Content uploaded by Ronit Kark
Author content
All content in this area was uploaded by Ronit Kark on Dec 30, 2016
Content may be subject to copyright.
Journal of Applied Psychology
!"#$%&'()*(+,#-./"0(12"(3#4"(56(!"#$"7(839
!"#$%"&&'($")*$+,)-.$/"01
2)#-)'$3-04.$567#-8".-,)($9"08:$;<($;=>>?$*,-@$>=?>=ABC"==;;DED
FG!H!G2I
%"&&'($!?($J$/"01($+?$K;=>>($9"08:$;<L?$M'"*-)N$7O$PQ"RS#'@$!:'$F"4'$,&$M'"*'0$2FT?$!"#$%&'
"()*++',-.)/0123"'"41?$H*U")8'$,)#-)'$S67#-8".-,)?$*,-@$>=?>=ABC"==;;DED
Leading by Example: The Case of Leader OCB
Tal Yaffe
Ben-Gurion University of the Negev
Ronit Kark
Bar-Ilan University
The importance of leading by personal example or role modeling for effective leadership has been
recognized in many leadership theories. However, leaders’ ability to influence group behavior through
exemplary behavior has received little attention in empirical work. This study explores leading by
example through theoretical development and empirical testing of a moderated mediation model of the
potential effects of leader organizational citizenship behavior (OCB). This model suggests that a leader’s
OCB may promote group OCB directly and indirectly by enhancing the group’s belief that OCB is
worthy. It also specifies the moderators of the direct and indirect effects of leader OCB on group OCB.
Data from 683 members of 67 intact work groups, 67 group managers, and their supervisors support the
hypothesized model. The theoretical and practical implications of these findings are discussed.
Keywords: leading by example, organizational citizenship behaviors, role modeling, role model,
group-level OCB
Leadership has been defined as “a process of social influence
through which an individual enlists and mobilizes the aid of others
in the attainment of a collective goal” (Chemers, 2001, p. 376).
The importance of leading by example or role modeling for effec-
tive leadership has been recognized in many leadership theories,
including theories of self-leadership (Manz & Sims, 1980), eco-
nomic leadership (Hermalin, 1998), self-sacrificial leadership
(Choi & Mai-Dalton, 1999), authentic leadership (Luthans & Avo-
lio, 2003), and ethical leadership (Trevino & Brown, 2005). Trans-
formational and charismatic leadership theories in particular posit
that role modeling of a contribution to the collective is a major
means by which effective leaders encourage followers to do the
same (e.g., Bass, 1985; Burns, 1978; Conger & Kanungo, 1987;
House, 1977; Shamir, House, & Arthur, 1993).
Although leading by example is a central premise of these
theories, there is a lack of field studies that directly examined
whether leaders’ contribution to the collective actually promotes
similar contributions by groups.
1
Laboratory studies, by contrast,
have clearly demonstrated this phenomenon (Gu¨th, Levati, Sutter,
& van der Heijden, 2007; Potters, Sefton, & Vesterlund, 2007), but
these studies inevitably omit essential features of real leadership,
work groups, and the relationships between them. A theoretical
development and empirical study of leading by example in real
organizations is still needed if we are to advance our understanding
of leadership, particularly at the group level. Zaccaro, Rittman, and
Marks (2001) argued that “despite the large literature on leadership
and groups/teams dynamics, we know surprisingly little about how
leaders create and handle effective teams” (p. 452).
In the current study we try to fill this gap by addressing the need for
conceptual models of collective performance that integrate leadership
influence and group processes in a natural organizational context
(Zaccaro et al., 2001). We focus on role modeling of organizational
citizenship behavior (OCB), which typically is not an enforceable
requirement of any specific formal job role but in aggregate promotes
organizational effectiveness (Organ, 1988; Organ, Podsakoff, &
MacKenzie, 2006; Podsakoff, MacKenzie, Paine, & Bachrach, 2000).
In particular, we examine OCB that is targeted at the entire group or
organization and demonstrates loyal and responsible involvement in
the organization (Van Dyne, Graham, & Dienesch, 1994).
There are several reasons why the focus on modeling of OCB by
leaders is important for the study of the process and the effective-
ness of leading by example. First, OCB should be particularly
affected by modeling, as organizations typically cannot rely on
formal systems of job description, training, or rewards to cultivate
OCB (Podsakoff et al., 2000).
2
Second, compared to leader be-
havior that is directed at individuals, leader OCB, which is ambient
and is directed at the entire group, is more likely to operate at the
group level and has the potential to motivate the group as a whole
(Hackman, 1992; Kark & Van Dijk, 2007; Shamir, Zakay, Breinin,
& Popper, 1998). Finally, both theory and cumulative research
suggest that group-level OCB contributes to organizational effec-
tiveness (for recent meta-analyses, see Nielsen, Hrivnak, & Shaw,
2009; Podsakoff, Blume, Whiting, & Podsakoff, 2009). Thus,
1
Studies of transformational leadership typically measure the degree to
which the leader is perceived as a role model to follow but not the actual
contributions of the leader (e.g., Richardson & Vandenberg, 2005).
2
Modeling may also affect enforceable role behaviors but probably to a
lesser degree. Enforceable role behaviors are typically explicitly defined
and shaped by formal role description, formal training, and explicit direct
rewards and punishments, and this tends to limit their variability within the
organization. By contrast, OCB is discretionary and is typically not sup-
ported by formal systems of role description, training, and rewards (Pod-
sakoff et al., 2000). Thus, employees are more likely to turn to their role
model to determine the desirable or normative level of OCB within the
group (Bommer, Miles, & Grover, 2003; Ehrhart & Naumann, 2004).
Tal Yaffe, School of Management, Ben-Gurion University of the Negev,
Beer-Sheva, Israel; Ronit Kark, Department of Psychology, Bar-Ilan Uni-
versity, Ramat Gan, Israel.
We thank Dikla Elisha and Boas Shamir for their thoughtful insights and
helpful advice and Gil Luria and Dina Van Dijk for their constructive
comments on an earlier draft of this paper.
Correspondence concerning this article should be addressed to Ronit
Kark, Bar-Ilan University, Department of Psychology, Ramat Gan 52900,
Israel. E-mail: karkro@mail.biu.ac.il
Journal of Applied Psychology © 2011 American Psychological Association
2011, Vol. ●●, No. ●, 000– 000 0021-9010/11/$12.00 DOI: 10.1037/a0022464
1
inducing groups to engage in OCB may be an important criterion
for leadership effectiveness.
Our aims in this study are threefold. First, we attempt to extend
existing knowledge on the leading-by-example effect by studying
the ways leaders’ OCB can enhance OCB of work groups. Second,
we seek to identify and empirically examine a potential mediator
and the potential moderators of the effect of leader OCB on group
OCB in a natural organizational setting. Third, we hope to con-
tribute to the emerging study of the antecedents of group-level
OCB. To achieve these goals, we present a theoretical moderated
mediation model that clarifies how and when a leader’s OCB is
likely to promote group-level OCB. This model suggests that a
leader’s OCB may promote group OCB directly and indirectly by
enhancing the group’s belief that OCB is worthy. It also specifies
features of the organizational context (i.e., leader– group distance),
the leader’s standing (i.e., the leader as a role model), and the
group (i.e., group consensus), each of which is expected to
uniquely moderate the direct and/or indirect influence processes
and thus the leading-by-example effects. This model is summa-
rized in Figure 1.
A Mediation Model of the Effects of Leader OCB
In the following section we define and review the literature on
leader OCB and group-level OCB and present the direct and
indirect (mediated) relationships between them.
Leader OCB
The construct of OCB refers to the efforts undertaken by em-
ployees to behave as good citizens within their organization
(Organ, 1988). There is no consensus on the definition and dimen-
sions of the construct (Podsakoff et al., 2000), but Organ’s (1988)
and Graham’s (1991) conceptualizations of OCB are probably the
most widely acknowledged. The early definition of OCB (Smith,
Organ, & Near, 1983) built on social psychology theory (Katz &
Kahn, 1978) and suggested that OCB was both extra-role and
organizationally functional. This definition was criticized for its
vague and subjective distinction between in-role and extra-role,
which varies across individuals, jobs, and organizations (Morrison,
1994; Van Dyne et al., 1994). Organ (1997, p. 88) thus recom-
mended excluding reference to extra-role behavior when defining
OCB.
3
However, in most organizations and jobs, OCB is still not
likely to be (a) an explicit part of the formal strict job description;
(b) guided by formal training; or (c) formally or explicitly re-
warded (Podsakoff et al., 2000).
Graham (1991) introduced an alternative approach to OCB,
which is theoretically independent of role requirements or expec-
tations and is based on the construct of active civic citizenship in
political philosophy. According to this approach, OCB refers to
loyal and responsible involvement in the organization through
balanced engagement in three interrelated behavioral dimensions
that make up the “active citizenship syndrome”: obedience, loy-
alty, and participation (Van Dyne et al., 1994). The operational-
ization of the dimensions of the construct was further developed by
Moorman and Blakely (1995) to include behaviors such as making
persistent efforts to attain organizational goals, promoting and
defending the organization’s image to outsiders and insiders, and
initiating constructive change in the workplace (Moorman &
Blakely, 1995).
4
These behaviors are the focus of this study.
Graham’s construct has major advantages for the study of the
effect of leader OCB. First, the meaning and dimensions of OCB
are theoretically similar across organizations, persons, and jobs
(Van Dyne et al., 1994), and thus should be similar for leaders and
followers. This point is important, because in order to explore the
effect of leader modeling of OCB on group OCB, it is vital to
assess leader and group OCB along the same dimensions (Ehrhart
& Naumann, 2004). Second, this construct is restricted to forms of
OCB that are targeted at the entire organization (OCBO), and
excludes forms of OCB that are targeted at individuals (OCBI),
such as helping (Williams & Anderson, 1991). Leader OCB which
is directed toward the group is more likely to be visible to all group
members, and more likely to affect group processes (Kark & Van
Dijk, 2007; Shamir et al., 1998), and thus the group-level OCB.
Previous studies on formal leader (i.e., manager) OCB have
suggested that it is both possible and important to study its effects
on their groups. First, Conway’s (1999) meta-analysis indicated
that managers’ OCB can be distinguished from in-role task behav-
iors in managerial jobs. Second, several studies showed that man-
3
Organ (1997) posited that OCB differs from task performance by its
degree of discretionariness and by its guarantees of systematic rewards.
That is, “OCB is less likely to be an enforceable job requirement and less
likely to be regarded by the performer as leading confidently to systematic
rewards” (p. 91). Organ et al. (2006) further clarified that the discretionary
nature of OCB is independent of an employee’s subjective definition of
OCB as in-role or extra-role: “People may feel that certain behaviors are
‘expected’ as part of the job even though they may believe that the
behaviors are discretionary and not formally rewarded by the organization”
(p. 143).
4
The active citizenship syndrome is multidimensional and includes both
affiliative and challenging OCBs. These may have different antecedents
(Van Dyne et al., 1994) but still share a common theoretical factor of loyal
and responsible organizational participation (Graham, 1991). Moreover,
responsible citizenship requires engagement in both, in order to balance
competing organizational demands for maintenance and flexibility, coop-
eration and innovation, and quality and efficiency (Miron, Erez, & Naveh;
2004; Quinn, 1988).
