”II III Iz-m
yr... IKI1I ItsMII

   

_:_fi_:_::____ _

      

 

      

       

 

‘1 }fi(\‘i‘ui<ll nlmu.“ n 1

IflL

(259/. Hgbvzu" ‘ ....y(.,

nun-n-unuuunuu

  
      

u
h
I—\
‘
1

 

='='"

    

 

 

 

 

 

  

 

MATHEMATICAL MODEL OPATTrruDECf .- I ~ ' : *
CHANGE IN A LONGITUDINAL STUDY 2

. 0F LARGE ORGANIZATIONS .

Thesis for the Degree of :Ph. D.
. MICHIGAN STATE UNIVERSITY
-, JIMLTARTER ’

 

 

 

 

   
     

‘Alllllllljlllllljlllllllllll L—

31 oo 99 9495

  

LIBRARY

Michigan State
University

This is to certify that the ,

thesis entitled
MATHEMATICAL MODELS OF ATTITUDE CHANGE
IN A LONGITUDINAL STUDY‘ OF LARGE ORGANIZATIONS

presented by

Jim L. Tarter

has been accepted towards fulfillment
of the requirements for

 

Ph . D. degree in Management

Adam/735:5

Nljor professor

Date November TO 1972

m

0-7 639

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ABSTRACT

MATHEMATICAL MODELS OF ATTITUDE CHANGE
IN A LONGITUDINAL STUDY OF LARGE ORGANIZATIONS

By

Jim L. Tarter

The study of managerial behavior uses a wide range of research
teduuques ranging from simple observation to sophisticated experimental
imnhods. Recently, there has been increased interest in longitudinal
snufies which facilitate the inference of causality. The research
rqnnted here represents one phase of a longitudinal field study designed
toincrease the understanding and generalizability of the attitude change
reunionships found in a Management By Objectives (MBO) system used in
two large organizations.

The specific research problem addressed by this project is best
(Escribed as an attempt to answer the question: "Are the changes in
nmnagerial attitudes observed in organizations utilizing an M80 system
mfnmnagement the result of real or apparent change?" The Chesser study
ofcmange relationships in the MBO system provided a foundation for this
Iesearch (Chesser, 1971). First, data were collected at two points in
‘Unw from a large organization and were used to replicate the Chesser
study. Then the data were reanalyzed and a revised research model of
(mange relationships was develOped. The replication and the revision of

the Chesser model produced contradictory conclusions about the change in

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

v.. ~
c“ 4.3.

\::'e\

 

 

 

Jim L. Tarter

the system. These contradictory results were evident when the data were
compared to the predictions of two change models.

The first, the ”general factor" model, assumes that there is a
general factor of real change which is represented by the variables in
the MBO system. It is not precisely defined but a close analysis of the
data shows that five of the seven variables in the MBO system are very
highly correlated and serve as estimates of this general factor. When
the data were compared to the predictions of this hypothesized model, it
was concluded that there was real change in the system. L

The second model, which predicted no real change in the true
scores for the managers, was called the ”mood" model. In effect, the
mood model assumes that there is some transient component in the observed

scores for the managers which produces apparent change in attitude.

 

Thus, under the assumptions for this model, the data did also fit its
predictions.

The dilemma posed by these obviously contradictory inferences was
attacked by analyzing a third set of data. It was concluded that there
was real change in the attitudes and perceptions of the subjects and that
the general factor of real change model was the most plausible and
appropriate theoretical formulation of that attitude change process.

The results reported here point to important considerations in
the implementing of organization change programs such as M150. More
importantly, a theoretical framework has been provided for further

longitudinal studies of organization change.

 

MATHEMATICAL MODEL OF ATTITUDE CHANGE
IN A LONGITUDINAL STUDY OF LARGE ORGANIZATIONS

By

Jim LQTTarter

 

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Management

1972

 

 

p. H
"‘ L. ..

'~
“.2...

 

“Mi. .

Mud... ,

-~.\

Kw-

v .. {um
i
i

'3.

‘ N

v
"has,

it acii

555m

“5hr.

 

‘I-; v
.4‘
it“)

 

 

{5} ACKNOWLEDGEMENTS

It is a pleasure to acknowledge those who contributed to this

thesis and say a Special "thank you" to each of them.

Dr. Henry Tosi, the chairman of my thesis committee, who
supplied the "Open door" for data collection and, more importantly, the

guidance for a genuine research eXperience.

Dr. John Hunter, who invested much time and energy toward
providing the methodological understanding upon which this thesis is

built.

Drs. John and Jeanne Gullahorn, who provided the encouragement

 

to continue through several research design preposals. Additionally,
I am very grateful for the financial support provided by the Gullahorn
administered National Institute of Mental Health research grant

entitled, "Testing Organization Theories by Computer Simulation."

Dr. Rod Chesser, who so willingly provided his personal support

and advice to initiate this research.

Mrs. Janet Hertzler, who by meticulous editing and very pro-
fessional typing, made the writing and producing of this thesis much

easier.

My wife, Ann, to whom I dedicate this thesis, for her love,

faith, understanding, and ever present encouragement.

ii

 

 

 

 

 

 

NA

.1

 

Chapter

1

TABLE OF CONTENTS

GENERAL RESEARCH PROBLEM, REVIEW OF THE LITERATURE,
RESEARCH STRATEGY, AND PRELIMINARY FINDINGS .

General Research Problem.
Review of the Literature.
The Chesser Study.
Research Strategy and Preliminary Findings.
Summary . . . . . . . . . . . . . . .

THE REPLICATION OF THE CHESSER STUDY AND A REVISED
RESEARCH MODEL. . . . . . . . . . . . .

The Replication of the Chesser Study. . .
Correlational Data for Replication Study .
Static Correlations .
Dynamic Correlations. .
Cross-Lagged Correlations .
Impact Correlations .
Effects Diagrams . . -
Contradictions to the Real Change Hypothesis .
Results of Revised Research Model Development .
Summary . . . .

MODELS THAT ASSUME REAL CHANGE.

The Endogenously Generated Real Change Model.
The Exogenously Generated Real Change Model .
General Factor Model. . -
The Basic Model and the Static Correlations.
The Cross- -Lagged Correlations.
The Dynamic Correlations .
The Impact Correlations. . .
The Nature of the General Factor .
Discussion.

MODELS THAT ASSUME NO REAL CHANGE .

A Model Based on Errors of Measurement.
The Transient Factor Model.

The ”Mood" Model.

Summary ,

iii

 

32

32
35
37
38
43
49
54
59
63

65

66
70
73
77

 

 

 

 

 

 

 

 

 

\'

AU“
"DHUJR

iv
Chapter Page

5 TEST OF THE REAL CHANGE HYPOTHESIS. . . . . . . . . . . . . 79

Transient Factors and Unreliability . . . . . . . . . . . . 79
Real or Cumulative Change . . . . . . . . . . . . . . . . . 84
The Test for Real Change. . . . . . . . . . . . . . . 93

The Firm A Third Administration Data. . . . . . . . . . . . 95
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 110

6 SUMMARY AND IMPLICATIONS OF THIS RESEARCH . . . . . . . . . 112

Implications of This Research . . . . . . . . . . . . . . . 116
Implications for Practice. . . . . . . . . . . . . 116
Implications for Further Research. . . . . . . . . . . 117

APPENDICES

 

APPENDIX A - MBO Study Questionnaire. . . . . . . . . . . . 119
APPENDIX B - Chesser Seven Scale Research Model . . . . . . 126

APPENDIX C - Other Findings in the Replication of the
Chesser MBO Study. . . . . . . . . . . . . . . 130

APPENDIX D - MBO Questionnaire Items, Scale Inter-
Correlations, and Loading Matrices for the
Revised Fifteen and SeVen Scale Research
Models . . . . . . . . . . . . . . . . . . . . 158

1
APPENDIX E — Other Findings in the Firm B MBO Study - - ° ' 179 1

APPENDIX F - Moderated Change Relationships in the M80
System . . . . . . . . . . . . . . . . . . . . 190

BIBLIOGRAPHY. . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

 

.. fl

.\ .\ .\

.. ..m of. "pm .4 s. A u . i Y
c u o . d a Q .P- .. \ds \|\ talk e h\ I
s . . . x . s .U... I... .9... . k Int. fit“ 11... Ti \H A L In. .\~ I-

I‘IK
O I . I\I '. I“ k I - \\~ lil“ -1 I I ‘1 ~ I
I I I I I D Q I I C \\ I Q
I I l I R I I I 0 . l I I I I I ‘1. I‘lxi ‘ I! ~‘\ ‘NI‘

 

 

Table
2-1
2-2
2-3
2-4

2-5

2-6

2-7

2-8

2-9

2-10

3-1

3-2

3-3

3-4

LIST OF TABLES

Description
Classification of Firm B Managers by Functional Area .
Classification of Firm B Managers by Organizational Level.
Classification of Firm A Managers by Functional Area .
Classification of Firm A Managers by Organizational Level.

Means and Standard Deviations for First and Second
Administrations of Questionnaire for Firm A and Firm B .

Reliabilities of Seven Scales. . . . .
Correlation Matrices for the First Administration, Second

Administration, and Change Scores in the Seven Scale
Research Model for Firm B Managers . . . . . . . . .

Means and Standard Deviations for Revised Seven Scale Model.

Internal Scale Reliabilities and Change Score
Reliabilities for Revised Seven Scale Model.

Correlation Matrices for the First Administration, Second
Administration, and Change Scores in the Rev1sed Seven
Scale Research Model for Firm B Managers .

Reordered Matrix of Averaged Static Correlations for the
Chesser Study Firm A First and Second Administrations.

Reordered Matrix of Averaged Static Correlations for the

Replication Study First and Second Administrations--Firm B .

Reordered Matrix of Averaged Static Correlations for the
First and Second Administrations--Revised Seven Scale
Research Model~~Firm B Managers. . . . - . - - .

Reordered Cross-Lagged Correlation Matrix for the Firsth
and Second Administrations—-Chesser Seven Scale Researc

Model-~Firm A Managers .

Page
11
ll
11

ll

13

14

17

28

29

3O

42

44

45

50

 

 

 

 

 

 

 

 

.__.
U m
. ...
bz.

 

 

 

I. a

‘ A:
r;
Pl

\l‘v

y.) Ll) ~41

Table

3-5

3-6

3-7

3-8

3-9

3-10

3-11

3-12

5-1

5-2

5-3

5-5

vi
Description

Reordered Cross—Lagged Correlation Matrix for the First
and Second Administrations—-Seven Scale Research Model
Replication Study—-Firm B Managers . . . . . . . . .

Reordered Cross-Lagged Correlation Matrix for the First
and Second Administrations-~Revised Seven Scale Research
Model--Firm B Managers . . . . . . . . . . .

Reordered Dynamic Correlation Matrix for the First and
Second Administrations--Seven Scale Research Model—-Firm
A Managers . . . . .

Reordered Dynamic Correlation Matrix for the First and
Second Administrations--Seven Scale Research Model
Replication Study--Firm B Managers . . .

Reordered Dynamic Correlation Matrix for the First and
Second Administrations--Revised Seven Scale Research
Mode1--Firm B Managers . . . . . . .

Reordered Impact Correlation Matrix for the First and
Second Administrations—-Seven Scale Research Model--
Firm A Managers. . . . . . . . . . . . . . . . .

Reordered Impact Correlation Matrix for the First and
Second Administrations-—Seven Scale Research Model
Replication Study--Firm B Managers . . . . . . .

Reordered Impact Correlation Matrix for the First and
Second Administrations-~Revised Seven Scale Research
Mode1--Firm B Managers . . . . . . . . . . . . . . .

Means and Standard Deviations for the First, Second, and
Third Administrations--Seven Scale Research Model—-
Firm A Managers. . . . . . . . . . . . . . . . .

Correlation Matrices for the First, Second, and Third
Administrations-~Seven Scale Research Model--Firm A
Managers . . . . . . . . . . . . . . . . . . .

Matrix of Averaged Static and CrossTLagged Correlations
for the General Factor During the First, Second, an
Third Administrations--Firm A Managers . -

Means and Standard Deviations for the General Factor and
Changes in the General Factor During the FlrSt, 539°“ 2
and Third Administrations--Firm A. ' '

General Factor Correlation Matrix for First, Second, and
Third Administrations--Firm A. - - - -

 

61

62

96

98

104

106

108

 

vii
Table Description Page
(Ll Fourteen Scale Model . . . . . .'. . . . . . . . . . . . . . 131

(%2 Correlation Coefficients Between Fourteen Scales for Firm
A First and Second Administrations and Firm B First and
Second Administrations . . . . . . . . . . . . . . . . . . . 133

C~3 Significant Differences Between Correlations for Firm A
and Firm B . . . . . .. . . . . . . . . . . . . . . . . . . . 134

(L4 Comparison Between Firm A and Firm B Reliabilities of
Fourteen Scales. . . . . . . . . . . . . . . . . . . . . . . 135

C-5 Scale Means and Standard Deviations of Fourteen Scale
Model for Firm B Managers at First and Second
Administrations of the Questionnaire . . . . . . . . . . . . 137

C-6 Seven Scale Research Model . . . . . . . . . . . . . . . . . 138
C-7 Correlation Coefficients Between Seven Scales for Firm A
First and Second Administrations and Firm B First and

Second Administrations . . . . . . . . . . . . . . . . . . . 140

C-8 Matrix of Raw and Corrected Correlation Coefficients
Between Change Scores for Firm A . . . . . . . . . . . . . . 142

 

(F9 Matrix of Raw and Corrected Correlation Coefficients
Between Change Scores for Firm B . . . . . . . . . . . . . . 143

I
C~10 Matrix of Differences Between Dynamic (Corrected) . . 1
Correlations of Variables of Seven Scale Model for Firm A I

and Firm B . . . . . . . . . . . . . . . . . . . . . . . . . 144

C-ll Inter-Scale Correlation Coefficients Between First and
Second Administrations of the Questionnaire for Firm A . . . 14S

C-12 Inter—Scale Correlation Coefficients Between First and

Second Administrations of the Questionnaire for Firm B . 146 I
(*13 Matrix of Differences Between Cross-Lagged Panel 148
Correlations for Firm A. . . . . . . . . . .
(L14 Matrix of Differences Between Cross-Lagged Panel 149
Correlations for Firm B. . . . . . . . . -
C‘15 Matrix of Dynamic (Corrected) Correlation Coefficients 150
for Pooled Firm A and Firm B Sample. . . . . . . .
(E16 Matrix of Inter—Scale Correlation Coefficients Between
First and Second Administrations of the Questionnaire 151

for the Pooled Sample.

