trait and behavioral theories of leadership: an ......(c) follower satisfaction with leader, and (d)...
TRANSCRIPT
-
PERSONNEL PSYCHOLOGY2011, 64, 7–52
TRAIT AND BEHAVIORAL THEORIESOF LEADERSHIP: AN INTEGRATION ANDMETA-ANALYTIC TEST OF THEIR RELATIVE VALIDITY
D. SCOTT DERUEStephen M. Ross School of Business
University of Michigan
JENNIFER D. NAHRGANGW.P. Carey School of Business
Arizona State University
NED WELLMANStephen M. Ross School of Business
University of Michigan
STEPHEN E. HUMPHREYSmeal College of Business
The Pennsylvania State University
The leadership literature suffers from a lack of theoretical integra-tion (Avolio, 2007, American Psychologist, 62, 25–33). This arti-cle addresses that lack of integration by developing an integrativetrait-behavioral model of leadership effectiveness and then exam-ining the relative validity of leader traits (gender, intelligence,personality) and behaviors (transformational-transactional, initiat-ing structure-consideration) across 4 leadership effectiveness crite-ria (leader effectiveness, group performance, follower job satisfac-tion, satisfaction with leader). Combined, leader traits and behaviorsexplain a minimum of 31% of the variance in leadership effectivenesscriteria. Leader behaviors tend to explain more variance in leadershipeffectiveness than leader traits, but results indicate that an integrativemodel where leader behaviors mediate the relationship between leadertraits and effectiveness is warranted.
Leadership is one of the most discussed and debated topics in thesocial sciences (Avolio, Sosik, Jung, & Berson, 2003; Bass, 1990;Bennis, 2007). Research on leadership began with a search for herita-ble attributes that differentiated leaders from nonleaders and explainedindividuals’ effectiveness as leaders (Galton & Eysenck, 1869). In effect,this early research was the beginning of the trait paradigm of leadership re-search. Subsequent studies have established that individual characteristics,
Correspondence and requests for reprints should be addressed to D. Scott DeRue,Assistant Professor of Management and Organizations, Stephen M. Ross School of Busi-ness, University of Michigan, 701 Tappan Street, Ann Arbor, MI 48109; [email protected]© 2011 Wiley Periodicals, Inc.
7
-
8 PERSONNEL PSYCHOLOGY
such as demographics, skills and abilities, and personality traits, predictleadership effectiveness (e.g., Eagly, Karau, & Makhijani, 1995; Judge,Bono, Ilies, & Gerhardt, 2002; Judge, Colbert, & Ilies, 2004; Mumford,Campion, & Morgeson, 2007).
Critiques of the leader trait paradigm (Jenkins, 1947; Mann, 1959;Stogdill, 1948) prompted scholars to look beyond leader traits andconsider how leaders’ behaviors predicted effectiveness. This led toresearch on initiation of structure and consideration (Hemphill &Coons, 1957; Stogdill, 1963), and established the behavior paradigm ofleadership research. The influence of the leader behavior paradigm can beseen across leadership theories, including Fiedler’s (1967) contingencymodel, Blake and Mouton’s (1964) managerial grid, and the work ontransformational and transactional leadership (the full range model ofleadership; Avolio et al., 2003; Bass, 1985; Podsakoff, MacKenzie,Moorman, & Fetter, 1990). Not only did the leader behavior paradigmprovide the basis for new theory, but meta-analytic evidence also suggeststhat leader behaviors are important predictors of leadership effectiveness(Judge & Piccolo, 2004; Judge, Piccolo, & Ilies, 2004).
Both leader traits and behaviors have been investigated in scores of re-search studies. Despite the theoretical and applied value of these studies,leadership research is plagued by a lack of integration. In fact, schol-ars dating back to Bennis (1959) and as recently as Avolio (2007) havelamented over the proliferation and lack of integration of leadership theo-ries and constructs. The primary criticism is that leadership scholars createnew theories of leadership without attempting to compare and contrast thevalidity of existing theories.
The lack of integration in leadership research is evident both withinand across the trait and behavior paradigms, as research within eachparadigm generally focuses on a single trait or behavioral perspective. Forexample, within the trait paradigm, Eagly, Johannesen-Schmidt, and vanEngen (2003) provided meta-analytic estimates for gender and leadershipeffectiveness, whereas Judge et al. (2002, 2004) did the same for person-ality and intelligence, respectively. None of these studies controlled for orcompared the effects of different traits, such as gender, personality, andintelligence concurrently. This lack of integration is problematic giventhat many of these studies found similar effect sizes across leader traits.For example, Judge et al. (2002) found absolute effect sizes ranging from.16 to .24 for personality and leadership effectiveness, whereas Judge et al.(2004) found an effect size of .21 for intelligence. However, because therewas no integration across traits, it remains unclear as to whether these areindependent effects.
Similarly, research within the leader behavior paradigm often focuseson a single behavioral perspective. For example, Judge and Piccolo (2004)
-
D. SCOTT DERUE ET AL. 9
meta-analyzed the literature on transformational and transactional leader-ship, and Judge, Piccolo, and Ilies (2004) did the same for initiating struc-ture and consideration. Neither of these studies integrated across leaderbehaviors or considered whether the effects were independent. Yet, ini-tiating structure and transactional leadership both focus on task-orientedleader behaviors, whereas consideration and transformational leadershipboth comprised relational-oriented leader behaviors (Bass & Bass, 2008;Fleishman, 1953). Given the conceptual similarity, it is not surprising thatseparate meta-analyses found similar effect sizes—for example, overallvalidities of .41 for consideration and .44 for transformational (Judge &Piccolo, 2004; Judge et al., 2004). Thus, the two leader behavior paradigmsthat have shaped leadership research for decades may not be independent,and even more importantly, it is unclear if one is a better predictor ofleadership effectiveness.
This article reviews and integrates the literature on leader traits andbehaviors, and takes a first step toward an integrative theory of how leadertraits and behaviors influence leadership effectiveness. To accomplishthis, we follow a three-stage process. First, based on a narrative reviewof the literature, we develop a conceptual model that organizes the cur-rent literature and models how leader traits and behaviors affect leadershipeffectiveness (see Figure 1). Second, we empirically test the relative valid-ity of select leader traits and behaviors using a combination of previouslypublished meta-analytic data and new meta-analyses. Third, we investi-gate an exemplary set of relationships from our conceptual model to seeif leader behaviors are one possible mechanism through which individualtraits influence leadership effectiveness.
Conceptualizing Leadership Effectiveness
Before presenting our integrative model, we first define the leadershipeffectiveness domain. Scholars often vary in their definition of leadershipeffectiveness (Avolio et al., 2003; Yukl, 2006), which is one reason theliterature is not well integrated. Based on our review of the literature, lead-ership effectiveness criteria can be conceptualized along three dimensions:(a) content, (b) level of analysis, and (c) target of evaluation. As shownin Figure 1, the content of leadership effectiveness can relate to task per-formance (e.g., individual or group performance), affective and relationalcriteria (e.g., satisfaction with the leader), or overall judgments of effec-tiveness that encompass both task and relational elements (e.g., overalleffectiveness of the leader). The level of analysis corresponds to whetherleadership effectiveness is conceptualized at the individual, dyadic, group,or organizational level. For example, some studies conceptualize leader-ship effectiveness as individual-level leader effectiveness, whereas other
-
10 PERSONNEL PSYCHOLOGY
Lea
der
Tra
its
& C
har
acte
rist
ics
Dem
ogra
phic
s
Task
Com
pete
nce
Inte
rper
sona
l Att
ribu
tes
Task
-ori
ente
d R
elat
iona
l-or
ient
ed
Lea
der
Beh
avio
rs
Att
ribu
tion
Pro
cess
es
Iden
tifi
catio
n P
roce
sses
Att
ribu
tion
s &
Rel
atio
nal
Ass
ocia
tion
s
Lea
ders
hip
Eff
ecti
ven
ess
Con
tent
Lev
el o
f A
naly
sis
Targ
et o
f E
valu
atio
n •I
mpl
icit
lead
ersh
ip
theo
ry
•Lea
der
prot
otyp
es
• Lea
der-
foll
ower
perc
eive
d si
mil
arit
y •I
dent
ific
atio
n w
ith
lead
er
•Gro
up id
enti
fica
tion
Cha
nge-
orie
nted
•Ini
tiat
ing
stru
ctur
e •C
onti
ngen
t rew
ard
•Man
agem
ent b
y ex
cept
ion-
acti
ve•B
ound
ary
span
ning
•D
irec
tive
•Con
side
rati
on
•Em
pow
erm
ent
•Par
tici
pati
ve
•Dev
elop
ing
•Ena
blin
g •S
erva
nt le
ader
ship
•Tra
nsfo
rmat
iona
l •C
hari
smat
ic,
insp
irat
iona
l
•Gen
der
•Age
•Eth
nici
ty
•Hei
gh•E
duca
tion
, soc
ial s
tatu
s
•Int
elli
genc
e •C
onsc
ient
ious
ness
•O
penn
ess
to e
xper
ienc
e •E
mot
iona
l sta
bili
ty
•Tec
hnic
al k
now
ledg
e •L
eade
rshi
p se
lf-e
ffic
acy
•Ext
rave
rsio
n• A
gree
able
ness
•C
omm
unic
atio
n sk
ills
•E
mot
iona
l int
elli
genc
e •P
olit
ical
ski
lls
•Ove
rall
•T
ask
(e.g
., pe
rfor
man
ce)
•Affe
ctiv
e/re
lati
onal
(e.
g.,
foll
ower
sat
isfa
ctio
n, le
ader
-m
embe
r ex
chan
ge)
•Ind
ivid
ual
•Dya
d •G
roup
•O
rgan
izat
ion
•Lea
der
•Oth
er (
e.g.
