You are on page 1of 12

The Leadership Quarterly 22 (2011) 881–892

Contents lists available at ScienceDirect

The Leadership Quarterly


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / l e a q u a

A two-study investigation of item wording effects on leader–follower


convergence in descriptions of the leader–member exchange
(LMX) relationship☆
Chester A. Schriesheim ⁎, Joshua B. Wu, Cecily D. Cooper
School of Business Administration, University of Miami, USA

a r t i c l e i n f o a b s t r a c t

Available online 10 August 2011 The leader–member exchange (LMX) theory of leadership focuses on the quality of
relationships built between leaders and subordinates; LMX measures are designed to assess
Keywords: the quality of these relationships. Since the leader and subordinate are jointly embedded in the
Leader-member exchange relationship, it is reasonable to assume that their ratings of their relationship will converge to
Leader-member exchange convergence some reasonable extent. However, leader–member convergence on the most widely used
measures, the LMX-7 and SLMX-7 scales, has historically been low. To address this, we conduct
two studies to investigate whether item wording and content might be a cause of poor LMX
convergence. The obtained results support the conclusion that the wording and content of
these measures are, in fact, introducing attributional and social desirability biases into the
rating process and causing some of the poor LMX convergence that is commonly seen.
Implications for extant and future LMX research are discussed.
© 2011 Elsevier Inc. All rights reserved.

1. Introduction

The field of leadership research has evolved considerably over the last two decades. Some frameworks, such as Path–Goal
Leadership Theory (Evans, 1970; House, 1971) have seen a substantial decline in research interest (Schriesheim & Neider, 1996),
while increased attention has been placed on the study of transformational approaches to leadership (Dumdum, Lowe, & Avolio,
2002; Yukl, 2010).
One approach in which interest has remained reasonably constant, however, is leader–member exchange (LMX) theory (Graen
& Uhl-Bien, 1995). Rather than assuming a leader has a general style that is used consistently across all subordinates, this
framework is distinguished by its focus on the unique relationships that develop between leaders and their subordinates. More
specifically, leader–member exchange theory assumes (and finds) that leaders develop relationships of varying quality with their
subordinates (i.e., that exchanges can range from being very high to very low quality). These relationships have important
consequences for outcomes such as subordinate job satisfaction, career development, and performance (Erdogan & Liden, 2002;
Gerstner & Day, 1997; Liden, Sparrowe, & Wayne, 1997; Schriesheim, Castro, & Cogliser, 1999a).
Itself evolving over time, the LMX approach now focuses on the relationship between leader and follower. The relational
emphasis also makes LMX unique to other approaches that typically focus on just the leader or the follower alone (Graen & Uhl-
Bien, 1995; Schriesheim, Castro, & Cogliser, 1999a). According to Graen and Uhl-Bien (1995), studying the relationship is now the
focus of LMX research, and the relationship is seen as something that involves both the leader and subordinate but exists
independently of them and is perceptually shared (Dansereau, 1995; Dansereau, Yammarino, & Markham, 1995).

☆ Paper accepted for publication at The Leadership Quarterly.


⁎ Corresponding author at: Management Department, School of Business Administration, University of Miami, 5250 University Drive, 414D Jenkins Building,
Coral Gables, FL 33146-6548, USA. Tel.: + 1 305 284 3758.
E-mail address: chet@miami.edu (C.A. Schriesheim).

1048-9843/$ – see front matter © 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.leaqua.2011.07.009
882 C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892

This new focus and specification of the relationship as being a shared phenomenon suggests that leaders and followers should
describe the relationship at least somewhat similarly. At a minimum, one can argue that, “It is reasonable to expect the two parties
to agree about something as important and salient as the quality of their relationship” (Yukl, 2010, p. 124). Thus, while divergence
in leader–member descriptions of some specific exchange elements might be expected, substantial evidence now shows a serious
lack of convergence1 in subordinate- and supervisor-provided descriptions of overall relationship quality (usually less than 10–
15% shared variance in these descriptions; Gerstner & Day, 1997; Schriesheim, Cogliser, Scandura, Lankau, & Powers, 1999b). In
fact, Schriesheim, Neider, and Scandura (1998), (p. 307) discuss and summarize several studies in which convergent correlations
ranged from .16 to .50. Gerstner and Day (1997) report a .29 meta-analytic average correlation across 24 samples that increased to
.37 when it was corrected for measurement error, and a follow-up meta-analysis by Sin, Nahrgang, and Morgeson (2009) of 64
LMX samples yielded the same .37 correlation.
Yukl (2010) notes that, “It is not clear whether the low correlation reflects measurement problems in the scales or actual
differences in perception…. More research is needed” (p. 124). Earlier, Gerstner and Day (1997) suggested that “leader–member
agreement [convergence] should be examined” (p. 835) and that “leader–member LMX agreement [convergence] is an interesting
and potentially valuable outcome variable in its own right” (p. 837). Poor convergence implies that either leaders or members
experience the same relationship in a much different way or that they experience the relationship similarly but that current
measures are not assessing this accurately. The latter is obviously problematic for researchers who are concerned with drawing
valid conclusions from their empirical studies. The current research seeks to better understand why this divergence exists, and, in
doing so, builds upon a small but burgeoning area within LMX research focusing specifically on issues of convergence (e.g.,
Cogliser, Schriesheim, Scandura, & Gardner, 2009; Gerstner & Day, 1997; Zhou & Schriesheim, 2009). We offer novel insight to
convergence research by conducting two studies to explore the potential effects of attributional and social desirability biases that
may be elicited by LMX measures, influences that scholars have not previously considered.

2. Sources of poor convergence in Leader–Follower LMX ratings

While research on this issue has been sparse, several reasons for poor LMX convergence have recently been studied. These
include the lack of measurement equivalence when the same items are completed by subordinates and supervisors (Schriesheim &
Zhou, 2002), the use of different factors by subordinates and supervisors to judge relationship quality, and differential weighting
(by subordinates and supervisors) of the factors used to judge relationship quality (Zhou, 2003; Zhou & Schriesheim, 2010).
However, these investigations into LMX convergence have not yielded significantly greater convergence between leader and
member LMX descriptions or improved measures of leader–member exchange quality. For example, in the Schriesheim and Zhou
(2002) study, eliminating those leader–member exchange items that did not have measurement equivalence across the leader and
follower sub-samples did not improve the supervisor–subordinate LMX correlation. It therefore appears reasonable to pursue
other possible causes of poor LMX convergence beyond differences in the rater's perspective.

2.1. Item content and wording

While the rater's perspective appears to affect LMX ratings, the actual content and wording of LMX items may be equally or
more responsible for low convergence; this has not been systematically investigated in previous research. If respondents are asked
to rate unobservable variables (such as another person's mental state), or if they are asked to describe themselves, attributional
and social desirability biases may make the obtained ratings of each dyad member more uniquely idiosyncratic and therefore yield
lower levels of LMX rating convergence (Nunnally & Bernstein, 1994).