 

 

 

 

 

 

 

 

 

 

A
.- . -_
.l 1..
‘ ~ ..:
.. ...
_. ...
- .s:
.-1 \-
“ ~M
. H ‘
.- v
.. A

‘l .
-;-. 3.
‘.

l. .
~ .§ §

 

 

 

viii

Table Description

(#17' Matrix of Differences Between Cross-Lagged Panel

D-I
0-2
0-3
D-4
0-5
0-6
0-7
D-8
D-9

D—IO

D-20
D-Zl
D-22
D-23

D—24

Correlations for the Pooled Sample .

Scale 501, Objective Feedback.

Scale 502, Interest in M80 .

Scale 503, Goal Relevance.

Scale 504, Boss Concern With Failure .
Scale 505, Influence Upward.

Scale 506, Need for Policy .

Scale 507, Satisfaction with Boss.

Scale 508, Boss Supportiveness .
Scale 509, Orientation Toward MBO.
Scale 510, Influence on Goals.
Scale 511, Goal Difficulty .
Scale 512, JOb Satisfaction.
Scale 513, Goal Success.
Scale 514, Performance-Reward Association.
Scale 515, Importance of Competence.
Scale 516, Residual.
Scale 501, Importance of Goals
Scale 502, Goal Setting Behavior .
Scale 503, SUperior—SUbordinate Relationship .
Scale 504, Utility of MBO.
Scale 505, Importance of Competence.
Scale 506, JOb Satisfaction. . . . . . . .
Scale 507, Performance-Reward Association.

Scale 508, Residual. . . . . . . . . .

 

Page

. 153

. 160

160

. 162

. 162

. 163

163

. 164

. 166

166

. 168

. 168

169
169
170
172
172
173
173
175
175
177
177
178

178

 

 

 

 

 

u"
.1.

 

 

 

 

 

 

 

Table

E-l

E-3

E-4

E-S

F-l

F-2

ix

Description

Fifteen Scale Research Model . . . .

Means and Standard Deviations for Fifteen Scale Model.

Internal Scale Reliabilities and Change Score
Reliabilities for Fifteen Scale Model. .

Revised Seven Scale Research Model .

Matrix of Raw and Corrected Correlation Coefficients
Between Change Scores for Revised Model. . . . . .

Inter-Scale Correlation Coefficients Between First and
Second Administrations of the Questionnaire~-Firm B
"High Cool” Managers . . . . . . . . . . . . . . .

Inter-Scale Correlation Coefficients Between First and
Second Administrations of the Questionnaire——Firm B
"Low Cool" Managers. . . . . . . . . . . . . . . .

Matrix of Dynamic (Corrected) Correlation Coefficients
for Firm B "High Cool" Managers. . . . . . . . . .

Matrix of Dynamic (Corrected) Correlation Coefficients
for Firm B "Low Cool” Managers . . . . . . . .

Matrix of Differences Between Cross-Lagged Panel
Correlations for Firm B "High Cool" Managers .

Matrix of Differences Between Cross-Lagged Panel
Correlations for Firm B "Low Cool” Managers.

 

Page
181

182

183

185

186

. 191

192

193

194

196

197

 

 

 

 

 

91'

III

' Q

‘ s

'r’
o
,__4

. ~¢t

(/1

 

 

 

Figure

1-1

2-1

5-1

5-2

5—3

C-1

C-2

C-3

E-l

F—l

F—2

LIST OF FIGURES

Description

Effects Diagram for Change Relationships in the M80
Behavioral System-~Firm A. . .

EffeCts Diagram for Change Relationships in the M80
Behavioral System--Firm B. . . .

Scatterplot for Time l,Time 3 versus Time l,Time 2
Cross-Lagged Correlations for the MEG Variables which
Estimate the General Factor-~Firm A Managers .

Scatterplot for Time l,Time 3 versus Time 2,Time 3
Cross-Lagged Correlations for the MBO Variables which
Estimate the General Factor—-Firm A Managers .

Scatterplot for Time l,Time 3 versus the Averaged

Time l,Time 2 and Time 2,Time 3 Off~Diagonal Cross-Legged

Correlations for the MBO Variables which Estimate the
General Factor--Firm A Managers. . . . . . . . . . .

Effects Diagram of Change Relationships in the MEG
Behavioral System—-Firm A. . . . . . . . . . . .

Effects Diagram of Change Relationships in the MBO
Behavioral System--Firm B. . . . . . . . . . . .

Effects Diagram of Change Relationships in the MBO
Behavioral System for Pooled Samples from Firm A and
Firm B . . . . . . . . . . . . . . . .

Revised Model of Change Relationships in the MBO
Behavioral System. . . . . . . . . . . . . . .

Effects Diagram of Change Relationships in the M80
Behavioral System for Firm B--"High Cool” Managers .

Effects Diagram of Change Relationships in the MBO
Behavioral System for Firm B—-"Low Cool” Managers.

 

Page

22

99

101

102

. 155

156

157

188

. 198

. 200

 

......

   

 

tit

 

 

 

 

CHAPTER 1
GENERAL RESEARCH PROBLEM, REVIEW OF THE LITERATURE,
RESEARCH STRATEGY, AND PRELIMINARY FINDINGS
General Research Problem

The study of managerial behavior uses a wide range of research
techniques ranging from simple observation to SOphisticated experimental
methods. Recently, there has been increased interest in longitudinal
studies which facilitate the inference of causality (Lawler, 1968;
Kavanaugh, 1971; Bentz, 1971). The research reported here is one phase
of a longitudinal causation - field study project designed to increase
the understanding and generalizability of the attitude change relation-
ships found in a Management By Objectives (MBO) System used in a large
organization.

M80 is a results-oriented subsystem directed at planning and
controlling the organization. The activities in the system carried out
by the superior and subordinate are (l) the interaction of superior and
subordinate to mutually establish and communicate organizational per-
formance and deve10pment goals; (2) the periodic review of the goals
and assessment of goal attainment; and (3) final review of the sub-
ordinate's performance using the established goals as a criteria for
evaluation.

The general research problem addressed by this project is an
attempt to answer the question: "Are the changes in managerial

attitudes observed in organizations utilizing an MBO system the result.

 

 

 

 

 

 

 

o
v ..
0- ‘; ~

I. "N l"

»o;O--.'; no.

.u-‘o‘VV' W

so.....,o.

\ ‘ .
0‘09”...-

v
..-..,..o. a
‘a \ -
"‘¢-.ovo.~

—,
I
..
~. _'
I‘ s:
‘H 0.5
-..

.ue‘v.‘

a
4...,

1 I;
M.)

('1'!

n.

 

 

 

2
of real or apparent changes in attitude?" This research effort Specifi-
cally seeks to describe empirically and account mathematically for the
attitude change processes (e'.g., changes in. goal clarification, superior-
subordinate interaction, performance-reward association, and job
satisfaction) found in two large industrial organizations which utilize

a Management By Objectives system.

Review of the Literature
This research seeks to extend the theoretical understanding of
M80. The theory underlying MBO has been drawn from related research into
goals and motivation (Locke and Bryan, 1966); participation in decision
making (Vroom, 1965; Likert, 1961); and organizational climate and
motivation (Litwin and Stringer, 1969). ;
Research into the effectiveness of M80 and the Goal Setting
Process has been performed by several groups of researchers. The first
study, a longitudinal one, was performed by the General Electric
research team of Meyer, Kay and French (Meyer, Kay and French, 1964).
The GE program, called Work Planning and Review, was an alternative to
the traditional performance appraisal system. They concluded that
those managers who had participated in the deve10pment of goals for
their job had a more favorable attitude toward the content and challenge
of those jobs.
A second longitudinal study was performed by Raia in Purex
Corporation, Ltd. (Raia, 1965, 1966). Immediately after the firm had
instituted an MBO program, there was an increase in productivity and a
greater awareness of organizational goals by the managers. However, in

a follow-up study, Raia found a decline in the effectiveness of MBO as

 

 

3
a motivator for job performance. There was a fourteen-month time Span
between the 1965 study and the 1966 study.

Mendleson studied managerial goal setting of superior—subordinate
pairs in eight different companies (Mendleson, 1967). This research
assessed the relationships between goal setting activity and (1) the
degree of job understanding between superior and subordinate, and (2)
ratings of subordinate effectiveness (present effectiveness and poten-
tial for promotion). Mendleson found a positive relationship between
the extent of goal setting activity and the superior‘s rating of his
subordinate's promotability.

Ivancevich, Donnelly, and Lyon examined the relationship between

satisfaction and different methods of implementing MBO (Ivancevich, et al.,

1970). In one company the personnel manager conducted the implementa-
tion of MBO. In the other, top-level executives handled the MBO
implementation. The major finding was that when MBO was implemented and
administered by the company executives, there was a significantly higher
level of perceived need satisfaction of the managers.

Research by Tosi and Carroll (1968, 1970) has centered on the
investigation of relationships between process and end—result variables.
The Process variables are considered fundamental to MBO and serve as
independent variables in their research model. They include character—
iStics of goals, feedback characteristics, and the nature of the

- . . . - n as the
suPeI‘lor-subordinate relationship. End-result varlables are see

dependent variables of the behavioral system. They include level 0f
goal achievement, effort expended, level of the goals set, and satisfac-

. . ‘ _ l
tlon With the MBO program. The results of the T051 and Carroll rescarci

- . - , rocess
Indicate that establishing clear and important goals (1.e , p

 

 

........ .

 

.ln-h.nk v.

 

4

variables) produced favorable results (i.e., favorable responses to the

muhqesult variables) especially for certain personality types.

The Chesser Study

Utilizing an instrument designed by Tosi and Carroll for a manu-

facunfing company (Firm A), data related to the Management By Objectives

(MBO) process were collected at two points in time, eighteen months

smart.

The forty-seven item, Likert-type questionnaire contained items

Unreasure the following: (See Appendix A.)

1. Goal Characteristics

2. Feedback Characteristics

3. Boss-subordinate Relationships

4. Job Characteristics

5. Orientation Toward Management By Objectives

The instrument was cluster analyzed and yielded seven (7) sub-

scales (see Appendix B). These were:

1.

SuPerior-SUbordinate Relationship——This variable represents a
Complex description of the overall relationship between the
SUperior and the subordinate. It incorporates three aSpects

of the relationship: (1) the frequency of interaction in goal-
oriented activity; (2) the subordinate's perception of the
usefulness of the interaction to job performance; and (3) the
subordinate's evaluation of his superior as a superior.

Goal Clarity and Relevance—-Represents a measure of the degreed
to whiCh the goals set reflect organizational and personal use 5.
It also reflects whether or not the goals were clearly state

and priorities established.

Orientation Toward MBO—-Is a measure of th? PerceivedIEtSSEZZSS:
the MBO program as experienced by the part1c1pants.. etin
the degree to which MBO is viewed as being helpful 1“ me g

job requirements.

. which
Performance-Reward Association-—Assesses the degr:?ohoare viewed
rewards in the form of salary increases and promszrmance
as being based on an evaluation of actual Job per

 

 

4 A15
. .. -I at - 0
s h i Q. A I4b | In If: I
h . a a BL ~..u mum A1,. OI» A
nub .- \ Q a all“ .0 I
- n-Ms s i k.\ n b “In C.» . It. a u‘ C»
I . . I ..I I
‘ o: u u”. .- Q. h ..K n.4- n ”a Ill IF cit Ht VI 1 h\ t t I 1
a . -\. .- .\|I PI ‘1‘ (I a: AV ~II\ shy R bi Hui

-. ..a -

 

 

 

 

5

5. Subordinate's Influence Over Goals--Measures the relative
influence shared by the subordinate and his superior in the goal

setting process .

6. Satisfaction With Job--Measures the subordinate's satisfaction
with the job situation.

7. Perceived Success-~Assesses the perceived relationship between
performance and established goals.

Chesser focused the research design directly on the dynamic rela-
tionships in the model (i.e., the behavior over time of the variables)
and the inference of causality. Using dynamic correlations and cross-
l'agged panel correlations, a causal set of change relationships in the
system was deve10ped (see Figure l-l). Chesser concluded that there
were changes in the variables over time which were statistically related
and that by analyzing the relationships across time, changes in certain
variables (e.g., changes in goal clarity and relevance, changes in
orientation toward MBO) were concluded to be causally related to changes
in other variables (e.g., changes in superior-subordinate relationship,
satisfaction with job) (see Chesser, pp. 105-110).

A close examination of the Chesser data which support his con-
clusions reveals seVeral reasons for concern. First, for the two
administrations of the questionnaire, the means and variances for each
0f the seven scales showed little change between administrations (see
Chesser, Table 2-8, p. 55). If there was, in fact, real change in the
system variables between administrations there would usually be a corres-
pending increase in the variance of the scales.

A second concern is found in the analysis of the 7 x 7 initial
score - change score correlation matrix for the Chesser data. Seventy
percent of these correlations were negative. The negatives along the

diagonal could be attributed- to regression toward the mean. This

 

 

 

Chesser, 1971, page 106)

MBO Behavioral System ~ Firm A

Figure 1-1.--'Effects Diagram for Change Relationships in the
(Reference:

 

 

 

new
spa; nonpowumaeem.

(I

 

.oma Seneca
noapmpnmflho,

 

SH newnono

 

 

A.

