Copyright
D. Anthony (David Anthony) Butterfield.

The Likert organizational profile: methodological analysis and test of System 4 theory in Brazil online

. (page 1 of 2)
Online LibraryD. Anthony (David Anthony) ButterfieldThe Likert organizational profile: methodological analysis and test of System 4 theory in Brazil → online text (page 1 of 2)
Font size
QR-code for this ebook


The Likert Organizational Profile:
Methodological Analysis and Test of System 4
Theory in Brazil



Dewe,





D. Anthony Butterfield and George F. Farris
Working Paper No. 608-72 July, 1972



The Likert Organizational Profile:
Methodological Analysis and Test of System 4
Theory in Brazil

D. Anthony Butterfield and George F. Farris
Working Paper No. 603-72 July, 1972




OCT 17 1972

M. 1. T. LIbKAKltS



Butterfield



The Likert Organizational Profile:
Methodological Analysis and Test of System 4 Theory in Brazil

Abstract
A 20-item Likert Organizational Profile (LOP) was administered twice
to 256 employees in 13 Brazilian development banks. Actual and ideal
bank profiles were similar to those found in the U.S. and elsewhere:
employees want participative-group management methods but say their organ-
ization uses autocratic or consultative methods. Factor analyses did
not yield the six dimensions predicted by Likert ' s theory; factors were
only partially consistent over time and for different hierarchical levels.
Retrospective scores were quite accurate and equally so for 6, 12, and
18 month time periods since original administration. Test-retest relia-
bility of the LOP as a whole was moderate. Bank LOP scores were unrelated
to objective measures of organizational effectiveness, but were positively
related to employee satisfaction. The LOP appears useful for current
and retrospective organization studies, but the theory of management systems
measured by it was only partially supported.



63G345



Butterfield



The Likert Organizational Profile:
Methodological Analysis and Test of System A Theory in Brazil-'

D. Anthony Butterfield^ and George F. Farris^

Alfred P. Sloan School of Management
Massachusetts Institute of Technology

The work of Rensis Likert and his associates at the Institute for
Social Research represents a major contribution to modern organization
theory. Likert' s two books (Likert, 1961, 1967) summarize much of the
work of the Institute and present his own theory of management. The
theory postulates four systems of management, and argues that system 4,
"participative-group", is the most effective. A considerable amount of
empirical research is presented in support of the theory. Most of the
research was done in organizations in the United States.

More recently, the theory has been tested in Yugoslavia and Japan,
using various versions of a "profile of organizational characteristics"
(Likert, 1969; Kavcic, Rus, & Tannenbaum, 1971). The profile is a ques-
tionnaire which describes an organization on the dimensions of Likert 's
theory, and has come to be knovm as the "Likert Organizational Profile",
or LOP. The LOP has also been used in the United States, frequently in
studies of change (e.g.. Marrow, Bowers, & Seashore, 1967; Golembiewski
& Carrigan, 1970a and b; Blumberg & Wiener, 1971).

Despite this use of the LOP in organization studies, it has not been
subject to much analysis as a measuring instrument. Little evidence has



Butterfield 3

been presented as to the LOP's reliability or factor structure. The
purpose of the present paper is to present a methodological analysis of
the LOP, and to examine the suitability of using the LOP to study organ-
izations in a country outside the United States — Brazil in the present
case. We shall also test support for Likert's theory in Brazil.

To meet these objectives we shall a) present profiles of Brazilian
organizations as seen by their members in terms of Likert's four management
systems; b) discuss factor analyses of the LOP; c) present a longitudinal
analysis of the profile; d) examine relationships between LOP scores and
organizational effectiveness and employee satisfaction.

Method
Instrument

The LOP used in the present study was Form S, distributed by the
Foundation for Research on Human Behavior (1967). Two items (nos. 2 - sub-
ordinates confidence in superiors - and 12 - subordinates knowledge of
organ:, ttional policies) were added at the request of the Brazilians, making
a tot.il of 20 items on the profile. Questions were translated from English
into Portuguese by bilingual Brazilians. These translations were then
checked and revised by the authors in collaboration with the translators
until all were satisfied that the translations were as accurate as possible.
An independent re-translation into English indicated that the Portuguese
and English versions were essentially identical.

