its mean and standard deviation values, this to find the pure effects of
the independent and mediator variables on LMX differentiation.
Testing for mediation
After examining the direct relations, the remaining potential mediation effects have been
tested. The test for mediation is executed with two mediation methods, with the procedures of
Baron and Kenny (1986) and the process-modeling program PROCESS developed by
Preacher and Hayes (2004). And both individual outcomes are checked by the normality
check, this is done by executing a Sobel-test (Sobel, 1982).
Results
In this chapter, first the correlation analysis will be discussed. Subsequently the results from
the regression analyses will be outlined. Direct relationships between the diversity in
personality traits among the team members and LMX differentiation are discussed in advance
before examining potential mediation effects.
Correlation analysis
An overview of the main variables’ means, standard deviations, and (bivariate) correlations
values are presented in table 1. For clarification, the role of a particular variable in this
research is given.
A first observation derived from the table, and especially for the dependent variables is
that all three variables (LMX differentiation measured by the leaders; LMX differentiation
measured by the members; LMX measured by the members) significantly correlate with
significantly correlate with each other (r = .290, p-value <.05). Although it is a small to
medium effect (Field, 2013), the finding itself is somewhat notable, because this indicates that
both variables are consistent to each other. In other words, if team members indicate that their
leader uses more LMX differentiation, the leaders do acknowledge this when judging/scoring
their own LMX differentiation higher. Thus a cause-effect relationship does not apply within
this correlation matrix (and with correlations in general), the reverse explanation is also true
and therefore no causal conclusions could be made around this relationship. It only indicates
that if one of the two values goes up/down, the other value either goes up/down. There should
be noted that correlating LMX ratings does not indicate interchangeability (Bliese, 2000). But
in general this result is contrary to some literature regarding the perception of the LMX
relationship between a leader and his or her follower. As stated by Schyns (2015), “In terms
of LMX, one of the stable results found in the literature is that the agreement between leader
and follower perspectives on their mutual relationship is rather low (Gerstner & Day, 1997;
Sin, Nahrgang, & Morgeson, 2009)”. Thus normally, as mentioned by those authors, there
should not be a (positive) correlation between LMX differentiation measured by the leaders
and LMX differentiation measured by the members.
Although the abovementioned finding is interesting, the lack of at least one (direct)
correlation between diversity in personality traits among team members and any of the three
forms of LMX differentiation is more important. This finding should already imply that
hypothesis 1, ‘Team diversity positively relates to LMX differentiation’, no longer can be
supported because there seems to be no relationship (positive or negative). But a closer look
into the correlation scores indicates that ‘diversity in neuroticism among team members’
(hereafter: diversity in neuroticism) and ‘diversity in openness among team members’
34
.05) with ‘LMX differentiation that is measured by the team members’ (hereafter: LMX
differentiation). Diversity in neuroticism does have a Pearson’s correlation coefficient of .255
and a p-value of .065, thus when the significant level is reduced to a level of .05, the
correlation becomes significant at the .10 significant level. When projecting this reducing in
significant level on diversity in openness and again in relation to LMX differentiation, this
correlation becomes also significant (r = .228, p-value = .061).
After meeting the first requirements, that some independent variables do have a
relationship with a dependent variable, research is done if those variables have a correlation
with one (or more) mediator variables.
Diversity in neuroticism is small to medium significantly correlated (r = .269, p-value
< .05) with the ‘diversity in personal needs among team members’ (hereafter: diversity in
personal needs). In first instance this could be the basis for a potential mediation effect of
diversity in personal needs in the relationship between diversity in neuroticism and LMX
differentiation. But based on the correlation analysis, it is shown that diversity in personal
needs is “totally” not correlating with LMX differentiation (r = -.057, p-value .664).
Diversity in openness is small to medium correlated (r = .206, p-value .092), when
reducing the significant level to .10, with ‘constraints for the leader in building (equal)
relationships among his/her team members’ (hereafter: constraints for the leader). And is,
constraints for the leader small to medium correlated with LMX differentiation (r = .267,
p-value < .05).
Based on the results of the correlation analysis, there could be concluded that diversity
in neuroticism is related to LMX differentiation, but that there is no significant mediation
effect possible, by one of the three mediators within this research. But constraints for the
35
and LMX differentiation. Therefore hypothesis 1 (team personality diversity positively relates
to LMX differentiation) and hypothesis 3 (the positive relationship between team personality diversity traits and LMX differentiation is mediated by constraints for the leader) could still
find support within this research. But hypothesis 2 (the positive relationship between team
personality diversity and LMX differentiation is mediated by diversity in similarity attraction)
and hypothesis 4 (The positive relationship between team personality diversity and LMX
differentiation is mediated by diversity in personal needs) could not be supported within this
research.
Thus as a result of reducing the significant level, and finding some correlations
between the main variables, it is still plausible that hypotheses 1 and 3 could find support.
Direct regressions have been done to further investigate these hypotheses.