, fol
low
er, g
roup
, or
gani
zati
on)
Pass
ive
Lea
ders
hip
•Man
agem
ent b
y ex
cept
ion-
pass
ive
•Lai
ssez
-fai
re
Lea
der
Tra
its
& C
har
acte
rist
ics
Dem
ogra
phic
s
Task
Com
pete
nce
Inte
rper
sona
l Att
ribu
tes
Task
-ori
ente
d R
elat
iona
l-or
ient
ed
Lea
der
Beh
avio
rs
Att
ribu
tion
Pro
cess
es
Iden
tifi
catio
n P
roce
sses
Att
ribu
tion
s &
Rel
atio
nal
Ass
ocia
tion
s
Lea
ders
hip
Eff
ecti
ven
ess
Con
tent
Lev
el o
f A
naly
sis
Targ
et o
f E
valu
atio
n •I
mpl
icit
lead
ersh
ip
theo
ry
•Lea
der
prot
otyp
es
• Lea
der–
foll
ower
perc
eive
d si
mil
arit
y •I
dent
ific
atio
n w
ith
lead
er
•Gro
up id
enti
fica
tion
Cha
nge-
orie
nted
•Ini
tiat
ing
stru
ctur
e •C
onti
ngen
t rew
ard
•Man
agem
ent b
y ex
cept
ion-
acti
ve•B
ound
ary
span
ning
•D
irec
tive
•Con
side
rati
on
•Em
pow
erm
ent
•Par
tici
pati
ve
•Dev
elop
ing
•Ena
blin
g •S
erva
nt le
ader
ship
•Tra
nsfo
rmat
iona
l •C
hari
smat
ic,
insp
irat
iona
l
•Gen
der
•Age
•Eth
nici
ty
•Hei
ght w
eigh
t•E
duca
tion
, soc
ial s
tatu
s
•Int
elli
genc
e •C
onsc
ient
ious
ness
•O
penn
ess
to E
xper
ienc
e •E
mot
iona
l Sta
bili
ty
•Tec
hnic
al k
now
ledg
e •L
eade
rshi
p se
lf-e
ffic
acy
•Ext
rave
rsio
n• A
gree
able
ness
•C
omm
unic
atio
n sk
ills
•E
mot
iona
l int
elli
genc
e •P
olit
ical
ski
lls
•Ove
rall
•T
ask
(e.g
., pe
rfor
man
ce)
•Affe
ctiv
e/re
lati
onal
(e.
g.,
foll
ower
sat
isfa
ctio
n, le
ader
-m
embe
r ex
chan
ge)
•Ind
ivid
ual
•Dya
d •G
roup
•O
rgan
izat
ion
•Lea
der
•Oth
er (
e.g.
, fol
low
er, g
roup
, or
gani
zati
on)
Pass
ive
Lea
ders
hip
•Man
agem
ent b
y ex
cept
ion-
pass
ive
•Lai
ssez
-fai
re
Fig
ure
1:A
nIn
tegr
ated
Mod
elof
Lea
der
Tra
its,
Beh
avio
rs,a
ndE
ffec
tive
ness
.
-
D. SCOTT DERUE ET AL. 11
studies focus on dyadic-level relationships, group-level performance, ororganizational performance (Kaiser, Hogan, & Craig, 2008). Finally, tar-get of evaluation refers to whether the leader is the target of evaluation(e.g., leader effectiveness, satisfaction with leader) or another outcomethat is within the domain of leadership effectiveness but not specific to theleader (e.g., group performance).
As Yukl (2006: 11) notes, “the selection of appropriate [leadershipeffectiveness] criteria depends on the objectives and values of the per-son making the evaluation, and people have different values. . . it isusually best to include a variety of criteria in research on leadershipeffectiveness.” In this study, we focus on four distinct leadership effective-ness criteria: (a) individual leader effectiveness, (b) group performance,(c) follower satisfaction with leader, and (d) follower job satisfaction. Wechose these criteria for two reasons. First, we wanted to cover a rangeof content dimensions, levels of analyses, and targets of evaluation. In-dividual leader effectiveness provides an individual-level, leader-focusedassessment of overall effectiveness. Group performance offers a group-level, other-focused assessment of task-related performance, and followersatisfaction (with the leader and job) provides an affective, individual-level, and other-focused assessment of leadership effectiveness. Second,given that we are using meta-analytic techniques, we can only includethose criteria that have been examined across a sufficient number of stud-ies. With these criteria, we meet both of these parameters and are able toexamine the relative validity of traits and behaviors across a diverse set ofimportant criteria.
Toward an Integrated Model of Leader Traits and Behaviors
Although prior research has established that leadership effectivenessis influenced by both leader traits and behaviors, it is not clear from thisresearch how leader traits and behaviors complement or supplement eachother, and how they can be incorporated into a more integrative modelof leadership effectiveness. Based on prior reviews (Avolio et al., 2003;Bass & Bass, 2008; Yukl, Gordon, & Taber, 2002) and our own reviewof the literature, most leader traits can be organized into three categories:(a) demographics, (b) traits related to task competence, and (c) interper-sonal attributes. Similarly, leader behaviors are often discussed in terms ofwhether the behavior is oriented toward (a) task processes, (b) relationaldynamics, or (c) change.
Drawing on this classification scheme, we develop a conceptual frame-work that organizes the current literature and models how leader traits andbehaviors affect leadership effectiveness (Figure 1). In this model, we in-corporate a wide range of leader traits and behaviors that were identified in
-
12 PERSONNEL PSYCHOLOGY
our narrative review. Our empirical tests focus on a subset of these leadertraits and behaviors. Specifically, we focus on those traits and behaviorsthat comprise most of the empirical research on leadership, and at leastone trait or behavior from each major category. Although we incorporateother, less commonly studied variables in our model, these traits and be-haviors have not been studied enough empirically to be included in ourmeta-analytic tests.
With respect to leader traits, we focus on gender, intelligence, and theBig Five personality traits (Costa & McCrae, 1992). Collectively, theseleader traits span the demographic, task competence, and interpersonal di-mensions. For leader behaviors, we focus on transformational leadership,specific dimensions of transactional leadership (e.g., contingent reward),initiating structure, and consideration. We also focus on leader behaviorsrelated to passive leadership, namely laissez-faire and management byexception-passive (MBEP). For the sake of clarity, we italicize in Figure 1those leader traits and behaviors that are examined in our empirical anal-yses.
Finally, an important aspect of our model is that we position leaderbehaviors as one possible mechanism through which leader traits influenceleadership effectiveness. In some cases, it might be that leader traits andbehaviors have independent effects on effectiveness, but we posit thatbehaviors can also serve as a key mediator in the relationship betweenleader traits and effectiveness. Considering that leader traits such as genderand personality are often discussed in terms of the behaviors associatedwith those traits, the idea that leader behaviors mediate the relationshipbetween leader traits and effectiveness seems especially plausible. Wealso posit that traits impact outcomes not through actual behavior butrather by how those traits are perceived by others and the attributions thatpeople make related to individual traits. Altogether, Figure 1 provides anintegrative account of research on leader traits and behaviors, and pointsto possible mechanisms linking traits, behaviors, and effectiveness.
The Leader Trait Paradigm
In reviewing trait theories of leadership, Bass (1990) proposed twoquestions: (a) Which traits distinguish leaders from other people, and(b) what is the magnitude of those differences? With respect to the firstquestion, leadership scholars have generally examined leader traits re-lated to demographics (e.g., gender, age, education), task competence(e.g., intelligence, Conscientiousness), or interpersonal attributes (e.g.,Agreeableness, Extraversion; Bass & Bass, 2008). Unfortunately, little tono research has systematically addressed Bass’ second question regardingthe relative magnitude of effects across leader traits.
-
D. SCOTT DERUE ET AL. 13
Understanding the relative validity of leader traits is important be-cause traits might not be independent. For example, there are biologicaland sociocultural reasons for why men and women score differently onpersonality and intelligence (Feingold, 1994; Halpern, 1997). The biologi-cal model posits that gender differences are a function of innate differencesbetween sexes, whereas the sociocultural model posits that social and cul-tural factors directly produce differences. A detailed discussion of thesemodels is beyond the scope of this article, but it is clear that gender differ-ences exist for both intelligence and personality (Feingold, 1994; Halpern,1997; Hedges & Nowell, 1995). In addition, meta-analyses on the rela-tionship between intelligence and personality suggest that Extraversionand Openness to Experience are related to intelligence (Ackerman &Heggestad, 1997; Judge, Jackson, Shaw, Scott, & Rich, 2007). This find-ing is especially interesting considering that Extraversion and Opennessto Experience are personality traits that have been shown to have strongrelationships with leadership effectiveness (Judge et al., 2002). Thus, it islikely that the effects of gender, intelligence, and personality are not in-dependent. In the following sections, we build a theoretical case for whycertain leader traits will be more predictive of leadership effectivenessthan other traits. Moreover, we specify how the relative validity of leadertraits will vary by leadership effectiveness criteria.