LOP questions are answered on a 20-point scale divided into four
sections, each section representing one of the four management systems
described by Likert's theory. Each question has a different set of four
descriptive alternatives, corresponding to the characteristics of the four
management systems. For example, question 1 asks, "How much confidence is



Butterfleld _ ,

shown in subordinates by superiors?" The four alternatives spread out
on the 20-point scale are, "Very little, Some, Substantial, Very much."
Responses more toward the right on the scale receive a higher numerical
value and indicate use of a management system more like the participative-
group system 4 seen by the theory as most effective. According to Likert
(1967), the fact tHat all the scales are in the same direction does not
influence the results.

The items in Form S are grouped into six theoretical dimensions of
organizational processes. Items 1-4 measure leadership; 5-7, motivation;
8 ~.12, communication; 13 - 15, decision making; 16 - 17, goal setting;
and 18 - 20, control. Both individual item scores (Golembiewski & Carrigan,
1970a and b; Kavcic, Rus & Tannenbaum, 1971) and index scores (Blumberg &
Wiener, 1971; Marrow, Bowers, & Seashore, 1967) on the basic dimensions
have been used in previous research.

Procedure

The LOP was administered in group sessions during visits to Brazilian
development finance institutions. On the first administration, instructions
were to answer each question twice, once to describe the organization as it
was actually, and a second time as it should be ideally. On the second
administration, six to eighteen months later, instructions were to describe
the organization as it was actually, and as it had been previously (at the
time of the first administration, although the first administration was not
referred to specifically).

Complete data are available from 13 organizations. One was privately
owned, the others were either development banks or development finance



Butterfield

companies, owned by state or regional governments. For two banks there
was a six-month time lag between administrations; for eight banks there was
a 12 month time lag; and three banks received the LOP 18 months after the
original administration. These differences in time were designed to look
at accuracy of retrospection as a function of the passage of time. It was
the authors' impression that the participants in this study were quite
interested in the task of completing the questionnaire; questionnaire
administration sessions seemed to go as well as they do in the United States
with similar respondents.

Respondents

Respondents were professional employees ranging from non-supervisors
to bank presidents. Sixty-seven per cent were between the ages of 25 and
34; 76% had more than 15 years of schooling; 53% had been with their organ-
izations four years or less; 80% were either lawyers, economists, engineers,
or accountants.

Total individual N's for the Time 1 and Time 2 administrations were
211 and 256, respectively. They represent a non-random sample of about 85%
of the professional staff involved with a loan program for small and medium
sized industrial firms. N's within banks varied from 7 to 41, the size
of the sample generally reflecting the size of the bank. Size did not
influence the results.

Analysis

Except for the factor analyses and hierarchical analyses, data were
analyzed using bank mean scores. The bank was selected as the unit of
analysis since the LOP is designed to measure organizational characteristics,
and our primary concern is with the management systems of organizations.



Butterfield

6

Results
Profiles

What management system are these Brazilian development banks actually
using, and what system would they like to be using ideally? Figure 1



Insert Figure 1 about here



presents the average actual and ideal profiles from the 13 banks, based on
the first LOP administration. It is evident that these 13 financial insti-
tutions were actually using a system 3 "consultative" style of management,
with system 2 attributes in the area of decision making and goal setting.
The ideal style was seen as system 4. Analyses of variance indicated
essentially no differences among banks on their ideal profiles, and signifi-
cant differences on their actual profiles. Similarly, analysis of variance
among three hierarchical levels in the banks (non-supervisory professionals,
supervisors, and top management) showed no difference on ideal profiles,
but frequent differences on the actual profiles. Higher organizational levels
tended to see their organization as more toward system 4 than did lower
levels. Thus there was substantial agreement among banks and organizational
levels that system 4, "participative-group", would be the ideal management
style, yet the actual system was seen as being more towards consultative or
autocratic.

The profiles in Figure 1 are similar to those found in Likert's studies.
They suggest that, in terms of Likert's theory, Brazilian development banks
may be using management systems not very different from those used by
organizations in the United States, or even Yugoslavia or Japan.