 

 

 

 

nowpmfloommd
enormmloodoanomnom

Ga momnmgo

 

QH oomnmmU

 

 

 

 

as

 

 

A.

 

 

 

_iAl

 

 

 

anamQOAPeHsm
opmnfiomopdminoflnmmam
qfi momqono

__. __a..a_. ... _a. __. .EI‘T

 

 

I

 

 

eds Apaaoao deco
Qfl momemflo

 

ooddboaom

 

“-

 

 

 

manom manganppa an
mmmoodw wobfloonom
Q...” moonmflo

 

 

 

 

 

 

no. '9‘ '

k" ‘ l
fito.
-

09.... ..- :-

’,1 -~

u..... I.
.

. . .
III. 90‘.
" l
“fig...
I

.
In 0 .0-
n.

a
'l-glo‘.‘

l- . ‘.'
“0h.\4

p
u
“o\.;

 

7

regression toward the mean implied that managers who rated high on a
variable in the first administration tended to decrease in score on the
same and other variables during the second administration, while
managers who initially scored low on a variable tended to increase. The
negatives down the diagonal of this matrix could also be due to unre-
liability. However, the off-diagonals (the initial score on a variable
correlated with changes in another system variable) would not be
Spuriously negative, since errors of measurement or unreliability are
Imcorrelated between measures.

These problems in the Chesser analysis (constant scale means and
variance across administrations, negative initial score - change score

correlations, and low change score reliabilities) force consideration of

 

a very interesting question: "Were the changes in the scales due to
real change or were they only 'apparent' changes?" There are two alter-
native answers to the question.

First, it can be assumed that there is real change in managerial
attitudes within the MBO system. The data will support this if it is
assumed that the absence of change in the scale variance is due to real
regression toward the mean (as shown by the negative initial score -
change score correlations) and that the real change is in part due to , ,
random factors.

The second explanation of the data is that there is no change in
the true scores of managers for the scales between administrations of
the questionnaire. Unreliable measures and their spurious effects could
be the cause of the negative diagonal correlations (the regression
toward the mean effect) in the initial score — change score correlation

matrix, but further assumptions would be required to account for the

 

 

 

 

 

 

 

 

 

“u.
«:95

“i,“
rm)

:3}
A. {c

..
ls.

A

m

in
yr

m

‘1:
i

8
negative off-diagonals. Such a model will be presented in Chapter
Three. This No Change Hypothesis assumes that the managers demonstrate
very stable di5positions toward their work and their relationships with
their boss. It is the pursuit of an answer to this question of real
versus apparent change that forms the basis of the research reported in

this dissertation.

Research Strategy and Preliminary FindirLg_s_

Using the Chesser study of change relationships in the MBO
system as a starting point, this research focuses upon a test of the
general hypothesis that these change relationships are the result of real
Changes in managerial attitUdes. To test this hypothesis, the Chesser
study has been replicated. An expanded version of the questionnaire
developed by Tosi and Carroll (see Appendix A) was administered to the
managers of a second large manufacturing organization (Firm B) at two
points in time, eighteen months apart.

The methodology used by Chesser to develop the effects diagrams
and to infer causal relationships was also applied to the Firm B data.
The results of this replication (see Chapter Two) are such that it is
not possible in the two time period study to conclude decisively that
the changes in managerial attitudes are real, and not apparent, changes.

To improve the scales of the research model, the Firm B data
collected with the revised instrument was cluster analyzed. Using the
same methodology as in the replication, the analysis of this data also
produced inconclusive results. Extensive reporting of the findings from
the replication study and the study using the revised scales is found in

Appendices C, D, E, and F.

 

 

 

 

 

 

 

 

; I
gnu.

- -
0p. "

I? . ‘i‘

“a".

.oé Me '
,:.. .L: o '

.
.
._.., ”p
‘ u
5-... to:
o

D... 0‘.
Wk -
Pb. Iii

. . ,

-. .-.

I' ul"
~ ”I...

l

| 9

I The analysis of the data from both organizations presented in
Chapter Two produced equivocal evidence of real change over the two time
periods. Several models which assume real change are considered in
Chapter Three. Models developed to explain the data in terms of the
phenomena of the hypothesis of no change are presented in Chapter Four.
One model in each set (real change and no change) was found to fit the

data for two time periods. However, a direct empirical test of real

 

 

change requires data from another administration of the questionnaire
(time 3). This data was made available for Firm A and the real change

hypothesis was further evaluated. These results appear in Chapter Five.

Summafl

The research reported here focuses upon the assessment of real

 

or apparent change in the attitudes of managers participating in the MBO _ ‘/
programs of two large organizations. Using the same methodology as

Chesser (1971), the data from these managers have suggested that the

changes in attitudes from the first to the second administration of the

questionnaire are not real changes, but only apparent. The results of

the replication study and the revision of the research model are pre—

sented in the next chapter.

 

 

 

'
'~ to.

-\ \aau-

.
V~O~
a -T
bah-I.
.

.
.
vnh.\.

.0
w-“r _

waking

. l
“cab ‘
“i=1“

:uv.5d

Aim

’2‘
31

r4 _
(J:
9.)

mg:
In

all

r'r’

CHAPTER 2
THE REPLICATION OF THE CHESSER STUDY
’ AND A REVISED RESEARCH MODEL

This chapter presents the major results of the replication of
the Chesser study, using data from Firm B. The first section of the
dunner describes that organization and the sample of managers who
participated in the MBO study. The next section presents the major
fhufings which are used to test the general research hypothesis. The
third section describes the development of a revised research model and
ruesents additional results for the test of the general research

hypothesis.

The Replication of the Chesser Study

The Tosi and Carroll questionnaire with some additional items
was administered to a group of Firm B managers (see Appendix A). Firm B
is a container manufacturing division of a large conglomerate and has an
cmgoing MBO program. The first administration of the questionnaire in
Mardh, 1970, produced 600 completed.questionnaires. From the second
zuhfinistration in August, 1971, S48 usable questionnaires were received.
The 117 managers who identified themselves in both administrations con-
stitute the sample for Firm B that will be used in the present study.
Table 2-1 and Table 2—2 show the classification of the Firm B managers
by organization level and functional area. Table 2-3 and Table 2—4
present similar statistics for the Firm A managers who participated in

both administrations.

10

 

 

 

 

.__.—-
-.. ,x V
233..

 

 

 

IQI/

2m
:lm
V

”int
I?“

hit

11

Table 2-1.-- Classification of Firm B Managers by Functional Area

 

 

 

Functional Area Number Percent
General Management 19 16%
Production 33 2 8
Sales 39 33
Staff 26 23

Total 117 100%

 

Table 2-2.-- Classification of Firm B Managers by Organizational Level

 

 

 

 

Organizational Level Number Percent
Senior Officer 1 1%
Major Area or Activity Manager 11 9
Plant or Special Staff Manager 19 16
Department Manager 29 25 /
Section Manager or Foreman 57 49

Total 117 100%

 

 

 

 

 

-.__4_w_——— _ 7

Dim

Vidd?

125's

 

11

Table 2-3.-- Classification of Firm A Managers by Functional Area

 

 

Functional Area Number Percent
Marketing 30 41%
5
Research and Development 20 27
Production 13 18
Administration 10 14
Total 73 100%

Table 2-4.-- Classification of Firm A Managers by Organizational Level

 

Organizational Level

u

 

Vice President
Directors
Middle Management

Lower Management

 

 

Total

Number Percent
3 4%
7 10
38 52
25 34
73 100%

 

 

 

 

12

The methodology utilized in the replication was the same as that
employed by'Chesser. The responses by the Firm B managers were scored
Lsing the seven scale research model developed for the Firm A managers.
Then, the Changes in the scales (second administration scale scores
ndnus first administration scale scores) were calculated and "dynamic
correlations" between the change scores for the Firm B sample were
cmtained. These dynamic correlations are Pearson Product-Moment corre-
lations for the changes in the variables (Vroom, 1966). The cross—
lagged panel correlation technique (Pelz and Andrews, 1964) was used to

infer the causal priority of the variables in the MBO system. These

 

results were then used to deve10p an effects diagram for the Firm B
nmnagers. The detailed deve10pment of the effects diagram for the Firm

B managers is contained in Appendix C. /

Findings
The means and standard deviations for the seven scales of
ChesSer's model are presented in Table Z-S. There is some change in the
scale means from the first administration to the second administration;
however, the standard deviations do not change significantly. This
relatively constant variance (the standard deviation squared) suggests

that the change calculated in the analysis of the data was not real

 

Change,tnuzspurious. Chesser also found no change in means or
‘wnfiances for Firm A (See Chesser, Table 2-8, p. 55).

The reliability of the change scores is very important in con-
sidering either the dynamic correlations or the initial score-change
score correlations. Table 2-6 shows both the internal scale reliabili—

ties and the Change score reliabilities for Firm A and Firm B subjects.

 

 

 

¢_ :f- ...- .\..--

t
r-.- duel-a..va vrrs..- .-.v nu... -.~.~ .at ... a..._v< _u.-nv.v.vmu. --- .nn.. —.- vhav-u Ann-Ayn: a: .-\u.uA- .-11 .

 

.41

 

 

 

13

 

 

 

 

 

 

mm. R. mm. om. o.m H.m H.m o.m mmooo:m wo>floopom .5
mm. mm. ow. Hm. m.m m.m N.m H.m DOW £94.“: Cowpommwflpww .0
om. om. mm. H.H w.m o.N o.m w.m mamoo ho>o
oesozamcH m.opm:flvuoasw .m
co. we. an. on. w.m m.m n.m w.m :oflpmfioomm<
pamzomuoocwshowhom .v
mm. mm. mm. ow. m.m n.m v.m o.m om: wawzoh :owpmpcowho .m
mm. NV. we. we. o.m o.m o.m w.m oocm>oaom mam prhmau Hmou .N
mm. mu. mm. mm. m.m m.m N.m H.m magmaowpmfiom
oumnfiwaondmuhoflwomsm .H
mm Hm N< H< mm Hm N< H< moawom

 

mcofiumfl>oo pudendum

 

mama:

 

m SHAW @:m < SHAME .HOM

ohfiwccoflumosd mo COfipMHpmficflEw< vacuum paw umhflm How mcofiumfi>oo pumwcmHm mam mane: u|.mum oanwb

 

 

 

 

 

 

14

Table 2-6.-- Reliabilities of Seven Scales

 

 

 

 

Internal Change Score

Reliabilitya Reliabilityc
Scale rxx rx_x

Firm Ab Firm B Firm A Firm B
1. Superior-Subordinate Relationship .96 .94 .94 .90
2. Goal Clarity and Relevance .90 .80 .87 .60
3. Orientation Toward MBO .80 .84 .50 .50
4. Performance—Reward Association .84 .70 .68 .25
5. Subordinate's Influence Over Goals .75 .72 .50 .50 ,/
6. Satisfaction With Job .58 .59 .35 .13
7. Success in Attaining Goals .65 .57 .30 .13

 

aCoefficient alpha scale reliabilities.
bChesser, page 47.

cSee formula on page 15.

 

 

 

 

 

 

V

fl; v».-
_,_, v.6»!
1 ,'.Z'
. 1
ID.-.

1"

I.b I

 

15
The change score reliability is sensitive to changes in the internal
reliability as well as the test—retest correlation of the scales.

Change score reliabilities are calculated using this formula:

r11 * r22 ' r12

 

= 2
rdd 1
‘ r12
where:
r = reliability of change score

r11 = internal reliability of scale at Time 1

r22 = internal reliability of scale at Time 2

r = correlation of the scale between Time 1 and Time 2

(McNemar, p. 157)

(fimen that the internal reliabilities for the scales are adequate
(approximately .60 or better), the test-retest scale correlation is the
key variable in the calculation of the change score reliability. If
there are no sabstantial changes in the scales between the administra-
timug then the correlation between the scales will approach the internal
scale reliability as a limit. That is, a.high test-retest correlation
nmans that real change on that variable is negligible. In the replica-
thng three of the seven scales have a change score reliability of .25
(u'less. These same three scales also have the lowest internal relia-
lfllities (eaCh of these scales has four or less items). One objective
mfthe revised research model presented below is to increase the number

mfitems in each scale and thereby increase the reliabilities.

 

 

 

 

 

. t . ;
'( ..I.l :
M .5 h

.
to» -n.
O J
8-0. u‘:
.

. .
‘0... .

neg...

'uc'

 

16
Correlational Data for Replication Study
The correlational data which resulted from the replication study
is found in Table 2-7, page 17. The legend at the lower right corner of

that page provides a guide for the location of relevant matrices.

Static Corre lations

 

There are two matrices in Table 2-7 which contain "static corre-

 

lations," i.e., the Pearson Product-Moment correlations between the seven
scales of the model at a single point in time. One matrix is the set of
correlations between the variables at the time of the first administra-
tion ("time 1 static correlations"). The second matrix is the set of
correlations between the variables at the time of the second administra-

tion ("time 2 static correlations"). These matrices may be compared in

 

order to note the similarities of the correlations (magnitudes and
signs) at each administration, i.e., the stability of the static corre-
lations across time. As expected in this data, the pattern of static
correlations was very similar for the Firm B managers during both
administrations .

