Professional Documents
Culture Documents
G - HuntLeadership-Style Effects at TwoManagerial Levels in A Simula
G - HuntLeadership-Style Effects at TwoManagerial Levels in A Simula
G - HuntLeadership-Style Effects at TwoManagerial Levels in A Simula
Hunt
There has been an increasing emphasis in recent textbook and theoretical literature on the
possible differences in leadership re- quirements at different managerial levels (Litterer, 1965;
Nealey and Fiedler, 1968; Pfiffner and Sherwood, 1960), but very few empirical studies. Mann
(1965) found in a study of hospitals that different leadership styles were required at different
managerial levels to satisfy subordinates. Nealey and Blood (1968), in a study of a Veteran's Ad-
ministration hospital, found that both satis- faction and favorable performance required different
leadership styles at different hier- archical levels. A laboratory study by Misumi and Shirakashi
(1966) attempted to examine performance and satisfaction under different first- and second-level
leadership- style combinations; however, their manipu- lation of the second-level manager's
leader- ship style was not completely successful.
The present study was designed to extend knowledge in this area by considering in a laboratory
setting (I) the effects of com- binations of different leadership styles across first and second
levels of management on performance and satisfaction; and (2) com- parison of these effects
with those which occur when first- and second-level managers' leadership styles are considered
separately. The first condition is here called two-level
knowledge, while the second is termed one- level knowledge, where level refers to managerial
level.Ì‚
The measure of leadership style used in the present study was Fiedler's (1967) measure of a
person's esteem for his least preferred coworker (LPC). A subject is asked to think of the person
with whom he has had the most difficulty in working and to describe him using a series of eight-
point bipolar adjective scales. High scores indicate that the respondent evaluates his least pre-
ferred coworker in a relatively favorable manner. Fiedler (1967, 1970) and others have shown
leaders' scores on this variable to be significantly related to group per- formance in a wide
variety of groups, the direction of the relationship being a func- tion of the situation in which the
leaders operate. Extensive efforts have been made by Fiedler (1967, 1970) to provide a mean-
ly This research was conducted at Southern IUi- nois University at Carbondale and is based on
re- finements of an earlier pilot study conducted at the Center for Management Science,
University of California at Berkeley as a part of the Workshop in Laboratory Experimental
Research in Business Administration and Economics. The author is in- debted to the Ford
Foundation, the sponsor of the workshop. He is also grateful to R. Bishop, J. Hill, and W. Vicars
for helpful comments on the paper and to J. Reaser and G. Yunker for assistance in the statistical
analysis.
476
ingful interpretation of the measure; how- ever, as Fiedler (1970) indieated, it has not been
correlated with most personality test seores and various attempts to relate the score to self-
descriptions, descriptions by others, or behavioral observations have led to complex or
inconsistent results. Thus, Fiedler (1970) has reinterpreted the meas- ure as more data have
become available. His 1967 interpretation was that a leader with a high LPC score was primarily
oriented toward interpersonal relations while a leader with a low LPC score was primarily
oriented toward successful task completion. His 1970 interpretation is that the LPC score reflects
a hierarchy of goals. Leaders with high LPC scores have the establishment and maintenance of
inter- personal relations as a primary goal with prominence and self-enhancement as a secondary
goal. A leader with a low LPC score has as his primary goal the achieve- ment of task and
material rewards and as his secondary goal the development of good interpersonal relations. A
person will try to achieve both primary and secondary goals in situations where his influence is
relatively great; whereas he will stress only his primary goals where he is in an unfavorable or
stressful situation where it is not possible to obtain both goals. Differences between leaders with
high and low LPC scores are also discussed by Mitchell et al. (1970).
METHOD
an introductory management course. All were juniors and seniors, and most were business-
school students. They were tested for leadership style and mental ability. Leadership style was
measured by the previ- ously described LPC scale. Mental ability was assessed by means of the
12-minute Wonderlic Personnel Test (Wonderlic, 1966).
perimental teams, each composed of 7 sub- jects. A second-level executive headed each team
with two Brst-level managers subordi- nate to him, each of whom supervised two
workers. Each team was assigned to one of four executive-manager LPC combinations on the
basis of pretest LPC scores. Com- binations were diose shown in Table 1.