Demographics. Among the possible demographics of leaders, genderhas received the most attention. Other demographics such as physicalcharacteristics (e.g., height; Judge & Cable, 2004), education (Howard &Bray, 1988), and experience (Fiedler, 1970) have been examined in priorresearch, but the amount of research on these other demographics palesin comparison to the research on gender and leadership. Most notably,Eagly and colleagues (Eagly & Johnson, 1990; Eagly, et al., 1995, 2003)meta-analyzed the relationship between gender and leadership and foundthat, although men and women exhibit some differences in leadershipstyle, men and women appear to be equally effective—thus drawing intoquestion gender as a valid predictor of leadership effectiveness.
Based on this research, we do not expect to see differences betweengenders in terms of leadership effectiveness. We also propose any dif-ferences that might exist are due to confounding relationships with otherleader traits such as intelligence and personality (Feingold, 1994; Halpern,1997). Thus, when examining gender in conjunction with these otherleader traits, we do not expect to observe a meaningful effect of genderon leadership effectiveness.
Task competence. Task competence is a general category of leadertraits that relate to how individuals approach the execution and perfor-mance of tasks (Bass & Bass, 2008). Although a variety of task-related per-sonality traits have been studied, leadership scholars most often describe
-
14 PERSONNEL PSYCHOLOGY
task competence in terms of four traits: intelligence, Conscientiousness,Openness to Experience, and Emotional Stability. Intelligence reflects ageneral factor of cognitive abilities related to individuals’ verbal, spatial,numerical, and reasoning abilities, and has been established as a consis-tent predictor of task performance (Hunter & Hunter, 1984). With respectto intelligence and leadership, Judge et al. (2004) meta-analyzed 151samples and found that intelligence was positively related to leadershipeffectiveness (rc = .21).
Beyond intelligence, Conscientiousness, Openness to Experience, andEmotional Stability are often used to describe how one approaches andreacts to task work (Barrick & Mount, 1991). Conscientiousness reflectsthe extent to which a person is dependable, dutiful, and achievement-oriented, and is often associated with deliberate planning and structure.Openness to Experience is commonly associated with being imaginative,curious, and open minded to new and different ways of working. EmotionalStability refers to a person’s ability to remain calm and not be easily upsetwhen faced with challenging tasks. In a meta-analysis of 73 independentsamples, Judge et al. (2002) found that Conscientiousness (rc = .16),Openness to Experience (rc = .24), and Emotional Stability (rc = .22)were all positively related to leadership effectiveness.
Interpersonal attributes. Interpersonal attributes is a general categoryof leader traits that relate to how individuals approach social interactions(Bass & Bass, 2008). These traits include the interpersonal plane of per-sonality (i.e., Extraversion, Agreeableness; Costa & McCrae, 1992), aswell as skills and abilities related to social functioning (e.g., commu-nication skills; Klimoski & Hayes, 1980). The most commonly studiedinterpersonal attributes of leaders are Extraversion and Agreeableness,with prior meta-analyses finding that both Extraversion (rc = .24) andAgreeableness (rc = .21) were positively related to leadership effective-ness (Judge et al., 2002).
Relative validity of leader traits. Research suggests that leader traitsrelated to task competence and interpersonal attributes are important pre-dictors of leadership effectiveness. Yet, we expect that the relative validityof these leader traits will vary depending on the effectiveness criterion. Inparticular, to the degree that the content of leadership effectiveness criteriafocuses on execution and performance, we expect that leader traits relatedto task competence will be particularly important. Highly intelligent andconscientious leaders, for example, will be especially adept at ensuringtheir followers have sufficient role clarity, structure, and goals to help facil-itate task performance. In contrast, to the degree leadership effectivenesscriteria focus on affective and relational elements, we expect that the in-terpersonal attributes of leaders, namely Extraversion and Agreeableness,will be important. For example, leaders who are especially extraverted or
-
D. SCOTT DERUE ET AL. 15
are highly agreeable are more likely to invoke strong emotional ties andbuild high-quality relationships with followers (Nahrgang, Morgeson, &Ilies, 2009), which should lead to improved scores on affective criteriasuch as follower satisfaction with the leader.
Hypothesis 1: Leader traits related to task competence will exhibita stronger, positive relationship with task performancedimensions of leadership effectiveness than leaders’ de-mographics or interpersonal attributes.
Hypothesis 2: Leader traits related to interpersonal attributes will ex-hibit a stronger, positive relationship with affectiveand relational dimensions of leadership effectivenessthan leaders’ demographics or traits related to taskcompetence.
Following this same logic, to the degree that leadership effectivenesscriteria rely on a more general or global assessment of effectiveness, weexpect that leader traits related to task competence and the interpersonalattributes of the leader will both be important in predicting leadershipeffectiveness. This is because overall effectiveness criteria encompass thedegree to which leaders facilitate task performance but also the degreeto which leaders develop relationships with and consider the welfare offollowers (Yukl, 2006).
Hypothesis 3: Leader traits related to (a) task competence and (b)interpersonal attributes will both be positively re-lated to overall leader effectiveness and more so thandemographics.
The Leader Behavior Paradigm
In their narrative review of the leader behavior literature, Fleishmanand colleagues (1991) identified 65 distinct classifications of leader be-havior, and subsequent reviews have only further highlighted the prolif-eration of leader behavior typologies and theories (Avolio et al., 2003;Pearce et al., 2003). Unfortunately, new leader behavior theories continueto be conceived without explicit comparison to or falsification of existingleader behavior theories.
One consistent theme in the literature is that behaviors can be fit intofour categories: task-oriented behaviors, relational-oriented behaviors,change-oriented behaviors, and what we refer to as passive leadership.In this section, we illustrate how two of the most studied theories of leaderbehavior, initiating structure-consideration (IS-C; Halpin, 1957; Stogdill,1963) and transformational-transactional (T-T; Bass, 1985; Burns, 1978),
-
16 PERSONNEL PSYCHOLOGY
can be arranged along these behavioral dimensions. We also discuss howthese theories of leader behavior overlap conceptually in ways that in-form our understanding of their relative validities in predicting leadershipeffectiveness.
Task-oriented behaviors. Initiating structure and select transac-tional leader behaviors, namely contingent reward and management byexception-active (MBEA), represent task-oriented behaviors. Initiatingstructure describes behaviors such as defining task roles and role rela-tionships among group members, coordinating group members’ actions,determining standards of task performance, and ensuring group mem-bers perform up to those standards. Similarly, transactional leaders makeclear what is expected in terms of task performance and the rewards formeeting those expectations (contingent rewards), anticipate task-orientedproblems, and take corrective action (MBEA). Both initiating structureand contingent reward describe leaders as being clear about expectationsand standards for performance, and using these standards to shape followercommitment, motivation, and behavior. Moreover, initiating structure andMBEA discuss dealing with deviations from those standards via the useof structure and routines.
Relational-oriented behaviors. Relative to initiating structure andtransactional leadership, consideration leader behaviors describe morerelational-oriented behaviors. In particular, leaders high on considerationshow concern and respect for individual group members, are friendlyand approachable, are open to input from others, and treat all groupmembers as equals (Bass, 1990). Similar relational-oriented behaviorsare described in research on empowering (Conger, 1989; Srivastava,Bartol, & Locke, 2006), participative (Kahai, Sosik, & Avolio, 1997),and democratic (Gastil, 1994) leadership. A common theme among theserelational-oriented behaviors is that the leader acts in ways that buildfollower respect and encourage followers to focus on the welfare of thegroup. It should be noted that certain aspects of transformational leaderbehaviors (e.g., individualized consideration) also consist of a relationalorientation, which is a point we revisit later in the manuscript. However,broadly speaking, transformational leadership is conceptualized as a set ofbehaviors designed to create and facilitate change in organizations, whichbrings us to our third category of leader behaviors.
Change-oriented behaviors. Leader behaviors oriented toward fa-cilitating and driving change in groups and organizations represent athird category of leader behaviors that is conceptually distinct fromtask and relational-oriented behaviors. According to Yukl et al. (2002),change-oriented leader behaviors encompass actions such as devel-oping and communicating a vision for change, encouraging innova-tive thinking, and risk taking. For example, transformational leaders
-
D. SCOTT DERUE ET AL. 17
(inspirational motivation) focus on communicating a compelling vision forthe future; in addition, transformational leaders (intellectual stimulation)seek different perspectives from group members, challenge assumptions,and take risks. These dimensions of transformational leadership conceptu-ally distinguish it from the research on task and relational-oriented leaderbehaviors.
