Professional Documents
Culture Documents
Leadership: The Multiple-Level Approaches: Fred Dansereau
Leadership: The Multiple-Level Approaches: Fred Dansereau
LEADERSHIP:
THE MULTIPLE-LEVEL APPROACHES
Fred Dansereau*
State University of New York at Buffalo
Francis J. Yammarino
State University of New York at Binghamton
Steven E. Markham
Virginia Polytechnic Institute and State University
INTRODUCTION
What would happen if we viewed the major approaches to leadership as described in
most academic writings and organizational behavior textbooks in light of multiple levels
of analysis? An answer to this question is the 13 articles that appear in this two-part
special issue titled: “Leadership: The Multiple-Level Approaches.” As editors, we wanted
the leading proponents of 13 major theories in the field to examine the implications of
multiple levels of analysis for their work. As a result, most readers of Leadership
Quarterly will recognize not only the theories but also the names of the authors.
We expected, obviously, the theoretical content of the articles to vary by approach;
we did not expect such a substantial variation in methods, but this is what we found.
At one extreme, Hunt and Ropo display the logic behind a qualitative grounded theory
* Direct all correspondence to: Fred Dansereau, School of Management, State University of New York
at Buffalo, Department of Organization and Human Resources, Jacobs Management Centre, Buffalo, NY
14260.
The three articles that fit under classical approaches include: (1) Schriesheim,
Cogliser, and Neider; (2) Ayman, Chemers, and Fiedler; and (3) Vroom and Jago. These
should be instantly recognizable by any student of organizational behavior because of
their resilience and longevity during the last 35 years.
Contingency Approach
The article by Ayman, Chemers, and Fiedler shows how the contingency model
clearly stated 30 years ago the multiple levels of analysis contained in its domain.
Although readers may agree or disagree with the arguments of the authors about the
meaning of LPC (least-preferred coworker), the clarity of specification of it and other
variables in terms of levels of analysis is a remarkable example of one of the first
multiple-level theories. This article makes the specification of levels of analysis clear
and unambiguous. Moreover, the article makes a clear distinction between the level
of the concepts and the level of measurement. In addition to the original specification,
the authors offer various extensions of the theory, for example, to the dyad level of
analysis. Although the extensions are just beginning to receive research attention, they
represent in our view a potentially useful direction for the future.
the relationship) between the follower and the leader and does not focus on the leader
or follower per se.
Charismatic Leadership
Charisma, for the Greeks, was the fire from the gods that anointed the leader. This
article is “on fire” as we see the clarity of House’s charismatic approach to leadership
where the level of interest is the dyad. But this article is much more than about just
linkages or dyads. Through its propositions, Klein and House specify how and why
charismatic dyads form and how they do or do not become widespread. There is little
ambiguity here about levels. Leaders have different characteristics (an individual
difference perspective) and followers respond as individuals. Clearly, this is a first step
in developing a concept of charismatic leadership not just from a dyadic perspective
but also from an individual difference perspective of leaders and followers. It is, in our
view, a major step in a new and exciting direction with testable propositions for future
research.
Transformational Leadership
Avolio and Bass take a narrower perspective by focusing on one construct-
individual consideration-and extending it to clarify‘its different meanings at different
levels of analysis: individual, group, and collective. They show how the concept changes
meaning depending upon which level of analysis is of interest. Moreover, they show
how, over time, the concept changes as organizations develop. Thus, changes in
theoretical meaning, measurement tools, and the nature of what happens over time are
all linked to levels-of-analysis issues. The authors also describe their current research
that focuses on these issues. The focus of the article on only one variable shows the
importance of clearly specifying the level of analysis for a particular variable because
the meaning of the variable changes depending on which level of analysis is of interest.
Leader-Member Exchange
The article by Graen and Uhl-Bien suggests to us that the leader-member exchange
(LMX) approach focuses solely on (leader-follower) relationships (or linkages). This
is an extremely important clarification of the LMX approach. The leader and the
follower per se are not of interest in this approach. The relationship between them is
of interest. The primary variable (LMX) is defined by them as involving mutual respect,
reciprocal trust, and mutual obligation, thereby making it a relational concept.
This new definition helps explain why the LMX approach, unlike other approaches,
is not specific about the appropriate level of analysis forfollowers or leaders. Followers
and leaders per se are not of interest; relationships are of most interest. Thus, the LMX
approach does not seem to specify levels for the leaders and followers as separate entities.
Graen and his associates appear to believe that any level of analysis (or all levels of
analysis) can be subsumed under the idea of relationships.