2.1.1. Attributional distortion


Looking at this issue in greater detail, it would seem that attributional distortions may be particularly pronounced when
respondents are asked to rate unobservable variables related to another person (such as that other person's mental state)
(Martinko, 2002; Martinko & Gardner, 1987). As Cascio and Aguinis (2005) note with respect to performance appraisals, raters, “…
infer common personality characteristics that go beyond that which is warranted” (p. 109) and that “idiosyncratic variance (i.e.,
variance due to the rater) has been found to be a larger component of variance in performance ratings than the variance attributable
to actual ratee performance…” (pp. 108–109). Findings such as these are apparently responsible for the often-repeated
recommendation that performance appraisals should focus on observable behavior and not on unobservables—such as mental
states or processes (Bernardin & Beatty, 1991; Cardy & Dobbins, 1994). Parallel arguments might be offered for non-performance
ratings, such as those collected to measure leader–member exchange quality. Additionally, with respect to LMX, Liden, Sparrowe
and Wayne (1997) suggest that measuring psychological states “…does not capture actual exchanges between leader and member”
and that such a focus “…has caused LMX research to drift from its initial foundation in social exchange theory” (p. 108).

1
In discussing this phenomenon, some authors have used the term “agreement” (e.g., Sin, Nahrgang, & Morgeson, 2009), whereas others have used
“convergence” (e.g., Schriesheim, Neider, & Scandura, 1998). We believe both terms can be useful and appropriate in discussing this issue but will rely solely on
the term “convergence” for the purpose of this discussion. Our goal is not to create leader and member response measures which are completely interchangeable,
as the term “agreement” would imply (James, Demaree, & Wolf, 1984), since “it is unlikely that problems of low agreement…will be completely eliminated with
…new measures” (Gerstner & Day, 1997: 837). Rather, our (more modest) goal is to achieve a higher correlation between the descriptions provided by the two
raters.
C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892 883

2.1.2. Social desirability bias


Similarly, the tendency to present oneself in a socially desirable manner may come into play when persons are asked to
describe themselves—especially their own internal mental states (Crowne & Marlowe, 1964; Edwards, 1990; Paulhus, 1985). In
fact, according to Nunnally and Bernstein (1994), “social desirability is certainly a major factor in self-description among normal
individuals…” (p. 385). Socially desirable responding may arise from a number of different sources, including a lack of self-
knowledge and a lack of frankness. As a result, it yields idiosyncratic rating variance that lowers self-other rating convergence
(Nunnally & Bernstein, 1994).

2.1.3. LMX-7 and SLMX-7


The subordinate-completed LMX-7 scale and its corresponding supervisor version, the SLMX-7 (Graen & Uhl-Bien, 1995),
currently dominate LMX research (Gerstner & Day, 1997) and are the only measures recommended by Graen and Uhl-Bien (1995)
for assessing leader–member relationship quality. Thus, we focus on the LMX-7 and SLMX-7 measures because of their dominance
in LMX research and because they are the basis of almost all extant research on lack of LMX rating convergence (Gerstner & Day,
1997; Schriesheim, Neider & Scandura, 1998). To our knowledge, a recent study by Sin, Nahrgang and Morgeson (2009) including
the LMX-MDM scale (Liden & Maslyn, 1998) is the only exception. Even in this study, however, the LMX-MDM only represented 9
of the 64 samples included in the meta-analysis.
When the LMX-7 and SLMX-7 measures are closely examined they can be seen to be a mixture of items calling for self-ratings of
both observable and unobservable variables, as well as ratings of observable and unobservable variables related to the other dyad
member (this is documented in greater detail below). Based on the widely accepted concept of Ockham's Razor (Ariew, 1976),
simple solutions to the problem of poor LMX rating convergence should be pursued before more complex ones. Therefore, the
current study was undertaken to examine the effects of rating referent (self or other dyad member) and rating attribute (overt
physical behavior or internal mental process) on convergence between supervisor SLMX-7 and subordinate LMX-7 ratings. Prior
studies of LMX convergence have not studied these intuitive, but likely, sources of bias.
The arguments summarized briefly above about attribution and social desirability effects attenuating leader–member
description convergence suggest that such effects are likely to be produced by current LMX-7 and SLMX-7 item content. Simply
put, assuming exchange relationships are balanced transactions, we believe that rating convergence will be least attenuated by
various distortions and biases (and therefore highest) when both raters are rating observable variables related to the other dyad
member. Correspondingly, we expect that rating convergence will be lowest when ratings are most affected by attributional and/
or social desirability biases—that is, when the raters describe unobservable variables related to the other dyad member (prone to
attributional bias) or when they describe themselves (prone to social desirability bias). Item content has not been previously
discussed or analyzed in the LMX literature. If results supporting these predictions are obtained, then future research will be
needed to develop revised LMX-7 and SLMX-7 measures so as to minimize attribution- and social desirability-produced
divergence in the LMX descriptions of subordinates and supervisors. Improved measures will allow a more rapid accumulation of
valid knowledge about the LMX relationship and also allow the more ready application of existing self-other rating convergence
frameworks (e.g., Atwater & Yammarino, 1997; Ostroff, Atwater, & Feinberg, 2004; Yammarino & Atwater, 1997) to LMX research.
In other words, once improved measures are available researchers can assume that if ratings of leader–member relationship
quality diverge then the divergence is the product of held perceptions and beliefs (and not artifacts produced by questionnaire
item content). Disparity or concurrence in perceptions can then be studied as a phenomenon in its own right. Studies focusing on
the consequences of perceptual differences in LMX have just started to appear in the literature (e.g., Cogliser, Schriesheim,
Scandura & Gardner, 2009) and much more research is needed in this domain.

3. Theoretical analysis of LMX-7 and SLMX-7 item content

The first half of Appendix A presents the actual items contained in the most frequently used (Gerstner & Day, 1997) and highly
recommended (Graen & Uhl-Bien, 1995) scales for measuring subordinate-perceived (LMX-7) and supervisor-perceived (SLMX-
7) leader–member exchange. These items are a diverse collection, differing not only in who is the rater (leader or follower) but
also in who is being rated (self or other dyad member) and the rating subject or focus (internal mental processes or overt physical
behaviors).
The first half of Table 1 presents the authors' assessment of the LMX-7 and SLMX-7 items with respect to (a) who is the
initiator, actor, or agent in the item, (b) who is the subject, recipient, or target, and (c) whether the item describes an overt
physical behavior or an internal mental process. All three authors assessed the scale items independently for each of these
attributes, and then came together to compare classifications. Any instances of disagreement, of which there were few, were
discussed until reaching consensus. Based on these assessments, it can be seen that both the LMX-7 and SLMX-7 scales are about
equally divided in terms of agent/target (3 with the follower and 4 with the leader as agent) and attribute being rated (4 with
mental process and 3 with physical behavior).
Based on the analysis shown in Table 1 and discussed above, it appears that there are actually eight different identifiable types
of items being used to measure the leader–member exchange relationship on the LMX-7 and SLMX-7 scales. Briefly, these are:

A. Follower descriptions of the leader's mental process (LMX-7-2 and 3).


B. Leader descriptions of the leader's mental process (SLMX-7-2 and 3).
C. Follower descriptions of the leader's behavior (LMX-7-4 and 5).
884 C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892

Table 1
Summary of item differences in four leader–member exchange (LMX) measures.