Passive leadership. In addition to task, relational, and change-oriented leader behaviors, many leader behavior taxonomies also includereference to leader inaction or passive leadership. For example, as part ofthe transactional model of leader behaviors, MBEP refers to how lead-ers only engage their followers when task-related problems or challengesemerge (Bass, 1990). When a problem does not exist or is not apparent tothe leader, the leader does not actively engage. Similarly, a common di-mension of leader behaviors is laissez-faire, which describes the absenceof leader behaviors (Avolio, Bass, & Jung, 1999).
Relative validity of leader behaviors. These models of leader be-havior, IS-C and T-T, have developed and evolved largely independentof each other. However, given the conceptual similarities between thesemodels, there is reason to question the independence of their effects andtheir relative validity as predictors of leadership effectiveness. Similar toour discussion of leader traits, we expect that task-oriented leader behav-iors will ensure that followers have specific goals, an established groupstructure with clear roles, and transparent metrics upon which to com-pare their performance. As a result, task-oriented leader behaviors shouldpromote greater task productivity in follower or group performance. Wealso expect change-oriented leader behaviors to be important predictors oftask performance. By establishing a vision for the future and challengingfollowers to not settle for the status quo, change-oriented leader behaviorsshould facilitate improvements in task productivity. Thus, to the degreethat the content of leadership effectiveness criteria focuses on task execu-tion and performance, we expect task-oriented and change-oriented leaderbehaviors will be important.
Hypothesis 4: Task-oriented and change-oriented leader behaviors willexhibit a stronger, positive relationship with task per-formance dimensions of leadership effectiveness thanrelational-oriented or passive leader behaviors.
In contrast, leaders who engage in relational-oriented behaviors areempathetic and skilled at sensing the needs of their followers; likewise,these leaders show concern for others and appeal to followers’ emo-tions. These leader behaviors should invoke a strong interpersonal connec-tion with followers and ultimately higher levels of follower satisfaction.Likewise, change-oriented behaviors can also enhance the attitudes and
-
18 PERSONNEL PSYCHOLOGY
satisfaction of followers. Prior research shows that individuals whofeel that they are growing, developing, and making improvements overtime feel more satisfied at work (Hackman & Oldham, 1980). Weexpect that change-oriented leader behaviors will enhance these feel-ings of growth and development. Thus, to the degree that leadershipeffectiveness criteria focus on affective and relational elements, weexpect relational-oriented and change-oriented leader behaviors to beimportant.
Hypothesis 5: Relational-oriented and change-oriented leader be-haviors will exhibit a stronger, positive relationshipwith affective and relational dimensions of leader-ship effectiveness than task-oriented or passive leaderbehaviors.
As noted above, global assessments of leadership effectiveness reflecta leader’s ability to facilitate task performance as well as manage re-lationships and interpersonal concerns within the group. Therefore, task-oriented, relational-oriented, and change-oriented leader behaviors shouldbe important predictors of overall leader effectiveness.
Hypothesis 6: Task-, relational-, and change-oriented leader behaviorswill be positively related to overall leader effectivenessand more so than passive leader behaviors.
With respect to passive leader behaviors, namely MBEP and laissez-faire leadership, it is clear that there is a negative relationship betweenthese leader behaviors and leadership effectiveness (Judge & Piccolo,2004). In terms of relative (absolute) validity, however, it is unclear howthese passive leader behaviors compare to more active task- and relational-oriented behaviors—and how this relative validity might differ acrossleadership effectiveness criteria.
Most research on leader traits and behaviors is embedded within aformal social structure whereby leaders hold formal positions that comewith an expected set of role behaviors (Biddle, 1979). When leaders donot actively engage in behaviors consistent with these role behaviors, theinaction becomes symbolic (Pfeffer, 1981) and likely renders the persona nonleader in the eyes of followers. Thus, to the degree leadership effec-tiveness criteria focus on the leader as the target of evaluation, we expectpassive leadership behaviors to be important predictors of effectiveness.Thus, we expect passive leadership to be a more important predictor foroutcomes such as leader effectiveness and satisfaction with leader, as op-posed to group performance or follower job satisfaction. In fact, given theimportance of action in leadership, even if that action is purely symbolic(Pfeffer, 1981), we propose that passive leadership will be as important
-
D. SCOTT DERUE ET AL. 19
in predicting leader effectiveness and satisfaction with leader as will theactive forms of leader behaviors.
Hypothesis 7: In comparison to more active forms of leader behavior,the relative validity of MBEP and laissez-faire leader-ship will be greater for criteria that have the leader asthe target of evaluation than for criteria that do not focusspecifically on the leader.
Leader Traits Versus Behaviors: A Test of Relative Validity
Thus far, we have focused purely on the predictive validities of leadertraits relative to other leader traits and leader behaviors relative to otherleader behaviors. However, in order to progress toward an integrativeunderstanding of leadership, we must simultaneously consider alternativetrait and behavioral explanations. It is this direct comparison of alternativeexplanations of leadership effectiveness that is absent from the currentliterature.
We offer several theoretical explanations for why leader behaviors willhave greater validity than leader traits in predicting leadership effective-ness. First, consistent with recent literature on the distal and proximalantecedents to leadership effectiveness (Van Iddekinge, Ferris, & Heffner,2009), we submit that leader behaviors are more proximal to the act ofleadership than are traits and, thus, will be more predictive of leadershipeffectiveness. Second, although traits reflect behavioral tendencies in peo-ple, the manifestation of those traits into behaviors can be affected by thesituation. Drawing from trait activation theory (Tett & Guterman, 2000;Tett & Burnett, 2003) and related research (Mischel & Shoda, 1995), traitsmanifest into the expected set of behaviors only when the situation makesthe need for that trait behavior salient. When situations do not call for aparticular trait, the trait does not manifest and its impact on outcomes ismarginalized. Given the complexity and ambiguity of leadership contexts(Pfeffer, 1977), it is likely that leadership situations vary with respect totrait relevance. In other words, leaders’ traits will not always manifest inways that impact leadership effectiveness. Contrast this with assessmentsof leader behavior, where the assessment measures actual, observed be-havior that has already manifested during the act of leadership, and wewould expect that leader behaviors will be more predictive of leadershipeffectiveness than leader traits.
Hypothesis 8: Leader behaviors will predict more variance in leader-ship effectiveness than leader traits.
-
20 PERSONNEL PSYCHOLOGY
Traits, Behaviors, and Leadership Effectiveness: An Integrated Model
As illustrated in Figure 1, we propose that there are two mechanismsthrough which individual traits affect leadership effectiveness. The firstinvolves actual behaviors that result as a function of the leader’s traits.The current literature on leadership has treated leader traits and behaviorsas independent explanations of leader effectiveness. However, if leadertraits and behaviors are not entirely independent, an alternative model isthat leader behaviors serve as one meditational mechanism. The secondmechanism through which leader traits might impact effectiveness is notthrough actual behavior but rather how followers attribute and identifywith the leader’s traits. Specifically, we propose that certain traits, espe-cially those that are highly salient to followers, have symbolic meaningand can be the basis upon which followers make judgments about thatleader that are independent of any actual behavior.
Leaders behaviors. The notion that leader behaviors mediate the rela-tionship between traits and effectiveness seems especially plausible con-sidering the conceptual and empirical links between individual traits andbehaviors that are apparent in much of the personality literature (Barrick& Mount, 1993) and research on gender in leadership (Eagly & Johnson,1990). In particular, the interpersonal attributes of leaders, such as Ex-traversion and Agreeableness, should predict the degree to which leadersengage in relational-oriented and change-oriented behavior. For example,extraverted individuals will be more inclined to seek input from followers,talk enthusiastically about the work, and be more comfortable setting adirection and vision for the group. Similarly, agreeable individuals willbe more friendly and approachable, likely to help followers develop theirstrengths, and respectful to followers. All of these behaviors are akinto those articulated in the consideration and transformational theories ofleader behavior.
Likewise, the traits related to task competence (intelligence, Consci-entiousness, Openness to Experience, and Emotional Stability) shouldpredict how leaders approach behaviors such as structuring task work,challenging assumptions, risk taking, and solving problems. For exam-ple, conscientious leaders, due to their preference for planned rather thanspontaneous behavior, will be inclined to initiate structure in leadershipcontexts (Humphrey, Hollenbeck, Meyer, & Ilgen, 2007). They will alsobe more inclined to actively monitor and intervene when problems arise,which are behaviors associated with MBEA. Similarly, leaders who arehigh in Openness to Experience are more likely to monitor their envi-ronment, challenge assumptions, recognize the possible implications ofexternal forces, and then intervene as appropriate. Finally, leaders who areemotionally stable will be more likely to remain calm, maintain order and
-
D. SCOTT DERUE ET AL. 21
structure, and be more comfortable with taking risks during challengingsituations.
Hypothesis 9: Relational-oriented and change-oriented leader behav-iors will mediate the relationship between leader-ship effectiveness and the interpersonal attributes ofleaders.
Hypothesis 10: Task-oriented and change-oriented leader behaviorswill mediate the relationship between leadership effec-tiveness and leader traits related to task competence.