Buttcrfield - ^

Factor Analysis of the LOP

The short form of the LOP was designed to measure six basic organi-
zational processes: leadership, motivation, communication, decision making,
goal setting, and control. Does it? Factor analyses of the LOP v^ere
carried out to examine its basic structure and to see if there was empirical
support for the proposed six dimensions.

Table 1 presents results of the factor analysis of the ideal responses
from individual respondents at all organizational levels. A principal



Insert Table 1 about here



components analysis was used, with varimax rotation for six factors.
Except for the first factor, which is clearly leadership, the factors are
at best mixtures of the dimensions proposed by the theory. We have given
them the tentative labels of "resistance, guidance, informed decision
making, dispersion of goal setting and control," and "motivation and
communication." In additional analyses it was more difficult to label
factors. Note that analysis should have stopped with five factors, since
the Eigenvalue for the sixth factor is less than 1.00. A careful examina-
tion of Table 1 also indicates than neither item 5 (type of motivation used)
nor item 7 (amount of teamwork) find a place in any of the six factors.

Additional factor analyses were carried out for the actual responses
from the two time periods. However, since results were much less inter-
pretable, they will not be presented here. Controls for organizational
level tended to complicate rather than clarify results. The most con-
sistent factors were leadership and resistance, but they emerged in a
different order and with a different make-up in each analysis. Usually
there were only three or four factors with Eigenvalues greater than 1.00.



Butterfield 8

Items 7 and 5 frequently did not appear.

The results in Table 1 represent in a sense the best of what the
factor analyses produced; that is, somewhat interpretable factors with
at least some relationship to the theory. Repeated analyses produced
only fair factor consistency over time, fair overlap between ideal and
actual factors, and fair to poor comparability across organizational levels.
This lack of consistency in factors across different analyses was the
characteristic outcome, not the emergence of a large and possibly Halo-
induced general factor. Other studies (Likert, 1967; Marrow, Bowers &
Seashore, 1967) have alluded to such a Halo factor, but none was found in
the present case. Since factors were not clear cut and consistent
across the different analyses, the creation of factor scores did not seem
justified. Subsequent analyses were based on individual items as well as
a total LOP score (overall average across the 20 items) . The latter seemed
justified in view of high intercorrelations among the twenty LOP items
and Likert 's (1967) assertion that the LOP measures relatively consistent
management systems.

4
Longitudinal Analysis of the LOP

How reliable are LOP scores over time? In particular, how accurate
are LOP scores when used retrospectively? The accuracy issue should be
of general interest since investigators frequently ask respondents on
questionnaires to remember how things were at some point in the past, with-
out knowing in fact how they really had been. In the present study we knew.

Accuracy of retrospection . We might expect that accuracy of retrospection
decreases as time increases due to forgetting. Thus in the present study
the banks in which the Time 2 administration followed the Time 1 adminis-



Butterfield ^

tration by six months should be the most accurate: retrospective scores
from the 18-month time lag banks should be the least accurate, and the 12
month group of banks should fall in between. However, if retrospection
were perfect, the actual and retrospective profiles within each time
grouping should be the same. Preliminary visual inspection showed re-
markable overlap between the actual (first administration) and retrospective
(second administration) profiles y/ithin each of the 3 time groups. T tests
done on an item by item basis, separately for each time group, confirmed
that there were few significant differences between actual and retrospective
item scores, for any time group. These results are summarized, using mean
total LOP scores, in Table 2. An F test on the average difference score



Insert Table 2 about here



for the three groupings confirmed that there was no difference in accuracy
among the three retrospective periods (F = 0.268).

Test-retest reliability . To look at "real change" as a function of
time, similar longitudinal analyses were carried out comparing Time 1
actual LOP scores with Time 2 actual scores. Results (not shown) were
the same as for retrospective scores: time had no effect. Since time
differences produced no differences in either retrospective accuracy or
change, simple rank order correlations were calculated using all 13 bank
means on total LOP. Time 1 actual scores correlated rho .63 with Time 2
retrospective scores, and rho .52 with Time 2 actual scores. With an N of 13',
both correlations are significant beyond the .05 probability level using
a one-tail test. The first rho (.63) might be considered the retro-
spective reliability of the LOP. The second (.52) is the test-retest



Butterfield

11

one another.