Executives Managers
Subjects were also matched according to Wonderlic scores so that mental ability did not differ
significantly in combination or within any given team.
This design, in conjunction with the sta- tistical technique discussed later, allows a comparison
of two-level versus one-level knowledge effects, that is, a comparison of the effects of both
executive and manager together versus the separate effects of each.
lated tape-recorder design department of the Acme Company. Figure 1 shows how the teams
were organized and the permitted communication channels for task information. Note that there
is unlimited interaction al- lowed within each of the two sections within a department, but that
communication be- tween sections must follow hierarchical lines. Also, each manager while
allowed communications of unlimited frequency and length with the executive, could not meet
with the other manager except for a 10- minute period during preliminary planning for the task.
While either the executive or manager could initiate interactions, all meet- ings were in the
executive's office separated from the managers' work areas.
Description of Task
Department
Amplifier Section
T Speaker designer
T
Manager Playback Head
Amplifier designer
Motor designer
The overall departmental task objective was to reach a set of numerical goals as- signed hy top
management, who was the experimenter, on each of three criteria of tape-recorder
performance—durability, sound quality, and appearance. The same numerical goals were
assigned to each of the 26 departments.
Each section functioned essentially as an engineering design unit with the objective of selecting
appropriate tape-recorder parts, which in combination with those from the other section would
yield the assigned de- partmental values for the three criteria. De- eisions concerning one part,
such as a speaker, could not he made without certain information concerning the other parts, in-
cluding those in the other section. Such in- formation was obtainable only hy means of the
previously described communications channels. Thus, there was a highly inter- dependent task
with a functional and geo- graphical division of labor requiring a great deal of coordination at
each of the two man- agerial levels.
The roles of the speaker designer and manager in the speaker and amplifier section
illustrate the experimental design and are similar to those of the other participants. The speaker
designer could design four types of component speakers—R, S, T, and U. Eaeh speaker type
interacted diflFerently when connected with the component am- plifier, playback head, and
motor types he- ing designed by others. The speaker designer worked with three suhcomponents:
dia- phragm, magnet, and fabric. There were a number of different types of each subcom-
ponent, for example, type a, b, c, and d dia- phragms, each interacting with the otlier
subcomponent types to produce various levels in pitch, tone, and volume. These levels were
expressed in numerical terms. A table was then used to convert these levels for pitch, tone, and
volume to speaker nu- merical contributions to the departmental goals of durabihty, sound
qualit)', and ap- pearance.
The selection of subcomponent types was restricted in that only certain types could be used to
design each of the four speaker types; for example, only diaphragm type a, b, or c could be
selected for speaker R. Similar restrictions held for magnet and fabric types. In turn, only certain
types of speakers, amplifiers, playback heads and motors could be combined for the tape re-
corder. The manager of the section had in- formation ahout which subcomponent types
were permissible for each speaker and am- plifier type. Information about which com- ponent-
type combinations were permissible tor the tape recorder was divided between the speaker
designer and amplifier designer.
The above restrictions, of course, had to be considered when arriving at numerical contributions
from each component. These numerical contributions were then averaged for each section and, at
the executive level, die section contributions were averaged to provide the departmental values
for the three criteria.
The executive was assigned the role of clarifying the nature of the task and pro- viding the
overall coordination necessary to accomplish departmental goals. He was the only one provided
with information about the overall task and the specific numerical goals. Everyone else was
given only partial information. The executive was also pro- vided with exhibits providing an
illustrative sample of the exact kind of detailed infor- mation each manager and designer was
given. His role then was to work with his managers to clarify the overall task, deter- mine how to
implement the policy direc- tives, and to provide a coordinative role. Coordination was
extremely important be- cause of the interdependency of the func- tions and the geographic
division of labor, with its accompanying restrictions on com- munications. He also monitored the
task performance of each section, since all infor- mation concerning component and subcom-
ponent combinations and contributions to departmental goals had to be submitted to him so that
he could summarize the infor- mation and turn it in to top management.