Attributions and identification processes. Drawing on leadership cat-egorization theory (Lord, 1985) and related research on attributional andidentification processes in leadership (Conger & Kanungo, 1987; Hogg,Hains, & Mason, 1998), we submit that leader traits can influence lead-ership effectiveness by way of attributions that followers make about theleader and perceived identification and similarity with the leader. For ex-ample, independent of leader behaviors, Cherulnik, Turns, and Wilderman(1990) found that physical appearance in terms of maturity and attractive-ness influenced attributions of leadership emergence and effectiveness.Similarly, numerous studies have found that gender is an important factorin shaping followers’ attributions of leadership and effectiveness (e.g.,Sczesny, Bosak, Neff, & Schyns, 2004). Finally, drawing from the simi-larity attraction paradigm (Byrne, 1971), research has consistently shownthat followers who perceive a leader to be similar to themselves reportstronger identification with the leader and grant that leader more favorableevaluations (Engle & Lord, 1997; Liden, Wayne, & Stilwell, 1993; Turban& Jones, 1988).
Hypothesis 11: Follower attributions and identification processes willmediate the relationship between leader traits andeffectiveness.
As noted earlier, we present our conceptual model of leader traits,behaviors, and effectiveness to help organize and integrate the existingliterature. In our empirical analyses and hypothesis testing, we examinean exemplary set of relationships from this model. Specifically, we focuson those relationships where meta-analytic data already existed or therewere sufficient primary studies for us to conduct our own meta-analyses.In some cases, such as with leader attributions and identification pro-cesses, there was insufficient data for us to examine these relationships.In our Discussion section, we return to these relationships, discuss theimplications of our findings for these other variables and processes, andoutline an agenda for future research that will be critical for developing amore integrative understanding of leadership.
-
22 PERSONNEL PSYCHOLOGY
Methods
To examine these hypotheses, we use previously published meta-analytic estimates, update previously published meta-analytic estimates,and conduct our own meta-analyses of primary studies. Table 1 providesthe sources of the intercorrelations among all study variables. In total,59 studies consisting of 13 existing meta-analyses and 46 primary stud-ies were included in our final analysis. Of the 143 bivariate relationshipsestimated in Table 2, 90 of the estimates were drawn from previously pub-lished meta-analyses, and 53 were estimated via our own meta-analysesof primary studies or via updating previously published meta-analyses.
The correlation matrix used as an input for our analyses (Table 2) wasconstructed via a two-step process. First, where meta-analytic estimates ofa relationship were available, those data were used directly. In some cases,it appeared the authors of the original study had meta-analytic estimatesrelevant to our study that were not reported in the original meta-analysis.In such cases, we contacted the authors directly and obtained the meta-analytic estimates. Second, where previous meta-analytic estimates werepublished before 2003, we updated the existing meta-analyses with recentprimary studies. Third, where meta-analytic estimates were unavailable,we conducted our own meta-analyses by collecting and examining primarystudies.
Literature Search
To identify previous meta-analyses, we searched the online databasesPsycINFO (1887–2008) and Web of Science ISI (1970–2008) using com-binations of the terms leader, leadership, or manager with the termsmeta-analysis or quantitative review. Our search was supplemented witha reference search of key articles in the area of leadership (e.g., Avolioet al., 2003; Day, 2000; Yukl, 1989). We identified 919 articles in our ini-tial search but narrowed this down to 79 studies based on a review of theabstracts. We then examined each of these studies to determine whetherthe study should be included in our analysis. To be included, a study had tobe a meta-analysis, contain a variable of interest per our hypotheses, andthe effect size for the bivariate relationship had to be reported in the study.In some cases, we identified multiple studies that reported meta-analyticestimates for the same bivariate relationship. In these cases, we used themeta-analytic estimate that was the most recent and had the largest samplesize.
For bivariate relationships where meta-analytic estimates were not al-ready available, we conducted our own meta-analyses of primary data. Toidentify the relevant primary studies, an additional literature search was
-
D. SCOTT DERUE ET AL. 23
TAB
LE
1So
urce
sof
Inte
rcor
rela
tion
sA
mon
gSt
udy
Vari
able
s
12
34
56
78
910
1112
1314
1.G
ende
r2.
Inte
llige
nce
LI0
43.
Em
otio
nalS
tabi
lity
LI0
5JJ
074.
Agr
eeab
lene
ssL
I05
JJ07
OV
965.
Con
scie
ntio
usne
ssL
I05
JJ07
OV
96O
V96
6.E
xtra
vers
ion
LI0
5JJ
07O
V96
OV
96O
V96
7.O
penn
ess
LI0
5JJ
07O
V96
OV
96O
V96
OV
968.
Initi
atin
gst
ruct
ure
DP8
6(U
)N
ewN
ewN
ewN
ewN
ewN
ew9.
Con
side
ratio
nD
P86(
U)
New
New
New
New
New
New
JPI0
410
.Tra
nsfo
rmat
iona
lE
J03
New
BJ0
4B
J04
BJ0
4B
J04
BJ0
4N
ewN
ew11
.MB
E-a
ctiv
eE
J03
New
BJ0
4B
J04
BJ0
4B
J04
BJ0
4N
ewN
ewJP
0412
.MB
E-p
assi
veE
J03
New
BJ0
4B
J04
BJ0
4B
J04
BJ0
4N
ewN
ewJP
04JP
0413
.Con
tinge
ntre
war
dE
J03
New
BJ0
4B
J04
BJ0
4B
J04
BJ0
4N
ewN
ewJP
04JP
04JP
0414
.Lai
ssez
-fai
reE
J03
New
BJ0
4B
J04
BJ0
4B
J04
BJ0
4N
ewN
ewJP
04JP
04JP
04JP
0415
.Lea
der
effe
ctiv
enes
sE
J03
JC04
JB02
(U)
JB02
(U)
JB02
(U)
JB02
(U)
JB02
(U)
JPI0
4JP
I04
JP04
JP04
JP04
PB06
JP04
16.G
roup
perf
orm
ance
–N
ewN
ewN
ewN
ewN
ewN
ewJP
I04
JPI0
4B
S06
JP04
JP04
PB06
–17
.Fol
low
erjo
bsa
tisfa
ctio
nD
P86(
U)
–N
ewN
ewN
ewN
ewN
ewJP
I04
JPI0
4JP
04N
ewN
ewJP
04JP
0418
.Sat
isfa
ctio
nw
ithle
ader
EJ0
3–
New
New
New
New
New
JPI0
4JP
I04
JP04
JP04
JP04
JP04
JP04
BJ0
4=
Bon
oan
dJu
dge
(200
4),B
S06
=B
urke
etal
.(20
06),
DP8
6(U
)=
Upd
ated
Dob
bins
and
Plat
z(1
986)
,EJ0
3=
Eag
ly,J
ohan
nese
n-Sc
hmid
t,an
dva
nE
ngen
(200
3),J
B02
(U)=
Upd
ated
Judg
e,B
ono,
Ilie
s,an
dG
erha
rdt(
2002
),JC
04=
Judg
eet
al.(
2004
),JJ
07=
Judg
e,Ja
ckso
n,Sh
aw,S
cott,
and
Ric
h(2
007)
,JP0
4=
Judg
ean
dPi
ccol
o(2
004)
,JPI
04=
Judg
e,Pi
ccol
o,an
dIl
ies
(200
4),L
I04
=Ly
nnan
dIr
win
g(2
004)
,LI0
5=
Lip
pa(2
005)
,O
V96
=O
nes,
Vis
wes
vara
n,an
dR
eiss
(199
6),P
B06
=Po
dsak
off,
Bom
mer
,Pod
sako
ff,a
ndM
acK
enzi
e(2
006)
.
-
24 PERSONNEL PSYCHOLOGY
TAB
LE
2M
eta-
Ana
lyti
cE
stim
ates
ofIn
terc
orre
lati
ons
Am
ong
Stud
yVa
riab
les
12
34
56
78
910
1112
1314
1.G
ende
r—
90%
CI
—K
,N—
2.In
telli
genc
e.1
6—
90%
CI
.15,
.18
K,N
10,9
631
—3.
Em
otio
nalS
tabi
lity
.24
.09
—90
%C
I.2
2,.2
7—
—k,
N4,
4746
61,2
1404
—4.
Agr
eeab
lene
ss−.
11.0
0.2
5—
90%
CI
−.14
,−.0
9—
——
K,N
4,47
5338
,111
9018
,369
0—
5.C
onsc
ient
ious
ness
−.09
−.04
.26
.27
—90
%C
I−.
12,−
.06
——
——
k,N
4,47
4656
,154
2926
,538
034
4,16
2975
—6.
Ext
rave
rsio
n−.
10.0
2.1
9.1
7.0
2—
90%
CI
−.13
,−.0
7—
——
——
k,N
4,47
3861
,216
0260
,109
2624
3,13
5529
61,2
1602
—7.
Ope
nnes
s.1
1.2
2.1
6.1
1.2
2.1
7—
90%
CI
.08,
.13
——
——
——
k,N
4,47
3846
,131
8221
,487
023
6,14
4205
46,1
3182
418,
2520
04—
8.In
itiat
ing
stru
ctur
e−.
01.1
4.1
0−.