Mediator
Moderators
c
H6
Leader
OCB
Group
OCB
Group Belief
about OCB
Leader
Distance
Role
Model
ab
H5
Group
Consensus
H7
H2 H3
H4*
H1
Figure 1. Moderated mediation model of the effects of leader organiza-
tional citizenship behavior (OCB) on group-level OCB. H !hypothesis.
H4 is the mediation hypothesis.
2YAFFE AND KARK
agers tend to display some type of OCB (e.g., change initiatives)
more than non-managers, but there are also sizeable differences
among managers in their level of such behaviors (e.g., Farh,
Zhong, & Organ, 2001; LePine & Van Dyne, 1998; Morrison &
Phelps, 1999), implying that they are not an enforceable part of a
manager’s job. Finally, managers’ OCB was found to be positively
related to managers’ performance evaluations by their supervisors
(MacKenzie, Podsakoff, & Paine, 1999) and peers (Conway,
1999). MacKenzie et al. (1999) also found that managers’ OCB
affects their performance evaluations as much as objective mea-
sures of performance, and that OCB better predicts evaluations of
managers than nonmanagers. A possible explanation for the am-
plified valuation of managers’ OCB may be its potential effect on
their groups’ attitudes and behaviors (Conway, 1999; MacKenzie
et al., 1999).
Group-Level OCB
Although OCB was originally conceptualized at the individual
level (Organ, 1988), scholars have recognized the need for theory
and study of OCB at the group level (e.g., George & Bettenhausen,
1990; Schnake & Dumler, 2003). Recent thinking on various
group cognitive, affective, and behavioral processes as emergent
group properties (Kozlowski & Ilgen, 2006) and on OCB as a
normative phenomenon within work groups in particular (Ehrhart
& Naumann, 2004) has set the stage for a study of OCB at the
group level. The construct of group-level OCB refers to the nor-
mative level of OCB enacted within the work group (Ehrhart,
2004). The essence and dimensions of OCB remain the same, but
the group-level construct includes interactive elements that are
not part of the individual-level construct. Although OCB does not
require coordination, it usually occurs within work groups that
provide a strong social context. This leads group members to
develop shared cognition, customs, and knowledge (Moreland &
Levine, 1989) as well as a shared social identity (Turner, 1985). A
group-level approach to OCB acknowledges the impact of this
social context, as well as the group dynamics and processes that
make the group more than a mere collection of individuals (Lewin,
1951). Group-level OCB presumably results from mutual learning,
mutual adjustment and assimilation, and mutual (informative and
normative) influence and is thus qualitatively different from
individual-level OCB (Ehrhart & Naumann, 2004). This implies
that group members may explicitly or implicitly coordinate and to
some extent synchronize their OCB (e.g., decide to stay together
after work hours in order to complete a task or use their lunch
break to elaborate on suggestions for improving work methods).
The few studies of the antecedents of unit-level OCB have
confirmed that leadership is an important antecedent of OCB at
this level. Transformational leadership (Richardson & Vanden-
berg, 2005), servant leadership (Ehrhart, 2004), supportive lead-
ership (Euwema, Wendt, & Van Emmerik, 2007), and ethical
leadership (Mayer, Kuenzi, Greenbaum, Bardes, & Salvador,
2009) have all been found to be related to OCB at the unit level.
The Relationship Between Leader OCB
and Group OCB
It has been argued that the influence of leaders on their follow-
ers stems from their use of personal idiosyncratic power (referent
and expert power) rather than their use of position power (legal,
coercive, and reward power; Conger & Kanungo, 1998). Accord-
ing to Yukl (1998), one way in which leaders exercise referent
power is through role modeling, which is likely to enhance fol-
lowers’ emulation of leaders’ behavior. The idea that modeling of
OCB is an important way by which leaders can enhance followers’
OCB has long been recognized by OCB researchers (e.g., Bolino
& Turnley, 2003; Conway, 1999; MacKenzie et al., 1999; Organ et
al., 2006; Smith et al., 1983). Modeling should be especially
important for enhancing behaviors that are not already supported
by formal systems of job description, training, or rewards. Nau-
mann and Ehrhart (2005) further argued that modeling of OCB by
a leader is likely to be particularly effective and that “given that the
leader of the group is a likely role model for group members,
his/her performance of OCB should have a large impact on the
overall unit-level OCB in the work group” (p. 151). This propo-
sition draws in part on Bandura’s (1977) social learning theory,
which posits that most human behaviors are learned by observa-
tion. Social learning theory suggests that individuals will strive to
emulate the behaviors of their role models, leader, and coworkers
to ensure that their behavior is in line with accepted norms (Bom-
mer, Miles, & Grover, 2003; Ehrhart & Naumann, 2004; Mayer et
al., 2009).
The literature on leadership (e.g., Manz & Sims, 1980, 1981;
Trevino & Brown, 2005) and organizational socialization alike (e.g.,
Moreland & Levine, 1989; Ostroff & Kozlowski, 1992; Weiss, 1977)
suggests that vicarious learning by observing leaders’ behaviors plays
akeyroleinshapingfollowers’conductandmisconduct.Further,the
literature on team leadership suggests that vicarious learning may be
an important group process (Kozlowski & Ilgen, 2006) by which
leader role modeling may enhance team contributions to the common
goal (e.g., Zaccaro et al., 2001). Thus, a leader’s observable OCB may
be emulated by group members and become an emerging group
property (Kozlowski & Klein, 2000). Similarly, through modeling
and mimicry, avoidance of OCB can aggregate, spill over, and be-
come a group property (Felps, Mitchell, & Byington, 2006). We
hypothesized as follows:
Hypothesis 1: Leader OCB is positively related to group
OCB.
The Mediating Role of Group Belief in the Value
of OCB
Although social learning is not limited to level of behaviors, one
of our goals is to differentiate the process of learning by behavioral
modeling and imitation from a presumably deeper, indirect process
that may involve the transformation of group-level OCB-related
beliefs. The term beliefs refers to convictions held to be true by
individuals and groups and can refer to various content, including
values, goals, and preferable modes of conduct (Bar-Tal, 2000;
Rokeach, 1973). Researchers acknowledge that beliefs, judgments,
and attitudes are socially constructed and are likely to be shared by
group members who have common experiences and social inter-
actions through social processes in which collective meaning is
established (e.g., Bar-Tal, 1990; Hardin & Higgins, 1996; Sch-
neider & Reichers, 1983).
We focus on the group-value-laden belief that OCB is worthy,
though not formally rewarded, as a potential mediator of the effect of
3
LEADING BY EXAMPLE
manager OCB on group OCB. Haworth and Levy (2001) addressed a
related individual-level belief, namely, the “belief that OCB is in
general worthwhile.” They built on Vroom’s (1964) expectancy the-
ory, arguing that an individual’s decision to enact OCB draws on the
belief that OCB yields a “valent outcome which may be instrumental
in achieving some greater goal” (p. 65).
5
Haworth and Levy distin-
guished this belief from the “belief that OCB is formally rewarded,”
reporting that the former is related to perceived indirect psychological
benefits and the latter is related to perceived direct tangible rewards.
We sharpen this distinction by emphasizing the evaluative and value-
laden nature of the belief that OCB is worthy (Rokeach, 1973)
6
and
conceptualize it as an emergent group property (Kozlowski & Klein,
2000). The belief that OCB is worthy is general enough to include any
discretionary contribution that may benefit the organization (allowing
team members the flexibility to deal effectively with unpredictable
situations and demands), and it does not specify the potential benefits
of OCB performance. The unspecified nature of returns for contribu-
tions characterizing this belief is a defining feature of the type of
relationship that promotes OCB (Graham & Organ, 1993). This
feature also makes this belief relevant to work groups, as group
members may come to value OCB due to its contribution to their
group and organization (Ehrhart & Naumann, 2004; Shamir et al.,
1993).
7
According to Schein (2004), deliberate role modeling is one of
the primary embedding mechanisms by which leaders create,
maintain, and sometimes change their group culture (i.e., its shared
basic assumptions, beliefs, and values). The leader’s assumptions
that help the group solve its problems of external adaptation and
internal integration gradually become shared assumptions and are
taught to new members as the correct way to perceive, think, and
feel in relation to those problems. Similarly, leadership theories
view role modeling as an important means by which work group
leaders transmit their belief systems to their groups (Dragoni,
2005). Because leaders are expected to represent group identity
and values in their personal behavior (Shamir et al., 1998), the
leader’s behaviors provide influential input for collective sense
making about what is really of value in the work group. Moreover,
leaders’ exemplary behaviors often aimed to actively foster shared
values that leaders wish their followers to adopt (House, 1977;
Shamir et al., 1998). Transformational and charismatic leadership
theory posits that role modeling is a major way in which leaders
transform followers’ values, goals, and aspirations (e.g., Bass,
1985; Conger & Kanungo, 1987; Shamir et al., 1993). Exemplary
behaviors refer to the leader’s display of a commitment to the
shared values, identity, and goals that aim to increase the intrinsic
valence of group efforts on behalf of the collective goal (Shamir et
al., 1993, 1998). We propose that leader OCB, which represents
loyal and responsible involvement in an organization (Graham &
Organ, 1993), suits this purpose well. A leader’s role modeling of
OCB may serve four interrelated functions, which together are
likely to enhance the group belief that OCB is worthy.
First, a leader’s OCB demonstrates commitment and loyalty to
the group and the organization, thus enhancing the feeling of unit
pride and the salience of collective identity and values (e.g., De
Cremer & van Knippenberg, 2005; Shamir et al., 1998). Group
members are more likely to value a contribution in the form of
OCB when unit pride and the salience of collective identity are
elicited (Turner, 1985; Tyler, 1999). Second, by modeling OCB
the leader displays goal dedication, communicating to the group
that the group’s mission is important and worthy of members’
extra efforts (e.g., De Cremer & van Knippenberg, 2005; Kark,
Shamir, & Chen, 2003; Shamir et al., 1993). Group members are
more likely to value OCB when they have a sense of mission and
when they recognize the importance of their task (Shamir et al.,
1998). Third, leader behavior “models the way” organizational/
group goals should be pursued (Kouzes & Posner, 2003). It clar-
ifies role expectations, norms of conduct, and the acceptable
means for accomplishing the group’s goals (e.g., Ostroff & Koz-
lowski, 1992). Thus, by modeling OCB a leader links shared
identity, values, and goals with the expected group behaviors (i.e.,
OCB) for goal attainment and group success (Schein, 2004;
Shamir et al., 1998). Group members are more likely to value OCB
when they realize the importance of OCB to group survival and
success (Ehrhart & Naumann, 2004). Finally, by setting a personal
example of the efforts the leader expects from the group, the leader
demonstrates credibility (Kouzes & Posner, 2003) and trustwor-
thiness (Mayer, Davis, & Schoorman, 1995), thereby building trust
in the leader (Organ et al., 2006; Rich, 1997) and the organization
he or she represents (Lapidot, Kark, & Shamir, 2007; Shamir &
Lapidot, 2003). Group members are more likely to value OCB
when they trust their leader and are not afraid of being exploited
(De Cremer & van Knippenberg, 2005). For all these reasons,
leader OCB should increase the likelihood that the group will
come to value OCB. We made the following hypothesis:
Hypothesis 2: Leader OCB is positively related to the group
belief that OCB is worthy.