There is also important information contained within each matrix
individually. These matrices show remarkable consistency between
variables in both time periods. For example, Job Satisfaction and the
SUperior—Subordinate Relationship are positively and significantly
correlated (time 1, r-_-.37, p < .01 and time 2, r=.36, p < .01). The
correlations shown in the static correlation matrices are necessarily
r ). That is, the static correlation is used to

1X6= x6xl
Specify the degree of relationship between two variables. The static

Symmetri cal (rX

correlation would be exactly the same when either of those two variables

fihflfi'ahU “thunk-huh!!!
uvmnhduu nfihV iUFNnN F a“ Iu-| u \
till ||\. -NI OVN‘fll‘IP

an pupal d A. :- avuh d 'VIVAUIL navVl- I! rhfittfiw- 9.9 I ah ha." nape) at“... cv-n y mi 1 “NH run: quit I. W pry-Li Harv «hall fiVrlvHrba
'0. an. nuns-I< Iva-II i104” I Q I yin-n. iiV filti.lAiHVal. IIRb. IIVIN bml IIIQIIIII< ulnhll. ‘NI lib-HE I‘NVI‘ IIUTIV .‘II Elhhl‘ AHPV‘NIV~IIVI~LLVATVI\V

Fifty-Audit. IIII‘ O..-

 

 

17

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

oHEmcmn uommaH
uHumum vowwmA-umouo «N. u Hm>oH Ho.
uommEH vmwmmA-mnouo oHumum mg. n Hm>oa no.
Hozmoma up we msHm> ucNUHMHame
ocH mH- HH vH N- m- H- Hm H oH HH NH H- sH He- «H v H- NA e NH HN
mH- oaH v vN NH oH mm NH- we H- o N- H n n cm- m- eH- mH- NH- He- 6N
HH c oH va- H- RN mm m m XNm o mu N- N- a- H- Mn- a- ,m- -nwu1mnn:11aa:
eH «N VH OOH 0N HH Hm oH mN n we m A HN n H H- Hq- HH- n- NH- NH
N- NH H- .oN ocH eH NH oH oH s- HH «e m HH NH N- N- H- mm- 3H- N- NH
m- 0H NN HH VH ccH Ne :H NN «am «H1 «H av-Ime- mH- H m- M. m nm- aw- nan
H- A» nN He NH oe 30H 0 SN .H DH H NN e. 3H m- NH- .H. H. \N- Hs- mH
Hu N.. n OH CH CH w osH o H- HN Hm «N Na Hm mH e- A HN H. mm ea.
H me m mm- CH NN, an Xe an m mv.ldN- NH .HW-Afi: Nm - N- NH a- -n NH
OH H- mm m o- mN VH H- e ocH HH QH- m H oH- m ma m a. sN- mH- NH
HH 6 a we HH OH oH HN me HH 09H on on mm DH an H OH NH .HH Hm HH
0. m- m- m av «H H Hm Hm» GH- 0&1 mad Nm- me NH Nn- m- 63 he an a. .IIdH:
H- H N- H m 0. MN VN HH n on He OHH om ON OH A He om av NN m
0H a N- HN HH AH ea n» on H am we or QHH 3N m e m. OH 4H A. N
N.- m N- m - NH AHT-aH (Hm c aH-xaH NH- 0N. aw amH N an a m a A slai-
oH om- H H N- n N- NH Nm m um NN oH Nm N 09H v mm mm HH an o
v m- Hm- NH- N- m- NH- a- N- an H m- N V n- a c-H NH N NH SH m
.- aH- e- He- H- H -aH- N [NN a .Na- am Hm mm min mm NH 34H NH NN we a
No NH- m- HH- mn- m o- ¢N NH v- N» as on em m an N we oHH NH H? m
e mH- mN- mu OH- nm- mm: ”H o- om- NH vN cc Hm a HH NH NN mm 29H «m m
NH HN- mN- cH- N- vm- Ha- pm- a mH- nmi mm- Na ha a NM hH mm. H- am H.H H
Ho an 3 A; NH A; .3 3- .HH NH HT NH .3 1m A a A. .- m N H
H 05.2. .- N 05.3. . N ofiwb H GEM-H

 

thHucu whommcwz

 

 

m Ehwm How mono: condoned onow Go>om one cw Aw oswh - H oswhu mohoow omcnau paw-.AN oEMBV

mmhoom cowumhumwcwfiv< caboom «AH ofiwhv mohoom CONHNHpmeHEv< umuwm ozu Mom moowhumz :owumaohhou -.m-N ofinwh

 

 

 

 

 

'"

no .
"‘il
In.

a

\; 1:.-

..u y.

o
‘9 Io-

“. nu

‘0...-

5.3.1:
5

u
N

V3“
"531-

 

17
is used as the predictor and the other the predicted variable. The net
result of this is that the off-diagonals of the static correlation

matrix are "reflections" of each other, i.e., symmetrical.

Dynamic Correlations

 

The lower right-hand matrix found in Table 2-7 contains the
"dynamic correlations" or the Pearson Product-Moment correlations
between change scores in the system variables. These change scores are
calculated by subtracting the time 1 score from the time 2 score (i.e.,
change score = time 2 score — time 1 score). As in the static correla-
tion matrices, the dynamic correlation matrices, the dynamic correlations
are necessarily symmetric about the diagonal.

The dynamic correlations are used in this study to determine the
significant relationships between changes in the system variables.

Using the same variables as before--Job Satisfaction and Superior-
Subordinate Relationship-~it can be seen that there is a significant
dynamic correlation (r = .35, p < .01) between Changes in Job Satisfac-
tion (variable 20 in the table) and Changes in Superior-Subordinate
Relationship (variable 15 in the table). This dynamic correlation
Suggests that Changes in Job Satisfaction are related to Changes in

Superior-Subordin ate Re lationship .

grgss-Lagged Correlations
There are two sections of Table 2—7 which contain the "cross-
lagged" correlations. These cross-lagged correlations are Pearson

Product-Moment correlations between the system variables at time 1 and

time 2.

 

 

 

 

18

The diagonals of this matrix are the test-retest correlations
for the MBO variables. If there were no real change, then this correla-
tion would be an estimate of the reliability for each of the variables.
It is the test-retest correlation which is used in the calculation of
the change score reliabilities discussed previously.

The off—diagonal correlations of this matrix are not necessarily
symmetric. That is, if x1 is the score on x at time 1, x:2 is the score
on x at time 2, and yl and y2 are the two measurements on y, then the
two matched correlations in this matrix are the two correlations involv-

ing x and y, i.e., Since the two correlations are

d .
rlez an rylxz
calculated for mathematically different variables, it would be possible
for the two correlations to be completely different. For example, let
Job Satisfaction be x and let Superior-Subordinate Relationship be y.

Then x is variable 6, x is variable 13, y1 is variable 1, and y2 is

l 2
variable 8. The two cross-lag correlations are thus rxlyz = 16,8 = .32
and r = r = .06. And indeed they are not equal but are

YIXZ 1,113

"asymmetri cal . "

As a matter of fact, it is the asymmetry of these correlations
that facilitates the inference of causal priority in the system. To
illustrate Chess-er's test for causality, consider again the two
variables, Job Satisfaction and Superior-Subordinate Relationship.

Job Satisfaction and Superior-Subordinate Relationship have significant

static correlations (rxl),l = r6 1 = .37, = r13 8 = ~36) and a

rx2Y2

Significant dynamic correlation (rAx A = r20 15 = .35). Thu-‘3 “0‘5 0T11Y

Y
is there a relation between the variables at one point in time, but a
Change in one tends to be accompanied by a change in the other. There-

fore, Chesser (following Vroom, 1966) concluded that these variables are

 

 

 

 

 

 

II
“no:
A; ‘ui

u.
5.. O

“O. F

33.51":

N

u‘,‘';

"l.\

H. y

n“...

‘-i§:
?
I
‘l
'1

 

19
causally related. Assume (for the moment) that this is so. What is the
direction of the causality? Suppose that Job Satisfaction exerts a
causal influence on Superior—Subordinate Relationship but not vice versa
(i.e., an arrow from x to y in the effects diagram). Then if x1 is large,

y will tend to increase while if x is small, y will tend to decrease.

1

This tends to create a "considerable" correlation between x1 and y2.

However, if SUperior-Subordinate Relationship exerts no causal influence
m1Jbb Satisfaction, then there will be no analagous influence on the

correlation between y1 and.x. Thus the assumption "x influences y and

2.

not vice versa" leads to the inference r i.e., asymmetric

I -
lez > ylxz
cross—lag correlations (Pelz and Andrews, 1964). In the present example

I = .32 h'l = .06. Th‘ d'ff ' ' 'f' t t th
XIYZ w i e rylxz is i erence is signi ican a e

.10 level and hence provides rather weak evidence for an asymmetric
caus a1 re lationship .

Chesser's synthesis of Vroom (1966) and Pelz and Andrews (1964)
cmrrumrbe succinctly stated: If x and y have significant static corre-

thons (r ) and a significant dynamic correlation (r
2

X1Y1’ rXzy Ax,Ay)’

then infer them to be causally related. If the cross-lag correlations

are asymmetric, then either r If

1' Off >I‘ .
x1y2 > Y1X2 Y1X2 X1Y2

r ' H .H I
lez > nylxz, then infer x causes y f rylxz > rxly2,

"y causes x.” If the cross—lag correlations are symmetrical, i.e., if

then infer

Igly2 = rylxz, then infer mutual causation. For small samples these

inequalities could be replaced by significance tests in the usual way.
This method010gy will be shown to be wanting on both empirical

grounds and theoretically in succeeding chapters.. On the other hand,

the cross-lag correlations are important in their own right. Large

cross-lag correlations mean that the value of one variable at one time

 

 

 

_—

til 1
113‘

W-

.MT.
.0

Spu:

30)

ii}

 

 

 

 

20
can predict the value of the other variable at a later time. Thus
significant cross-lag correlations imply the existence of a lasting bond
between the variables and rule out many of the Spurious static relation-

ships that can arise from response sets, demand characteristics, etc.

Impact Corre lations

 

The only remaining sections of Table 2-7 which contain informa-
tion relevant to this study are "impact correlation" matrices. The
impact correlations are Pearson ,r's, the initial or time 1 score corre-
lated with change scores for the system variables.

The diagonals of this matrix contain the correlations of a
system variable with changes in that same variable. Since the diagonals
of this matrix are negative, two possibilities suggest themselves: one,
the.true regression of change score on initial score and, two, the
spurious negative effect of unreliability. Because of the low change
score reliabilities. and because of problems with transient factors that
will be explained below, the Spurious negative effect is known to be
large but cannot be estimated and corrected for.

The off-diagonals do not contain the spurious component of
unreliability since errors of measurement are uncorrelated. The high
percentage of negative off-diagonals contradicts many of the conclusions
represented by the effects diagram. While the dynamic correlation for
Changes in Job Satisfaction and Changes in Superior-Subordinate
Relationship indicates a positive relationship (rxAy = .35, p < .01),
the impact correlations for those same variables are : erAY = -.31

and rX6Ay = -.08. Both of these indicate that initial score and change

 

 

 

 

21
score are inversely related rather than directly related as implied by

a positive causal relation. This contradiction will be addressed below.

Effects Diagrams

The analysis of the data for the Firm B managers produced a
number of significant dynamic correlations. These dynamic correlations
provide the major support for acceptance of the Real Change Hypothesis.
These correlations between change scores are, for the most part, positive
and significant. This means that changes in one of the variables
explain a major portion of the variance in changes of the other. Used
in conjunction with the cross-lagged correlations, these statistical
relationships for the Firm B managers have been interpreted in the form
of an effects diagram of the change relationships between variables of
the MBO system (Figure 2-1).

This effects diagram was developed in the same manner as Chesser
(Chesser, p. 105). When there was a significant dynamic correlation, the
two variables in the relationship are connected by a straight line. When
the cross-lagged correlations are symmetrical, the relationship is
defined as mutually reinforcing and is indicated by an arrowhead at both
ends of the connecting line. For an asymmetrical relationship, e.g.,
Changes in Job Satisfaction and Changes in Superior-Subordinate Rela~
tionship, the arrow is unidirectional and indicates the causal relation-
ship.

On balance, the data from the Firm B managers produced results
quite similar to those found in the study by Chesser. The effects
diagram for the replication study does show two significant differences

when compared to the Chesser effects diagram (see page 6). First, the

 

 

 

 

 

 

 

22

Figure 2-1.-- Effects Diagram of Change Relationships fer the

117)

MBO Behavioral System - Firm B Managers (n

 

 

‘i

 

 

now cHHz
:oHuomMmHHmm
:H mmmamnu

 

 

 

 

 

 

4‘

.coHHmHoomm<
ransom-oocNEHOMHom
:H momcmao

 

 

om: vamzoe
:OHHmpcoHHo
:H mowcmzo

 

 

 

 

Ha

 

 

 

 

 

{A

 

i

 

 

H 1.,

J

 

mflcmcoHHmHom
opmcHwHoasm-HOHHoQSm
:H mow:mno

 

 

 

 

mHmou mchHNHH<
CH mmooosm wo>Hmo
-Hom :H momqmgu

 

 

w

 

 

i

H

 

 

 

oocm>o~m¢

w HHHHeHu Haoo
cw momcmzo

 

 

 

 

 

23
relationship between Changes in Goal Clarity and Relevance and Superior-
Subordinate Relationship is reversed in the two diagrams. Second, the
"driver" or causal variable for the Firm B managers is Changes in Job
Satisfaction while for the Firm A managers it is Goal Clarity and Rele—
vance. One possible explanation for this is that for the Firm A
managers there was a change agent present between administrations of the
questionnaire. That change agent was the introduction of MBO as a
management policy. Since one purpose of MBO is to improve the subor-
dinate's exPectations, Goal Clarity may have taken on added importance.
For the Firm B managers, MBO was an on-going program prior to the first
administration of the questionnaire. The Firm B managers had already
been utilizing MBO and thus job satisfaction was a key attitude in that
firm.

Given similar results in both studies, a decision was made to
pool the samples. This was accomplished simply by averaging the corre-
lations between similar variables from both studies. With the larger
sample, there were a larger number of significant dynamic correlations;
however, the pattern of cross-lagged panel correlations did not change.

The significant asymmetrical relationships required for causal
inferences were not present. (See Appendix C for a more detailed dis— ,
cuSSion of the methods and results of the pooling technique.)

The fact that a statistically more reliable matrix showed pg
asymmetric relations causes some concern. After all, what this
strongly suggests is that all the asymmetries in the smaller samples for
Firm A and Firm B separately were the product of sampling error. But
this would imply that all the causal relations in the MBO system are

mutually reinforcing. This not only contradicts existing theories but

 

 

 

24
seems rather implausible on the face of it. These "contradictory"
results for the pooled sample cross—lag correlations tended to make the

negative impact correlations that muCh more salient.