01.2
4.1
9.0
2—
90%
CI
−.02
,.02
−.24
,.52
−.03
,.23
−.27
,.25
.13,
.36
.11,
.26
−.16
,.19
—k,
N9,
5315
2,18
44,
635
4,63
54,
635
4,63
55,
843
—
cont
inue
d
-
D. SCOTT DERUE ET AL. 25
TAB
LE
2(c
ontin
ued)
12
34
56
78
910
1112
1314
9.C
onsi
dera
tion
−.02
−.07
.15
.26
.30
.33
.05
.17
—90
%C
I−.
02,.
02—
.01,
.30
−.08
,.59
.20,
.40
.13,
.52
−.06
,.16
.12,
.22
—k,
N9,
5315
1,68
4,63
54,
635
4,63
54,
635
4,63
518
1,26
295
—10
.Tra
nsfo
rmat
iona
l−.
06.1
6.1
7.1
4.1
3.2
4.1
5.4
4.7
1—
90%
CI
−.07
,−.0
5.1
0,.2
3.1
4,.2
0.0
8,.2
0.0
8,.1
8.2
1,.2
7.0
8,.2
2.3
3,.5
5.6
2,.8
1—
K,N
44,2
9806
6,82
618
,338
020
,391
618
,351
620
,369
219
,388
710
,130
88,
1074
—11
.MB
E-A
ctiv
e.0
7−.
05−.
02−.
11−.
02−.
03−.
04.3
1.0
4.1
7—
90%
CI
.05,
.09
−.18
,.08
−.07
,.03
−.15
,−.0
7−.
06,.
02−.
08,.
02−.
08,.
00.1
1,.5
2−.
18,.
26.0
6,.2
7—
K,N
16,1
1993
2,37
17,
1532
6,14
696,
1469
5,12
156,
1469
2,84
2,84
60,1
2600
—12
.MB
E-p
assi
ve.1
6−.
07−.
05−.
12−.
11−.
09.0
4−.
26−.
11−.
20−0
.05
—90
%C
I.1
5,.1
7−.
17,.
03−.
09,−
.01
−.17
,−.0
7−.
17,−
.05
−.16
,−.0
2.0
0,.0
8—
—−.
30,−
.10
− .23
,.12
—K
,N18
,122
492,
371
8,16
277,
1564
7,15
646,
1310
7,15
641,
501,
5050
,109
2813
,276
2—
13.C
ontin
gent
rew
ard
−.08
.11
.10
.17
.02
.14
.03
.28
.58
.80
0.19
−0.0
5—
90%
CI
−.09
,−.0
6.0
1,.2
1.0
4,.1
6.0
6,.2
8−.
02,.
06.0
9,.1
9−.
01,.
07.0
7,.4
8.4
0,.7
6.7
8,.8
3.0
7,.3
1−.
21,.
11—
k,N
21,1
5227
2,37
17,
1532
7,16
226,
1469
5,12
156,
1469
2,84
2,84
87,2
2369
20,4
795
17,4
253
—14
.Lai
ssez
-fai
re.0
9−.
16−.
05−.
12−.
11−.
09.0
4−.
30−.
47−.
65−0
.51
0.24
−0.3
8—
90%
CI
.08,
.11
−.26
,−.0
6−.
09,−
.01
−.17
,−.0
7−.
17,−
.05
−.16
,−.0
2.0
0,.0
8−.
43,−
.16
−.59
,−.3
4−.
73,.
57−.
76,−
.26
.09,
.39
−.45
,−.3
2—
k,N
16,1
2947
2,37
18,
1627
7,15
647,
1564
6,13
107,
1564
2,18
62,
186
25,5
674
5,10
753,
690
6,12
93—
15.L
eade
ref
fect
iven
ess
−.03
.15
.24
.08
.28
.31
.24
.39†
.52†
.64†
0.21
−0.1
90.
51−0
.54
90%
CI
−.06
,.01
.11,
.19
.23,
.26
.07,
.10
.26,
.29
.30,
.32
.22,
.26
——
—.1
1,.3
1− .
37,−
.01
.43,
.59
−.68
,−.4
0k,
N10
,U34
,336
951
,896
045
,105
0739
,100
5663
,126
4039
,776
220
,196
020
,160
527
,541
514
,211
78,
2627
26,4
418
11,1
920
16.G
roup
perf
orm
ance
—.0
4−.
03.2
0.3
1.0
0.1
3.3
5†.2
8†.3
80.
16−0
.18
0.24
—90
%C
I—
−.07
,.16
—−.
02,.
41.1
3,.4
9−.
17,.
17−.
05,.
32—
—.3
2,.4
4.0
6,.2
6−.
29,−
.08
.15,
.33
—K
,N—
6,29
11,
502,
845,
203
3,13
52,
117
17,1
242
11,1
019
25,1
291
7,11
805,
1115
19,1
361
—17
.Fol
low
erjo
bsa
tisfa
ctio
n−.
04—
.02
.01
−.08
.07
.00
.22†
.46†
.58†
0.02
−0.3
10.
64−0
.28
90%
CI
−.03
,.03
—−.
10,.
13−.
15,.
17−.
19,.
03−.
04,.
18−.
11,.
11—
——
−.18
,.22
−.48
,−.1
5.4
7,.8
0−.
44,−
.12
k,N
8,38
24—
2,30
02,
300
2,30
02,
300
2,30
072
,103
1776
,113
7418
,527
93,
414
3,41
46,
1933
2,39
218
.Sat
isfa
ctio
nw
ithle
ader
.00
—.0
8.2
2−.
03.0
3.0
3.3
3†.7
8†.7
1†0.
24−0
.14
0.55
−0.5
890
%C
I−.
03,.
03—
.02,
.15
.11,
.33
−.09
,.03
−.03
,.09
−.07
,.13
——
—.1
1,.3
6−.
37,.
10.4
4,.6
7−.
79,−
.37
k,N
7,U
—3,
1078
2,30
03,
1078
3,10
783,
400
49,8
070
49,7
871
23,4
349
11,2
272
8,32
5514
,407
65,
838
Not
e.†C
onfid
ence
inte
rval
ofco
rrel
atio
ndo
esno
tcro
ssze
ro.
U=
sam
ple
size
unsp
ecifi
edin
the
orig
inal
met
a-an
alys
is.
-
26 PERSONNEL PSYCHOLOGY
conducted using PsycINFO (1887–2008) and Web of Science ISI (1970–2008). The search terms used to identify primary studies for each bivariaterelationship where meta-analytic estimates were not available includedthe following key words: gender, cognitive ability, intelligence, mentalability, Big Five, five factor model, Openness, Emotional Stability, Neu-roticism, Agreeableness, Conscientiousness, Extraversion, extroversion,introversion, introversion, initiating structure, consideration, transforma-tional, management by exception, contingent reward, laissez-faire, groupperformance, group effectiveness, team performance, team effectiveness,satisfaction, leadership, leader, and manager. Our searches yielded 729articles, which were reviewed for inclusion in the meta-analysis.
Procedures for Meta-Analytic Estimates
We created a comprehensive list of codes for the variables reportedacross all studies. The studies were divided among the authors who thencoded the meta-analytic estimates for the variables of interest from eachstudy. When an individual was not clear about what code to assign to aspecific variable, the authors reviewed the study and discussed the variablein question until consensus was reached regarding the appropriate code.
Where possible, we coded the meta-analytic estimate that had beencorrected for measurement error in both the predictor and criterion scores.For some relationships, a corrected meta-analytic estimate was not avail-able (e.g., gender–personality). In these cases, we took the uncorrectedmeta-analytic estimate and corrected for measurement error based onexisting reliability data. For example, for the gender–personality rela-tionships, we corrected for error in the measurement of the personalitydimensions using existing reliability data from the NEO Manual (Costa& McCrae, 1992). We assumed perfect reliability in the measurement ofgender.
Some studies selected for inclusion reported effects in terms of thestatistics rather than correlations. For these studies, the data were trans-formed from effect sizes into correlations (see Lipsey & Wilson, 2001).We also coded the 90% confidence interval from previous meta-analysesbased on information provided in the meta-analyses. For 11 relationships(e.g., initiating structure-leader effectiveness), the previous meta-analysesdid not provide enough information for us to calculate the 90% confidenceinterval but did indicate that the 90% confidence interval did not cross zero.We have noted this in Table 2.
For meta-analytic estimates that were published before 2003, we up-dated the estimates using procedures outlined by Schmidt and Raju (2007).For conducting our own meta-analyses of primary studies, we used theSchmidt–Hunter psychometric meta-analysis method (Hunter & Schmidt,
-
D. SCOTT DERUE ET AL. 27
2004). We first corrected the primary estimates for measurement error inboth the predictor and criterion scores. The majority of primary studiesprovided the reliabilities of the measured scores. However, for studiesmissing this coefficient, we used the average reliability coefficient fromthe other studies also reporting data for those variables of interest (Hunter& Schmidt, 2004) or from publishers of the measures (e.g., NEO Manual;Costa & McCrae, 1992). For each of these estimates, we calculated 90%confidence intervals using a random effects estimate of the standard errorof the mean corrected correlation (Burke & Landis, 2003). All of the pri-mary studies used in our meta-analysis are marked with an asterisk in thereference section.