The last four measures in Table 3 are items taken from another
questionnaire administered in the banks at the time of the second LOP
administration. They are the employees' ovm perceptions of how effec-
tive their bank is, how satisfied they are with their own supervisor,
the extent to which the bank's employees are satisfied rather than dis-
satisfied, and their evaluation of their bank as a place to work compared
with other organizations they know. Since organizational level had a
significant effect on these items (higher levels reporting more favorable
responses) , the results in Table 3 are based on LOP and questionnaire data
from non-supervisory professionals only.

Table 3 shows that bank scores on the LOP were not related to objective
measures of bank effectiveness, but they were related in the predicted
manner to three of the four questionnaire measures of effectiveness and
satisfaction. Contrary to theory, system 4 methods of management were
not associated with better organizational performance, but they were
associated with employees' feelings about performance and with employee
satisfaction. The theory is thus only partially supported.

Note the nature of the one satisfaction item in Table 3 not related
to LOP. Apparently, employees who described their organization as more
towards system 4 were not more satisfied with their own supervisor (nor
were they more dissatisfied). One's immediate supervisor may not be more
satisfying in a system 4 organization, yet employees do feel that the entire
organization is more effective, is a better place to work, and has more
satisfied employees. It appears that the LOP can distinguish between more
general organization-wide and more immediate work-group characteristics.



Butterfield 12

Discussion

Development banks are probably among the more modern of Brazilian
organizations. They are relatively new, and hire a high percentage of
young and well educated employees. Under these conditions, the LOP appears
to be a useful instrument for organization studies in Brazil. Questionnaire
administration went quite smoothly, and profiles were remarkably similar
to those obtained in the United States. Common stereotypes of Latin
American culture would p'-edict that in more traditional Brazilian organ-
izations we might encounter more difficulties in administration of the LOP,
and we might find both actual and ideal profiles toward the authoritative
end of the scale. These stereotypes merit empirical test.

The factor structure of the LOP does not correspond well to the six
hypothesized organizational processes. This is not a serious drawback,
but it does suggest that index scores on the sub-scales should be used
with great caution. Likert himself recognizes the difficulty in obtaining
factors corresponding exactly to the proposed dimensions (Likert, 1967).
He argues that either ideal scores or actual scores from within a single
organization will yield cleaner factors. In the present study we did not
have sufficient data to justify factor analysis from within a single
organization. Factor scores aside, much can still be done with individual
items and especially with a total LOP score. Using ideal and actual
profiles in a feedback program for organizational development has been
useful in both Brazil (Butterfield, 1970) and the United States (Blumberg &
Wiener, 1971).

The reliability of the LOP over an extended period of time is a complex
question. On the one hand, it seems quite clear that organizational members



Butterfield

13

do a fairly accurate job in remembering the state of their organization ■

,-, 6
6, 12, or 18 months into the past. Furthermore, these differences in

time have no effect on the accuracy of retrospection. Using the LOP

retrospectively thus appears to be quite legitimate, at least up to a

period of 18 months. On the other hand, total scores on the LOP are

only moderately stable over time. The test-retest reliability (.52)

obtained in the present study, though statistically significant, is

certainly not high by traditional testing standards. However, given the

small number of organizations (13) and the rather dynamic state of the

development finance field in Brazil, perhaps we ought not to have expected

the test-retest reliability to have been much higher.

Results of the correlations between LOP and dependent variables are
of particular theoretical interest. As predicted, Brazilian development
bankers were more satisfied with participative-group management systems
(although such satisfaction was not dependent on satisfaction with the
immediate supervisor), and said such systems would be the ideal. Contrary
to theory, organizations tending to use such systems were not more
effective in terms of objective performance.

System 4 theory would explain these different results by invoking
a time-lag argument. The benefits of using system 4 methods are already
apparent in the greater satisfaction of organization members. Such
satisfaction usually precedes changes in organizational end-product variables.
Improved performance is on the way; it is only a matter of time. Unfor-
tunately, although we do have longitudinal data in the present study, we
do not have sufficient data over a sufficient time period to test adequately
the time-lag hypothesis.