The managers, in turn, were required to provide coordinative activities within their section. Their
function differed from tliat of the executive in its narrower scope and in their constant contact
with both of their de- signers whose work needed to be coordi- nated. Their role was
considerably facili- tated if the executive demonstrated a comprehensive grasp of the overall task
and provided a systematic approach to overall goal accomplishment; if not, the managers and
designers were at a considerable dis- advantage because of the incompleteness of
The task was iterative in nature, and, therefore, four trials were allowed, the first 2 hours long,
the second 1 hour, and tbe third and fourth each 30 minutes. At the end of each trial, a
department was required to submit its calculated numerical values, to- gether with substantiating
data. It was tben to refine these figures for the next ti-ial. If a department reached tbe assigned
goals be- fore the end of a trial period, it was di- rected to submit results at that time. New goals
were assigned to any department achieving the assigned goals before the end of four trials.
The original intent was to base the per- formance scoring on tbe amount of time required to meet
the original goals and any subsequently assigned goals. However, only two teams reached the
goals and were as- signed new ones before the end of the last trial, so this procedure could not be
used. Instead, a scoring procedure was devised which recognized tiie iterative nature of the task
and rewarded teams for their stand- ing at the end of the first trial and for prog- ress toward the
goals for each subsequent trial, even though they did not actually achieve the assigned goals. The
procedure also provided a bonus for those teams which did achieve the original goals and made
progress toward achieving a second set of goals. The assumption behind this scoring procedure
was that progress toward the goals indicated a systematic problem-solving approach as opposed
to one which might allow a team to come relatively close in an early trial and then not be able
eitber to reach the assigned goals or move closer to reaching the goals in later trials.
A team's standing at the end of the first trial and its progress toward the goals was based on three
considerations: (1) Was a team able to submit a complete set of goals? (2) If a goal set was
submitted, did any goal violate any of the constraints men- tioned earlier? and (3) If an
acceptable set of goals was submitted, how close were they in numerical terms to the goals
assigned? The last of these considerations was ex-
pressed as the sum of tbe absolute distance of the goals from the sum of tbe assigned goals.
However, ability to submit a com- plete set of goals or constraint violation could be expressed
only in discrete yes or no terms, since data-gathering methods did not allow for differential
weighting of errors.
Performance scores to reflect the variety of requirements above were assigned as follows:
1. At tbe end of tbe first trial, a team turning in a complete set of goals with no violation of
constraints scored 1. If it did not submit a complete set of goals or if there was a constraint
violation it scored 0.
2. For each of the three remaining trials, a team received -f-l if it progressed closer to the goals
than it was at the end of the previous trial, 0 if it stayed the same, or — 1 if it got worse. For
example, a team that did not submit an acceptable set of goals for the first trial and was able to
submit such a set for the second, would be given H-l as would a team that progressed
numerically closer to the assigned goals between the first and second trials. The values assigned
in steps 1 and 2 above were then summed to give an overall score, the measure of per- formance.
3. The two teams which reached the goals during the fourth period and were able to submit
figures for the next set of goals assigned were each given a bonus of -1-1 to be added to their
score. With the consideration of these bonus points the per- formance scores for the 26 teams
ranged from —1 to +4 with a mean of 1.54 and a standard deviation of 1.33.
It should be noted that the relationships between LPC and performance reported in the present
study are, with one exception, substantially the same whether or not bonus points are assigned or
scoring explicitly in- cludes standings in the first trial or does not include them but uses them as
a basis for the calculation of progress toward the goal scores for later trials. The exception
pertains to the relationship between manager LPC and performance which differs because in
some cases it is significant at slightly better than the .05 level while in others it does not quite
meet this level.
cate this study with a more refined scoring system. Such a scoring system would prob- ably be
relatively complex and might in- corporate differential weightings for errors committed, and so
forth. If the task were simplified, time to reach assigned goals might be used with some provision
for er- rors. However, results from such a simplified task might differ from those here because
the task would be changed.
Experimental Procedures
1. The representative of top management took the executive and his two managers to an
orientation room, administered a pretest measure of anxiety to them, and gave them an
orientation to the simulation by reading from an experimenter's orientation script.
2. The executive and his managers were each given a separate written set of instriic- tions. The
executive was sent to his oflBce, and each manager was sent to his work area.
3. The designers were taken to the orien- tation room, oriented in a manner similar to the
executive and managers, and given their written instructions.
4. Designers were sent to their work areas and allowed to meet with their man- ager for 5
minutes.
5. Both managers were sent to the execu- tive's ofiice to meet for 10 minutes.
6. Each manager was sent back to his work area and throughout the remainder of the simulation,
the earlier discussed com- munication rules were in force.