The study of individual-level OCB suggests that individual display
of OCB is correlated with individual OCB-related beliefs, such as
instrumentality that links OCB to valued outcomes (Haworth & Levy,
2001), self-efficacy belief (Bandura, 1986), and perceived role
breadth (Morrison, 1994; see McAllister, Kamdar, Morrison, & Tur-
ban, 2007). However, because group-level OCB involves interactive
elements that are not included in the individual-level construct (Ehr-
hart, 2004), it is more likely to be guided by group-level constructs,
such as shared beliefs (Karam & Kwantes, 2006; Leung & Bond,
2004) and particularly by behavioral beliefs or attitudes (Ajzen, 1991)
that specifically address the value of OCB.
5
Haworth and Levy (2001), building on Vroom’s (1964) expectancy
theory, referred to OCB as an instrumental belief, but the construct ad-
dresses the valence of the behavior and thus may represent a behavioral
attitude. The theory of planned behavior (Ajzen, 1991) defines attitude
toward the behavior as the degree to which the performance of the
behavior is positively or negatively valued.
6
Rokeach (1973) defined value as “an enduring belief that a specific
mode of conduct or end state of existence is personally or socially pref-
erable to an opposite or converse mode of conduct or end state of exis-
tence” (p. 5).
7
The evaluative nature of this belief (i.e., the focus on what is worthy of
doing) distinguishes it from a group climate that promotes OCB, because
the climate construct is mainly descriptive (i.e., describes “how things are
done”; Ehrhart & Naumann, 2004). Moreover, Ehrhart and Naumann
(2004) argued that the climate construct addresses formal organizational
practices, procedures, and rewards but that “OCB generally occurs outside
the formal requirements and rewards in the organization” (p. 962).
4YAFFE AND KARK
Shared beliefs and attitudes serve important functions for individ-
uals as group members and for the group as a whole (Bar-Tal, 2000;
Hardin & Higgins, 1996). Festinger (1950) argued that people expe-
rience their personal beliefs as being valid when these beliefs are
shared by others who are sufficiently similar to themselves. Bar-Tal
(2000) clarified that shared beliefs exist in the individual’s mind, but
the recognition that they are shared tends to arouse high confidence in
their content. He further argued that shared beliefs play an important
role in the formation and maintenance of social identities as well as
collective reality. Shared beliefs may provide the basis for a sense of
unity, solidarity, and interdependence, and they may guide, support,
and coordinate group-level actions. In general, when a behavior is
within their control, group members tend to act upon their shared
beliefs (Bar-Tal, 1990, 2000). Thus, the shared belief that OCB is
worthy is likely to be positively related to a group-level display of
OCB. Ehrhart and Naumann (2004) argued that when OCB is aggre-
gated over time, group members tend to view OCB as a standard
behavior in the group and may develop prescribed norms to regulate
OCB if they realize the contributions of OCB to the group. In sum,
leader OCB may enhance the level of group OCB indirectly by
enhancing the group belief that OCB is worthy. We hypothesized as
follows:
Hypothesis 3: Group belief that OCB is worthy is positively
related to group OCB.
Hypothesis 4: Group belief that OCB is worthy mediates the
relationship between leader OCB and group OCB.
Moderated Mediation Model of the Effect of Leader
OCB
Figure 1 depicts the two influence processes discussed above. It
also specifies the moderating role of leader distance, group belief
about the leader as a role model, and group consensus about the
value of OCB, each of which is expected to uniquely shape the
direct and/or indirect effects of leader OCB. The direct effect of
the leader’s OCB (Path c) presumably represents observational
learning of the leader’s OCB and thus should be affected by the
conditions for such learning (Bandura, 1977). The leader’s dis-
tance should reduce the opportunity to observe the leader’s behav-
iors, whereas the group belief that the leader is a worthy role model
should boost the motivation to emulate him or her. The indirect
effect of the leader OCB involves enhancement of the value that
the group ascribes to OCB. This process requires a significant
effect of leader OCB on the group’s belief in the value of OCB
(Path a), as well as a significant effect of this belief on group OCB
(Path b). The former effect (Leader OCB3Group belief about
OCB) is more likely to be evident when the group believes that the
leader is a highly worthy role model, whereas the second effect
(Group belief3Group belief about OCB) is more likely to be
evident when there is a high level of consensus in the group about
the value of OCB. We clarify these propositions below.
The Leader’s (Physical) Distance
MacKenzie et al. (1999) argued that one reason for the amplified
importance of OCB at the managerial level “has to do with the fact
that managers are in highly visible positions, affecting more people
as role models and leaders” (p. 399). We intentionally distinguish
leader visibility from the potential impact of the leader as a role
model for two reasons. First, distance leadership theory suggests
that visibility is neither sufficient nor essential for the leader to be
perceived a role model (e.g., Antonakis & Atwater, 2002; Shamir,
1995). Second, visibility is closely associated with social learning
processes (Bandura, 1977) that underpin the direct effect of leader
OCB on group OCB, while the status of the leader as a worthy role
model is closely associated with the transformational process that
underlies its indirect effect.
In the organizational context, the leader’s visibility is highly
related to physical leader– group distance.
8
The theory of substi-
tutes for leadership regards the leader distance as a neutralizer of
leadership, arguing that physical distance creates “circumstances
in which effective leadership may be impossible” (Kerr & Jermier,
1978, p. 396). Antonakis and Atwater (2002), in contrast, argued
that the dynamics of influence differ depending on the leader–
follower distance and that “leader–follower distance can contribute
to or detract from leader effectiveness” (p. 697). Although groups
may perceive distant leaders as role models, Shamir (1995) con-
cluded that the influence of distant leaders relies more on their
vision and rhetorical skills, and the influence of close leaders
derives to a greater extent from their personal example and ob-
servable behaviors. Thus, the direct effect of leader OCB should be
more salient for close than for distant leaders. Further, if observa-
tional learning is involved in the effect of the leader OCB, as
suggested by Naumann and Ehrhart (2005), the direct effect of
leader OCB on group OCB may not be evident for distant leaders,
as an opportunity to closely observe a model’s behavior is a crucial
condition for social learning (Bandura, 1977). In other words, a
leader’s physical proximity increases his or her visibility and thus
the likelihood that the level of leader OCB will be more adequately
emulated by the group. However, the social learning process that
presumably underlies this effect does not necessarily involve trans-
formation of OCB-related values and beliefs.
Leaders are prominent role senders in the role set of group
members (Katz & Kahn, 1978), their behaviors convey role ex-
pectations and norms of conduct. Thus, group members may feel
expected to perform the OCB the leader exhibits (Organ et al.,
2006) or simply mimic these behaviors without recognizing their
value to the group. In terms of Kelman’s (1958) theoretical model
of social influence processes, a leader’s physical proximity is
likely to encourage compliance but is not a sufficient condition for
identification and internalization of the leader’s values and beliefs.
The ability of the leader’s OCB to influence group beliefs and
values may thus be dependent on other factors, which are dis-
cussed in the following section. Therefore, we contend that the
leader’s physical distance should mainly reduce the direct effect of
leader OCB. In Edwards and Lambert’s (2007) terms, the moder-
ating effect of distance should conform to a pattern of a direct
effect moderated mediation model (i.e., applied to Path c). We
hypothesized as follows:
Hypothesis 5: Leader distance reduces the direct effect of
leader OCB on group OCB.
8
There are settings in which distant leaders are highly visible, for
instance, on TV or the Internet (e.g., political leaders).
5
LEADING BY EXAMPLE
Group Belief That the Leader Is a Worthy
Role Model
The term role model draws on the concept of role, the human
tendency to identify with and emulate other people occupying
important social roles (Katz & Kahn, 1978), and the concept of
behavioral modeling (Bandura, 1977). Gibson (2004, p. 134) de-
fined a role model as a cognitive construction based on the attri-
butes of people in social roles to which an individual desires to
increase similarity by emulating those attributes. This definition
emphasizes the distinction between the cognitive construction of
role model and the role model’s behaviors. Gibson’s review sug-
gests that the cognitive construction of role models involves iden-
tification and social-comparison processes and that it serves the
interrelated psychological functions of learning, inspiration and
motivation, and self-concept definition.
Follower perception of the leader as a worthy role model is
conceptually different from exemplary behaviors that the leader
may exhibit, including OCB (Organ et al., 2006). Thus, although
setting a personal example should facilitate the social learning
process (Manz & Sims, 1981), it should have a stronger effect on
unit-level behavior when group members share the belief that the
leader is a worthy role model. From a social learning perspective,
this belief is likely to enhance both the group’s attention to the
leader’s behaviors and the group’s motivation to emulate them. In
line with this argument, Jaussi and Dionne’s (2003) laboratory
study showed that the interaction between leader creative behavior
and follower perceptions of the leader as a role model for creativity
contributes to the prediction of follower creativity.
The study of the cognitive construction of role models implies
that mere organizational status does not ensure that a leader will be
perceived by the group as a role model. In fact, depending on
individual needs and aspirations, each group member may identify
with various role models inside and outside the organization (Gib-
son, 2004). However, leaders who are viewed by their groups as
worthy role models should have profound inspirational effects on
their groups’ values and beliefs, because they do not merely
convey role expectations but also represent what people would like
to be and to achieve (Gibson, 2004). Thus, internalizing leaders’
values and beliefs as well as emulating leaders’ behaviors helps
group members define and develop aspects of their self-concept,
including their role identity and social identity (Ashforth, 2001;
Kark & Shamir, 2002). This idea is a main theme in charismatic
and transformational leadership theories, which posit that the in-
fluence of leaders on followers’ identity, values, beliefs and be-
haviors is to a large degree dependent on their ability to earn the
status of role models “worthy of identification and imitation” (e.g.,
Conger & Kanungo, 1987; Kark & Shamir, 2002; Kark et al.,
2003). In Kelman’s (1958) terms, the influence of a highly worthy
role model relies on identification and internalization processes.
The social identity theory of leadership (Hogg, 2001; Hogg & van
Knippenberg, 2003) addresses this phenomenon by focusing on group
members’ collective self-conception and social identity processes,
which underlie the leadership influence in salient groups. According
to the theory, social influence in salient groups reflects and is driven
by prototypicality, the degree to which a member of the group defines
the group as a whole (van Knippenberg, van Knippenberg, De Cre-
mer, & Hogg, 2005). In the terms of this theory, leaders who are
viewed by their groups as role models worthy of identification are
likely to be highly prototypical; that is, they embody central and
desirable attributes of the group and are representative of group
members’ collective identity (van Knippenberg et al., 2005). Because
these leaders are seen as the best representatives of the group’s
attributes and shared social identities, they are better able to redefine
group identity and thereby influence group members to internalize
and act on new values, beliefs, and norms (Reicher, Haslam, &
Hopkins, 2005). Thus, the group belief that the leader is a worthy role
model should enhance the effect of leader OCB not only on the group
OCB but also on the value the group ascribes to OCB. In Edwards and
Lambert’s (2007) terms, the moderating effect of this belief should
conform to a pattern of a first stage and direct effect moderated
mediation model (applied to Paths c and a). We hypothesized as
follows:
Hypothesis 6: The group belief that the leader is a worthy role
model enhances the effect of leader OCB both on group OCB
and on the group belief that OCB is worthy.