Item number Source of Item content (A) Initiator, actor, (T) Subject, recipient, Describes overt physical behavior
description or agent or target or internal mental process

LMX-7-1 Follower A understands T Follower Leader Mental


LMX-7-2 Follower A has empathy for T Leader Follower Mental
LMX-7-3 Follower A recognizes T's potential Leader Follower Mental
LMX-7-4 Follower A uses power to support T Leader Follower Behavior
LMX-7-5 Follower A uses power to protect T Leader Follower Behavior
LMX-7-6 Follower A defends T's decisions Follower Leader Behavior
LMX-7-7 Follower A assesses relationship Follower Leader Mental
SLMX-7-1 Leader A understands T Follower Leader Mental
SLMX-7-2 Leader A has empathy for T Leader Follower Mental
SLMX-7-3 Leader A recognizes T's potential Leader Follower Mental
SLMX-7-4 Leader A uses power to support T Leader Follower Behavior
SLMX-7-5 Leader A uses power to protect T Leader Follower Behavior
SLMX-7-6 Leader A defends T's decisions Follower Leader Behavior
SLMX-7-7 Leader A assesses relationship Follower Leader Mental
ALMX-1 Follower A understands T Leader Follower Mental
ALMX-2 Follower A has empathy for T Follower Leader Mental
ALMX-3 Follower A recognizes T's potential Follower Leader Mental
ALMX-4 Follower A uses power to support T Follower Leader Behavior
ALMX-5 Follower A uses power to protect T Follower Leader Behavior
ALMX-6 Follower A defends T's decisions Leader Follower Behavior
ASLMX-1 Leader A understands T Leader Follower Mental
ASLMX-2 Leader A has empathy for T Follower Leader Mental
ASLMX-3 Leader A recognizes T's potential Follower Leader Mental
ASLMX-4 Leader A uses power to support T Follower Leader Behavior
ASLMX-5 Leader A uses power to protect T Follower Leader Behavior
ASLMX-6 Leader A defends T's decisions Leader Follower Behavior

D. Leader descriptions of the leader's behavior (SLMX-7-4 and 5).


E. Follower descriptions of the follower's mental process (LMX-7-1 and 7).
F. Leader descriptions of the follower's mental process (SLMX-7-1 and 7).
G. Follower descriptions of the follower's behavior (LMX-7-6).
H. Leader descriptions of the follower's behavior (SLMX-7-6).

Based on our discussion of attributional and social desirability biases, these mixed item wordings in the LMX-7 and SLMX-7
scales are likely to trigger raters' cognitive biases which lead to a low level of convergence on the rated relationship between
leaders and subordinates. Item types A, C, F, and H above would be expected to elicit responses confounded by various
attributional biases (with item types A and F showing the strongest effects), while item types B, D, E, and G above would be
expected to elicit socially desirable responses (with item types B and E showing the strongest effects). Consequently, if this
theoretical content analysis is subsequently supported by empirical evidence, reconsidering the items that comprise the current
LMX-7 and SLMX-7 scales would appear warranted on the basis of improved psychometrics. We therefore conducted two studies
to document the existence of mixed item wordings in the LMX-7 and SLMX-7 scales (Study 1) and its effects on LMX-7-SLMX-7
convergence between leaders and followers (Study 2).

4. Study one: A test of LMX item content

Although we believe that our characterization of LMX-7 and SLMX-7 items, as shown in Table 1, is accurate, we also believe that
it is better to ensure classification accuracy by empirical validation processes. Accordingly, we decided to employ the content
adequacy assessment procedure developed by Schriesheim, Powers, Scandura, Gardiner, and Lankau (1993) and Schriesheim,
Cogliser, Scandura, Lankau and Powers (1999) to verify our characterization of the LMX-7 scale items. However, rather than using
factor analysis to examine the data (as suggested by Schriesheim and associates), we used nonparametric analyses (chi-square and
binomial tests; Siegel, 1956) that are similar in approach to the analysis of variance analytic strategy suggested by Hinkin and
Tracey (1999). This approach eliminates the use of subjective judgments in making determinations about item content, as the chi-
square and binomial tests provide a direct empirical assessment for classifying each item. Additionally, the chi-square and
binomial tests are very simple and straightforward. Finally, this technique can be utilized with small sample sizes, which are
clearly desirable for assessing statistical differences when small panels of “item judges” are employed (cf., Hinkin & Tracey, 1999;
Siegel, 1956). To obtain the judgment data for analysis, the scale items are presented to respondents, along with theoretical
definitions of each attribute for which the items are to be judged. We provide more detail on this process in the Procedure section
that immediately follows below.
C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892 885

4.1. Sample and procedure

4.1.1. Sample
The sample consisted of 50 MBA students at a medium-sized southern U.S. university. After deleting two incomplete or
unusable responses, the final sample consisted of 48 respondents. Participation in the study was voluntary. The average age was
26.3 years old and the sample was comprised of 60% males. All had prior work experience of at least 2 years, and 27 were currently
employed (average hours/week worked = 24.7). As noted in Schriesheim, Powers, Scandura, Gardiner and Lankau, (1993);
Schriesheim, Cogliser, Scandura, Lankau and Powers (1999), the requirements to complete a task such as this are sufficient
intellectual ability to rate the items and the lack of any theoretical biases. As such, the use of MBA students who were unfamiliar
with LMX theory was considered appropriate. The researchers administered questionnaires during normal class time, taking
approximately 15 min to complete. Explicit written and verbal instructions were provided prior to administration, and the
responses were anonymous.

4.1.2. Procedure
First, the respondents rated each of the LMX-7 items on whether the actor or agent is the leader, the employee, and/or some
other (other was defined as “Someone else, there is no agent, or the agent cannot be determined”). Then, the respondents rated the
items again, this time for the attribute(s) that were being described (overt physical behavior, internal mental process, and/or
neither). The definitions of agent and attributes that were employed read:

Agent: An agent is a person whom an item describes as thinking, acting, feeling, or doing something toward a target. In other
words, a person is an agent when that person thinks, feels, acts or displays behavior for, to, or about a target person.
Attribute: Overt physical behavior occurs when an agent acts or does something; an internal mental process occurs when an
agent thinks or feels something.

It should be noted that the respondents were instructed to place one or more X's in columns on the rating forms to indicate who
the agent is in each LMX-7 item (first survey rating page) and what the agent is portrayed as doing (second page). In other words,
the judges were permitted to rate the items as measuring a combination of agents (e.g., both the leader and the employee; both the
employee and other; the leader, the employee, and other; etc.) and a combination of activities (e.g. overt physical behavior and
internal mental processes; overt physical behavior and neither; etc.). Using a classification assignment approach to scoring
(Schriesheim, Powers, Scandura, Gardiner & Lankau, 1993), a respondent's score for a dimension on an item could be 0.00 (if the
item was not seen as reflecting that rating dimension), 0.33 (if an item was seen as reflecting all three rating dimensions), 0.50
(if an item reflected two rating dimensions), or 1.00 (if an item was seen as reflecting only one dimension—e.g., having only the
supervisor as agent or as being concerned only with overt physical behavior).