7. The top management representative notified the executive 15 minutes and again 2 minutes
before first-trial results were due.
8. First-trial figures were collected, checked for errors, and the executive was informed as to
whether there was an error or not.
9. The executive was left to communicate trial results as he pleased, subject to the general
restrictions on interaction, and the procedures in steps 7 and 8 were repeated.
10. A posttest measure of anxiety was ad- ministered, and everyone completed a post- session
questionnaire.
11. Participants were debriefed.
Postsession Questionnaires
Participants were asked to complete a number of different items concerning their perceptions of
the simulation. Those of rel- evance for the present study were concerned with satisfaction.
Various scales of the Job Descriptive Index (JDI) developed at Cor- nell University (Smith et al.,
1969) were used. The executives completed JDI scales about their satisfaction with the work and
their managers; managers about their work, the executive, and their subordinates the de- signers;
subordinates with the work, the manager, and coworkers. Each of these was an 18-item scale.
Reliability and validity results are discussed in Smith et al. (1969).
It should be noted that both executive satisfaction with managers and manager satisfaction with
subordinates were mea- sured, with a change in directions, by the JDI scale designed to tap
satisfaction with coworkers.
Analysis of Data
While this study produced ordinal rather than interval performance data, it was felt tliat the
research questions could be most readily answered using parametric multi- variate analysis rather
than a nonparametric technique. Anderson (1961) and Boneau (1961) provided considerable
justification for tliis conclusion. Also, parametric tech- niques have traditionally been used with
JDI satisfaction scales.
The approach used for the analyses was the multiple linear regression analog to a parametric
analysis of variance (Bottenberg and Ward, 1963; Kelly et al, 1969), a tech- nique especially
suitable for answering the kind of research questions posed in this study.
Since the performance criterion was a team rather than individual score and there were 26 teams,
LPC scores of the two man- agers on each team were averaged to pro- vide n = 26. Satisfaction
measures at both the manager and subordinate level were also treated in this way. Thus, all
analyses are based on an n = 26.
Results are based on a series of compari- sons of various full-regression models against various
restricted models for performance
and each satisfaction criterion (Kelly et al, 1969). All the models are based on an initial full
model, which takes the form:
Y = E + M-\-EXM where:
The E + M term of the model, here called additive effects, includes both executive and manager
main effects. The E X M term of the model refiects the effects of the four executive-manager
LPC combinations dis- cussed earlier.
1. The full model was compared against a restricted model which included only the additive
effects. This showed whether LPC combination accounted for significantly more criterion
variance then additive effects alone; that is, whether there was a significant executive-manager
interaction with respect to LPC scores, which would indicate that interactive two-level
knowledge is an im- portant predictor.
2. The additive model was compared against zero, to determine whether there were additive two-
level effects where inter- active two-level effects were not significant.
3. The additive two-level model was compared, in turn, against an executive LPC model and a
manager LPC model. If the additive model was significant in step 2, this would tell whether it
predicted signif- icantly better than either executive or man- ager one-level knowledge, which
would in- dicate the extent to which each kind of one-level knowledge accounted for uniquely
different variance.
4. Executive and manager LPC scores were separately compared against zero to see if one-level
knowledge was significant.
Two-level
One-level
Executive-manager combination, that is, variable 2, did not account for a signifi- cantly larger
amount of variance in team performance than executive-manager addi- tive effects R] ^ = .316,
RÌ‚ = .315Ì‚ DF = 1 and 22, F = .02, p = .878; therefore, it was concluded that executive-manager
LPC in- teraction effects were not significant and that additive effects accounted for virtually aU
of the variance.
The data further showed that this vari- ance in additive effects was significantly different from
zero iRÌ‚ = .315, R^ = .561, DF = 2 and 23, F = 5.28, p = .013; addi- tive two-level effects
significantly predicted team performance.