Group Consensus About the Value of OCB
The indirect effect of the leader’s OCB is dependent not only on the
impact of the leader on the group belief that OCB is worthy but also
on the impact of this belief on group OCB. Because the impact of
group belief on group behavior is dependent on its “sharedness”
(Kameda, Tindale, & Davis, 2003), the belief– behavior relationship
at the group level should increase with the level of the group consen-
sus about the value of OCB. Kameda et al.’s (2003) review of the
group decision-making literature suggests that groups tend to define
“correctness” by consensus and that shared information, beliefs, and
preferences within groups exert an extraordinary influence on group-
level decision processes and outcomes. Ehrhart and Naumann (2004)
addressed the importance of consensual beliefs (i.e., norm strength)
and argued that the level of within-group agreement about descriptive
OCB norms (i.e., shared recognition that OCB is considered to be
standard or “correct” within the group) should enhance the impact of
the group norm on group members’ OCB.
Note that sufficient within-group agreement about the belief in
the value of OCB is a prerequisite for analyzing this belief as a
group-level property when the consensus model is applied (Chan,
1998; Kozlowski & Klein, 2000).
9
Yet, even groups with sufficient
agreement (indicated by adequate statistical measures) are still
likely to vary in their degree of consensus about the value of OCB,
due to between-group differences in social interaction among
group members (Klein, Conn, Smith, & Sorra, 2001), group co-
hesiveness (Luria, 2008), and the group development stage (Tuck-
man, 1965). Because the development of consensual beliefs about
the value of OCB is assumed to be the process underlying the
indirect effect of the leader’s OCB on group OCB, the model
clearly suggests that the group should reach high rather than
sufficient consensus about the value of OCB. In other words, the
consensus about the value of OCB should be high in order for
OCB to become a collective— explicitly or implicitly—
coordinated action (instead of a mere aggregation of individuals’
9
The requirement for within-group agreement depends on the aggrega-
tion model being used (Chan, 1998). When an additive model rather than
a consensus model is assumed, it is not necessary to demonstrate within-
group agreement. (For an example, see Colquitt, Noe, & Jackson, 2002.)
6YAFFE AND KARK
OCB). In Edwards and Lambert’s (2007) terms, the moderating
role of group consensus should show a pattern of a second stage
moderated mediation model (i.e., applied to Path b).
Hypothesis 7: Group consensus about the value of OCB
moderates (enhances) the effect of the belief that OCB is
worthy on the unit-level OCB.
Method
Sample and Procedure
The study was carried out in a large Israeli communication
organization. The sample was composed of 67 work units from
three departments: service (37 units), technical (21 units), and
sales (9 units). Basic task activities in these three departments
included responses to customer calls, equipment installations and
repairs, and sales of communication packages.
The department managers supported the study and provided a
letter requesting the employees’ cooperation and assuring the
confidentiality of responses. To avoid a common source bias,
we collected data from three different sources (roughly at the same
time). First, work unit managers were rated by their immediate
(regional) managers for their level of OCB. Second, group mem-
bers (employees) of these work units rated their beliefs that OCB
is worthy and their beliefs that the leader is a worthy role model.
Last, unit managers rated the level of OCB of their group as a
whole. Participation in the study was voluntary, to encourage
honest responses, and surveys were returned directly to the re-
search staff.
Usable surveys were obtained from 683 employees, representing
an average response rate of 73% within units. Response rate varied
considerably between groups (38%–100%) and tended to be lower
for shift employees. However, we did not find any significant
relationship between response rate and the study variables, except
for a positive relationship with group age. An average work unit
had 14 employees (SD !3.2). For managers, average age was 30.5
years (SD !5.0), average organizational tenure was 3.4 years
(SD !.74), and average job tenure was 2.0 years (SD !1.2).
About 33 percent of the managers were women. For employees,
average age was 26.6 years (SD !2.4), average job tenure was 1.3
years (SD !0.64), and 42.5% were women.
10
Measures
All the scales in this study were 7-point scales (ranging from
1!strongly disagree or not typical at all to7!strongly agree
or very typical).
Leader OCB. Area managers rated the degree to which each
of their subordinate group managers engaged in OCB directed
toward the organization. The construct of OCB included three
scales that were chosen based on theoretical considerations (i.e.,
OCB that is likely to be observed by group members) as well as
their psychometric qualities. Two 5-item scales from Moorman
and Blakely’s (1995) instrument were used to assess managers’
“loyal boosterism” and “personal industry.” Loyal boosterism re-
fers to active defending and promoting of the organizational image
to outsiders and insiders (e.g., “Defends the organization when
other employees criticize it,” “Shows pride when representing the
organization in public”). Industry refers to performance above and
beyond the call of duty and high attention to quality (e.g., “Always
meets or beats deadlines for completing work,” “Performs his/her
duties with extra-special care”). Change-oriented OCB, also re-
ferred to as innovative OCB (Moon, Van Dyne, & Wrobel, 2005),
was assessed by Morrison and Phelps’ (1999) Taking Charge
Scale, which is somewhat more comprehensive than the parallel
Individual Initiative Scale developed by Moorman and Blakely.
Taking charge refers to employees’ initiatives to bring about
functional change in the workplace (e.g., “Tries to correct a faulty
procedure or practice,” “Makes constructive suggestions for im-
proving how things operate within the organization”). On the basis
of a pilot study, which is described in the preliminary analysis
section, we used 8 items out of the 10 of the Taking Charge Scale.
Work group OCB. Following the procedure applied by
Ehrhart (2004) and Richardson and Vandenberg (2005), group
managers rated the level of their groups’ OCB along the same
scales, which were refined to address the group as a whole rather
than the individual’s OCB.
Group belief about OCB. Group members’ beliefs that OCB
is worthy were assessed by three items adapted from Haworth and
Levy (2001). The items were modified to reflect group belief and
to focus on the worthiness of OCB and not on OCB being worth-
while (i.e., “Behaviors above and beyond the job requirements are
worthy although not formally rewarded,” “Performance above and
beyond the formal job requirements is in general valuable,” and “It
is worthy to devote efforts far beyond job requirements, even if
these efforts are not paid for”; "!.84). The focus on the group
was aimed to make sure that all unit members would address the
same referent, and it is consistent with the concept of shared
beliefs as an emergent group property (Kozlowski & Klein, 2000)
and with Chan’s (1998) referent-shift consensus model.
Leader distance. The firm’s policy dictated uniform com-
munication procedures aimed at enhancing efficient management
of its highly geographically dispersed sites. Thus, there were
considerable structural differences between the departments with
regard to leaders’ physical distance and the frequency of leader–
group interactions, which were mainly determined by the firm’s
procedures. In the service department (37 work units), employees
worked daily in the vicinity of their leader in the same office. In
the other departments (30 work units), employees worked in the
field in dispersed geographical locations and met their manager
once a week at the work unit meeting held at the firm’s headquar-
ters. Thus, leader distance was coded as a dummy variable (distant
leader !1; close leader !0).
Group belief about the leader. Group beliefs that the leader
is a worthy role model were assessed with three items from Rich’s
(1997) Role Model Scale and Podsakoff, MacKenzie, Moorman,
and Fetter (1990) “providing an appropriate model” scale. In line
with the referent-shift consensus model (Chan, 1998), the items
were modified to reflect the belief about the degree to which the
group leader is a worthy model for the group as a whole (i.e.,
10
Work units had between 7 and 23 employees. For managers, age
ranged from 23 to 50 years, organizational tenure ranged from1.0 to 5.0
years, and job tenure ranged from 0.4 to 4.0 years. For group members,
mean age ranged from 23 to 50 years and mean job tenure ranged from 0.4
to 4.0 years.
7
LEADING BY EXAMPLE
“provides us a good model to follow,” “leads us by example,” and
“acts as role model worthy of imitation”; "!.91).
Group consensus. In line with the conceptualization of
within-group variability as indicative of the strength of group
beliefs (Chan, 1998) and with previous studies on group consensus
(see Luria, 2008), the level of within-group agreement/
disagreement on the value of OCB was assessed by the within-
group standard deviation in the scores of this belief (higher values
signify lower consensus). Note that in this study we were inter-
ested both in the average level of belief about OCB and in
within-group variability in this belief. Thus, in Chan’s terms we
first applied a consensus model, in which sufficient level of agree-
ment is a prerequisite for the emergence of group property and
shared meaning, and then a dispersion model, in which the level of
within-group agreement is the group characteristic of interest.
Controls. Our main analysis method, structural equation
modeling (SEM), and the sample size limited the ability to control
for potential covariates of OCB. However, in our complementary
regression analysis we controlled for leader and group tenure,
which are theoretically related to the development of relationships
that enhance OCB (e.g., Van Dyne et al., 1994), and for group size,
which may affect social integration, social coordination, or social
loafing (e.g., Karau & Williams, 1993) and consequently group-
level OCB.
Analytical Approach to Hypothesis Testing
We examined measurement adequacy and the study hypotheses
by SEM analysis, using AMOS 7.0 software (Arbuckle, 2006).
SEM is the preferred method for mediation analysis (Baron &
Kenny, 1986). Unlike regression analysis, SEM estimates the
relationship between the theoretical (latent) constructs in a struc-
tural path model that takes into account the imperfect reliability of
the study measures and corrects for random measurement error. In
addition, SEM enables the simultaneous estimation of all the paths
in a mediation model and the degree of fit of the entire mediation
model, by comparing it to the fit of alternative plausible models.
By contrast, regression analysis provides only separate and partial
tests of model components conducted on an equation-by-equation
basis. AMOS software also provides bootstrap estimates for all the
parameters in a path model and significance tests of the direct,
indirect, and total effects based on bias-corrected confidence in-
tervals. The bootstrap method (Efron, 1979) is regarded the most
powerful test of the size and significance of indirect effects (Edw-
ards & Lambert, 2007; MacKinnon, Lockwood, Hoffman, West, &
Sheets, 2002; Preacher, Rucker, & Hayes, 2007).
11
We began with preliminary tests of the adequacy of our mea-
surement model and then applied nested structural model compar-
isons and bootstrap analysis to draw conclusions about the direct
and indirect effects of leader OCB on unit-level OCB and the
mediating role of the belief about the value of OCB. We then used
SEM multigroup analyses and the bootstrap method to test the
specific moderating role of each potential moderator of the effect
of leader OCB. Given the limitation of the sample size and the
complexity involved in estimating interactions with continuous
latent variables in structural equation modeling (Preacher et al.,
2007), each moderation hypothesis was examined in a different
SEM multigroup analysis, and continuous moderators were cate-
gorized by the median split. This procedure has two limitations.
First, the separate testing of each moderator without adequate
controls hinders conclusions about unique contribution of the
hypothesized moderator. Second, dichotomizing quantitative mea-
sures results in a loss of information and havoc with regard to
estimation and interpretation of relationships among variables
(MacCallum, Zhang, Preacher, & Rucker, 2002). We therefore
applied a complementary analysis using Preacher et al.’s SPSS
(regression-based) macro, to assess the combined effect of all the
hypothesized moderators and the unique contribution of each
continuous moderator, when other interactions and potential cova-
riates of OCB are controlled for.