The LMX approach departs markedly from the contemporary approach of Klein
and House, where leaders are viewed as individuals and followers are viewed as
individuals and the linkages are dyadic. The LMX approach provides little guidance
as to how to view leaders or followers from a levels-of-analysis perspective. Likewise,
one measure (LMX) is used to reflect relationships apparently at all levels of analysis,
Introduction 101
which differs from the approach of Avolio and Bass who suggest that measures and
concepts change as one changes levels.
We believe the LMX and vertical dyad linkage (VDL) approaches are very different.
The original VDL model, as described in this article, makes assertions about followers
within groups and processes within leaders and dyadic linkages within groups. This
is clear from this article and is the position taken in the later article by Dansereau et
al. In contrast, the more recent LMX model seems to focus on relationships. We believe
a focus on relationships is indeed important and the article by Graen and Uhl-Bien
develops this perspective in detail. But, to us, a focus on leaders and followers per se
also seems as important as a focus on relationships if one is interested in understanding
leadership.
Although the alternative approaches of the three articles in this section seem to take
a very different tack than the classic or contemporary views, these articles show that
the distinction is quickly becoming blurred. Thus, Hall and Lord’s discussion of
information processing expands beyond a simple individual difference perspective.
Similarly, Podsakoff and MacKenzie refocus the substitutes for leadership area from
individual- to group-level effects. Why the convergence? Simply stated, when levels of
analysis are taken seriously, narrow approaches that previously focused solely on
followers become broader by including leaders and their linkages with followers. The
article by Meindl is an exception to this generalization in that it clearly maintains a
focus on followers.
with how to separate out such effects with contextual analysis. In our view, there is
not yet a satisfactory solution to this controversy.
From our perspective, the results of their study suggest that substitutes for leadership
make some contribution to criterion variables over and above leadership variables. The
individual-level results based on individual reports make some sense to us. In contrast,
the conclusions about the groups (based on group means) seem very tenuous. To
examine group effects, they apparently ignore over 50% of the total variance as well
as all within-group covariance. Why, then, did we publish an article that we consider
to be problematic? We believe the debate about a key issue has only begun, and this
article will add to it. The key issue is: what are the criteria that researchers should use
to decide whether there is a group effect or an individual effect, or both? Substantial
discourse may still be necessary in this area because this issue is far from resolved. We
believe that the Podsakoff and MacKenzie approach uses the absolute minimum criteria
one could use to assert a group effect. Based on more rigorous and conservative criteria,
we do not see group effects in their data; rather, we see largely individual self-reports
that weakly correlate with some criterion variables when considered separately. (This
can be viewed as a type of null result with some weak effects at the individual level.)
The article, however, illustrates an approach that might be useful under the following
condition. When the results show significant effects based on the raw score correlations
(individual differences) and significant group effects (based on group averages), then
one could ask, how much does the group add beyond the individual (raw score effects)?
Podsakoff and MacKenzie’s procedures might be helpful in this circumstance, but the
results they present seem too weak to support such an analysis given, in our view, the
weak group-level results.
Nevertheless, the article illustrates, along with the article by Schriesheim and his
colleagues, some key empirical indicators to use when analyzing data. Both articles
include analyses of within and between variation and covariation to assess levels issues-
obviously we agree with the interpretation by Schriesheim et al. but not with Podsakoff
and MacKenzie’s. Moreover, the article shows that the substitutes-for-leadership
approach, like the information-processing views, is clearly open to testing at multiple
levels regardless of differences of interpretation about the strength of effects. In other
words, the substitutes-for-leadership approach allows for linkages between followers
and leaders but also suggests that there is something beyond the leader’s style.
Romance of Leadership
Meindl’s article about the romance of leadership suggests that most classical and
contemporary approaches to leadership ignore followers. We simply disagree. In our
view, most approaches focus on leaders, followers, and their linkage. Meindl, however,
prefers to focus on followers. He views followers as independent of leaders and has
little, if any, interest in the possible link between followers and leaders or in leaders
per se. According to this view, leadership can be viewed as potentially a construction
that individuals can make almost regardless of what leaders actually do. Thus, leadership
can be studied independently of leaders.
This article differs markedly from the article by Hall and Lord, who view dyads and
groups as a possible basis for cognitive and affective effects in information processing.
Likewise, it differs from the work of Podsakoff and MacKenzie where leadership
Introduction 103
measures are considered in addition to substitutes for leadership. We believe that the
romance-of-leadership approach would be advanced by an incorporation of the thinking
in these other alternative approaches. Clearly, the information-processing view of Hall
and Lord and the substitutes-for-leadership approach of Pod&off and MacKenzie
allow for leaders to influence subordinates.