4.2. Analysis

The first step in the analysis is to calculate the number of assignments or classifications for each item on each of the rated
dimensions (leader, employee, and other; behavior, mental process, and neither). Then, an overall comparison of these
frequencies is conducted for each item across the three dimensions using a chi-square test. Finally, a directional (one-tailed)
binomial test is employed to determine that each item is statistically significantly higher on the dimension to which it had been
previously classified (see Table 1) as compared to each of the two other dimensions.

4.3. Results and discussion

The results of the judge content ratings are shown in Table 2. Interestingly, none of the respondents rated any item as having
more than one agent. However, several items were rated as involving both overt physical behavior and internal mental processes.
As can be seen from Table 2, both the pattern and statistical significance of the classifications of all seven LMX-7 items mirror
those that we had developed a priori and presented in Table 1. Specifically, LMX-7 items 2, 3, 4, and 5 are classified as having the
leader as the actor or agent, while items 1, 6, and 7 have the follower as actor or agent. LMX-7 items 4 and 5 involve the rating of an
overt physical behavior, while items 1, 2, 3, and 7 involve rating an internal mental process. Item 6 is seen as involving both overt
physical behavior and internal mental process, but is more oriented toward overt physical behavior.
Having not empirically refuted our earlier (Table 1) theoretical analysis of LMX-7 items, there is no reason to believe that we
mischaracterized the SLMX-7 items, since they are essentially mirrored reflections of the LMX-7 scale (Graen & Uhl-Bien, 1995).
Consequently, we now advanced to Study Two, designed to empirically test the degree to which leader–member convergence is
affected by the agent (leader or follower) and attribute (behavior or mental process) that were being described. (Our analyses
below use the classification of LMX item 6 as being focused on overt physical behavior but repeating the analyses using that item as
also involving internal mental processes yielded virtually identical results.)
886 C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892

Table 2
Content assessment frequency results for LMX-7 items.

Item number Actor or agent Attribute being rated

Overt Internal
Leader Follower Othera X2 Physical behavior Mental process Neither X2

LMX-7-1 09 39⁎⁎ 00 52.13⁎⁎ 03.5 43.5⁎⁎ 00.0 74.56⁎⁎


LMX-7-2 43⁎⁎ 05 00 69.13⁎⁎ 03.0 42.0⁎⁎ 01.0 69.70⁎⁎
LMX-7-3 46⁎⁎ 02 00 84.50⁎⁎ 13.0 32.0⁎⁎ 02.0 29.40⁎⁎
LMX-7-4 43⁎⁎ 03 02 68.38⁎⁎ 30.0⁎ 15.0 02.0 25.06⁎⁎
LMX-7-5 44⁎⁎ 02 02 73.50⁎⁎ 32.5⁎⁎ 13.5 01.0 32.12⁎⁎
LMX-7-6 03 43⁎⁎ 02 68.38⁎⁎ 24.0⁎⁎ 20.0⁎⁎ 03.0 15.87⁎⁎
LMX-7-7 00 45⁎⁎ 03 79.13⁎⁎ 01.5 44.5⁎⁎ 01.0 79.61⁎⁎
2
Note. All chi-square (X ) tests have df = 2 and are non-directional. Significance levels shown for the actor or agent and attribute frequencies are the result of
directional (one-tailed) binomial tests of that frequency compared (pair-wise) to the two other frequencies. For example, the follower as actor or agent frequency
for LMX-7-1 (39) is significantly (p b .01) greater than both the leader as actor (09) and other as actor (00) frequencies.
a
Defined on the rating form as “Other = Someone else, there is no agent, or the agent cannot be determined.”
⁎⁎ p b .01.
⁎ p b .05.

5. Study two: A test of item content impact on LMX rating convergence

Recasting the expectations that we presented earlier, Study Two was designed to test the following four hypotheses concerning
LMX-7/SLMX-7 item types A through H noted above. The first hypothesis is essentially a building block hypothesis predicting
overall convergence effects. The second hypothesis examines attributional effects by having each rater describe the other dyad
member (and not themselves). It is also based upon the fact that behavior ratings of each dyad member by the other should be less
affected by attributional errors than ratings of mental processes (due to their being observable and not inferential). The third
hypothesis tests for social desirability effects by comparing self-descriptions of behavior with self-descriptions of mental
processes. Here, it is expected that the behavior descriptions should show greater convergence since they are observable and
therefore less likely to be as strongly affected by social desirability. Finally, the fourth hypothesis also tests for social desirability
and controls for attributional effects by having each rater describe observable behaviors (and not mental processes). It is also
based upon self-descriptions being more susceptible to social desirability bias than are descriptions of others.

Hypothesis 1. Leader–member convergence on relationship quality should be significant and positive for corresponding rating
stimuli:

(a) Rating the same dyad member's behavior (rCD and rGH).
(b) Rating the same dyad member's mental processes (rAB and rEF).

Hypothesis 2. Leader–member convergence on relationship quality should be significantly higher when each rater rates the other
dyad member's behavior (rCH) as compared to when each dyad member rates the other dyad member's mental processes (rAF).

Hypothesis 3. Leader–member convergence on relationship quality should be significantly higher when each rater rates their
own behaviors (rDG) as compared to when each dyad member rates their own mental processes (rBE).

Hypothesis 4. Leader–member convergence on relationship quality should be significantly higher when both raters rate the other
dyad member's behavior (rCH) as compared to when each dyad member rates their own behavior (rDG).

5.1. Sample and procedure

Survey data were collected from 135 full-time employed executive MBA students at a medium-sized southern U.S. university. A
recruitment advertisement was posted for all 143 students enrolled in the four classes, and 139 signed up voluntarily to participate
in this study. Later, a survey link was sent via electronic mail to all the participating students (employees) who also were asked to
forward another survey link to their direct supervisors (leaders). Both the participating students and their superiors submitted
their survey responses online. Respondents identified themselves using their names in the surveys in order to match employees'
responses with leaders'. Strict confidentiality was assured and a small extra credit inducement was given to the participating
students. After matching the responses from employees and supervisors, the final dataset consisted of 135 leader–member dyads
(a 94.4% response rate). Within the 135 employees, 62.2% are males and the average age is 32.75 years old. All of them have full-
time jobs with an average organizational tenure of 4.81 years, and they have been working under their current superiors for about
2.67 years (average). As for the 135 leaders, 70.4% are males. Their average age and organizational tenure are 41.87 and 9.23 years,
respectively.
C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892 887

Table 3
Eight sets of Leader–Member Exchange (LMX) description items.