Results
Preliminary Analysis: Examining the Adequacy
of the Study Measures
Group beliefs. Group members’ (n!683) ratings of their
beliefs about the leader and about OCB were subjected to confir-
matory factor analysis (CFA), using the maximum likelihood
method. The hypothesized two-factor model fit the data well. The
chi-square was 14.3 (df !8, ns), the comparative fit index (CFI)
was 0.99, the Tucker–Lewis index (TLI) was 0.99, and the root-
mean-square error of approximation (RMSEA) was .03. All scale
loadings on their intended factors were high (.73–.91) and signif-
icant ( p#.001), and the correlation between the factors was
moderate (r!.37). This model was superior to a one-factor
model, which significantly decreased the model fit ($%
1
2
!150.4,
p#.001). Following Kozlowski and Klein (2000), the interrater
agreement coefficient (r
wg(J)
; James, Demaree, & Wolf, 1984) and
interclass correlation coefficient (ICC(1) and ICC(2); Bliese,
2000) were examined to determine whether group members’ be-
liefs could be analyzed as group-level constructs.
The r
wg(J)
index compares the observed variance in group rat-
ings to the expected variance of random response. Leadership
researchers have concluded that a slightly negatively skewed dis-
tribution is a reasonably good approximation of a random response
to leadership and attitude questionnaires (e.g., Schriesheim, Cog-
liser, & Neider, 1995; Shamir et al., 1998). Such a distribution is
theoretically defensible, given the common social desirability and
leniency biases in attitude studies (James et al., 1984). Our de-
scriptive statistics indeed indicated mildly negatively skewed re-
sponses in the item ratings. We thus computed the r
wg(J)
index
based on expected distributions possessing a small skew. For the
belief about the leader, the r
wg(J)
values ranged from .71 to .99, and
the median r
wg(J)
value was .90. For the belief about OCB, the
r
wg(J)
values ranged from .58 to .98, and the median r
wg(J)
value
was .85. Although two groups showed an interrater agreement
below .70 with regard to the belief about OCB, r
wg(J)
values
generally indicated a meaningful within-group consensus regard-
ing both beliefs.
11
This resampling procedure creates multiple independent subsamples
from the original database, which operates as an empirical rather than
theoretical sampling distribution, and is free from restricting assumptions
inherent to classic inferential statistics. In contrast to other tests of medi-
ation, the bootstrap method does not rest on the assumption that the indirect
effect is normally distributed, which is rarely true (Preacher et al., 2007).
8YAFFE AND KARK
We also looked at the range of r
wg
values when calculated
against other plausible null distributions, as recommended by
James et al. (1984) and LeBreton and Senter (2008). For the belief
about the leader, median r
wg(J)
values ranged between .93 when
the uniform null distribution applied and .72 when the heavy skew
null distribution applied. For the belief about OCB, median r
wg(J)
values ranged between .90 when the uniform null distribution
applied and .52 when the heavy skew null distribution applied.
12
We regard both extremes as a less reasonable approximation for
the expected random variance. A uniform distribution neglecting
common biases most likely leads to overestimation of agreement,
whereas a heavy skew null distribution most likely provides an
underestimation of agreement. Moreover, applying this null distri-
bution resulted in negative r
wg(J)
in several groups, which often
indicates that the random response null distribution has been
incorrectly specified (LeBreton & Senter, 2008).
ICC(1) values for the belief about the leader (ICC(1) !.26, p#
.001), and for the belief about OCB (ICC(1) !.16, p#.001),
indicated that a significant percent of the variance in both beliefs
could be explained by membership in a particular work group.
ICC(2) values (.67 and .78, respectively) indicated that for both
beliefs, the group means were fairly stable. Taken together, these
results suggested sufficient justification for the aggregation of both
beliefs to the group level.
OCB measures. On the basis of Graham’s (1991) theory, we
treated OCB as a second-order latent construct reflecting an active
citizenship syndrome, indicated by the three dimensions (first-
order factors) of loyalty, industry, and constructive initiative (i.e.,
taking charge). Because the number of groups in this study (J!
67) was not suitable for CFA at the item level, we obtained
supporting evidence by using a different sample of 1,387 employ-
ees from the same industry who rated their coworkers along these
scales. Following exploratory factor analysis we omitted two items
of the Taking Charge Scale that had low loadings on their latent
factor. After omission of these items, a second-order-factor model
fit the data rather well (CFI !0.94, TLI !0.93, RMSEA !.08).
CFA indicated adequate loadings of scale items on their latent
dimensions, the first-order factors (.74 –.91), and adequate load-
ings of the dimensions on the second-order factor, presumably
representing OCB (loyalty !.86, industry !.85, initiative !.78).
The dimensions were separable, but a large portion of their vari-
ance (61%–74%) was explained by the shared general OCB con-
struct.
Although we could not analyze the current data in the same
manner, the moderately high correlations among the dimensions
(.67–.69 and .64 –.70 for the leader and the group OCB, respec-
tively) supported the assumed second-order latent factor. We thus
assessed a simplified measurement model that represents the
leader and group OCB as two distinct but related second-order
latent factors, each reflected by three indicators (i.e., the three
scale means). The reliabilities of the dimensions for the leader
OCB (initiative, "!.91; loyalty, "!.88; and industry, "!.89)
and for the group OCB (initiative, "!.81; loyalty, "!.84; and
industry, "!.86) minimize the risk that parceled scales masked
problems with individual items. CFA results showed that this
model fit the data well (%
8
2
!7.1, ns, CFI !1.0, TLI !1.0,
RMSEA !.00). All the scale loadings on their intended factors
were high (.73–.87) and significant ( p#.001). We further com-
pared this general measurement model to several restricted models,
in order to establish the invariance of measurement across the
ratings of leaders and group OCB, to control for similarity of
meaning of the two latent factors (see Byrne, 2001). Nested model
comparisons indicated that (a) imposing equality constraints on the
factor loadings of the three scales on their intended latent factors—
work-unit OCB and leader OCB— did not decrease the measure-
ment model fit to the data ($%
2
2
!4.1, ns, and $%
2
2
!0.51, ns),
respectively, implying that the three scales can be considered
parallel measures of both latent constructs; (b) equality constraints
on the loadings of parallel scales of the same leader and group
behaviors (e.g., leader loyalty and group loyalty) on their respec-
tive latent factors, leader and group OCB, did not decrease the fit
of the measurement model to the data ($%
2
2
!2.7, ns); and (c) a
restricted model that enforced the three scales to be parallel mea-
sures of the latent constructs of leader and group OCB, as well as
invariance across the ratings of group and leader OCB, seemed
useful. Both fit indices (%
12
2
!11.7, ns, CFI !1.0, TLI !1.0,
RMSEA !.00), and nested model comparisons ($%
4
2
!4.6, ns),
suggested that this model adequately combined parsimony and
fit. These results support the construct validity and the similar-
ity of the factorial structure of the leader and the group OCB.
13
Descriptive statistics and correlations among variables are
shown in Table 1.
Tests of Mediation
Given that at least in some conditions, the belief about OCB
may not fully account for the effect of the leader OCB on the group
OCB, our basic model (see Figure 1) is a partial mediation model.
In this model the total effect of the leader OCB (X) on the unit
OCB (Y) is the sum of the direct effect (c&!'
yx.m
) and the indirect
effect of the leader OCB through the mediator (M; group belief
about OCB), which is equal to the product of the a and b paths
('
mx
('
ym.x
). The measurement part of this model, where all
structural paths between the three variables are free (but the leader
and the group OCB have a similar factorial structure), fit the data
well (%
28
2
!32.9, ns, CFI !0.98, TLI !0.98, RMSEA !.05).
Hypotheses 1– 4 concerning the mediating role of the belief
about OCB were examined by comparing three structural models.
The direct effect model included only the direct effect of the leader
OCB on group OCB, omitting its indirect effect through the group
belief about the value of OCB. The full mediation model included
the indirect effect of leader OCB, omitting its direct effect. These
two models are nested in the partial mediation model, which
included both the indirect and direct effects of leader OCB. The
direct effect model indicated a significant effect of leader OCB on
group-level OCB ('!.34, p#.01), implying that there is a
relationship between these variables that may be mediated (Baron
12
Expected variances for a 7-point scale were adapted from LeBreton
and Senter (2008). The median r
wg
values for the belief about the leader
and about OCB were .93 and 90 for the uniform null distribution ()
2
!4);
.90 and .85 for the slightly skewed distribution ()
2
!2.90); .85 and .77 for
moderately skewed distribution ()
2
!2.14); .84 and .76 for the triangular
distribution ()
2
!2.10); and .72 and .52 for the heavily skewed distribu-
tion ()
2
!1.39).
13
Although our hypotheses refer to the general construct of OCB, it is
worth noting that the correlations of the OCB dimensions with the other
study variables, although not identical, are generally in the same direction.
9
LEADING BY EXAMPLE
& Kenny, 1986), and supported Hypothesis 1. Although some fit
indices showed a reasonable model fit (CFI !0.94, TLI !0.93),
other fit indices indicated that the fit of the direct effect model to
data is deficient (%
30
2
!48.9, p#.02, RMSEA !.09).
The partial mediation model fit the data well (%
28
2
!32.9, ns,
CFI !0.98, TLI !0.98, RMSEA !.06), and showed signifi-
cantly better fit than the direct effect model ($%
2
2
!16, p#.001).
However, although yielding somewhat better fit indices, it did not
show significantly better fit than the full mediation model ($%
1
2
!
2.0, ns;$%
29
2
!34.9, ns; CFI !0.99, TLI !0.98, RMSEA !.05),
implying that the more parsimonious full mediation model was
preferable (Mathieu & Taylor, 2006). Inspection of the path coef-
ficients in the partial mediation model indicated a significant effect
of leader OCB on group belief that OCB is worthy ('!.38, p#
.005) and a significant effect of this belief on group OCB ('!.39,
p#.05), thus supporting Hypotheses 2 and 3, respectively. How-
ever, the direct effect of leader OCB on group OCB, which was
significant in the direct effect model, became nonsignificant when
the effect of the belief about OCB was introduced in the partial
mediation model ('!.20, ns). Moreover, bootstrap estimates
(means estimates across 2,000 independent samples drawn from
the original data) revealed that the indirect effect of leader OCB on
group OCB was significant ('
mx
('
ym.x
!.15, SE !.07, p#
.01), as was the total effect (.35, SE !35, p#.05), but the direct
effect of leader OCB was not ('
yx.m
!.20, SE !.16, ns).
Together, these results satisfy Baron and Kenny’s criteria for
establishing full over partial mediation, thus partially supporting
Hypothesis 4. See Figure 2.