Testing Relative Importance and Validity
To determine the relative importance of predictors, researchers of-ten examine regression coefficients or zero-order correlations with thecriterion. When predictors are uncorrelated, these indices are appropriatebecause they are equivalent, and the squares of the indices sum to R2. Thus,relative importance can be expressed as the proportion of variance eachvariable explains. When predictor variables are correlated, however, theseindices are considered inadequate for determining the relative importanceof predictor variables because the indices are no longer equivalent, donot sum to R2, and take on different meanings (Budescu, 1993; Johnson,2001; Tabachnick & Fidell, 2001).
The use of epsilon is one way of determining relative importance whenpredictors are correlated (Johnson, 2000). The estimates derived from ep-silon, often labeled relative weights, sum to the model R2. Thus, the relativeweights represent the proportionate contribution each predictor makes toR2, considering the predictor’s direct effect and its effect when combinedwith other predictors. Researchers can also calculate the percentage of R2
explained by each predictor by dividing the relative weight of each pre-dictor by the total R2. Because of these attributes, epsilon is the preferredstatistic for computing relative importance (Johnson & LeBreton, 2004;LeBreton, Binning, Adorno, & Melcher, 2004). Thus, we use the epsilonstatistic in the current study to examine the relative predictive validity ofleader traits and behaviors.
Testing the Full Model
For testing Hypotheses 9–10, we utilized EQS 6.1 (Bentler, 1995)to model the relationships among the leader traits, leader behaviors, andleadership effectiveness. We created a complete mediation model, which
-
28 PERSONNEL PSYCHOLOGY
tested the extent to which the leader traits were completely mediated bythe leader behaviors specified in the hypotheses.
After examining the results of this model, we also used a fully satu-rated model where the leader traits were directly related to all of the leaderbehaviors and leadership effectiveness, whereas the leader behaviors werealso directly related to leadership effectiveness. We chose to use a fullysaturated model because doing so allowed us to examine the extent towhich all leader traits were mediated by each behavior. Using this ap-proach, we were able to identify any nonhypothesized relationships thatshould be explored in future research. Because the model is fully saturated,the model fit statistics are not informative and thus are not reported.
Results
Table 2 provides the meta-analytic estimates of the intercorrelationsamong the study variables. In this table, we present several pieces of infor-mation about the population correlation estimates, including the correctedcorrelation (rc) estimates, the 90% CI for the corrected correlation (rc),the number of studies included in determining the correlation (k), and thetotal number of participants in the studies (n). It is important to note thata few of our estimated effects are based on a relatively small number ofstudies (e.g., the relationship between intelligence and leader behaviors).Although other meta-analyses both inside and outside of the leadershipdomain have relied on similar numbers of original studies (e.g., Baas,De Dreu, & Nijstad, 2008; Balkundi & Harrison, 2006; Bono & Judge,2004; De Dreu & Weingart, 2003; Gerstner & Day, 1997; Ng & Feldman,2009), it is important that we address this potential threat to the validity ofour results. To address this concern for the relationships where we claimstatistical support, we conducted a series of fail-safe k analyses, whichindicate how many similarly sized studies with null findings would needto be conducted before the estimated effect would lose statistical support(Hunter & Schmidt, 2004). The average fail-safe k for the relationshipsbetween leader traits and effectiveness outcomes is 145, 186 for the rela-tionships between leader behaviors and effectiveness outcomes, and 26 forthe intercorrelations between leader traits and behaviors. Based on theseresults, a substantial number of studies with null findings would have tobe conducted before our reported effects would lose statistical support.
Testing Relative Validity
Hypotheses 1–3 were concerned with the relative validity of leadertraits. Table 3 presents the direction of the relationship (positive or neg-ative) and the percentage of R2 explained by the specific leader trait for
-
D. SCOTT DERUE ET AL. 29
TAB
LE
3R
elat
ive
Impo
rtan
ceof
Lea
der
Trai
tsin
Pre
dict
ing
Lea
ders
hip
Effe
ctiv
enes
sO
utco
mes
Lea
der
effe
ctiv
enes
sG
roup
perf
orm
ance
Follo
wer
job
satis
fact
ion
Satis
fact
ion
with
lead
er
Lea
der
trai
ts+/
−%
R2
+/−
%R
2+/
−%
R2
+/−
%R
2
Gen
der
−1.
4−
13.6
Nul
l0.
4In
telli
genc
e+
8.2
+1.
5E
mot
iona
lSta
bilit
y+
14.3
−7.
6+
7.2
+7.
4A
gree
able
ness
+.9
+22
.0+
1.8
+81
.0C
onsc
ient
ious
ness
+27
.0+
61.5
−50
.6−
9.4
Ext
rave
rsio
n+
35.1
Nul
l.3
+25
.9+
0.6
Ope
nnes
sto
Exp
erie
nce
+13
.2+
7.1
Nul
l1.
0+
1.2
Tota
l R2
.22
.14
.02
.06
-
30 PERSONNEL PSYCHOLOGY
each leadership effectiveness criteria. Overall, leader traits explain be-tween 2% and 22% of the variance in leadership effectiveness criteria.For group performance, a task performance dimension of leadership ef-fectiveness, the most important predictor is Conscientiousness, which ispositively related to group performance and accounts for 61.5% of thetotal explained R2. This percentage is almost three times that of the nextmost important predictor of Agreeableness, which accounts for 22.0% oftotal explained R2. Overall, the leader traits explain 14% of the variancein group performance, and traits related to task competence account for77.7% of the total explained R2. Thus, Hypothesis 1 was supported.
For the affective and relational dimensions of leadership effectiveness,Extraversion and Agreeableness are both positively related to follower jobsatisfaction and account for 25.9% and 1.8% of total explained R2, respec-tively, whereas Conscientiousness accounts for 50.6% of total explainedR2. Leader traits only explain 2% of the total variance in follower jobsatisfaction, and interpersonal attributes account for only 27.7% of thetotal explained variance. For satisfaction with leader, Agreeableness isthe most important trait predictor, is positively related to satisfaction withleader, and accounts for 81.0% of total explained R2. Overall, leader traitsexplain 6% of the variance in satisfaction with leader, with traits relatedto interpersonal attributes accounting for 81.6% of the total explained R2.Thus, Hypothesis 2 was supported for follower satisfaction with leaderbut not follower job satisfaction.
Finally, for overall leader effectiveness, the most important leader traitsare Extraversion and Conscientiousness. These traits, which span acrosstask competence and interpersonal attributes, are both positively relatedto effectiveness and account for 35.3% and 27.6% of the total explainedR2, respectively. In total, leader traits explain 22% of the variance inoverall leader effectiveness, and traits related to task competence andinterpersonal attributes explain 98.6% of the total explained R2. Thus,Hypothesis 3 was supported.
Hypotheses 4–7 were concerned with the relatively validity of leaderbehaviors. As shown in Table 4, leader behaviors explain between 20%and 70% of the variance in leadership effectiveness criteria. The mostimportant leader behavior for predicting group performance is initiat-ing structure, which is positively related to group performance and ac-counts for 32.9% of total explained R2. Combined with contingent rewardand MBE-active, which are also positively related to group performance,task-oriented leader behaviors account for 47.6% of total explained vari-ance. Next, change-oriented transformational leader behaviors account for28.5% of the total explained variance in group performance. In contrast,consideration behaviors and passive leader behaviors account for 16.6%and 7.3% of total explained variance, respectively. Thus, Hypothesis 4 was
-
D. SCOTT DERUE ET AL. 31
TAB
LE
4R
elat
ive
Impo
rtan
ceof
Lea
der
Beh
avio
rsin
Pre
dict
ing
Lea
ders
hip
Effe
ctiv
enes
sO
utco
mes
Lea
der
effe
ctiv
enes
sG
roup
perf
orm
ance
Follo
wer
job
satis
fact
ion
Satis
fact
ion
with
lead
er
Lea
der
beha
vior
s+/
−%
R2
+/−
%R
2+/
−%
R2
+/−
%R
2
Initi
atin
gst
ruct
ure
+12
.9+
32.9
+2.
5+
4.8
Con
side
ratio
n+
19.5
+16
.6+
13.6
+44
.9T
rans
form
atio
nal
+22
.8+
28.5
+21
.0+
19.8
MB
E-a
ctiv
e+
3.0
+5.
7+
1.1
+3.
3M
BE
-pas
sive
−2.
3−
7.3
−14
.1−
0.6
Con
tinge
ntre
war
d+
17.4
+9.
0+
43.9
+12
. 0L
aiss
ez-f
aire
−22
.1−
3.8
−14
.6
Tota
lR2
.47
.20
.51
.70
-
32 PERSONNEL PSYCHOLOGY
supported in that task and change-oriented behaviors were most importantfor performance-related criteria, whereas relational-oriented and passiveleader behaviors were less important predictors of group performance. Intotal, leader behaviors explain 20% of the variance in group performance.