Rater

Follower Leader

Actor and attribute being rated Leader LMX-7-2 (Item Set A) SLMX-7-2 (Item Set B)
Mental LMX-7-3 SLMX-7-3
Process ALMX-1 ASLMX-1

Leader LMX-7-4 (Item Set C) SLMX-7-4 (Item Set D)


Behavior LMX-7-5 SLMX-7-5
ALMX-6 ASLMX-6

Follower LMX-7-1 (Item Set E) SLMX-7-1 (Item Set F)


Mental ALMX-2 ASLMX-2
Process ALMX-3 ASLMX-3

Follower LMX-7-6 (Item Set G) SLMX-7-6 (Item Set H)


Behavior ALMX-4 ASLMX-4
ALMX-5 ASLMX-5

5.2. Measures

The full set of LMX-7 and SLMX-7 items from Graen and Uhl-Bien (1995) were employed (see Appendix A). Employees
responded to the 7-item LMX-7 scale, while their leaders completed the 7-item SLMX-7 scale (using a 5-point response scale—see
Appendix A).
To test leader–member convergence on the same item content, two alternative versions of the LMX-7 and SLMX-7 scales
(ALMX and ASLMX) were created by rewording the first six items to reflect changes in rating actor or agent (alternate versions of
LMX and SLMX item 7 were not created or employed so that each of the item sets would have three items. As a result, differences in
correlations would not be due to differences in the number of items in each item set).
For example, LMX-7 item 2 (“How well does your leader understand your job problems and needs?”) was changed to ALMX-2,
“How well do you understand your leader's job problems and needs?”—representing a shift in agent from the leader to the
follower. SLMX-7 item 2 (“How well do you understand your follower's job problems and needs?”) was changed to ASLMX-2,
“How well does your follower understand your job problems and needs?”—here, shifting agent from the leader to the follower
since the leader was the respondent. (See Appendix A for the detailed wordings of all the new items).
Data on the new ALMX and ASLMX items also were collected from the participating students and their superiors. Ultimately,
eight summated scores (totals of 3-item sets) were created by grouping response items from the original and alterative scales,
based on the actors and attributes being rated (see Table 3).

Table 4
Descriptive statistics, raw and corrected correlations of LMX scales and items sets (N = 135 dyads).

Scale and Item set Mean S.D. 1 2 3 4 5 6 7 8 9 10

1 Leader mental process rated by follower (Item Set A) 3.86 .77 (.75) .46⁎⁎ .84⁎⁎ .35⁎⁎ .98⁎⁎ .51⁎⁎ .79⁎⁎ .55⁎⁎ 1.00⁎⁎ .44⁎⁎
2 Leader mental process rated by leader (Item set B) 4.22 .50 .28⁎⁎ (.49) .45⁎⁎ .65⁎⁎ .42⁎⁎ 1.00⁎⁎ .32⁎ .78⁎⁎ .44⁎⁎ 1.00⁎⁎
3 Leader behavior rated by follower (Item set C) 3.80 .77 .65⁎⁎ .28⁎⁎ (.80) .44⁎⁎ .84⁎⁎ .52⁎⁎ .95⁎⁎ .71⁎⁎ 1.00⁎⁎ .52⁎⁎
4 Leader behavior rated by leader (Item set D) 4.24 .62 .25⁎⁎ .38⁎⁎ .33⁎⁎ (.70) .38⁎⁎ .57⁎⁎ .28⁎ .66⁎⁎ .37⁎⁎ 1.00⁎⁎
5 Follower mental process rated by follower (Item set E) 4.00 .61 .66⁎⁎ .23⁎⁎ .59⁎⁎ .25⁎⁎ (.61) .53⁎⁎ .94⁎⁎ .71⁎⁎ .99⁎⁎ .45⁎⁎
6 Follower mental process rated by leader (Item set F) 4.00 .58 .34⁎⁎ .67⁎⁎ .36⁎⁎ .37⁎⁎ .32⁎⁎ (.60) .45⁎⁎ .82⁎⁎ .51⁎⁎ .99⁎⁎
7 Follower behavior rated by follower (Item set G) 3.96 .75 .59⁎⁎ .19⁎ .73⁎⁎ .20⁎ .63⁎⁎ .30⁎⁎ (.74) .92⁎⁎ .91⁎⁎ .39⁎⁎
8 Follower behavior rated by leader (Item set H) 3.97 .62 .42⁎⁎ .48⁎⁎ .56⁎⁎ .49⁎⁎ .49⁎⁎ .56⁎⁎ .70⁎⁎ (.78) .61⁎⁎ .86⁎⁎
9 LMX-7 (Item set L) 3.99 .70 .88⁎⁎ .29⁎⁎ .86⁎⁎ .29⁎⁎ .73⁎⁎ .37⁎⁎ .74⁎⁎ .51⁎⁎ (.89) .48⁎⁎
10 SLMX-7 (Item set S) 4.26 .45 .33⁎⁎ .75⁎⁎ .40⁎⁎ .76⁎⁎ .30⁎⁎ .66⁎⁎ .29⁎⁎ .65⁎⁎ .39⁎⁎ (.74)

Note. Scale reliabilities are shown in the parentheses at the diagonal. Raw correlations are presented below the diagonal. Reliability-corrected correlations are
presented above the diagonal.
⁎⁎ p b .01.
⁎ p b .05.
888 C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892

5.3. Analyses

Bivariate Pearson correlations were computed to investigate the level of association among the LMX-7, SLMX-7, and eight item
sets and to test the four hypotheses. Descriptive statistics (scale means and standard deviations) and coefficient alpha (α) internal
consistency reliability estimates of the LMX-7, SLMX-7 and eight item sets also were computed and are presented in Table 4. Due
to differences in reliability, the raw correlations were also corrected for scale unreliability (r*; Nunnally & Bernstein, 1994) so as to
equate the relationships for differences in internal consistency reliabilities among the variables under examination. Finally,
correlations among the item sets and scales were compared by correlated t-tests for differences in correlations that used data from
the same respondents (i.e., both correlations coming from only the subordinates or only the supervisors) and Z-tests were used
when one of the correlations being compared was based on the subordinate data and the other was based on the supervisor data
(see McNemar, 1969, pp. 157–158).

5.4. Results and discussion

The raw and corrected correlations are presented in Table 4, along with the means, standard deviations, and reliability
estimates. As can be seen in Table 4, the reliabilities for leader mental process rated by the leader (item set B; α = .49), follower
mental processes rated by the follower (item set E; α = .61) and follower mental processes rated by the leader (item set F; α = .60)
were below the .70 level suggested by Nunnally and Bernstein (1994) as desirable for basic research instruments. Consequently,
comparisons among the relationships were made using both raw (r) and reliability-corrected (r*) correlations.
Our first hypothesis appears strongly supported as all four correlations are positive and significant, using both raw correlations
(rCD = .33, rGH = .70, rAB = .28, and rEF = .32, all p b .01 or better) and reliability-corrected correlations (r*CD = .44, r*GH = .92,
r*AB = .46, and r*EF = .53, all p b .01 or better).
Hypothesis 2 predicted significant attributional effects on rater agreement. Here, the hypothesis is supported as rCH is
significantly greater than rAF (z = 2.27, p b .05) and a significant difference also holds when the reliability-corrected correlations
are examined (z = 2.63, p b .01).
The first social desirability effects hypothesis, Hypothesis 3, is not supported. Both the raw and the corrected correlations for
each rater rating their own behaviors (rDG and r*DG) are not significantly greater than the corresponding correlations for each rater
rating their own mental processes (rBE and r*BE).
Finally, Hypothesis 4, also predicting significant social desirability effects, is supported. Convergence in rated relationship
quality was higher when both dyad members rated the other dyad member's behavior (rCH) than when they each rated their
own behavior (rDG) (z = 3.46, p b .01). Again, this also holds employing the reliability-corrected correlations as well (z = 3.87,
p b .01).
Additional analyses comparing the item sets with the original LMX-7 and SLMX-7 scales are also informative. Looking at
Table 4, it can be seen that this sample evidenced a high initial level of LMX-7 and SLMX-7 convergence (rLS = .39, r*LS = .48), at
least as compared to those samples typically reported in the literature (Gerstner & Day, 1997; Schriesheim, Neider & Scandura,
1998; Sin, Nahrgang & Morgeson, 2009). However, the leader-provided description of follower behavior was more highly related
to LMX-7 scores (rHL = .51, r*HL = .61; p b .01 for both correlations) than was the leader-provided description of follower mental
processes (rFL = .37, r*FL = .51; p b .01 for both correlations) (t = 1.99 and 2.41, respectively; both differences significant at p b .05).
Similarly, the follower-provided description of leader behavior was more highly related to SLMX-7 scores (rCS = .40, r*CS = .52;
p b .01 for both correlations) than was the follower-provided description of leader mental processes (rAS = .33, r*AS = .44, p b .01 for
both correlations) (t = 1.05 and 1.90, respectively, the second difference being significant at p b .05). Again, these results indicate
the benefits of having respondents' rate other (rather than self) attributes and, more specifically, rate others' behaviors rather than
mental processes.