Contradictions to the Real Change Hypothesis
To this point in the replication study, the development of these

effects diagrams was based on the assumption of real change in the

 

system. This section will summarize the various pieces of evidence
which are counter to the assumption of real change in a multivariate
system suCh as that depicted in the effects diagrams. The first piece

of evidence is found in the means of the seven scales. They did not

 

increase. This could happen only if the positive changes produced by
high values on the MBO system variables were exactly balanced by nega- /
tive Changes produced by the low values on the MBO system variables.
But this precise balance can happen only if all seven means are exactly
at the zero effects point of the system. Furthermore this must be true
of both firms! .Although this eXplanation cannot be absolutely ruled
out, it is so unlikely on a priori grounds as to be highly implausible.
The second finding which challenges the assumption of real
change in the data is the fact that the variance of each variable was

lmchanged from time 1 to time 2. Real positive causal influences

 

normally produce a sharp increase in variance. This is particularly
true if high scores are producing positive changes while low scores are
producing negative changes. That is, if the managers who score high at
time 1, score higher at time 2, and the managers who score low at time 1,
score lower at time 2, then the variance will necessarily increase. And

this is precisely what is implied by positive causal relations without a

 

 

 

 

 

 

 

-
“.-

fu-
H...

"m

n.-

 

 

*1

25
change in the means. In the data the variances did not increase; they
stayed the same. Therefore if there is real change, it can only be real
regression to the mean, i.e., a negative relation between initial score
and change score. Furthermore for the variance to stay exactly the same,
the decrease in variance produced by regression to the mean must be
exactly balanced by the increase in variance due to change produced by
factors outside the system. While such perfect balance is not impossible,
there are no sound a priori grounds for such a finding and it must there-
fore be viewed with some suspicion. Furthermore it should be noted that
this balance must be assumed for each of the seven variables and in both
Firm A and Firm B.

Table 2-7contains evidence which suggests that there was no real
change in the system between administrations of the questionnaire. If
real change in attitudes had taken place, the static correlations
between the system variables should have exhibited different patterns in
both time periods. This did not happen. An examination of the correla—
tions between the variables for both administrations reveals that they
are not statistically different (see Table C—7 in Appendix C).

The impact correlation matrix has almost all negative correla—
tion. All of the diagonals are negative which implies an observed
"regression to the mean" effect. A sizable portion of this observed
regression to the mean is definitely known to be a spurious artifact of
unreliability in the measuring instruments. And since certain other
sPurious effects cannot be estimated in the data for two time measure-
ments, it is possible to disregard the negative diagonals altogether;

i.e., the diagonals for perfect measures might have been positive.

 

 

 

actually
nously d
553 53:15
anal a:
:me 55
related

U16 Sta?

standar
sim wh
mien
mange
change
outcm

additi

W35 3
hpm

mode

EST
for

re:

r___—7

r__ _ ._

26

The negative off-diagonals are not as easily explained and are
actually a contradiction to the assumption of real change. As pre-
viously discussed, the positive dynamic correlation between Changes in
Job Satisfaction and Changes in Superior-Subordinate Relationship implies
a real and positively related change. The off—diagonal impact correla-
tions suggest that changes in either of these variables are negatively
related to the initial score on the other. This is not consistent since
the status, cross-lags, and dynamic correlations are all positive.

When coupled with the finding of no change in the means and
standard deviations for the scales, the evidence points toward a conclu—
sion which rejects the general hypothesis of MBO as a multivariate
system of causally related variables. Could it be that there is no real
change in the actual MBO variables? That is, could the evidence of
change be due to some artifact of measurement? This would be a drastic
outcome indeed and it seemed unwise to consider such hypotheses when the
additional items existed to improve the Chesser scales.

In order to provide more distinct results using the Firm B data,
an attempt was made to construct improved scales. Chesser's methodology
was again employed to build a revised research model to test the
hypothesis of real change. The results of this revision of the research

model are reported in the next section.

Results of Revised Research Model Development
In an effort to increase the internal scale reliability and to
assess the influence of the new items in the questionnaire, the responses
for the total sample of Firm B managers were cluster analyzed. The

results of the cluster analysis and the revised scales are found in

 

 

r7

lfitfil D.

1201 difn

11235-4;
Smite
Seals-«‘5
1:535 in

32.3.;

hpmve
Stone 1
mm,
scales
reduce

[here

 

 

fi———7 , ____..

27
Appendix D. The new scales are very similar to the Chesser scales. The
major differences between the Chesser model and the revised model are:
(1) the Goal Clarity and Relevance scale was replaced by two scales
called Importance of Goals and Goal Setting Behavior; (2) Orientation
Toward MBO was renamed Utility of MBO; (3) a scale derived from new
items--Importance of Competence—-was formed; and (4) two Chesser scales——

Subordinate's Influence Over Goals and Perceived Success in Attaining

 

Goals—-were meshed into larger scales. Since only seven additional
items were available for the analysis, this similarity was not unex—
pected. (See Appendix E for other results of this analysis.)

Table 2-8 contains the means and standard deviations for the
Firm B responses to the revised scales. Again the data show a very con-
sistent pattern of no change across time.

Table 2—9 presents the reliabilities, test—retest correlations,
and change score reliabilities for each of the seven scales in the
improved research model. Several of the revised scales have poor change
score reliabilities. While the internal scale reliabilities maintain an
acceptable level, the test-retest correlation for each of the four
scales is very close to the internal reliability for that scale. This
reduces the change score reliability to essentially zero and suggests I
there is little or no change on the four scales in question.

The static, dynamic, and cross—lagged correlations show patterns
of relationships similar to those for Chesser (Table 2—10). The dynamic
correlations also indicate about the same number of significant rela—
tions between change scores. The cross—lags, for the most part, are
Symmetrical about the diagonals. Significant differences are found at

the .20 level or better for only four correlations.

 

 

,_.. .

AJU In“ w
. A! 1 1 .
D a «L K.“ at L ”mm... m 0 e a\
u. . . a t» .\J adIA 11V 1 C J P k

lulu I \M l Ql \.H Ikl\.| !, ‘0

 

 

 

 

r___i

28

| Table 2-8.-- Means and Standard Deviations for Revised
Seven Scale Model

 

 

Means Standard Deviations
Scale Description Time 1 Time 2 Time 1 Time 2

1. Importance of Goals 2 . 94 2. 97 0 .42 0. 41
2. Goal Setting Behavior 2.85 2.91 .44 .46
3. Superior- Sub ordinate

Relationship 3. 31 3. 32 . 39 . 34
4. Utility of MBO 3.59 3.53 .64 .61
5. Importance of

Competence 4. 36 4. 30 .54 .55
6. Job Satisfaction 3.30 3.34 .89 .88
7. Performance—Reward

Association 3.50 3. 38 0.59 0.63

 

 

E‘i‘
:pvxrhyull 15Wfl>31
il |RH|N .thufl.NT

1L. .a— witwubv.

.5732 O.u...JV.
L... r... . . . _._.. . .2: {LCQI 1.1.2:.3 1C: 3....UW ufiEUulu

 

 

29

 

 

 

.mofigflaflnwwdoh ofimom «gnaw p:o«0flwmooum

 

 

oo. mm. mm. mm. coapmHUOmw<
phmzomuouquHOMHom .5
Na. mm. mm. om. :ofluomMmfipmm new .0
oo. om. mm. me. mucouomsou
mo oonmphomEH .m
cm. me. am. am. cm: mo sunfish: .e
em. mm. mm. mm. magm20HumHom
oumcfiwHondmuuowHogsm .m
HH. we. em. me. Hoa>aemm weauuom Haou .N
am. mm. ma. ma. maaoo mo ooemuaoaeH .H
fleets amass flames N made adage a made godpaatumoa oaaom
spnaaaeaaom coauaaonaou .mspaaanefiaom HmeaoueH

macaw omzmzu N oEwHuH oEwH

Howoz oamom co>om womfl>om

you mmnpaaaeeaaom onoom «mango wee mmauaaaamwuom maeum HmeaopeH --.m-m magma

 

 

thwulnvcudl I 1.1.!
avin‘ I Ah QEVI.\
lUU~hU~NE |l.~v~|N Quilt!

.Ah-H II CV
.- - Au cg...“ N Hindu“ V

J .‘h J I i r- WE-.»<

:flv.‘ U Ivfi ”Inigo

1:“ u >vv: 1:. .-

I-Uhl‘ahlnv: that?!“ aurv>ak:
u.....fld.u... I A- Ill-.FV Isa-Lavau 1...

p-A.-JI\ . Jl .n.‘...-c<

 

 

 

30

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

oaamcho uumeH «N. n H96H Ho.
owumum vowwmgummouo mH. u Hm>mfi no.

pmwwmqummouo wfiumum "szuma "H mo 03Hm> ucmuwwaawam
ma ooH ma am vm 0 ma a: mv n a Ha oz mu NH: am. on mm: am: may «1 am
mm me cad mm mm ma Na em N nu ma NH em an e m- mvt «3 mm: w- «H: ma
mm HM NN ooa ov on em « ma n mm cm on ma mm: mo mac we: mm: ado vmo ma
vw v» MN 9: cod na 5N 0N mm m mm mm m ww r. on on may «fit A oz RH
mu ,0 ma on ma sea we No. a: a c N: or Q n N- own ma: NA: .ms ”.0 m
.o« ma mm cm RN ac cod nwiiow ow em Na an as m m n: ma: mw: mfio um: n
.3 .T em 3 nm N- am 3: mm r... am 3. a .3 .3 T.” S S a .4; mm X:
no wv h m: mN an 3H on oma ma mm m H mm um mm m“ ma «- .m m m
mm m w¢ m ma a ”a an a as” um cm a 0m 0w 0H om HM H: r mm ma
mm H ma nm am o «N mm mm \m osa He 0“ ow we mm mm mm mm «A we Ha
mm NH Na am Nm N: N: am mp on He sea N am cw AN m aw KN ,m an ad
m o: va 0H ma om an a a a ma N oga 3N n m Ma: m mmv.wu mu m
my M: ed ma ma w 0? em mm om or an no mad as aw m me n mm mm m
a?! Nfil V er w! n 0 Mb WH M NV wH m H? 34H cm w m& «N 4! am w
0 am: m: m: a: mu m an mm am on am 5 a .w cufi mu om mm «a ma m
N or me: mat on OH: N: 9% ma or mm m mwu me mm mm 95H on ma w mm m
HI mm! v: v?! was ma: 0H! xv ow am we vm C 0v :m mm M mma Mm ”MW 00 q
0H- am- mas mm: we: NH: mm: 0 fit an ma mm owl C en mw mfi mm nmwumu on m
vw ma- «t 0a: m cm: NH: NH m a ma m we an H- v“ v mm mo naa a- m
a: cam «a: ¢mu m: an- mm: nm N ma He an m: mm am mfl mw cm mm .Hm cad a
.fim cm oa..mfl NH 0H my .ve ma nu NH. 0H 0 m x a m v m .m a

a 05:. u N mafia. N mag. H mag.

.ANHH u av muowmcmz m Baum uom Hopoz

soumommm oamom ao>mm pomw>mm osu cw AH mafia u N «SHE v mmuoom mwdwno cam .AN mbHv mmuoom
Gowuwuumfiflwavd vacuum .AH meHv mwuoom GowumuumHGfiEp¢ umuwh mnu How wmowuumz souumHoHuoo as.oHuN manna

 

zatio
05861
\‘arie
the 1
shin

9001

 

 

r—7

31
The impact correlations also show a pattern similar to that of
the Firm A data. Forty-one of the possible forty—nine correlations (83%)
found in the initial score and change score combination matrix are

negative.

Summary

The replication of the Chesser study produced inconclusive
results regarding the real change hypothesis. The replication did pro-
duce scale means, variances, and scale clusters which were quite similar
to those of the original Chesser study. The dynamic correlations were
also similar for both organizations. However, several of the cross—
lagged correlations were different and resulted in some differences in
the effects diagrams.

A close examination of the correlation matrices for both organi-
zations produced some contradictions to the hypothesis of real change
observed in the system. The dynamic correlations suggest that the
variables of the system are changing and are positively related, while
the majority of the impact correlations indicate a negative relation-
ship. Also, the constant variances, constant static correlations, and
poor change score reliabilities cast doubt on the assumption of real
Change in a multivariate system.

The next chapter will consider several models that assume real
change in the MBO System. The "general factor" model will be shown to
fit the data. The chapter following that will present several models
which assume that there is no real change in the MBO system. The ”mood"
model will be shown to fit the data. These contradictory results will
then be discussed in the chapter on the critical importance of a third

administration of the questionnaire.

 

 

 

 

system.
brough
is der

IMO]

ma]

Rla‘
NU
Vafi

etc,

 

 

 

CHAPTER 3

MODELS THAT ASSUME REAL CHANGE

The intent of the replication of the Chesser study, as well as
the development of the revised research model, has been to assess real
changes in the MBO study. The data for all three analyses have suggested
that the hypothesis of real change is not well supported. To more
explicitly illustrate this, three models of real change are presented
along with appropriate elements from the data. The first model is
actually that underlying the Chesser interpretation. It suggests that
real changes in the variables are endogenously generated within the MBO
system. The second model assumes that changes within the system are
brought about by some factor exegenous to the system. The third model
is derived by assuming that there is a single variable or _"general
factor" that produces all the observed correlations among the explicit

variables of the MBO system.

The Endogenously Generated Real Change Model

The effects diagram which was derived by Chesser for the change
relationships in the MBO system is representative of a model which
assumes that these changes are generated within that system. Each
variable (Superior-Subordinate Relationship, Goal Clarity and Relevance,
etc.) is identified as a separate entity within the system. It is
implicit in the description of this model that the boundaries of the
system are identified. Then the critical assumption that the variables
of the system are all positively and causally related can be made. Model
descriptions and assumptions such as these are common in traditional

Path analysis and cybernetic model studies.