For follower job satisfaction, contingent reward is the most importantbehavioral predictor, accounting for 43.9% of total explained R2. Transfor-mational and consideration behaviors are positively related to follower jobsatisfaction and explain 21.0% and 13.6% of total variance in follower jobsatisfaction, respectively. In comparison, passive leader behaviors (MBE-passive, laissez-faire) combine to account for 17.9% of total varianceexplained. Overall, leader behaviors explain 51% of the variance in fol-lower job satisfaction. In predicting satisfaction with leader, considerationleader behaviors account for 44.9% of total variance explained, whereastransformational behaviors account for 19.8% of total variance explained.Collectively, task-oriented and passive leader behaviors account for 20.1%and 15.2% of total explained variance. In total, leader behaviors explain70% of the variance in satisfaction with leader. Thus, Hypothesis 5, whichpredicted that relational-oriented behaviors would be most important foraffective criteria, was supported for follower satisfaction with leader butnot follower job satisfaction.
Finally, task-oriented leader behaviors are positively related to overallleader effectiveness and account for 33.3% of total explained R2. Trans-formational and consideration behaviors are also positively related tooverall leader effectiveness and account for 22.8% and 19.5% of totalexplained R2, respectively. Passive leader behaviors are negatively relatedto leader effectiveness and explain 24.4% of total explained R2. In total,leader behaviors explain 47% of the variance in leader effectiveness. Thus,Hypothesis 6 was supported in that task, relational, and change-orientedleader behaviors were important predictors of overall leader effectiveness.
For leader effectiveness and satisfaction with leader, which both havethe leader as the target of evaluation, passive leader behaviors account for24.4% and 15.2% of total explained R2, respectively. For criteria that do notfocus specifically on the leader, passive leader behaviors account for 7.3%and 17.9% of total explained variance in group performance and followerjob satisfaction, respectively. In general, the portion of total varianceexplained by passive leader behaviors is less than the variance explained bymore active leader behaviors. Thus, although there is evidence suggestingthat laissez-faire behaviors are an important predictor of leader-centriccriteria, Hypothesis 7 was generally unsupported.
Table 5 provides the results of our analyses examining the relativeimportance of leader traits and behaviors concurrently. In Hypothesis 8,we predicted leader behaviors would be more predictive of leadershipeffectiveness than leader traits. Combined, leader traits and behaviors
-
D. SCOTT DERUE ET AL. 33
TAB
LE
5R
elat
ive
Impo
rtan
ceof
Lea
der
Trai
tsan
dB
ehav
iors
inP
redi
ctin
gL
eade
rshi
pE
ffect
iven
ess
Out
com
es
Lea
der
effe
ctiv
enes
sG
roup
perf
orm
ance
Follo
wer
job
satis
fact
ion
Satis
fact
ion
with
lead
er
Lea
der
trai
tsan
dbe
havi
ors
+/−
%R
2+/
−%
R2
+/−
%R
2+/
−%
R2
Gen
der
+.3
−.2
+0.
1In
telli
genc
e+
1.2
+.4
Em
otio
nalS
tabi
lity
+3.
9−
4.7
+.2
+0.
4A
gree
able
ness
+1.
0+
9.1
+1.
3+
3.3
Con
scie
ntio
usne
ss+
6.0
+17
.9−
3.7
−6.
3E
xtra
vers
ion
+7.
0N
ull
2.3
+.7
+4.
3O
penn
ess
toE
xper
ienc
e+
6.2
+3.
3N
ull
.2+
.5In
itiat
ing
stru
ctur
e+
7.8
+19
.6+
3.0
+6.
3C
onsi
dera
tion
+11
.9+
8.4
+15
.5+
41.7
Tra
nsfo
rmat
iona
l+
14.5
+19
.6+
17.7
+15
.1M
BE
-act
ive
+2.
8+
4.4
+1.
5+
2.4
MB
E-p
assi
ve−
1.7
−4.
1−
13.6
−0.
5C
ontin
gent
rew
ard
+15
.8+
6.3
+38
.7+
8.3
Lai
ssez
-fai
re−
20.0
−3.
7−
10.7
Tota
lR2
.58
.31
.56
.92
-
34 PERSONNEL PSYCHOLOGY
explain between 31% and 92% of the variance in leadership effectivenesscriteria. For leader effectiveness, leader behaviors account for 74.5% oftotal explained R2. In total, leader traits and behaviors explain 58% of thevariance in leader effectiveness. For group performance, leader behaviorsaccount for 62.4% of total explained R2. Overall, leader traits and be-haviors explain 31% of the variance in group performance. For followerjob satisfaction, leader behaviors account for 93.7% of total explainedR2. Overall, leader traits and behaviors explain 56% of the variance infollower job satisfaction. Finally, leader behaviors account for 85.0% oftotal explained R2 in satisfaction with leader. Together, leader traits andbehaviors explain 92% of the variance in satisfaction with leader. Thus,Hypothesis 8 was supported.
Testing an Integrated Model
Hypothesis 9–11 predicted the validity of an integrated model wherebyleader behaviors and follower attributions and identification processesmediate the relationship between leader traits and the four leadershipeffectiveness criteria. Because there was insufficient data on followerattributions and identification, we were unable to test Hypothesis 11.
Table 6 presents the path coefficients for the completely mediatedmodel. First examining Hypothesis 9 (i.e., relational and change-orientedbehaviors will mediate the effect of interpersonal traits), we find thatAgreeableness was mediated by consideration (but not transformational)behaviors, whereas Extraversion was mediated by both consideration andtransformational behaviors. Thus, the results show partial support forHypothesis 9.
Turning to Hypothesis 10, which predicted that task and change-oriented leader behaviors would mediate the effect of task competence-related leader traits, the results demonstrate that the effect of intelligenceon leadership effectiveness was mediated by initiating structure, transfor-mational leadership, and contingent reward. Similarly, Conscientiousnesswas mediated by initiating structure and transformational leadership. Incontrast, Emotional Stability was only mediated by contingent reward,whereas openness was only mediated by initiating structure. Thus, theresults show partial support for Hypothesis 10.
We next examined a fully saturated model relating all of the traits inour model to all of the behaviors and ultimately leadership effectiveness.Examining a fully saturated model allows us to observe the full extentto which leader behaviors mediate the relationships between leader traitsand effectiveness, as well as identify nonhypothesized relationships thatare present in our data. Table 7 presents the direct, indirect, and totaleffects for the full model among leader traits, behaviors, and leadership
-
D. SCOTT DERUE ET AL. 35
TAB
LE
6Pa
thC
oeffi
cien
tsfo
rH
ypot
hesi
zed
Com
plet
ely
Med
iate
dM
odel
Var
iabl
eIn
itiat
ing
stru
ctur
eC
onsi
dera
tion
Tra
nsfo
rmat
iona
lM
BE
-act
ive
Con
tinge
ntre
war
dL
eade
rshi
pef
fect
iven
ess
Inte
llige
nce
→.1
64.1
42−.
044
.104
(.10
0,.2
28)
(.07
8,.2
06)
(−.1
11,.
023)
(.03
7,.1
71)
Em
otio
nalS
tabi
lity
→.0
31.0
74−.
008
.091
(−.0
33,.
095)
(.01
0,.1
38)
(−.0
75,.
059)
(.02
4,.1
58)
Agr
eeab
lene
ss→
.210
.058
(.14
7,.2
73)
(−.0
06,.
122)
Con
scie
ntio
usne
ss→
.256
.086
−.01
4.0
02(.
192,
.320
)(.
022,
.150
)(−
.081
,.05
3)(−
.065
,.06
9)E
xtra
vers
ion
→.2
94.2
03(.
231,
.357
)(.
139,
.267
)O
penn
ess
toE
xper
ienc
e→
−.07
7.0
47−.
026
−.00
8(−
.141
,−.0
13)
(−.0
17,.
111)
(−.0
93,.
041)
(−.0
75,.
059)
Initi
atin
gst
ruct
ure
→.1
21(.
072,
.170
)C
onsi
dera
tion
→.1
57(.
106,
.208
)T
rans
form
atio
nal→
.497
(.44
6,.5
48)
MB
E-a
ctiv
e→
.087
(.03
6,.1
38)
Con
tinge
ntre
war
d→
−.04
9(−
.100
,.00
2)
Not
e.St
atis
tics
prov
ided
are
path
coef
ficie
nts
with
90%
confi
denc
ein
terv
als
inpa
rent
hese
s.
-
36 PERSONNEL PSYCHOLOGY
TAB
LE
7E
ffect
sD
ecom
posi
tion
for
Inte
grat
edM
odel
:L
eade
rshi
pE
ffect
iven
ess
Indi
rect
Indi
rect
effe
ctIn
dire
ctIn
dire
ctIn
dire
ctIn
dire
ctef
fect
Indi
rect
(med
iate
def
fect
effe
ctef
fect
effe
ct(m
edia
ted
effe
ctef
fect
(med
iate
d(m
edia
ted
(med
iate
d(m
edia
ted
effe
ct(m
edia
ted
Dir
ecte
ffec
tth
roug
hef
fect
effe
ctef
fect
effe
ctth
roug
hef
fect
Indi
rect
(unm
edia
ted
initi
atin
gth
roug
hth
roug
hth
roug
hth
roug
hco
ntin
gent
thro
ugh
effe
ctTo
tal