6. General discussion

The evidence reported above is based upon two very different studies but the results appear consistent and clear. The findings
would seem to strongly support the idea that convergence between the LMX-7 and SLMX-7 measures may be attenuated due to
the presence of attributional and social desirability biases (elicited when respondents describe unobservables related to the other
dyad member and when they describe themselves).
Since much of the extant LMX research has not collected assessments of exchange quality from both dyad members (Gerstner
& Day, 1997; Schriesheim, Neider & Scandura, 1998), our findings further support issues that were raised earlier (e.g., Schriesheim,
Castro, & Cogliser, 1999a) concerning exactly what we really know about relationships between leader–member exchange quality
and various dependent variables (such as job and supervisory satisfaction, organizational commitment, and job performance). We
would not go so far as to suggest that all previous research is invalid or uninformative, but clearly future research should attempt
to measure the leader–member relationship more effectively, using improved measures that are less likely to suffer from various
biases and response distortions.
Although our discussion has focused on convergence, incorporating these suggestions into a programmatic scale development
research effort, incidentally, may also result in an LMX measure with improved construct validity. Notably, Liden, Sparrowe and
Wayne (1997) and others (e.g., Gerstner & Day, 1997) have previously called for just the type of behavior- and exchange-based
measures which this study indicates would be more effective in measuring the leader–member relationship: “…the focus on
C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892 889

perceptions has caused LMX research to drift from its initial foundation in social exchange theory” (Liden, Sparrowe & Wayne, 1997,
p. 108) or, more succinctly, “the exchange process [in LMX] is inferred but not directly measured” (Gerstner & Day, 1997, p. 838).
Stated simply, from a psychometric or measurement validity perspective, the current evidence supports the earlier suggestions of
other LMX researchers.
Following the scale development procedure suggestions of Hinkin (1995, 1998) and others, a good starting place would
be to theoretically define exactly what behaviors are to be described (cf. Schriesheim, Cogliser, Scandura, Lankau &
Powers, 1999, Schriesheim, Powers, Scandura, Gardiner & Lankau, 1993). Based upon previous work in this area, both
positional (formally- or organizationally-mandated) and personal (discretionary) leader and member behaviors would
seem best included so as to capture the full range of relationship quality, from low or “out-group” to high or “in-group”
relationships. Both subordinate (LMX) and supervisor (SLMX) versions of the instrument would need to be developed and
validated.
Granted, alternative measures of LMX have appeared in the literature since the LMX-7 and SLMX-7 were introduced.
Mentioned earlier, the LMX-MDM scale (Liden & Maslyn, 1998), which conceptualizes exchange relationships as multi-
dimensional, has been used in some research, although this measure has not eclipsed the LMX-7 as a standard in the literature.
More recently, a measure of leader–member social exchange (LMSX) (Bernerth, Armenakis, Field, Giles, & Walker, 2007) has been
introduced to the literature, and it has been asserted that it addresses the construct validity concerns that are mentioned above
regarding explicit assessment of the exchange process. Since our primary focus is on issues of convergence, commenting on the
general merits of the LMX-MDM and LMSX measures compared to the LMX-7 is beyond the scope of this discussion. That said,
since convergence of leader and member ratings, like construct validity, is an important criterion by which to assess measures of
LMX, we encourage future research to examine convergence of these scales as we have here with the LMX-7 and SLMX-7.
(Parenthetically, as of 2007, Bernerth and colleagues had not even developed a leader version of the LMSX so efforts would need to
start there.)
In light of our findings, LMX scholars also need to examine the validity issue of LMX research, particularly about linking either
member's or leader's LMX perception with outcome variables. The biases due to attributional errors and social desirability might
pose threats to the construct validity of an LMX measure from one source (i.e., either member or leader). As a result, associating
these single-source LMX measures with outcome variables might produce misleading results, especially for outcome variables that
are obtained from the same source as the LMX measure. For example, although past research has found a significant positive
relationship between LMX and job performance, this effect size is generally modest and heterogeneous (Gerstner & Day, 1997).
This variability in effect size could be a result of unidentified moderators or it could be a measurement artifact. If LMX is rated by
the member and performance by the leader in some studies and by the follower in others, differences in response biases
(attributional versus social desirability) may be leading to the variability of results.
That said, we must also concede that many types of response biases can operate in any rating domain. A long history of research
on the existence of bias in questionnaire and performance ratings has indicated the existence of escalation bias (Schoorman,
1988), halo bias (Becker & Cardy, 1986), and primacy and recency effects (Steiner & Rain, 1989), among others. Thus, one
limitation of the current research is that we only examine two possible biases, those stemming from attributions and social
desirability. These biases were the two that seemed most relevant to the LMX convergence issue, but other sources of bias could
certainly have effects on LMX rating convergence. Further research on other possible biases is clearly warranted. A second
limitation of this research is that the survey measures in Study 2 were not counter-balanced because we did not expect order
effects. However, because we did not counter-balance we cannot be sure that order effects were not present.
Echoing the arguments of Gerstner and Day (1997) and Schriesheim, Neider and Scandura (1998); Schriesheim, Castro &
Cogliser (1999) information on LMX needs to be collected from both leaders and followers so that the quality of the
relationship can be accurately operationalized. The results from our study further suggest that not only do we need to
measure LMX from both parties, their convergence might also be a meaningful factor influencing employee work outcomes
(cf. Gerstner & Day, 1997). Outcomes associated with both leaders and followers should obviously be highest when both
parties have similar perceptions of the LMX relationship and both parties rate the relationship as being high (rather than
low) in quality. Outcomes associated with leaders and followers may be very different, however, when ratings diverge. If
leader ratings are higher, these relationships might be characterized by high member performance (as perceived by the
leader) but low job satisfaction (as felt by the member, who perceives a poor managerial relationship). If member ratings
are higher, these relationships might be characterized as having dynamics contrasting those just mentioned. Obviously, if
improved measures are used, these types of questions may be investigated by future research and might lend themselves
to testing through the use of analysis of variance (e.g., Cogliser, Schriesheim, Scandura & Gardner, 2009), within and
between entities analysis (WABA; Dansereau, Alutto, & Yammarino, 1984; see Schriesheim, Castro, & Yammarino, 2000, for
an example) and polynomial regression techniques (Edwards, 1994; Edwards & Parry, 1993), thus offering a new venue for
useful applications of these methods.
In conclusion, Schriesheim, Castro and Cogliser (1999a) asserted that the current commonly-used measures of LMX and
SLMX were developed and promulgated without adequate consideration being given to both theoretical issues and to
psychometric refinement, testing, and validation. Whether this assertion is true or not, the current research clearly
suggests that a new global measure of overall leader–member exchange relationship quality is needed. Perhaps by
developing an improved measure, knowledge about fundamental leader–member exchange relationships will continue to
advance and this approach will continue to remain one of the actively employed frameworks for the study of leadership
phenomena.
890 C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892