32

 

 

 

 

 

 

scores
he to
data f

con-e1

stati

tion.

ship

gem

cor

33

This model can be interpreted mathematically as:

xi = T + e
where: xi = observed score for the ith system variable
T = true score for the ith system variable
6 = measurement error

(See note below)
From this model, all of the correlations for the variables (static,
dynamic, cross-lagged, and impact) can be predicted.

For the static correlations, the covariance of the observed
scores during each administration will have a strong positive component
due to the true scores. There will be no influence due to error. The
data for all of the studies show a large number of significant positive

correlations.

 

The cross—lagged correlations will be a little smaller than the
statics assuming there is real change in the variables and a subsequent
change in the variance fer the variables. The data support this predic-
tion.
The dynamic correlations will be similar in pattern of relation-
ships to the static correlations. Positive dynamic correlations are
generally found between all the variables that have significant static ,
correlations in this study.
This model does not run into difficulty until the off-diagonals

of the impact matrix are examined. For the assumed real change model,

 

Note: All equations and models presented in this study will be
stated in standard score form rather than raw score form. This faci1i~

tates the calculation of predicted correlations from the models.

 

 

 

 

*7“.-

je off-d
side the
ad C12.
has fa

octets

34

the off-diagonal correlations should be positive. As an example, con-
sider the assumed causal relationship between Goal Clarity and Relevance
mulChanges in Superior-SUbordinate Relationship. The dynamic correla-
tions for change in these two variables are positive and the cross-lags
indicate that Changes in Goal Clarity is the driver or causal variable
in the relationship in both the Firm A and Firm B data.

Consider a manager who feels that his goals have been stated
with very high clarity. If his superior-subordinate relations are
already high, then perhaps there will be only a slight increase. But if
his relations had been low, then there should be a considerable increase.
In any case, for the managers who are high on Goal Clarity, then there
should be an average increase in their Superior—Subordinate Relations.
0n the other hand, consider a manager who feels that his goals are very
vague. If his relations with his boss are already poor, then perhaps
his frustration over goal setting will not lead to a decrease in
Superior-Subordinate Relations. But if his relations had been good, then
there should be considerable frustration and a sharp decrease in his
positive regard for his boss. In any case, for the managers who score
hmton Goal Clarity, there should be an average decrease on Superior—
Subordinate Relations. Considering all the managers t0gether, the model
assumption that Goal Clarity exerts a positive causal influence on
Superior-SUbordinate Relations clearly predicts that managers who are
high on Goal Clarity will increase on Superior-Subordinate Relations
while managers who are low on Goal Clarity will decrease on Superior-
Subordinate Relations (or increase by less if "low” means only ”rela—
tively low"). Thus the positive arrow from Goal Clarity to Superior-

Subordinate Relations in Chesser's model implies a positive correlation

 

 

35
between initial score on Goal Clarity and Change in Superior Subordinate
Relations. The dynamic correlations and cross-lags suggest that the
impact correlation of this assumed relationship should be positive.
However, the impact correlation is negative in the data. In fact, over
seventy percent of these correlations in the impact matrix for each of
the three studies are negative. The model which assumes that the
changes in the system are real and are endogenously generated is thus

contradicted by this data.

The Exggenously Generated Real Chanie Model

 

If the changes that are observed in the dynamic correlations are
not the result of causal influences from within the set of variables
identified as the MBO system, then what else could cause the changes?
One possibility is an "exogenous" factor or variable, i.e., some variable
that influences the system from outside the identified system boundaries.
Thus, factors such as external economic factors may be influencing each
of the participants in the study. This influence could be different or
the same during each administration of the questionnaire but the impor-
tant assumption is that it affects each participant in the sampleiin
some way.

This model is also represented mathematically as

x- = T + e

and in general is quite consistent with the data. The critical assump—
tion in this model is that real change in the system is the result of a
factor outside the MBO system.

The static correlations, just as in the previous model, would be A

Positive and stable across administrations. The cross-lagged

 

 

 

 

gmlatlw

is ex

t

..,
35L

fiiagm

hen

 

 

36
correlations would also be positive and would be of the same general
magnitude as the statics. This is due to the assumption that the
observed change is not the result of a variable within the system.

The dynamic correlations would be significant due to the effect
of the exogenous factor upon the system. The large number of positive
dynamic correlations would be interpreted in this model as implying that
it is 223 exogenous factor instead of several that causes change within
the system.

The negative impact correlations are only partly explained by
the exogenous factor. The diagonals are negative and could well be the
result of spurious correlation of measurement error. For the off-
diagonals in this matrix, the correlation is

2(T + e)(AU + Ac)
rx A = ~——-—————————-—————
’ y nquAy

M

where: y U + a

Ay = AU + As

y = observed score for an MBO variable different from x

U = true score for variable y

e = measurement error for y
then by assuming that the errors of measurement are uncorrelated and that
the changes in variable y are not related to the initial score for x but
are the result of the exogenous factor, the off—diagonal impact correla—
tion is:

r = EEAX__ = 0
X’Ay no- 0
x Ay
which again is not consistent with the data. This model would not pre-
dict the several large negative correlations which are present in the

data. It is further contradicted by the constant variance and the

 

rem

,,
is ,‘

ht

“Si:

 

 

37
constant static correlations. If there was real change in the system as
the result of the exogenous factor, there would be an increase in the
variance for the variables of the system. The data indicates constant
variance for all three studies (see Tables 2-5 and 2-8). An increase in
the variance would also show up as an increase in the static correla- '7
tions. In other words, the static correlations for the second adminis-
tration would be different than those for the first administration and
there is no significant difference between the static correlations for
the two administrations of the questionnaire for Firm A or Firm B.
Since the data do not fit the predictions of this model, it must be

rejected as an explanation of the changes observed in the MBO system.

General Factor Model

 

The possibility exists that instead of developing and soliciting
responses to a number of independent factors related to the MBO system,
the questionnaire is actually the explication of a single underlying
global factor. The model which considers this possibility assumes that
there is some general factor (G) which underlies the entire MBO system.
This single factor is responsible for all of the correlations between
the variables of the system. If the variables of the system represent
this general factor, then the static, cross—lagged and dynamic correla-
tion matrices will all have the distinguishing characteristic of a
"Spearman Rank-One Matrix" (Spearman, 1904). That is, the variables
which estimate and are highly correlated with the general factor will

themselves be highly intercorrelated.

 

 

1
Elm;
W515i.

ii: at

for a

the i

38
The Basic Model and the Static Correlations

Spearman's (1904) general factor model assumes that there is one
central variable which underlies some given domain, such as the MBO
system. The correlational significance of any particular variable in
his model is solely a function of the extent to which that particular
variable correlates with the general factor. In a path diagram or
effects diagram this means that the only arrows associated with the
observed variables are the arrows from the general factor to the
observed variable. More specifically, let x be the observed variable, G
the general factor, and let y be any variable outside the domain. Then
Spearman's theory assumes that the partial correlation between x and y
with G removed is zero, i.e.,

rxy.G = 0 ‘
for all x and y.

Thus the general factor model is a very strong theory in which
the individual observed variables are robbed of all significance. That
is, this model assumes that the difference between the observed variable
and the general factor can be partitioned into two sets of essentially
trivial determinants: the usual error of measurement and a "specific"
factor. The specific factor is that part of the true score for the ’
observed variable which is left when the general factor is partialled
out. Thus in this model the Specific factors are necessarily uncorre—
lated with any variable except themselves. The only statistical dif-
ference between "specific factors” and "errors" is that specific factors
may be stable over time. Some of the constituents of the specific
factor would be idiosyncratic semantic factors such as the peculiar

elements of the job situation that a manager assigns to the words,

 

 

wwage:
155';me

dime:

SIaIt‘C

39
"management by objectives system," idiosyncratic meanings for the
response categories such as "often” in a given context, individual
differences in the importance assigned by managers to minor features of
the goal setting process, etc. The critical test of a hypothesized con—
stituent of a Specific factor is: would it be uncorrelated with (l)
the general factor, (2) all other specific factors, and (3) all variables
outside the MBO system? This is a harsh criterion and if the general
factor model holds, then there is little significance to the specific
factors.

Mathematically, Spearman's general factor model is easily

stated:
xi = G + Si + ei
where: Xi = observed score

G = the general factor

S1 = the Specific factor for variable i or the
residual of the true score for variable xi
when G is partialled out

ei = the measurement error for variable i

The Specific factors, Si’ are recognized as separate but not very
significant components of the model. It is assumed that each specific
factor is uncorrelated with the general factor or with the Specific
factor of any other system variable.

If all the variables had been measured at only one point in time,
then the only test for the general factor model would be a test of the
predicted relations between the static correlations. This test was pro-

posed by Spearman in 1904 and is classic.

 

 

4
MED
in

t
tact

 

 

4O

2(G + Si + 6i)“; + Sj + ej)

 

Xi Xj no Xio xj

= 2(G + Si)(G + Sj) + Zeixj + inej

2(6 + si)(c + sj)

_ n0
xlxj

because the errors of measurement are uncorrelated. This formula is
actually quite familiar once it is pointed out that G + Si is the true
score Ti for the ith system variable. To continue,

Z(GG + st + sic + sisj)

r . =
Xli' no

 

0'
X1 XJ.

2
EC + ZGSj + 2816 + ZSiSj

 

no ,0 _
x1 xJ

_ ZGZ
nCXin

where strong use has been made of the assumptions that (l) the specific

factors Si and Sj are uncorrelated with G and (2) that the Specific

factors are uncorrelated with each other. Finally the formula can be

rewritten

l'mlnt
arisen

its: \‘2

501101

node
car

201

 

 

41

If Spearman's general factor model holds in the data, then the
correlation matrix must have a Special form. If the variables are
ordered from high to low on the basis of their average correlation with
other variables, then the variables will also be ordered from high to
low on the general factor. The general fit of the model can then be
tested as follows. If the strongest variable is listed first, then the
highest correlations should be in the top left—hand corner. Moving from
left to right, the correlations should all decrease in magnitude (to
within sampling error). Moving from the top down in the matrix, the
correlations should decrease in magnitude (to within sampling error).
Thus, by moving from the tOp left—hand corner of the matrix to the
bottom right—hand corner of the matrix, the correlations should decrease
from the highest in the matrix to the lowest (to within sampling error).
Spearman called this "hierarchical structure"; the modern term is "rank
mm" matrix.

The static correlation matrices for the Chesser Firm A study,
the replication of the Chesser study in Firm B, and the revised research
model all demonstrate the presence of a general factor; i.e., the static
correlations form a "rank one" matrix. For each study, the static
correlations for both administrations have been averaged. This was done
because the two time periods have shown very stable patterns across and
within administrations. In Table 3—1 the averaged static correlations
for Firm A managers studied by Chesser are presented. The data has been
reordered to place the strongest variable, Superior—Subordinate Rela-
tionship, in the top left-hand corner of the table, then the next
highest, Orientation Toward MBO, and so on for all seven variables. It

is seen that the ”rank one" characteristic is reasonably strong for five

 

 

Tile

 

 

42

Table 3-1.—- Reordered Matrix of Averaged Static Correlations for the
Chesser Study Firm A First and Second Administrations——
(n=73)a

 

Averaged First and Second Administration
Static Correlations

 

 

Variables l 3 6 4 2 7 S

l. Superior-Subordinate

Relationship 1.00 .45 .35 .33 .31 .27 .05
3. Orientation Toward MBO .45 1.00 .20 .14 .27 .27 .03
6. Satisfaction With Job .35 .20 1.00 .42 .16 —.03 .00
4. Performance—Reward

Association .33 .14 .42 1.00 .18 .06 -.08
2. Goal Clarity and

Relevance .31 .27 .16 .18 1.00 -.01 .01

I Perceived Success .27 .27

5. Subordinate's Influence

Over Goals .05 .03 .00 _,08 .01 —.06 1.00

 

aCorrelations for this table are taken from Table 2-6, page 48
(Chesser, 1971),

Significant value of r:
.05 level = .23

.01 level = .30

 

 

ham

"1‘ ‘
.m

van

an

n.
:1.

43
of the seven variables. The variables, Perceived Success and Subor-
dinate's Influence, fall outside the rank one matrix.

Table 3-2, the reordered matrix of averaged static correlations
for the Firm B managers, has been constructed in the same manner as
described above. VAgain the rank one characteristic is quite evident.
Also, it is the variables, Perceived Success and Subordinate's Influence,
which fall outside the rank one matrix.

The scales of the revised research model also show the rank one
characteristic (Table 3— 3). In the revised model, the scale develop-
ment concentrated on making the variables both more reliable and more
distinct within the system. In so doing, the items for the variables,
Perceived Success and Subordinate's Influence, were absorbed into better
scales. Thus in the reordering of this matrix, six of the seven
variables show the strong hierarchial structure. Thus, all three
averaged static correlation matrices Show the rank one pattern charac—

teristic of a general factor.

The Cross—Lagged Correlations

What does the general factor model predict for the cross—lag
correlations? The critical assumption in answering this question is
determined by examining the behavior of the specific factors. If the
general factor model holds at time 1 and holds again at time 2, then the
factors that are specific factors at time 1 are still specific factors
at time 2. What this means is that the specific factors are not only
uncorrelated with each other, but they do ESE causally interact over
time either. Again, the essentially trivial character of the specific

factors is evident in their correlational behavior.

 

 

 

e1.