Nevertheless, we agree with the romance-of-leadership notion that followers are very
important and that there is a need to study how they perceive the world (see, e.g., the
article by Hall and Lord). Certainly, followers may bring about the emergence of
leadership. But it is not clear to us that leadership resides only in followers without
at least some referent about which individuals are perceiving (i.e., leaders). There is
probably a philosophical position underlying Meindl’s approach because the article,
as the title suggests, is a social constructionist view of leadership.
Interestingly, the social constructionist perspective of interest seems to allow for the
notion of a contagion effect in which individuals share perceptions as a group-an
individual and group model involving follower-follower linkages. This point is
illustrated in the article. Indeed, we think that this multiple-level view is an important
positive step. We believe, however, that the romance of leadership could be enhanced
by focusing on leaders in addition to followers as well as by focusing on leader-follower
linkages in addition to follower-follower linkages.
The four articles of this section offer more recent and different perspectives on aspects
of leadership that can be analyzed from a multiple-levels perspective. First, the article
by Markham, Markham, and Neck illustrates some of the problems when moving from
an individual view of self-leadership and self-management to a self-directed work team
vantage point. Second, the article by Kim and Yukl offers the first empirical test at
multiple levels for Yukl’s multiple-linkage theory. Third, Hunt and Ropo use a
qualitative approach to identify how one individual’s leadership style may impact an
entire corporation, in this case, GM. Finally, the Dansereau, Yammarino, Markham,
Alutto, Newman, Dumas, Nachman, Naughton, Kim, al-Kelabi, Lee, and Keller article
takes a different direction by moving away from past VDL models and toward a
tripartite model that specifies a different view of the leader, the follower, and the leader-
follower relationship as necessary parts of a dyadic exchange process. Importantly, they
report data from a number of studies that replicate their findings.
Self-leadership
Self-directed leadership implies that leaders develop followers who can function
independently of the leaders, basically on their own. The ultimate development is self-
managed work teams. As Markham, Markham, and Neck show, this is a fertile area
for theoretical development and research. They suggest that the key levels of interest
in this area in prior research have been the differences between leaders as individuals
and entire work groups. They also note the role of individual differences in how self-
leadership is inculcated. Finally, they enumerate a number of research issues that need
to be addressed vis-bvis levels of analysis.
104 LEADERSHIP QUARTERLY Vol. 6 No. 2 1995
Multi-Level Theory
Hunt and Ropo’s article analyzes the case of Roger Smith and General Motors to
further develop Hunt’s multi-level theory. (Multi-level in this case includes the
organizational level.) Using a grounded theory approach, they generate propositions
and suggest various ways that these propositions may contain multiple levels of analysis.
It is an interesting article in that it suggests the potential importance of levels of analysis
even when using a qualitative grounded theory research approach.
individualized leadership
The article by Dansereau et al. has a somewhat different focus than the other new
wave articles. It moves away from the vertical dyad linkage approach that focused on
(1) the differences within followers, (2) differences within leaders, and (3) differences
among dyads within groups. Instead, it views followers as independent of each other
or as individuals, and leaders as linking with subordinates on this individual basis.
Leaders then respond to the situation (i.e., each follower) and link dyadically with
followers. In this new approach to leadership, leaders first provide support for the sense
of self-worth of followers as unique individuals who are independent of other individuals
with whom they interact. Second, in exchange, followers then perform in ways that
satisfy the leader. Third, as a result, leaders and followers link together in dyads where
there is consistency and agreement, yet differences, between these independent dyads
(i.e., agreement is not dependent on work group membership). This theory is tested
in a number of studies, and nearly identical effects were found in all studies. There
are two key points that distinguish this approach from the old VDL approach. First,
the failure of a key variable in the vertical dyad linkage approach, negotiating latitude
to show an effect. Second, the individualized leadership approach focuses on dyads
independent of groups.
This approach would be enhanced by considering some of the features of the other
new wave approaches. An increase in the number of variables of interest in future
research seems appropriate as well as a comparison with the variables described by
Kim and Yukl. Observations in the field of how support for self-worth, the key variable
of interest, might be manifested in multiple contexts using a qualitative perspective
similar to the one used by Hunt and Ropo seem useful. Finally, a consideration of
how the individualized approach might culminate in self-leadership as described by
Markham et al. seems appropriate.
Introduction 105
There are three very different views of leaders that emerge from this set of articles.