The question still remained as to whether additive two-level knowledge would predict
significantly better than one-level knowl- edge. The data showed that additive two- level
knowledge did indeed predict better than one level knowledge RÌ‚ = .315, RÌ‚ = .170, DF = 1 and
23, F = 4.88, p = .037 and Rl = .315, RÌ‚ = .145Ì‚ DF = 1 and 23, F = 5.69, p = .026; therefore,
LPC scores of executives and managers each appeared to contribute a relatively unique source of
variance. The results also indicated that even though two-level knowledge predicted better than
one-level knowledge, executive knowledge alone would predict significantly better than zero R|
= .170, Rg = - -412, DF = 1 and 24, F = 4.90, p = .037; while manager alone did not quite reach
the .05 significance level RÌ‚ = .145, RÌ‚ = .381, DF = 1 and 24, F = 4.08, p = .055. Figure 2
throws further light on these findings. It may be read in the same way as an inter- action diagram
often shown when analysis of variance has been used. It is significant that the curves for
managers with high and
low LPC scores are parallel to each otlier rather than being nonparallel or crossing each other as
they would if there were a significant interactive effect. Also, there is a positive relationship
between LPC scores and performance at the managerial level and a negative relationship at the
executive level.
Neither two-level nor one-level knowl- edge signifieantly predicted executive satis- faction with
either the work or managers. Likewise, manager satisfaction with the work and with subordinates
was not signif- icantly related to LPC scores. However, manager satisfaction with the executive
was significantly related to LPC scores. Consid- eration of the effects of all the models leads to
the conclusion that one can predict man- ager satisfaction with the executive as well with
manager LPC scores alone as with two- level knowledge RÌ‚ = .165, RÌ‚ = .161, DF = 1 and 23,
F = .11, p = .746, and that managers with high LPC scores are signif- icantly more satisfied with
the executive than managers with low LPC scores RÌ‚ = .161, RÌ‚ = .401, DF= 1 and 24, F =
4.61, p = .042. At the same time, executive LPC scores alone had no significant effect on
manager satisfaction with the executive RÌ‚
Leadership and Subordinate Satisfaction The effects of LPC scores were not signif-
icantly related to either subordinate satis- faction with the manager or coworker. Sub- ordinate
work satisfaction was, however, significantly related to executive LPC score alone RÌ‚ = .221, R
= .470, DF = 1 and 24, F = 6.80, p = .015 while none of the other LPC effects quite reaches
the .05 significance level.
Summary For no criterion did the interaction be-
tween executive and manager LPC scores add a significant amount of variance beyond that
accounted for by executive-manager additive effects. Two-level additive knowl- edge predicted
team performance signifi-
2.4
2.2
2.0
1.8
score high
ID
i 1-2 u
:!
.8
.6
.4
.2 0
Low High
Performanee was the only criterion for whieh two-level knowledge predicted better than one-
level knowledge. Manager LPC scores alone were significantly positively re- lated to manager
satisfaction with the execu- tive while executive LPC scores alone were significantly positively
related to subordinate
work satisfaction. No other satisfaction mea- sure was significantly related to LPC scores.
These results suggest that in the present organization, one should employ managers with high
LPC scores and executives with low LPC scores to improve performance. This would, of course,
be the same hiring decision that would be made if one-level knowledge had been significant and
two- level additive knowledge had not. However, here one can also increase performance by
pairing managers with high LPC scores with either kind of executive (see Figure
2); likewise, one can pair executives with low LPC scores with either kind of man- ager, because
each makes a relatively unique contribution to the variance accounted for. Thus, one can now
improve the performance of a poorly performing group with an ex- ecutive having a high LPC
score by pairing him with a manager with a high LPC score, whereas, if two-level knowledge
were not significant, he could not do this. Similarly, one can improve performance of groups led
by managers with low LPC scores by pair- ing these managers with executives with low LPC
scores.
However, companies which place a heavy emphasis on internal promotion from the first to the
second managerial level would have a mucb greater problem than those where the first-line
supervisory position was largely terminal, because high LPC scores at the first level, which
provide for the best performance in conjunction with low sec- ond-level LPC scores, are the least
success- ful second-level style. If LPC scores were used as the promotion criterion, the com-
pany would be in the paradoxical position of promoting those managers with inap- propriate
leadership styles, low LPC scores, to executive positions where their styles would now be
appropriate.
The previous discussion, has considered only LPC scores as predictors; however, in reality other
variables besides leadership style might be considered, so that the orga- nization could have
additional fiexibility in promoting a manager with a high LPC score.