Sig

 

44

Table 3—2.-- Reordered Matrix of Averaged Static Correlations for the
Replication Study First and Second Administrations-—
Firm B (n=ll7)

 

Averaged First and Second Administration
Static Correlations

 

Reordered Variables

l. Superior-Subordinate
Relationship 1.00 .52 .42 .52 .37 .23 .09

2. Goal Clarity and Relevance .52 1.00 .37 .29 .14 .17 .09

3. Orientation Toward MBO .42 .37 1.00 .41 .25 .17 -.04
4. Performance-Reward

Association .52 .29 .41 1.00 .41 .15 .12
6. Satisfaction with Job .37 .14 .25 .41 1.00 .04 .04
l Perceived Success .23 .17 .17 .15 .04 1.00 —.05

5. Subordinate's Influence
Over Goals .09 .09 —.04 .12 .04 —.05 1.00

 

 

Significant values of r:
.05 level = .18

.01 level = .24

 

hm

 

 

45

Table 3—3.—— Reordered Matrix of Averaged Static Correlations for th
First and Second Administrations — Revised Seven Scale
Research Model — Firm B Managers (n = 117)a

e

 

Averaged First and Second Administrati
Static Correlations

on

 

Variables l i 1 _5_ i E 3
1. Importance of Goals 1.00 .66 .42 .35 .33 .20 .30
4. Utility of MBO .66 1.00 .55 .37 .38 .38 .22
I Performance-Reward
Association .42 .55 1.00 .40 .31 .32 .04
5. Importance of Competence .35 .37 .40 1.00 .17 .19 .03
3. Superior—Subordinate
Relationship .33 .38 .31 .17 1.00 .30 .00
6. Job Satisfaction .20 .38 .32 .19 .30 1.00 .08
2. Goal Setting Behavior .30 .22 .04 .03 .00 .08 1.00

 

aThese correlations are taken from Table 2—7 , page 17.
Significant values of r:
.05 level = .18

.01 level = .24

 

dfferent
entries (
ietue

sitar :1‘

Means.

Yariab

 

Since

fattg

 

46

Formulas for the cross-lagged correlations must provide for two
different cases: the test-retest correlations that form the diagonal
entries of the cross—lag matrix and the off-diagonal entries that involve
the time 1 score on one variable and the time 2 score on another. Con—
sider first the test—retest correlations. Since there is only one
observed variable, denote it by x and use the subscripts for time 1 and
time 2. Thus

x1 = G1 + Sl + e1

x2 = G2 + $2 + e2

The test—retest correlation will be

2(G1 + 81 + el)(G2 + 82 + e2)

 

r =
xlx2 n0x10x2

2(61 + Sl)(62 + 82)

noxlon

because the errors of measurement are not correlated with any other
variable.
ZCGIGZ + 6182 + 8162 + 5182)

r =
x1X2 noxlox

 

2

 

nOX10x2

n
OXIOXZ
. - ' eral
since the specific factors do not interact over time with the gen

factor. To continue

 

 

here the
mnelatic
x: are the
nines of
wnelati:

tions by

iaractex

Tandy]:

The cros

hem]:

met;

 

47

06162 05152

r ______. ______.
x x
1 2 oxlon oxlox2

where the first term is for comparison with the off-diagonal cross—lag
correlations and the second term represents the fact that since x1 and
x2 are the same variable at two times, 51 and 82 are the two successive
values of the same specific factor. Thus rslsz#0 and the test—retest
correlations should be larger than the off—diagonal cross-lag correla-
tions by precisely the amount of the second term.

The prediction of the off-diagonal cross-lags is similar in
character although notationally more difficult. Consider two variables
x and y measured at two different times. Then

x1 = 61 + 8X1 + e1
Y2 = G2 + Sy2 + 82
The cross-lag correlation will be

_ 2(61 + SX1

+ e (G + S +
r _ 1) 2 y2 82)
x172

nox1°Y2

2(61 + sxl)(c2 + 5X2)

 

n
OxlOyZ
because the errors of measurement are uncorrelated. Continuing,

25 G + 26 S + 28 G + XS S
TX = 1 2 1 y2 X1 2 X1 y7
1Y2 “0x 0

1 Y2

26162
nOx1°>’2

because the specific factors do not interact causally with either the

general factor or with each other. This in turn can be rewritten

 

 

53M of \
Conelat

lltlon n

Thus

eSsem

mm.
110“

11am

 

 

48
0G G r o o
r = 1 2 = 6162 G1 62
x1Y2 o a o 0
x1 y2 x1 y2
oG 0G2

- 1' .
G G ' -- --
1 2 OX1 Oyz

‘ rclcz'a‘z‘.‘ :7..-

1 1 Y2 2

= r6162 ' rxlol ' ryzc2
This triple product can be broken up into two parts. The first part is
rGle’ the test—retest correlation of the general factor and is indepen-
dent of which variables are taken to be x and y. Thus the test-retest
correlation acts as a general multiplier of the entire cross lag corre-
lation matrix. The second part of the triple product is the product
rlel . ryzcz. If the total amount of change in the system is not
large, the ratio of the variance of the specific factor to the variance
of the general factor will not be greatly changed over time. If this
is true (and it is exactly true of the data in both firms), then
12'szerle That is
rx1G1 rYZngs rx151 ry161 = rx1Y1

Thus if r then the second part of the triple product is

Y2G2Q" rchl’
essentially the static correlation at time 1, i.e.,

r r r

x1Y2~ 6162 le1
In particular, if the static correlations remain about the same from
time 1 to time 2 (as they do in this study), then the cross-lag correla—
tion matrix is obtained by simply multiplying the static correlation

matrix by the test-retest correlation of the general factor.

1h
cross-lag
that eke e
hierardii‘
em. sa

ie 2mm

ihree s1
accord]
lattice
for the
for em

lions,

lagged
cones

clear

 

mode

 

 

49

Thus if the test-retest correlations in the diagonal of the
cross-lag matrix are ignored, then the general factor model predicts
that the cross—lag correlation matrix will show exactly the same
hierarchical order as did the static correlation matrix. That is, to
within sampling error, the magnitudes will simply have been reduced by
the multiplicative constant r6162.

These derivations for the cross—lagged correlations predict that
diagonals of the matrix (the test—retest correlations) will be greater
than the off—diagonals by an amount equal to the correlation of specific
factors for that particular variable at two points in time (i.e.,
r5182 # 0). The data for all three studies support this.

To illustrate, the cross-lagged correlation matrices for the
three studies (Firm A, Firm B, and revised Firm B) have been reordered
according to the hierarchical structure found for the static correlation
matrices (Table 3-4 for Firm A, Table 3—5 for Firm B, and Table 3-6
for the revised Firm B model). As predicted, the diagonal correlations
for each of these matrices are larger than the off—diagonal correla-
tions.

In each case, the off-diagonal correlations in these cross-
lagged correlation matrices show the same rank one pattern found in the
corresponding averaged static correlation matrix. This is particularly

clear for the revised scales in Firm B.

The Dynamic Correlations
The predictions for the dynamic correlations follow from the
derivation of the formulas for the static correlations. That is, the

model assumes real change in the general factor which accounts for the

 

 

Table :

3. Urie

.1 .

San

1. Peri
Assc

3. Goal
Re 1e

7- Peri

(he

1COriel
(Q1635

hm? d:
Variei

Varia

Signi

 

50

Table 3-4.-- Reordered Cross-Lagged Correlation Matrix for the First

and Second Administrations - Chesser Seven Scale

Research Model — Firm A Managers (n=73).a

 

 

Second Administration

 

 

First Administration 1 3 6 4 2 7 5
l. Superior-Subordinate b

Relationship .46 .26 .35 .26 .01 .32 .14
3. Orientation Toward MBO .39 .51 .29 .19 .09 .08 .10
6. Satisfaction With Job .05 .02 .48 .33 -.05 -.01 .07
4. Performance—Reward

Association .20 .03 .23 .42 .00 .12 .23
2. Goal Clarity and

Relevance .23 .36 .17 .13 .23 -.O6 .04
7. Perceived Success .31 .20 19 08 07 37 .19
5. Subordinate's Influence

Over Goals —.10 .05 .08 .02 —.ll —.07 17

 

aCorrelations in this table are taken from Table 3-2, page 67

(Chesser, 1971).

b . .
The diagonal entries are the test-retest correlations between

variable measured at time 1 and time 2.

correlations between one variable measured at time 1 and a second

variable measured at time 2.
Significant value of r:
.05 level = .23

.01 level = .30

Off—diagonal entries are

each

 

Idle

. Goal
Rele

3. 01h

.1.

Per
Ass

. P61

4..

0v.

 

 

51

Table 3-5.-- Reordered Cross-Lagged Correlation Matrix for the First
and Second Administrations - Seven Scale Research Model
Replication Study - Firm B Managers (n=ll7)

 

Second Administration

 

 

First Administration 1 3 6 4 2 7 S
1. Superior-Subordinate
Relationship .46a .22 .38 .30 .06 .25 -.13

2. Goal Clarity and

Relevance .31 .46 .24 .17 —.06 .13 —.20

3. Orientation Toward MBO .36 .39 .66 .32 .12 .24 —.04

4. Performance-Reward

Association .38 .31 .36 .60 .22 .07 .08
6. Satisfaction With Job .32 .16 .27 .37 .52 18 .05
7. Perceived Success .20 .26 .12 .10 .06 .51 —.16

5. Subordinate's Influence
Over Goals .04

 

8The diagonal entries are the test—retest correlations between each
variable measured at time 1 and time 2. Off-diagonal entries are
correlations between one variable measured at time 1 and a second

variable measured at time 2.
Significant value of r:
.05 level = .18

.01 level = .24

 

 

 

Table

3'

»_.
~51

."
pa m
245

time
Vari;

Sign

52

ble 3—6 .-— Reordered Cross—Lagged Correlation Matrix for the First
and Second Administrations — Revised Seven Scale
Research Model — Firm B Managers (n = 117)a

Second Administration

 

 

 

First Administration 1 4 7 5 3 6 2
Importance of Goals .53b .41 .28 .22 .31 .02 —.03
Utility of MBO .49 .68 .47 .31 .24 .16 .06

Performance-Reward
Association .41 .42 .58 .33 .18 .13 .05

 

Importance of Competence .27 .25 .30 .56 .08 .13 —.13

Superior—Subordinate

Relationship .06 .13 .06 —.01 .27 -.01 -.19
Job Satisfaction .24 .36 .33 .10 .21 .52 .07
Goal Setting Behavior .20 .13 .12 .00 .05 .02 .44

 

se correlations are taken from Table 2-10, page 30.

diagonal entries are correlations between a variable measured at
e l and time 2. Off-diagonal entries are correlations between a
iable measured at time 1 and a second variable measured at time 2.
nificant values of r:

.05 level = .18

.01 level = .24

 

 

observer:

n11 be 6

derived ;

then the

becaus.
variab

over t

Speci

that

53
erved dynamic correlations in the system. Thus, since there is
nge in this model, the dynamic correlations will not be spurious but
1 be evidence of that change.
Mathematically, the equation for the dynamic correlations is

ived as follows:

Given, Ax = x2 — x1 = AG + ASi + Aei

and Ay=y2-y1=AG+ASJ-+Aej
where: y = an MBO variable different from x
c = error of measurement in variable y

1 the dynamic correlations for the general factor model are given as

_ 2(AG + A5! + Ae-)(AG + A8- + Ae.)
I‘AX’AY - 1 l 1 1

 

nOAX OAy
ZAGZ + ZAGASj + ZAGASi + zASiASj‘
nOAx UAy
use the errors of measurement are not correlated with any other
able. When the assumption that the Specific factors do not interact
'time with the general factor is used, the equation becomes
ZAGZ + zAsiAsj

r :
AXAY noAony

additional assumption that the specific factors at time 1 are still
ific at time 2 and do not causally interact with each other implies

A51 and ASj are uncorrelated. Thus

ZAG2 CZAG
1‘ = —— = W
AXAY DOAXOAY AX Ay

1 yields a Positive dynamic correlation. This can also be written
= ' ' 's e uation with observed varia-
I rAxAG 'rAyAG which is Spearman q

Ax and AV and a general factor of AG. Thus the dynamic correlatlons

Ld also form a rank one matrix.

general
Iable 3-
images.

Silo.“ '

is:

. v,
\omlc

This

 

“hi.

to;

 

54

Once again, the prediction derived by assuming the presence of a
eral factor is corroborated in the data for all three studies. In
le 3-7 the reordered dynamic correlation matrix for the Firm A
agers and in Table 3-8 the matrix for the Firm B managers show
)ng positive dynamic correlations between the five variables which
a assumed to correlate highly with the general factor. This same
tern of strong positive dynamic correlations is seen in the revised
31 (Table 3—9 ). Each of the three matrices also shows the same
rarchical structure as did the corresponding static and cross-lagged

relations.

The Impact Correlations
The derivation of the off—diagonal impact correlations in the

eral factor model is straightforward.

V

r 3 2(5 + si + 2i)(AG + ASj + A2.
X’Ay nOxOAy

ZGAG + )ZGASj + ZSiAG + ZSiASj

 

 

 

nUXCAy

= ZGAG

noony
_ OGAG _ OGOAG
‘ ‘ TGAG ' To“

oony x Ay

in turn can be rewritten
OZG OZAG

rxAy = rGAG' .