Tests of Moderated Mediation
Given the limitation of the sample size and the complexity involved
in estimating interactions with continuous latent variables in structural
equation modeling (Preacher et al., 2007), each moderation hypoth-
esis was examined in a different SEM multigroup analysis, and
continuous moderators were categorized by the median split. Note
that although the full mediation model better fit the data from the
entire sample, the test of the hypotheses on the specific role of each
moderator required the assessment of the saturated (partial mediation)
model at both levels. Following Byrne’s guidelines (2001), each
moderation analysis began with tests of the invariance of the factorial
structure of the study measures across the levels of the moderator, and
invariant measures were constrained equal while subsequent tests of
the structural parameters were conducted. The results of these pre-
liminary tests indicated a reasonable similarity in the measurement
model across the levels of the three moderators, and thus the moder-
ating hypotheses were examined in a restricted measurement model of
group invariance. The potential moderating effects of each moderator
were examined by comparing the unconstraint structural model to
three nested models; each constrained a different path coefficient in
the mediation model to equality across the levels of the moderator (see
Table 2). A significant decrease in the unconstrained model fit to the
data implies a significant difference between the levels of the mod-
erator in the path coefficient that was constrained to equality, thus
indicating that the foci variable moderates the path of interest (Kline,
1998).
We further looked at the effects at each level of the moderator, to
examine whether the moderating pattern was consistent with our
hypotheses. Figure 3 shows the basic mediation model at the two
levels of each moderator (factor loadings are constrained to equality,
but structural paths are free of constraints). Note that for each level of
the moderator, the path (a, b, c) coefficients in the mediation model
represent conditional effects. That is, the magnitude of each path
coefficient in the mediation model is conditioned (or dependent) on
the particular level (high or low) of the moderator. Hence, although a
significance test of the difference between equivalent path coefficients
at the two levels of the moderator investigates whether the effects in
the mediation model vary systematically as a function of the moder-
ator, the magnitude and direction of these conditional effects indicate
how they vary. Bootstrap estimates of these conditional paths as well
as the conditional indirect and total effects for each moderator are
presented in Table 3.
Leader distance. Figure 3a depicts the unconstrained struc-
tural mediation model for close leaders who have daily direct
interactions with their groups (n!37) and for distant leaders who
interact with their groups in a formal meeting once a week (n!
31). Nested model comparisons (see Table 2) supported Hypoth-
esis 5, showing that the only significant difference between close
and distant leaders is in the direct path from the leader OCB to the
Table 1
Descriptive Statistics and Correlations
Variable M SD 12345678
1. Leader OCB 5.32 0.84 —
2. Work group OCB 4.65 0.63 .35
!!
—
3. Group size (n) 14.01 3.32 .12 *.30
!!
—
4. Leader tenure 1.98 1.19 .06 .24
!
.02 —
5. Group tenure 1.34 0.64 *.17 .33
!!
*.44
!!
.05 —
6. Distance
a
.07 .38
!!
*.34
!!
.19 .21 —
7. Role model 6.21 0.50 .40
!!
.09 .06 .14 *.23 *.03 —
8. Group consensus (belief variability)
b
0.97 0.24 *.20 *.29
!
.25
!
.05 *.09 *.12 *.12 —
9. Belief about OCB 6.00 0.47 .37
!!
.46
!!
*.16 .11 *.06 .23 .38
!!
*.32
!!
Note. N !67 work units. Latent variables are represented by maximum likelihood factor scores. Descriptive statistics are based on raw scores. SD !
standard deviation; OCB !organizational citizenship behavior.
a
Distant leader !1; Close leader !0.
b
Group consensus is measured by the within-group standard deviation in the belief about OCB; the higher the
score the lower the consensus.
!
p#.05.
!!
p#.01.
10 YAFFE AND KARK
group OCB. This implies a pattern of a direct effect moderated
mediation model (Edwards & Lambert, 2007). That is, enforcing
the equality constraint on the direct path from leader OCB to group
OCB across the levels of leader distance significantly decreases
the model fit to the data ($%
1
2
!5.6, p#.05). However, equality
constraints on the path from the leader OCB to the group belief
about OCB ($%
1
2
!.06, ns), and from the group belief to the group
OCB ($%
1
2
!.38, ns), did not decrease the model fit to the data
(see Table 2). In line with the role of visibility in a social learning
process, the direct effect of leader OCB was significant only for
close leaders. Figure 3a and Table 3 show that close leaders’ OCB
had a significant direct effect (.48, p#.001) as well as a signif-
icant indirect effect (.12, p#.001), which together produced a
significant total effect on group-level OCB (.60, p#001). Distant
leaders’ OCB had a significant indirect effect on group OCB
through the group belief about OCB (.13, p#.001), but this effect
did not yield a significant total effect.
Group belief that the leader is a worthy role model. Fig-
ure 3b depicts the unconstrained structural mediation model for
leaders who were seen by their groups as highly worthy role
models to follow (n!34; factor M!0.75) and for leaders who
were rated less favorably on this leadership quality (n!33; factor
M!*0.77). We use relative terms to describe the two groups
because the raw score rating of the leaders in our sample were
generally favorable (ranging between 4.46 and 6.97), consistent
with the typical leniency bias in leadership studies (e.g.,
Schriesheim et al., 1995; Shamir et al., 1998). Thus, the less
favorable role models still rated rather high on this quality (raw
score M!5.8), although meaningfully lower that the highly
worthy role models (raw score M!6.6).
R
2
=.22
Group
OCB
Model: Full Mediation
.55
Industry
e9
.74
.63
Loyalty
e8
.78
Initiative
e7
.88
Leader
OCB
.73
Initiative
e1
.67
Loyalty
e2
.65
Industry
e3
.86
.82
.81
R
2
=.15
Group belief
about OCB
.81
Value3
e6
.65
Value2
e5
.80
Value1
e4
.90
.81
.89
Z1 Z2
.38**
(s.e.=.13)
.47*
(s.e.=.15)
.80
.00
Chi-square = 34.97; df = 29; p < .22; CFI = .98; GFI =.90; TLI = .98; RMSEA = .05; RMR = .06
R
2
=. 25
Group
OCB
Model: Partial Mediation
.55
Industry
e9
.74
.63
Loyalty e8
.78
Initiative
e7
.88
Leade
r
OCB
.73
Initiative
e1
.68
Loyalty
e2
.65
Industry
e3
.86
.82
.80
R
2
=.14
Group belief
about
OCB
. 81
Value3
e6
.65
Value2
e5
.80
Value1
e4
.90
.81
.89
Z1 Z2
.38**
(s.e.=.13)
.39*
(s.e.=.15)
.80
.20
(s.e.=.16)
Chi-square = 32.94; df = 28; p < .24; CFI = .99; GFI = .90; TLI = .98; RMSEA = .05; RMR = .04
Figure 2. Partial and full mediation models of the leader organizational citizenship behavior (OCB). Standardized
path coefficients are the means of bootstrap parameter estimates from 2,000 independent samples drawn from the
original sample. Bootstrap standard errors are shown in parentheses. All factor loadings are significant at the .005 or
.001 level. Two-tailed significance is determined by bias-corrected confidence intervals. s.e. !standard error; df !
degrees of freedom; CFI !comparative fit index; GFI !goodness of fit index; TLI !Tucker–Lewis index;
RMSEA !root-mean-square-error of approximation; RMR !root-mean-square residual.
!
p#.05.
!!
p#.01.
11
LEADING BY EXAMPLE
Nested model comparisons (see Table 2) indicated significant dif-
ferences across the levels of this moderator in the effects of leader
OCB both on group belief and on group behavior, implying a pattern
of a first stage and direct effect moderated mediation model (Edwards
&Lambert,2007).Thatis,equalityconstraintsbothonthepathfrom
leader OCB to group belief about OCB ($%
1
2
!5.6, p#.05), and on
the path from leader OCB to group OCB ($%
1
2
!3.8, p#.05), across
the role model levels, significantly decreased the model fit to the data,
but the equality constraint on the paths from group belief to group
OCB did not ($%
1
2
!0.84, ns). Figure 3b shows that the effect of
group belief that OCB is worthy on group OCB was significant for
both higher and lower role models ('!.41, p#.01 and '!.35, p#
.01, respectively). However, the effects of leader OCB on belief about
OCB and on group OCB were significant only for those leaders who
were seen by their groups as highly worthy role models ('!.72, p#
.001 and '!.50, p#.005, respectively). These results support
Hypothesis 6. Table 3 shows that for leaders who were seen by their
groups as highly worthy role models, leader OCB had a significant
direct effect (.50, p#.005) as well as a significant indirect effect (.30,
p#.005), which together produced a significant total effect on group
OCB (.79, p#.005). Modeling of OCB did not seem to exert any
(direct or indirect) effect on group behavior for leaders who were not
seen by their groups as highly worthy role models.
Group consensus about the value of OCB. Figure 3c depicts
the unconstrained structural mediation model for groups with high
(n!34; mean SD !1.16) and relatively low (n!33; mean SD !
.78) levels of consensus about the value of OCB. The group consensus
was determined based on group within-group variability (i.e., standard
deviation) in the group belief that OCB is worthy. Nested model
comparisons (see Table 2) indicated that the only significant differ-
ence between the levels of group consensus was in the effect of the
group belief on group behavior, implying a pattern of a second stage
moderated mediation model (Edwards & Lambert, 2007). That is,
equality constraints on the path from group belief about OCB to
group-level OCB across the levels of group consensus significantly
decreased the model fit to the data ($%
1
2
!8.6, p#.01), whereas
equality constraints on the path from leader OCB to group belief
about OCB ($%
1
2
!1.6, ns), and to group OCB ($%
1
2
!.00), did not.
In support of Hypothesis 7, Figure 3c indicates that the positive effect
of the group belief that OCB is worthy on group OCB was significant
when the group consensus about the value of OCB was high ('!.77,
p#.001) rather than merely sufficient ('!.00). Table 3 also shows
that when the group consensus was high, leader OCB had a significant
indirect effect (.34, p#.01), thus yielding a significant total effect
(.48, p#.05) on group OCB. When the group consensus about the
value of OCB was lower, leader OCB had no significant effect on
group OCB.
Although we did not formulate hypotheses about the depen-
dence of indirect effects and the mediation pattern on the
moderators, Table 3 reveals several interesting findings. First,
the direct effect of the leader was significant in two conditions:
when the leader was close and when the leader was seen by the
group as a highly worthy role model. In both cases, the leader’s
OCB had significant direct and indirect effects, implying that
the group belief about OCB only partially mediates the effect of
the leader’s OCB. Second, the indirect effect of the leader’s
OCB was not significant in two conditions: when the leader was
not seen by the group as a highly worthy role model and when
the consensus about the value of OCB was low. Third, the
criteria for establishing full mediation were satisfied only in the
high consensus condition. That is, this is the only case in which
leader OCB had significant indirect and total effects but a
nonsignificant direct effect (see Mathieu & Taylor, 2006).
Thus, although the mediational analysis of the entire sample
data indicated that the belief about OCB may fully account for
the effect of leader OCB on group OCB, the moderated medi-
ation analyses suggested that the extent to which the belief in
the value of OCB fully or partially mediates the effect of leader
OCB is dependent on the moderators in our model.
Complementary Analyses
Despite the advantages of SEM, the multigroup analysis has two
shortcomings. First, each moderator was examined separately
without controlling for the other moderators; thus, the unique
contribution of each moderating effect to group OCB is unclear.
Table 2
Nested Model Comparisons Testing the Differences Between Path Coefficients Across the Levels of the Moderators
Model
Moderator
Distance Role model Group consensus
1. Unrestricted measurement model %
48
2
!56.1, ns, CFI !0.97, TLI !