Butterfield



14



An alternative explanation is the "manpower competence-performance
feedback loop hypothesis" (Farris, 1969; Farris & Butterfield, 1971).
The effectiveness of participative methods depends on having competent,
experienced employees. When such a manpower base does not exist, as is
the likely case in developing countries, participative raetl^ods are not
likely to be more effective. Greater relative competence resides at the
supervisory level, so that closer- methods of supervision may be more
appropriate, as well as "lore accepted (Farris & Butterfield, 1971). In the
present study, for example, satisfaction with immediate supervisor was
not significantly related to being in a more participative organization.

These are alternative rather than competing hypotheses. Results of
the present study suggest that the LOP may be a useful instrument for
exploring them further, even if system 4 theory is only partially supported
in Brazil.



Butterfield 15

References

Blumberg, A., & Wiener, W. One from two: facilitating an organ-
izational merger. Journal of Applied Behavioral Science, 1971,
l_y 87-102.

Butterfield, D. A. Pesquisa de agao em desenvolvimento organ: zacional
na America Latina pelo metodo de levantamento de feedback. Revista
de Administragao Publ ica, 1972, 6^, (1), 89-99.

Farris, G. F. Organizational factors and individual performance: a

longitudinal study. Journal of Applied Psychology , 1969, _52' 87-92.

Farris, G. F., & Butterfield, D. A. Are current theories of leadership
culture-bound? An empirical test in Brazil. Invited paper presented
at the symposiiim, "Contemporary Developments in the Study of Leader-
ship", Southern Illinois University at Carbondale, April, 1971.

Foundation for Research on Human Behavior. Sample kit: Profile of
organizational characteristics. Ann Arbor, Michigan: Institute
for Social Research, 1967.

Golembiewski, R. T. , & Carrigan, S. B. The persistence of laboratory-
induced changes in organization styles. Administrative Science
Quarterly , 1970, 15^, 330-3A0.

Golembiewski, R. T. , & Carrigan, S. B. Planned change in organization

style based on the laboratory approach. Administrative Science Quarterly ,
1970, 15^, 79-93.

Kavcic, B. , Rus, V., & Tannenbaum, A. S. Control, participation, and

effectiveness in four Yugoslav industrial organizations. Administrative
Science Quarterly, 1971, 16, 74-86.



Butterf ield



16



Likert, R. New patterns of management . New York: McGraw-Hill, 1961.

Likert, R. The human organization . New York: McGraw-Hill, 1967.

Likert, R. The relationship between management behavior and social
structure - improving human performance: Better theory, more
accurate accounting. Symposium (C-3) presented at the meeting of
the Conseil International Pour L' Organisation Scientif ique , Tokyo,
1969.

Marrow, A. J., Bowers, E G., & Seashore, S. E. Management by partici-
pation . New York: Harper & Row, 1967.



Butterfield

17

Footnotes

1 This study was supported by a grant from the Ford Foundation. The
authors wish to thank members of the Grupo de Trabalho Misto for their
help, and Dr. Rensis Likert for his permission to use the LOP.
Special acknowledgement goes to Cecilio Berndsen and Eldon Senner

for their assistance in gathering and processing the data,

2 Now at the School of Business Administration, University of Massachusetts,
Amherst, Massachusetts.

3 Requests for reprints should be sent to George F. Farris, Sloan School

of Management, Massachusetts Institute of Technology, Cambridge, Massachu-
setts, 02139.

4 All longitudinal analyses were carried out separately for each organi-
zational level. Results were essentially identical. Data presented
in the longitudinal section are from respondents at all organizational
levels combined.

5 This ratio includes development loans of all types, not just small and
medium industrial projects. The rank ordering of banks on financial
data was so highly correlated from year to year in prior years that
data on this ratio were not obtained for 1970.



Butterfield



X8



TABLE 1
Factor Analysis of Ideal Likert Organizational Profile



Factor



Item



"Leadership"

Eigenvalue 5.689
Cumulative prop, of
total variance



1) Confidence in subordinates .799

2) Confidence in superiors .75A


1

Online LibraryD. Anthony (David Anthony) ButterfieldThe Likert organizational profile: methodological analysis and test of System 4 theory in Brazil → online text (page 1 of 2)