Direct effects
Through the conclusions based on the correlation analysis there are a view relationships that
could find support and support the hypotheses, in this part there is analysed if the
relationships are in the way they are hypothesized. Specific, does diversity in neuroticism
have some predicting power in the variance of LMX differentiation, does diversity in
openness have some predicting power in the variance of LMX differentiation, does diversity
in openness have some predicting power in the variance of constraints for the leader, and does
constraints for the leader have some predicting power in the variance of LMX differentiation.
36
First the direct relationships are tested without controlling for any variable, this to
show the raw influence of the predicting variables on the outcome variables. As presented in
table 2, the directions of the direct relationships are in the same direction as the hypotheses.
Diversity in neuroticism is positively related to LMX differentiation (B = .122, p-value =
.065) and explains 5.1 percent of the variance in LMX differentiation. Diversity in openness
significantly (at a .10 significant level) predicts both LMX differentiation (B = .123, p-value =
.061) and constraints for the leader (B = .108, p-value = .092), and have 5.2 percent (LMX
differentiation) and 4.2 percent (constraints for the leader) predicting power in their variance.
Furthermore, constraints for the leader significantly predicts LMX differentiation (B = .123,
p-value = .061) for 6.4 percent of its variance.
Table 3. Direct regression, diversity in neuroticism to LMX differentiation, with control variables.
After showing the raw effects of diversity in neuroticism on LMX differentiation,
table 3 displays the effects of diversity in neuroticism on LMX differentiation, when
controlling for multiple variables. There is controlled for team related variables, this is
because the data is gathered in a team context and team characteristics does have influence on
37
theoretical background section. Besides the control for team related variables, there is
statistically controlled for the mean value (and or standard deviation value) of the independent
variable (or moderator variable), for example when testing for diversity in neuroticism (SD
value) in relation to LMX differentiation there is controlled for the potential effects of the
mean value of neuroticism. This to prevent the potential blurring effects of the mean values of
the independent variables (as “being’ the SD value), and be able to find the clearest/finest
effects of the independent and or mediator variables.
Diversity in neuroticism is, when controlling for the control variables, more positively
related to LMX differentiation and becomes almost significant at the .05 significant level (B =
.137, p-value = .051), and explains 11.0 percent of the variance in LMX differentiation.
Table 4. Direct regressions, diversity in openness to constraints for the leader and LMX differentiation, with control variables.
After showing the regression of diversity in neuroticism to LMX differentiation, table
4 presents the direct regressions with the independent variable ‘diversity in openness’.
Diversity in openness again does significantly (at a .10 significant level) predicts LMX
differentiation (B = .130, p-value = .066), and has 11.2 percent explanatory power in the
variance of LMX differentiation. In comparison to the first direct regression (thus without the
38
implies that when controlling for the control variables the explanatory power in the variance
of LMX differentiation is increased, but that the possibility that changes in diversity in
openness are resulting in changes in LMX differentiation is less likely.
When looking at the predicting power of diversity in openness to LMX differentiation,
when adding the control variables into the model, all values have been increased in
comparison to without control variables. Diversity in openness significantly (at a .10
significant level) predicts constraints for the leader (B = .123, p-value = .073), and has 11.3
percent predicting power in the variance of LMX differentiation. This implies that when there
is controlled for those control variables, diversity in openness does have more explanatory
power and that the likelihood that changes in diversity in openness leads to changes in LMX
differentiation is bigger.
Table 5. Direct regression, constraints for the leader to LMX differentiation, with control variables.
The last regression that is executed, including the control variables, is constraints for
the leader to LMX differentiation (see table 5). Constraints for the leader do significantly
39
variance. As these results shows, constraints for the leader do significantly predict a nice part
of the variance in LMX differentiation and is the possibility statistically high.
Based on the results of these executed direct regressions there could be concluded that
both diversity in neuroticism and diversity in openness does have predicting power in the
variance of LMX differentiation. Additionally the potential mediation effect, between
diversity in openness and LMX differentiation, by constraints for the leader is still possible. In
the upcoming section is the test for mediation explained and executed.
Mediation effects
After executing the correlation analysis and running the direct regressions, hypothesis 3 could
still be supported because the statistical finding does support the potential mediation effect of
constraints for the leader, between the relation diversity
in personality traits (in this case diversity in openness
among team members) and LMX differentiation (in this
case LMX differentiation measured by the followers).
To test this potential mediation effect there is made use
of the mediation procedures by Baron and Kenny
(1986), and is PROCESS by Preacher and Hayes (2004)
used to check if this method does find consistent results.
Mediation by the procedures of Baron and Kenny (1986)
Baron and Kenny (1986) describes a procedures to test for mediation, it consist of four
“requirements” that have to be met to show mediation. First there should be three significant
predictions, (1) IV predicts DV, (2) IV predicts MED, (3) MED predicts DV (IV and MED
are included in the same regression. After these regression there should be met with the last
requirement for having a mediation effect, (4) when IV and MED are included in the same
regression, the effect of IV on DV is less than the “original” effect of IV on DV (regression