0.96, RMSEA !.05
%
48
2
!54.7, ns, CFI !0.98, TLI !
0.97, RMSEA !.05
%
48
2
!52.3, ns, CFI !0.98, TLI !
0.98, RMSEA !.04
2. Unrestricted structural model
a
%
58
2
!62.7, ns, CFI !0.99, TLI !
0.98, RMSEA !.04
%
58
2
!63.1, ns, CFI !0.98, TLI !
0.97, RMSEA !.04
%
58
2
!61.8, ns, CFI !0.99, TLI !
0.99, RMSEA !.03
$%
10
2
!6.60 $%
10
2
!8.42 $%
10
2
!9.43
3. AL !AH
b
$%
1
2
!0.06 $%
1
2
!5.6
!
$%
1
2
!1.6
4. BL !BH $%
1
2
!0.38 $%
1
2
!0.84 $%
1
2
!8.6
!!
5. CL !CH $%
1
2
!5.6
!
$%
1
2
!3.8
!
$%
1
2
!0.00
Note. ns !nonsignificant; CFI !comparative fit index; TLI !Tucker–Lewis index; RMSEA !root-mean-square error of approximation.
a
In the unrestricted measurement model, all the parameters are free to vary. In the unrestricted structural model, structural path coefficients are free to vary,
but the measurement model is restricted in two ways. First, as in the mediation test, at each level of the moderator the three scales are enforced to be parallel
measures of organizational citizenship behavior (OCB) and invariant across the rating of the group and the leader OCB. Second, the factor loadings in the
latent factors are enforced to be equal across the level of the moderator.
b
In Models 3–5, the structural paths (a, b, c) on the mediation model are
constrained to equality across the two levels of the moderator (L !low, H !high).
!
p#.05.
!!
p#.01.
12 YAFFE AND KARK
Second, continuous variables were dichotomized in a common but
arbitrary way (i.e., median split).
14
To overcome these shortcom-
ings, we applied Preacher et al.’s (2007) SPSS macro to compute
bootstrap estimates for a conditional indirect effect in a moderated
mediation model in which the indirect paths to and from the
mediator (a and b) were moderated by different variables. This
macro incorporates Baron and Kenny’s (1986) stepwise procedure
with the bootstrap method. We used this macro to test the signif-
icance of the indirect effect of leader OCB, considering its con-
tingency both on the status of the leader as a role model and on the
group consensus about the value of OCB (treated as continuous
variables), while controlling for the moderating effect of leader
distance. In this analysis we also controlled for group size, leader
tenure, and group tenure, which might theoretically affect group-
level OCB. Consistent with the conceptualization of OCB as a
common factor underlying its dimensions and with the SEM
14
Dichotomizing quantitative measures results in a loss of information
and havoc with regard to estimation and interpretation of relationships
among variables (MacCallum et al., 2002). A simulation of the methods
used by Bissonnette, Ickes, Bernstein, and Knowles (1990) to study mod-
erating effects showed that the dichotomization approach yielded higher
Type I error rates, indicating detecting spurious interactions, and less
power to detect interactions that were actually present in the population,
than did the standard regression approach.
C.1. Low Consensus
LOCB GOCB
GBAO
= .00
(s.e.= .24)
= .21
(s.e.= .25)
R
2
=.04
R
2
=.04
C.2. High Consensus
LOCB GOCB
GBAO
= .77**
(s.e.= .16)
= .14
(s.e.= .17)
= .45**
(s.e.= .14)
R
2
=.67
R
2
=.20
= .48**
(s.e.=.15)
A.2. Close Leaders
GOCB
GBAO
=. 34**
(s.e. = .15)
= .36**
(s.e.= .16)
R
2
= .46
LOCB
R
2
= .13
= .21
(s.e.=.27)
Chi-square=62.94; df = 58; p < .32; CFI = .99; GFI= = .85; TLI = .98; RMSEA= .04 ; RMR = .05
A.1. Distant Leaders
LOCB GOCB
GBAO
= .38**
(s.e .= .16)
= .36**
(s.e .= .19)
R
2
= .12
= -.11
(s.e. = .25)
R
2
= .13
GOCB
GBAO
B.1. Low Role Models
= .35**
(s.e.= .20)
= -.04
(s.e.= .17)
= .00
(s.e.= .19)
R
2
=
.12
LOCB
R
2
= .00
Chi-square=61.76; df = 58; p < .42; CFI = .99; GFI=.84; TLI = .99; RMSEA= .03 ; RMR = .05
LOCB GOCB
GBAO
B.2. High Role Models
= .41**
(s.e.= .24)
=.50**
(s.e.= .26)
= .72**
(s.e.= .14)
R
2
= .71
R
2
= .51
Chi-square=63.12; df = 58; p < .30; CFI = .98; GFI=.84; TLI = .98; RMSEA= .04; RMR = .06
Figure 3. Conditional effects of leader organizational citizenship behavior (OCB). Conditional effects were tested
in an unconstrained structural model but a restricted model of measurement invariance across groups (i.e., factor
loadings are constrained to equality, but structural paths are free of constraints). Bootstrap standardized paths
coefficients are the means of parameter estimates from 2,000 independent samples drawn from the original sample.
Bootstrap standard errors are shown in parentheses. Two-tailed significance is determined by bias-corrected confi-
dence intervals. GBAO !group belief about OCB; LOCB !leader OCB; GOCB !group-level OCB. s.e. !
standard error; df !degrees of freedom; CFI !comparative fit index; GFI !goodness of fit index; TLI !
Tucker–Lewis index; RMSEA !root-mean-square-error of approximation; RMR !root-mean-square residual.
!!
p#.01.
13
LEADING BY EXAMPLE
analysis, the latent variables in the model were represented by
maximum likelihood factor scores. The results of this analysis
(shown in Table 4) indicated that when considered simultaneously,
all the moderating effects were significant and support the
hypothesized unique contribution of each moderating effect.
The macro also provides regions of significance of the bootstrap
estimates of the conditional indirect effect (assuming a normal
distribution of the indirect effect), as well as bootstrap estimates
based on biased-corrected and accelerated confidence intervals.
These outputs suggest that the probability of an indirect effect
increases with the enhancement both in group consensus about
the value of OCB and in group belief that the leader is a highly
worthy role model. The indirect effect of leader OCB was the
strongest when both of these are high but was also present when
the level of at least one of these moderators was high and the
level of the other was not low.
It is noteworthy that when we considered all the moderators
simultaneously, the effects of leader OCB and the belief about
OCB were still significant, and group belief about OCB apparently
only partially mediated the effect of leader OCB on group OCB.
Besides these variables, only leader tenure and group tenure had
significant main effects on group-level OCB. However, group size
had a marginally significant negative effect on group belief that
OCB is worthy.
Finally, because distance in this study was defined as between-
department differences, it might have confounded other differ-
ences between the departments, such as task type. We thus rean-
alyzed the data with the same macro, excluding the distance and
controlling for the department (recoded as two dummy variables).
The results were similar, indicating that the indirect effect was
dependent both on group consensus and on the belief that the
leader was a highly worthy role model. Although the level of group
OCB varied between departments, there was no significant depart-
ment effect when group size and tenure variables were controlled
for.
Discussion
The current study explored “leading by example” in work or-
ganizations, focusing on the effect of leader OCB on work-group
OCB. The results support the hypothesized moderated mediation
model and enhance our understanding of the processes underlying
leading by example effects, as well as their boundary conditions.
This study underscores the importance ascribed to leading by
example in leadership theories by providing evidence that exem-
plary leader behavior may promote group-level behaviors that
enhance organizational effectiveness (Organ et al., 2006). The
significant effect of leader OCB on group OCB implies that by
setting a personal example of contribution to the organization,
leaders can promote similar contributions from their groups.
We looked at two processes that may explain the leading-by-
example effect: a direct process, which presumably involves mere
emulation of the leader’s behaviors, and an indirect process, by
which the leader’s OCB enhances the group’s belief that OCB is
worthy and in turn enhances group OCB. The SEM results sup-
ported the mediating role of the group belief that OCB is worthy,
in the relationship between leader OCB and group OCB. They also
indicated that the extent to which the belief about OCB fully or
partially mediates the effect of leader OCB is dependent on the
moderators in our model. Furthermore, the moderated mediation
patterns indicated that both the direct and the indirect processes
may be viable when essential conditions are met. That is, condi-
tions that promote social learning of leader behavior moderated the
direct effect of leader OCB, and conditions that promote transfor-
mational process through group belief about OCB moderated the
indirect effect of leader OCB.
In particular, leader proximity and group belief that the leader is
a highly worthy role model facilitated the direct effect of leader
OCB, presumably by providing preconditions of social learning
(i.e., the ability to observe the behaviors and the motivation to
emulate them; Bandura, 1977). Leader OCB had no direct effect
Table 3
Conditional Effects of the Leader OCB
Moderator
Stage Effect
First (a) Second (b) Direct (c&) Indirect Total
Distance
Low (J!37) .36 (.16)
!!
.34 (.15)
!!
.48 (.15)
!!
.12 (.08)
!!
.60 (.12)
!!
High (J!30) .36 (.19)
!!
.38 (.16)
!!
*.11 (.25) .13 (.08)
!!
.03 (.27)
Role model
Low (J!33) .00 (.19) .35 (.20)
!!
*.04 (.17) .00 (.07) *.04 (.19)
High (J!34) .72 (.14)
!!
.41 (.24)
!!
.50 (.26)
!!
.30 (.20)
!!
.79 (.11)
!!
Group consensus
Low (J!34) .21 (.25) .00 (.24) .21 (.27) .00 (.10) .21 (.23)
High (J!33) .45 (.14)
!!
.77 (.16)
!!
.14 (.17) .34 (.12)
!!
.48 (.17)
!
Note. Bootstrap standardized estimates are the means of parameter estimates based on 2,000 independent
samples drawn from the original sample. Bootstrap standard errors are shown in parentheses. The number of
work units (J) was determined by the median split, except for the distance moderator, for which the levels were
determined by the organization structure. Indirect effects are approximately the multiplication of the bootstrap
estimates for indirect paths from the independent variable to the mediator (Stage 1) and from the mediator to the
dependent variable (Stage 2). Group consensus is measured by the standard deviation; the higher the score the
lower the consensus. Two-tailed significance is based on the bias-corrected confidence intervals.
!
p#.05.
!!
p#.01.
14 YAFFE AND KARK
on group OCB when the leader was distant and when the group did
not consider the leader to be a highly worthy role model.
Both the belief that the leader is a highly worthy role model and
a high consensus about the value of OCB facilitated the indirect
effect of leader OCB. The status of the leader as a highly worthy
role model enhanced the effect of leader OCB on the group belief
that OCB is worthy, and high consensus about the value of OCB
enhanced the effect of the group belief that OCB is worthy on
group OCB. Leader OCB had no indirect effect on group OCB
when the group did not view the leader as a highly worthy role
model and when the group consensus about the value of OCB was
not as high. Although social learning may also be involved in
transmitting OCB-related values and beliefs (Bandura, 2003), this
process is more closely addressed by transformational and charis-
matic leadership theories (e.g., Bass, 1985; Shamir et al., 1993).