Appendix A. Item content for four leader–member exchange (LMX) measures

Graen & Uhl-Bien (1995), (Table 3) recommended LMX-7 items

LMX-7-1. Do you know where you stand with your leader …do you usually know how satisfied your leader is with what you do?
Rarely Occasionally Sometimes Fairly often Very often

LMX-7-2. How well does your leader understand your job problems and needs?
Not a bit A little A fair amount Quite a bit A great deal

LMX-7-3. How well does your leader recognize your potential? (How well do you recognize)
Not at all A little Moderately Mostly Fully

LMX-7-4. Regardless of how much formal authority s/he has built into his/her position, what are the chances that your leader would use his/her power to help
you solve problems in your work?
None Small Moderate High Very high

LMX-7-5. Again, regardless of the amount of formal authority your leader has, what are the chances that s/he would “bail you out,” at his/her expense?
None Small Moderate High Very high

LMX-7-6. I have enough confidence in my leader that I would defend and justify his/her decision if s/he were not present to do so?
Strongly disagree Disagree Neutral Agree Strongly agree

LMX-7-7. How would you characterize your working relationship with your leader?
Extremely ineffective Worse than average Average Better than average Extremely effective

Graen & Uhl-Bien (1995), (Table 3) recommended SLMX-7 items

SLMX-7-1. Does your follower usually know where he/she stands with you …does your follower usually know how satisfied you are with what s/he does?
Rarely Occasionally Sometimes Fairly often Very often

SLMX-7-2. How well do you understand your follower's job problems and needs?
Not a bit A little A fair amount Quite a bit A great deal

SLMX-7-3. How well do you recognize your follower's potential?


Not at all A little Moderately Mostly Fully

SLMX-7-4. What are the chances that you would use your power to help your follower solve problems in his/her work?
None Small Moderate High Very high

SLMX-7-5. What are the chances that you would “bail out,” your subordinate at your expense?
None Small Moderate High Very high

SLMX-7-6. Your follower would defend and justify your decision if you were not present to do so.
Strongly disagree Disagree Neutral Agree Strongly agree

SLMX-7-7. How would your follower characterize his/her working relationship with you?
Extremely ineffective Worse than average Average Better than average Extremely effective

Alternative LMX items

ALMX-1. Does your leader know where s/he stands with you …does your leader usually know how satisfied you are with what s/he does?
Rarely Occasionally Sometimes Fairly Often Very Often

ALMX-2. How well do you understand your leader's job problems and needs?
Not a bit A little A fair amount Quite a bit A great deal

ALMX-3. How well do you recognize your leader's potential?


Not at all A little Moderately Mostly Fully

ALMX-4. Regardless of how much formal authority you have built into your position, what are the chances that you would use your power to help your leader
solve problems in his/her work?
None Small Moderate High Very high

ALMX-5. Again, regardless of the amount of formal authority you have, what are the chances that you would “bail him/her out,” at your expense?
None Small Moderate High Very high

ALMX-6. My leader has enough confidence in me that he/she would defend and justify my decision if I were not present to do so?
Strongly disagree Disagree Neutral Agree Strongly agree
C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892 891

Appendix A (continued)

Alternative SLMX items

ASLMX-1. Do you know where you stand with your subordinate …do you usually know how satisfied your subordinate is with what you do?
Rarely Occasionally Sometimes Fairly often Very often

ASLMX-2. How well does your follower understand your job problems and needs?
Not a bit A little A fair amount Quite a bit A great deal

ASLMX-3. How well does your follower recognize your potential?


Not at all A little Moderately Mostly Fully

ASLMX-4. Regardless of how much formal authority s/he has built into his/her position, what are the chances that your follower would use his/her power to
help you solve problems in your work?
None Small Moderate High Very high

ASLMX-5. Again, regardless of the amount of formal authority your follower has, what are the chances that he/she would “bail you out,” at his/her expense?
None Small Moderate High Very high

ASLMX-6. I have enough confidence in my follower that I would defend and justify his/her decision if s/he were not present to do so.
Strongly disagree Disagree Neutral Agree Strongly agree