First, there is the more traditional individual-difference view. The focus here is on
differences between leaders. Second, there is the view that leaders differentiate among
individuals and become leaders for only some followers. The focus in this case is upon
differences within leaders (a vertical dyad linkage approach). Third, there is the view
that leadership is based on the situation and, in this sense, independent of the leader.
The focus in this instance is not on the leader but rather on the situation. These three
views follow the ideas of Dansereau, Alutto, and Yammarino (1984) and Klein,
Dansereau, and Hall (1994): (1) homogeneity or “whole entities,” (2) heterogeneity or
“parts of entities,” and (3) independence or equivocality. These three perspectives are
different views of one level of analysis-in this case, of the leader as a person. This
focus on only the leader, however, is very deceiving because, in fact, the approaches
focus on followers per se and leader-follower linkages. We now consider each set of
approaches.
Leader-Member Exchange (LMX): Levels are open; Linkages may be individuals in groups, leader-followers, dyads, dyads in groups, and so forth.
Introduction 107
A Follower-Only Approach
The romance of leadership (Meindl) is clearly multi-level in the sense that individuals
or groups can cogitate about leaders. Moreover, in our view, the notion that followers
108 LEADERSHIP QUARTERLY Vol. 6 No. 2 1995
are important is an implicit part of all the approaches described thus far. The notion,
however, that followers alone influence their perceptions of the leaders without regard
to a target person is probably a new idea but it is not very persuasive for us. We do
believe the idea that followers’ perceptions may have no link with leaders and may simply
reflect followers’ individual differences is a very valid point made by the romance-of-
leadership approach. We define this as a case of no leadership, where there is a lack
of dyadic or person-group link between leaders and followers. Likewise, we believe
leadership may emerge from followers. We, however, believe, as apparently do most
theorists in this area, that at least sometimes there is a link or interdependence between
leaders and followers. We believe the romance-of-leadership approach could be
enhanced by incorporating a ‘more formal and explicit consideration of both leaders
and leader-follower linkages.
A Linkage-Only Approach
The LMX approach (Graen and Uhl-Bien) suggests a focus not on the leader or
follower but rather on their relationship (linkages in our terminology). We do not believe
that relationships exist entirely independently of the actors who compose them. While
Graen and Uhl-Bien describe different relationships that may evolve between leaders
and followers, there is a lack of specificity about the actors involved in this process.
For example, we are told that relationships may reflect dyads or dyads in groups and
that dyads may be about just any two entities. Clearly, this is a multi-level statement
that needs substantial clarification. For example, Klein and House refer to a dyadic
relationship in which the leader and follower are viewed as individual persons. In
contrast, Dansereau et al. refer to a dyadic relationship in which the follower is an
individual but the leader responds to just the follower (the situation) as such. Likewise,
several other approaches focus on links between the leader as a person and a group.
We are not convinced that a single LMX measure will assess all of these different aspects
of the linkage between leaders and followers. We believe that the power of the LMX
approach would be enhanced by focusing on leaders and followers and by making
predictions about them as do most other approaches. Finally, we distinguish the LMX
from the VDL approach because the latter approach makes predictions about all three
components of leadership: leaders, followers, and their linkage.
CONCLUSION
We believe that all the approaches to the study of leadership in this issue make a
contribution because they view leadership as a multiple-level-of-analysis phenomenon.
In fact, these well-developed approaches to leadership suggest that individuals who are
viewed as prominent and famous leaders probably (1) differ from other individuals,
(2) make distinctions among people, and (3) are influenced by the situation (the
followers). This general three-fold perspective endorsed by 11 of the 13 articles in this
special issue recognizes that followers are important (as pointed out by Meindl) and
that they form linkages with leaders (as pointed out by Graen and Uhl-Bien) but that
leaders, as a separate research focus, are important as well.
Introduction 109
We believe that there is a lack of empirical data about the variables that (1) distinguish
leaders from nonleaders, (2) capture how leaders differentiate among people, and (3)
assess how situations influence leaders. In other words, the theories (particularly as
described in the articles in this issue) go far beyond existing empirical results in the
extant literature. We hope the articles presented in this special issue, and in particular
the extension of the approaches in terms of levels of analysis, will encourage data
collection and research that include multiple levels of analysis. We view the articles
in this special issue not as an ending but as a beginning for theory development and
data analysis.
REFERENCES
Dansereau, F., Alutto, J.A., & Yammarino, F.J. (1984). ZZkeory testing in organizational
behavior: The vurient approach. Englewood Cliffs, NJ: Prentice-Hall.
Klein, K.J., Dansereau, F., & Hall, R.J. (1994). Levels issues in theory development, data
collection and analysis. Academy of Management Review, 19(2), 195-229.