The effect of leaders’ OCB on the value that groups ascribe to
OCB supports these theories’ propositions that role modeling of
contributions to the collective is a major means by which effective
leaders transform followers’ values and attitudes and change fol-
lowers’ focus from self-interest to collective interest (e.g., Bass,
1985; Shamir et al., 1993). The indirect effect of leader OCB on
group OCB, through the group belief that OCB is worthy, is
consistent with the process described by Shamir et al. (1993,
1998). In this process, a leader’s exemplary contributions enhance
the group’s collectivistic orientation, so that group members’ con-
tributions to the group become self-expressive and intrinsically
satisfying.
The moderators examined here advance our understanding of
the processes underlying the leading-by-example effects and of
their boundary conditions. As discussed below, each moderator
uniquely shaped the effects of leader OCB.
Leader distance significantly reduced the direct effect of leader
OCB, supporting the importance of visibility for social learning
process. Close leaders’ OCB had both direct and indirect effects on
group OCB, but the direct effect explained a large portion of the
total effect. Distant leaders’ OCB had only a minor significant
Table 4
Regression Results for Conditional Indirect Effect
Independent variable B SE 'tp
Mediator variable model (belief about OCB)
Group size (n)*0.06 .03 *.22 *1.78 .08
Leader tenure 0.03 .09 .03 0.29 .77
Group tenure *0.20 .18 *.14 *1.13 .26
Distance (D) 0.22 .22 .12 1.03 .31
Role model (RM) 0.30 .11 .30 2.59 .01
Leader OCB (LOCB) 0.31 .12 .30 2.64 .01
LOCB (D 0.25 .22 .12 1.13 .27
LOCB (RM .28 .09 .37 3.29 .01
Dependent variable model (group OCB)
Group size (n)*0.02 .03 *.06 *0.61 .55
Leader tenure 0.14 .07 .18 2.05 .05
Group tenure 0.28 .14 .19 2.00 .05
Distance (D) 0.14 .17 .08 0.83 .41
Role model (RM) *0.03 .09 *.03 *0.31 .76
Belief variability (BV) *0.38 .35 *.10 *1.07 .29
Leader OCB (LOCB) 0.30 .10 .30 3.03 .01
OCB is worthy 0.27 .11 .28 2.54 .01
LOCB (D*0.46 .17 *.23 *2.63 .01
LOCB (RM 0.18 .07 .25 2.54 .01
BAO (BV *1.16 .42 *.25 *2.73 .01
Role model
Belief
variability
Boot indirect
effect
Boot
SE
Boot
z
Boot
p
95% bias–corrected CI
Lower Upper
Conditional indirect effect effects at specific values of the moderators, M+1SD
*1SD *1SD .02 .08 0.23 .82 *.19 .31
*1SD M .01 .04 0.21 .82 *.10 .17
*1SD ,1SD .00 .02 *0.01 .99 *.07 .06
M*1SD .17 .08 2.02 .04 .02 .44
MM .08 .05 1.77 .08 .01 .23
M,1SD .00 .04 *0.05 .96 *.12 .07
,1SD *1SD .32 .13 2.48 .01 .12 .73
,1SD M .16 .08 2.01 .04 .04 .39
,1SD ,1SD .00 .08 *0.05 .96 *.18 .13
Note. Preacher et al.’s (2007) macro output is based on 5,000 bootstrap samples. SD !standard deviation; CI !confidence interval.
15
LEADING BY EXAMPLE
indirect effect, which was not sufficient to produce a significant
total effect on group OCB.
15
The sizable effect of the proximal
leader on group-level OCB challenges the assumption that physi-
cal proximity and frequent direct interactions limit the outcomes of
close leaders to the individual level (Antonakis & Atwater, 2002).
Close leaders apparently can improve group-level outcomes by
role modeling behaviors that are directed to this level (Hackman,
1992; Kark & Shamir, 2002; Kark & Van Dijk, 2007). On the other
hand, because distance did not moderate the indirect effect of the
leader OCB, our results also imply that physical proximity per se
cannot explain the transformational influence that leaders may
have on their group OCB-related values and beliefs. In light of the
increasing prevalence of “leading at a distance” in virtual organi-
zations, multinational firms and domestic companies with widely
dispersed sites (Howell, Neufeld, & Avolio, 2005), this issue
merits further study.
Group belief that the leader is a highly worthy role model
enhanced the direct effect of leader OCB on group OCB as well as
on the group belief that OCB is worthy and, as a result, on the
indirect effect of leader OCB. Leader OCB had significant direct
and indirect effects on group-level OCB and thus led to a remark-
able total effect for leaders who were highly worthy role models
for their groups. Leader OCB had no (direct or indirect) effect on
group OCB for leaders who were less admirable role models. This
finding suggests that hierarchical status per se does not ensure any
impact on group-level OCB. Instead, leaders must be perceived by
their groups as highly worthy role models (rather than just good
role models) in order to influence not only group-level behaviors
but also values and beliefs related to group-level OCB. The OCB
of leaders who are not viewed by their group as highly worthy role
models is apparently far less informative with regard to appropri-
ate behavior for group members, and it does not have a symbolic
value that may shape the group valuation of OCB. As posited by
the theory of social identity of leadership, leadership is fundamen-
tally a process of group influence, and a formal leader is really a
leader only to those who are willing to follow (Hogg, 2001). In line
with the theory of distant leadership (Antonakis & Atwater, 2002),
the status of the leader as a role model was not related to the leader
distance. It was, however, related to leader OCB, implying that by
modeling OCB, leaders may enhance their worthiness as role
models, thereby increasing their impact on their groups. Note,
however, that despite this positive relationship, the effect of lead-
ers who are viewed by their groups as highly worthy role models
also implies emulation of avoidance of low level of the leader
OCB (Bolino & Turnley, 2003).
The group consensus moderated the effect of the group-level
belief that OCB is worthy on the group-level OCB and, as a result,
the indirect effect of leader OCB on group OCB. These effects
were significant for groups with a high consensus about the value
of OCB but were nonsignificant when the group consensus was not
as high. This finding is important for several reasons. First, it
suggests that the influence process at the group level is dependent
not only on the leader’s impact on individuals but also on the
leader’s ability to facilitate group processes (Zaccaro et al., 2001).
Second, it supports the idea that the development of a consensual
belief in the value of OCB is a process that underlies the indirect
effect of leader OCB on group OCB. Third, the consensual nature
of the forces that drive OCB in work groups implies that group-
level OCB may explicitly or implicitly be a collective, coordinated
action (rather than a mere aggregation of individuals’ OCB) and
thus be qualitatively different from individual-level OCB (Ehrhart
& Naumann, 2004). Moreover, the results showed that group
beliefs about OCB had a significant effect on group OCB, regard-
less of the level of leader distance and the group belief about the
leader as a role model. However, this belief fully mediated the
effect of leader OCB only when the group developed a high
consensus about the value of OCB. Thus, the results are indicative
about the potential as well as the limitations of modeling behavior
to inspire a transformational process that targets collective perfor-
mance (Zaccaro et al., 2001). Future studies could explore how and
when leaders enhance consensual beliefs in their groups by con-
sidering structural features beyond the leader’s control that shape
the frequency and quality of interactions among group members
(Luria, 2008).
Practical Implications
There is a lack of consensus about the role of formal mecha-
nisms promoting OCB at the organizational level. Some research-
ers have argued that any behavior that is important for organiza-
tional functioning should be included in a formal job description
and appraisal system (e.g., Tansik, 1990). Others have pointed out
that this could be impossible and is probably undesirable. Many
role behaviors, particularly in service firms, are unique ways of
problem solving and thus cannot be fully specified in advance
(Katz & Kahn, 1978; Morrison, 1996). Moreover, rigid job de-
scriptions, rules, and regulations may reduce intrinsic motivation
(Deci, 1975) and may impede initiative and flexibility in the face
of unforeseen contingencies (Smith et al., 1983). For this reason,
informal influence processes, such as a leader’s role modeling,
seem particularly important in cultivating OCB.
The results of this study suggest that by modeling OCB, leaders
can educate work groups about the importance of contributions in
the form of OCB and at the same time make leaders deserving of
such contributions (Bolino & Turnley, 2003). Thus, organizations
seeking to promote OCB should devote special attention to human
resource management practices that are likely to enhance OCB
through the organization’s leaders. To begin with, organizations
can adopt selection practices that consider the tendency of appli-
cants to engage in OCB (Bolino & Turnley, 2003). Promoting
leaders’ OCB also requires investment in leadership development
programs aimed at competence building. Ready and Conger (2003)
observed that many leadership development efforts fail due to their
“product-focused quick-fix mentality” instead of a willingness to
invest in building leadership competencies. Our results imply a
high return on investment that educates leaders to lead by example
in general and to set a personal example of OCB in particular.
Using high-potential role models as instructors in the leadership
development process can help foster exemplary leadership
throughout work units and the organization (Day & Halpin, 2001).
15
The indirect effect of distant leaders may indicate that even if group
members do not meet their leader daily, they may develop a shared
understanding of the leader’s OCB-related values and norms with time and
cumulative interactions (Shamir & Lapidot, 2003). Alternatively, the leader
may communicate his or her citizenship-related values and beliefs via
mediated communication (Avolio, Kahai, Dumdum, & Sivasubramaniam,
2001).
16 YAFFE AND KARK
Finally, Morrison’s (1996) model suggests that selection, social-
ization, and other human resource management practices are more
apt to ensure leaders’ OCB if they emphasize broadly defined
obligations and mutual long-term commitment, which are the
cornerstones of social-exchange relationships.
Strengths and Limitations
The contribution of the current study should be considered in
light of its limitations. Most obvious is the study design, which
prevents firm causal conclusions. The effect of the leader’s con-
tribution on the group-level OCB is implied in charismatic lead-
ership theory (e.g., Shamir et al., 1993) and has been reported in
laboratory studies (Gu¨th et al., 2007; Potters et al., 2007). How-
ever, in natural organizations the relationship observed between
leader and group OCB is likely to be reciprocal. Such a reciprocal
influence is consistent with Burns’s (1978) view of transforma-
tional leadership as a process by which “leaders and followers
raise one another to higher levels of morality and motivation” (p.
20). Longitudinal field study should explore the potential effects of
work groups’ OCB on leaders.
Although using data from a single organization eliminates many
potentially confounding factors, such as type of industry, organi-
zational culture, and other systemic factors that may affect OCB
(MacKenzie et al., 1999), it may limit the results’ generalizability
(Ehrhart, 2004).
We followed previous studies (e.g., Farh et al., 2001; LePine &
Van Dyne, 1998; MacKenzie et al., 1999; Morrison & Phelps,
1999) in using the same measure to assess the level of managers’
and employees’ OCB. This procedure is vital for the study of
modeling OCB (Ehrhart & Naumann, 2004) and was justified by
the similar factorial structure of the measures of leader OCB and
group OCB. Nevertheless, we cannot disregard the possibility that
at least some OCBs, such as taking charge, are more likely to be
expected from managers than employees. This calls for two com-
ments. First, taking charge was regarded by the focal organization
as manager OCB according to Organ’s (1988) criteria: It was not
an explicit part of their job description; they were not trained by
the organization to initiate change and were not formally rewarded
for doing so. Second, the broad definition of OCB as proactive
responsible participation in an organization is theoretically inde-