References

Ariew, R. (1976). Ockham's razor: A historical and philosophical analysis of Ockham's principle of parsimony. Champaign-Urbana: University of Illinois.
Atwater, L. E., & Yammarino, F. J. (1997). Self-other rating agreement: A review and model. Research in Personnel and Human Resources Management, 15, 141–164.
Becker, B. E., & Cardy, R. L. (1986). Influence of halo error on appraisal effectiveness: A conceptual and empirical consideration. Journal of Applied Psychology, 71,
662–671.
Bernardin, H. J., & Beatty, R. W. (1991). Performance appraisal: Assessing human behavior at work (2nd ed.). Boston, MA: Kent.
Bernerth, J. B., Armenakis, A. A., Field, H. S., Giles, W. F., & Walker, H. J. (2007). Leader–member social exchange (LMSX): Development and validation of a scale.
Journal of Organizational Behavior, 28, 979–1003.
Cardy, R. L., & Dobbins, G. H. (1994). Performance appraisal: Alternative perspectives. Cincinnati, OH: South-Western.
Cascio, W. F., & Aguinis, H. (2005). Applied psychology in human resource management (6th ed.). Upper Saddle River, NJ: Pearson/Prentice Hall.
Cogliser, C. C., Schriesheim, C. A., Scandura, T. A., & Gardner, W. L. (2009). Balance in leader and follower perceptions of leader–member exchange: Relationships
with performance and work attitudes. The Leadership Quarterly, 20, 452–465.
Crowne, D. P., & Marlowe, D. (1964). The approval motive: Studies in evaluative dependence. New York: Wiley.
Dansereau, F. (1995). A dyadic approach to leadership: Creating and nurturing this approach under fire. The Leadership Quarterly, 6, 479–490.
Dansereau, F., Alutto, J. A., & Yammarino, F. J. (1984). Theory testing in organizational behavior: The “varient” approach. Englewood Cliffs, NJ: Prentice Hall.
Dansereau, F., Yammarino, F. J., & Markham, S. E. (1995). Leadership: The multiple-level approaches. The Leadership Quarterly, 6, 251–263.
Dumdum, U. R., Lowe, K. B., & Avolio, B. J. (2002). A meta analysis of transformational and transactional leadership. In B. J. Avolio & F.J. Yammarino (Eds.),
Transformational and charismatic leadership: The road ahead (pp. 35–66). Oxford, UK: Elsevier Science.
Edwards, A. L. (1990). Construct validity and social desirability. American Psychologist, 45, 207–209.
Edwards, J. R. (1994). The study of congruence in organizational behavior research: Critique and a proposed alternative. Organizational Behavior and Human
Decision Processes, 58, 51–100.
Edwards, J. L., & Parry, M. E. (1993). On the use of polynomial regression equations as an alternative to difference scores in organizational research. Academy of
Management Journal, 36, 1577–1613.
Erdogan, B., & Liden, R. C. (2002). Social exchanges in the workplace: A review of recent developments and future research directions in leader–member exchange
theory. In L. L. Neider & C.A. Schriesheim (Eds.), Leadership (pp. 65–114). Greenwich, CT: Information Age Publishing.
Evans, M. G. (1970). The effects of supervisory behavior on the path–goal relationship. Organizational Behavior and Human Performance, 5, 277–298.
Gerstner, C. R., & Day, D. V. (1997). Meta-analytic review of leader–member exchange theory: Correlates and construct issues. Journal of Applied Psychology, 82,
827–844.
Graen, G. B., & Uhl-Bien, M. (1995). Relationship-based approach to leadership: Development of leader–member exchange (LMX) theory of leadership over
25 years: Applying a multi-level multi-domain perspective. The Leadership Quarterly, 6, 219–247.
Hinkin, T. R. (1995). A review of scale development practices in the study of organizations. Journal of Management, 21, 967–988.
Hinkin, T. R. (1998). A brief tutorial on the development of measures for use in survey questionnaires. Organizational Research Methods, 1, 104–121.
Hinkin, T. R., & Tracey, J. B. (1999). An analysis of variance approach to content validation. Organizational Research Methods, 2, 175–186.
House, R. J. (1971). A path–goal theory of leader effectiveness. Administrative Science Quarterly, 16, 321–338.
James, L. R., Demaree, R. G., & Wolf, G. (1984). Estimating within-group interrater reliability with and without response bias. Journal of Applied Psychology, 69, 85–98.
Liden, R. C., & Maslyn, J. M. (1998). Multidimensionality of leader–member exchange: An empirical assessment through scale development. Journal of Management,
24, 43–72.
Liden, R. C., Sparrowe, R. T., & Wayne, S. J. (1997). Leader–member exchange theory: The past and potential for the future. In G. R. Ferris (Ed.), Research in personnel
and human resources management, Vol. 15. (pp. 47–119)Greenwich, CT: JAI Press, Inc.
Martinko, M. J. (2002). Thinking like a winner: A guide to high performance leadership. Tallahassee, FL: Gulf Coast Publishing.
Martinko, M. J., & Gardner, W. L. (1987). The leader/member attribution process. Academy of Management Review, 12, 235–249.
McNemar, Q. (1969). Psychological statistics (4th ed.). New York: Wiley.
Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill.
Ostroff, C., Atwater, L. E., & Feinberg, B. J. (2004). Understanding self-other agreement: A look at rater and rate characteristics, context, and outcomes. Personnel
Psychology, 57, 333–375.
Paulhus, D. L. (1985). Self-deception and impression management in test responses. In A. Angleiter & J.S. Wiggins (Eds.), Personality assessment via questionnaire:
Current issues in theory and measurement. New York: Springer-Verlag.
Schoorman, F. D. (1988). Escalation bias in performance appraisals: An unintended consequence of supervisor participation in hiring decisions. Journal of Applied
Psychology, 73, 58–62.
Schriesheim, C. A., Castro, S. L., & Cogliser, C. C. (1999). Leader–member exchange (LMX) research: A comprehensive review of theory, measurement, and data-
analytic practices. The Leadership Quarterly, 10, 63–113 (a).
892 C.A. Schriesheim et al. / The Leadership Quarterly 22 (2011) 881–892

Schriesheim, C. A., Castro, S. L., & Yammarino, F. J. (2000). Investigating contingencies: An examination of the impact of span of supervision and upward
controllingness on leader–member exchange using traditional and multivariate within- and between-entities analysis. Journal of Applied Psychology, 85,
659–677.
Schriesheim, C. A., Cogliser, C. C., Scandura, T. A., Lankau, M. J., & Powers, K. J. (1999). An empirical comparison of approaches for quantitatively assessing the
content adequacy of paper-and-pencil measurement instruments. Organizational Research Methods, 2, 140–156 (b).
Schriesheim, C. A., & Neider, L. L. (1996). Path–Goal theory: The long and winding road. The Leadership Quarterly, 7, 317–321.
Schriesheim, C. A., Neider, L. L., & Scandura, T. A. (1998). A within- and between-groups analysis of leader–member exchange as a correlate of delegation and as a
moderator of delegation relationships with performance and satisfaction. Academy of Management Journal, 41, 298–318.
Schriesheim, C. A., Powers, K. J., Scandura, T. A., Gardiner, C. C., & Lankau, M. J. (1993). Improving construct measurement in management research: Comments and
a quantitative approach for assessing the theoretical adequacy of paper-and-pencil and survey-type instruments. Journal of Management, 19, 385–417.
Schriesheim, C. A., & Zhou, X. (2002, Aug.). Using confirmatory factor analysis and item response theory to ensure measurement invariance in supervisor and
subordinate descriptions of leader–member exchange. Paper presented at the annual Academy of Management Meeting, Denver, CO.
Siegel, S. (1956). Nonparametric statistics for the behavioral sciences. New York: McGraw-Hill.
Sin, H., Nahrgang, J. D., & Morgeson, F. P. (2009). Understanding why they don't see eye to eye: An examination of leader–member exchange (LMX) agreement.
Journal of Applied Psychology, 94, 1048–1057.
Steiner, D. D., & Rain, J. S. (1989). Immediate and delayed primacy and recency effects in performance evaluation. Journal of Applied Psychology, 74, 136–142.
Yammarino, F. J., & Atwater, L. E. (1997). Do managers see themselves as others see them? Implications of self-other rating agreement for human resources
management. Organizational Dynamics, 25, 35–44.
Yukl, G. (2010). Leadership in organizations (7th ed.). Upper Saddle River, NJ: Pearson/Prentice-Hall.
Zhou, X. (2003). An exploration of supervisor-subordinate agreement on leader–member exchange. Unpublished doctoral dissertation, School of Business
Administration, University of Miami.
Zhou, X., & Schriesheim, C. A. (2009). Supervisor–subordinate convergence in descriptions of leader–member exchange (LMX) quality: Review and testable
propositions. The Leadership Quarterly, 20, 920–932.
Zhou, X., & Schriesheim, C. A. (2010). Quantitative and qualitative examination of propositions concerning supervisor–subordinate convergence in descriptions of
Leader–Member Exchange (LMX) quality. The Leadership Quarterly, 21, 826–843.

You might also like