If one were concerned with satisfaction without regard to performance, he need only consider
manager satisfaction with the ex- ecutive and subordinate satisfaction with work since none of
the other results are significant. To improve manager satisfaction with the executive, managers
with high LPC scores would be used without regard to ex- ecutive LPC scores. Subordinate
satisfac- tion with work would be improved by using executives with high LPC scores without
regard to manager LPC scores. Fortunately, here LPC scores required to optimize one kind of
satisfaction will not lower the other kind of satisfaction. In this sense then, LPC score
requirements are consistent.
mance and satisfaction, he can use managers with high LPC scores with executives having low
LPC scores and improve performance and manager satisfaction with the executive, since for this
measure of satisfaction it does not matter which executive leadership style is used. However, this
combination, because of the low executive LPC scores, will lead to the lowest subordinate work
satisfaction.
Results such as these, if found outside a laboratory setting, have important implica- tions for
management development, because they now require consideration of very dif- ferent
requirements at the two levels rather than simply treating a second-level manager as an improved
version of a first-level man- ager. Nealey and Fiedler (1968) discussed tbis issue along with
related ones in some detail.
Nothing has been said here about chang- ing leadership style through training, which if
successful would, of course, minimize some of the previously mentioned problems. However,
Fiedler (1967) did not consider training particularly appropriate where LPC scores were used as
a leadership measiu-e because he conceived of LPC as a trait rather than a behavioral measure
and pre- sented evidence that it was relatively stable so that training designed to change the
leadership style would probably be diflBcult to implement. Fiedler (1967) argued, in- stead, that
rather than trying to change a leader's style to fit a given situation, one might alter the situation to
fit the style.
These findings add to a growing body of literature indicating tbat leadership re- quirements at
different management levels may change markedly. It seems likely that leadership knowledge of
two levels will often be more fruitful tban knowledge of one level, but the specific kind of two-
level knowledge is quite likely to be infiuenced by a number of situational variables.
Thus, it seems important to conduct future studies under different kinds of organiza- tional
situations. While this obviously needs to be done in the field, the present experi- mental design
and task offer several possi- bilities under relatively controlled condi- tions. For example, the
division of labor
could be varied, the role and power of the executive could be changed, communica- tions
networks could be changed. The task would be especially useful if a more refined scoring system
were developed.
There is a need to determine some of the process variables that will give us a better
understanding of why a given leadership style or combination of styles is more ef- fective than
another. Why are higher-level managers able to influence the performance of workers when they
have very little if any contact with them? In the present study, for example, executive LPC had a
significant influence on team performance and subordi- nate work satisfaction while manager
LPC did not, even though the executive never directly contacted a designer. One obvious area for
examination is how the higher-level manager structures the options available to those below
through his influence on policy formulation and interpretation. Influence may also be exerted by
the way in which a higher-level manager's leadership modifies the behavior of a lower-level
manager and vice versa.
REFERENCES
Bottenberg, Robert A., and Joe H. Ward, Jr. 1963 Applied Multiple Linear Regression.
Washington, D.C: U.S. Department of Commerce Office of Technical Ser- vices, AD 413128.
Behavior of High and Low LPC Per- sons. Seattle: Organizational Research Laboratory,
University of Washing- ton.
1969 Research Design in the Behavioral Sciences: Multiple Regression Ap- proach. Carbondale:
Southern Illinois University Press.
Misumi, Jyuji, and Sanshiro Shirakashi 1966 "An experimental study of the effects
of supervisory behavior on productiv- ity and morale in a hierarchical orga- nization." Human
Relations, 19: 297- 306.
Mitchell, Terence P., Anthony Biglow, Cerald R. Orcken, and Fred E. Fiedler
1970 "The contingency model: criticism and suggestions." Academy of Man- agement Journal,
13: 253-267.
Nealey, Stanley M., and Milton R. Blood 1968 "Leadership performance of nursing
supervisors at two organizational levels." Journal of Applied Psychol- ogy, 52: 414-422.
Nealey, Stanley M., and Fred E. Fiedler 1968 "Leadership functions of middle man-
Pfiffner, John M., and Frank P. Sherwood 1960 Administrative Organization. Engle-
Minimum Occupational Scores for the Wonderlic Test. Northfield, 111.: Won- derlic.
Radiya Muhammed
Greetings everyone,
Benjamin Werner