OXOG OAYOAG

' rGAG rxG rAyAG
h is a triple product much like that found for the cross—lagged

is inde endent
elations. Again the first part of the PTOdUCt’ rGAG’ p

 

 

Table

 

a
9

55

Table 3-7.—- Reordered Dynamic Correlation Matrix for the First and
Second Administrations — Seven Scale Research Model -
Firm A Managers (n=73)a

 

 

Changes in Variable _1_ g 6 4 2 _7_ _5__
Superior—Subordinate
Relationship 1.00 .24 .34 .23 .30 -.16 .14
Orientation Toward MBO .24 1.00 .23 .08 .05 .21 .07
Satisfaction With Job .34 .23 1.00 .37 .11 —.13 —.O6
Performance—Reward
Association .23 .08 .37 1.00 .12 —.10 —.01 ;
Goal Clarity and
Relevance .30 .05 .11 .12 1.00 .06 .08
Perceived Success -.16 .21 -.13 -.10 .06 1.00 —.07
Subordinate's Influence
Over Goals .14 .07 —.06 -.01 .08 -.07 1.00

 

rrelations for this table are taken from Table 3-1, page 65
Lesser, 1971).

nificant values of r:
.05 level = .23

.01 level = .30

 

Table

Signi

 

56

'able 3-8.—- Reordered Dynamic Correlation Matrix for the First and
Second Administrations - Seven Scale Research Model
Replication Study — Firm B Managers (n=ll7)

 

Changes in Variable

Superior—Subordinate
Relationship

Goal Clarity and
Relevance

Orientation Toward MBO

Performance-Reward
Association

 

Job Satisfaction
Perceived Success

Subordinate's Influence
Over Goals

h—t

1.00

.48

.12

.37
.35

-.01

.23

Z 2
.48 .12
1.00 .14
14 1.00
.11 .20
.19 .12
—.OS - 02
.27 -.01

lb

.37

.20

.00

.24

IO"

.35

.12

.24

1.00

—.15

.04

.23

.27

—.01

.04

.11

1.00

 

ificant value of r:
.05 level = .18

.01 level = .24

 

 

lrille

l9

Inge

. Util

7. Per:
.355:

3These

Sign:

 

 

57

able 3-9.-- Reordered Dynamic Correlation Matrix for the First and
Second Administrations - Revised Seven Scale

 

 

 

Research Model — Firm B Managers (n = 117)a
Changes in Variable I .4 2_ §_ §_ 6_ 2
Importance of Goals 1.00 .54 .19 .22 .27 .IS .41
Utility of MBO .54 1.00 .27 .22 .40 .31 .30
Performance-Reward
Association .19 .27 1.00 .22 .34 .18 -.08
Importance of Competence .22 .22 .22 .00 .23 .15 .18
Superior-Subordinate
Relationship .27 .40 .34 .23 .00 .34 .13
Job Satisfaction .15 .31 .18 .15 .34 .00 .06
Goal Setting Behavior .41 .30 -.08 18 .13 .06 1.00

 

se correlations are taken from Table 2-10, p. 30.

nificant value of r:
.05 level = .18

.01 level = .24

 

 

 

 

of which

which

rank (

Slatil

be th
3le

15 (i!

 

 

58

hich variables are represented by x and y. Thus rGAG is a general
iplier of the entire impact correlation matrix. The second part of
triple product is the product rxG rAyAG and is more complex than the
uct for the cross-lag correlations. This term will be large to the
nt that the initial score in question is correlated with the general
or and the change score in question is correlated with the change in
eneral factor. If all the specific factors changed to the same
e, then the rank order of the rAyAG's would be the same as the rank
of the ryG’s, i.e., the same as the rank order of the static
lations. However, if some Specific factors change much more than
s, then this rank order could be modified. A rough estimate of the

cted rank order of the rhyAG'S is given by the reliabilities of the

,e scores.
Thus going down a column of the impact correlation matrix, the
order of the numbers should be the same as the rank order of the
c correlations (to within sampling error). Going across a row of
mpact correlation matrix, the rank order of the correlations should
5 same as the rank order of the change score reliabilities. The
a1 magnitude and the sign of the entire set of impact correlations
:ermined by the impact correlation of the general factor, TGAG-
The most important point of this derivation is in the last sen—
If the general factor undergoes real regression to the mean, then
S negative and the general factor model predicts that every entry
impact correlation matrix will be negative (to within sampling
. And thus if regression to the mean in the general factor is

d, then there is at least qualitative fit for the predictions of

neral factor model and the impact correlations found in the data.

 

 

 

Var:

let

 

59
The predicted pattern of impact correlations for the three
as does show a hierarchical structure (see Table 3—10 for Firm A,
3—11 for the Firm B replication study, and Table 3—12 for the Firm
Lsed study). Although not as distinct as in the static, cross—
1, and dynamic correlation matrices, the data do support the exis-

of a general factor.

The Nature of the General Factor

The nature of the general factor is unknown. On one hand, it
the manager's general attitude toward work—-a generalized atti-
f job satisfaction. Alternately, it could be more general.
5 the general factor in this system is the manager's general level
f-esteem or self—confidence. On the other hand, the general

may be less general than it appears to be. Suppose the general
is simply how well the manager thinks the boss likes him. If the

defines competence, performance, etc., almost exclusively in

f his ability to influence the boss; and if he defines the

of MBO and other aspects of goal setting solely in terms of pro-
an opportunity to get to the boss in an atmosphere of solemn
ation; then a single narrow attitude could account for all the
tion. The Specific factors in this case would be the "trivial"
ve aspects of the work situation that are brought to the

's attention by the Specific words in the specific questions.

The definition of the general factor is actually a function of
tem boundaries. There is evidence that five of the seven MBO

es are estimates of the general factor. However, until it can be

ed whether the general factor is completely within the present

 

 

 

 

Table

 

6O

‘able 3-lO.-- Reordered Impact Correlation Matrix for the First and
Second Administrations - Seven Scale Research Model —
Firm A Managers (n=73)a

 

Change Score

Initial Score 1 _3 _6_ 4 _2_ _7_ §_
Superior-Subordinate
Relationship —.46 -.15 .02 .02 —.32 —.03 —.03
Orientation Toward MBO -.03 -.39 .21 .13 -.19 —.16 -.19
Satisfaction With Job —.23 -.1O -.61 -.ll -.21 —.O6 —.05
Performance—Reward
Association —.04 -.07 -.27 —.53 -.12 .01 —.15
Goal Clarity and
Relevance -.13 .08 .04 .11 —.62 —.23 —.04
Perceived Success .05 .03 .26 .00 .01 -.57 -.02
Subordinate's Influence
Over Goals —.12 —.O8 .04 .08 .00 .10 —.76

 

 

relations for this table are taken from Table 2—7, page 54,
esser, 1971).

nificant values of r:

.05 level = .23

.01 level = .30

 

 

 

Table

Sign

 

61

ile 3-ll.—— Reordered Impact Correlation Matrix for the First and

Second Administrations - Seven Scale Research Model

Replication Study - Firm B Managers (n=ll7)

Initial Score .l
iperior-Subordinate
elationship —.61
)al Clarity and
alevance —.27
rientation Toward MBO —.09
:rformance—Reward
isociation —.l4
itisfaction With Job —.08
:rceived Success .10

ordinate's Influence
er Goals -.12

.34

.58

.03

.07

.03

.15

.05

Change Score

3

.02

.10

.38

.07

.02

.12

.08

i

.18

.05

.ll

.41

.01

.03

.13

6

.31

.18

.18

.50

.03

.05

Ix)

.04

.22

.16

-.47

.04

.25

.28

.05

.04

.01

.08

.57

 

icant values of r:

 

 

 

Tail

Signi

62

1e 3-12.—- Reordered Impact Correlation Matrix for the First and
Second Administrations — Revised Seven Scale
Research Model - Firm B Managers (n=ll7)

 

Change Score

Initial Score 1 4 1 i _3 _6_ _2_
portance of Goals -.52 —.34 -.Ol —.11 -.08 —.16 -.31
ility of MBO -.19 —.44 —.01 —.04 -.13 —.23 —.18
rformance—Reward
sociation .09 —.12 -.40 .04 -.08 —.17 .05

portance of Competence —.07 —.15 .02 —.45 -.06 —.06 -.16

perior—Subordinate

lationship —.32 -.28 —.19 —.15 —.68 -.30 —.17
b Satisfaction .05 —.05 .06 —.08 -.09 —.50 —.07
a1 Setting Behavior —.12 —.16 .14 -.04 .05 —.12 -.SO

 

 

icant values of r:
)5 level = .18

)1 level = .24

 

systen,

present

enint

e.‘
Llenal

devi:

mane
high
who

Exit

63
:em, is part in and part out, or is a factor completely outside the

:ently defined system, a sound definition cannot be established.

Discussion
Several models were developed to explain the results of the two

rical studies. Only the general factor model adequately explained
data and then only with some very delicate assumptions. First of
there is the assumption that a single general factor produced all
observed correlations in the system. This of course is completely
radictory to all the theory in the literature. If in fact the
ries that posit MBO as a multivariate system are right, then the
dity of the general factor model in the present studies would amount

n invalidation of the separate variables of the questionnaire used

Second, there is the finding of no change in means, standard
ations, and correlations over time. The fact that the variances
t change means that the impact correlation is negative; i.e.,
gers who start 10w increase by more than the managers who start
. The fact that the mean change is zero implies that the managers
start high actually decrease over time. Again this contradicts
:ing beliefs regarding organizational development programs.

Two other facts in the data inhibit the plausibility of the
ral factor model. First, the fact that the mean change is zero
.res the balance point between increase and decrease on the general
)r to be exactly the general factor mean for that firm. Second, the
that the variance does not change means that the decrease in

. . . , . t1
mce produced by the regression in managerial attitudes 15 exac y

 

 

balance

systen.

xeliab
{actor
hate 2
Store
test--
i818
then

Cum

pm
sta

dis

 

64

iced by the increase in variance produced by factors outside the
em. Furthermore, this exact balance must be assumed in both firms.

Another area of concern in the data is the poor change score
ability for several of the scales. Since the change in the general
or makes a contribution to every change score, no change score can
zero reliability. However, it is possible that the low change
e reliability is the result of sampling error. If the observed
,—retest correlation (the diagonals in the cross-lagged matrix) rxle
thigher than the population correlation because of sampling error,
lthe reliability of the change score for that variable would be
:omitantly underestimated.

Thus, there are enough problems with the general factor model to
t its plausibility. However, the general factor model cannot be
cted on the basis of the two time period study.

Although the data from both organizations indicate the presence
general factor, this is not proof of real change. Indeed, the
lems in the general factor model arise from the constancy of means,
dard deviations, and static correlations. All these problems would
ppear if there were no change in underlying MBO variables. The next
:er will present several models that assume no change in the MBO
3m. In particular, the "mood" model will be shown to fit the data \
‘11 as the general factor model without making so many special /’

t balance" assumptions.

 

produc
min
mm;
were

best ,
for i
the i

tiea‘

is t
he

ass

CHAPTER 4

MODELS THAT ASSUME NO REAL CHANGE

The analysis of the two time period data for both organizations
mced several contradictions to the hypothesis of real change in M80
tudes as a multivariate system. In an effort to explain these
;radictions, several models based on the assumption of real change
: developed in the previous chapter. The model which fit the data
5 the general factor model, required several delicate assumptions
its development. Because of these, other models which account for
pattern of results were deve10ped and evaluated. These models are
med in this chapter.

The basic problem in the data from a theoretical point of view
hat the means, variances, and correlations are unchanged over time./

the only model which fit the data and assumed real change had to
me that the MBO variables had no separate identity and that the
c pattern of change was regression to the mean. There is an alter-
ve assumption which fits the finding of no change in the means,
iard deviations, and correlations very nicely: assume that there
.n fact no change in the fundamental MBO variables. That is, assume
the questionnaire was basically valid but the managers are simply
able at the time of life measured that there is no measurable

e.

If there is no actual change, then why do the observed

Jles show change? This chapter presents three answers to this
ion, i.e., three agents which can produce "apparent" change in the

'ed variables. If each agent is paired with the assumption of no

65

 

 

 

change

obsem

tend
and s
sion
the (
were

mgr

will
the

CIO

b‘é’

66
: in the MBO variables, then a different model to predict the
red data is obtained. Each can then be tested.
The first section of this chapter describes the effects of
.ability and the second section analyzes transient effects upon the
;. The final section presents a model called the "mood" model

explains the two time period data quite well.

A Model Based on Errors of Measurement

 

For this project, the questionnaire (Appendix A) was adminis—
to the same managers in each organization on two or more occasions
:ored for direction and amount of change. The most certain conclu-
:hat could be reached was that the change scores, which represented
>served difference between administrations of the questionnaire,
mreliable. One result of these unreliabilities is spurious
sion toward the mean (Hunter and Cohen, 1972).
If there is no change in the underlying attitudes, then there
e no change in the means, variances, or static correlations between
served variables. Furthermore if there is no change, then the
lagged correlations will be substantial. Thus a model which
5 no change automatically fits much of the data. Where does the
§E£_of change come from? Consider unreliability. If there is no
in the underlying true score, then the model can be derived as:
X1=T+el
x2 = T + 62
where
x1 = observed score at time 1

x2 = observed score at time 2

 

he ch:

Sir

an:

67

T = true score for variable x
el = measurement error at time 1
e2 = measurement error at time 2
tange score is given by
AX = x2 - x1 = (T + e2) — (T + 61)
= T _ T + e2 _ el
= e2 - e1 # 0
Ls, in the change score, the two identical true scores cancel out
1e two different errors of measurement do not. Thus the observed
; show change where there is none.
The finding which posed the greatest difficulty for Chesser's
lology was the negative impact correlations. Could they be
)us? Consider first the correlation between the initial score, x,
ie change score, Ax

r. . “T + W62 - 6.)
1,Ax n
OXIOAX

 

n
OXIOAX

the errors of measurement are not correlated with each other or
ier score, the correlation is
_ 2612

r =

x ————————
1AX

noxloAx

s a spurious negative correlation due to errors of measurement.
However, this argument for the negative rx,nx applies only to‘*
gonal impact correlations. What about the off—diagonals? The
tion of initial administration variable x with the change in

variable (y) is shown as:

 

 

Sine

68

x1 = T + el
yl = U + 61
Y2 = U + £2

Ay = (U + 52) - (U + 51) = 52 - 51
where: U = the true score for variable yi

5i = the error of measurement for variable yi

then,
2(T1 + el)(52 - 61)
r = ._______________.___

ZTlcz + Eelez — ZTlel - Eels1

 

nOony

errors of measurement are not correlated with any other score, they
at correlated with each other. Hence,
rxAy = 0

:imple unreliability predicts that the off-diagonal entries of the
: correlation matrix should be zero. That is, simple unreliability
,ot predict a spurious negative correlation between the initial
on one variable and the change score on another. Hence, simple
ability does not account for the off—diagonal negative impact
ations found in both firms.

How would simple unreliability affect the correlation between

5 in two system variables, x and y? For a starting point, assume:

X1=T1+61 Y1=U1+€1
x2 = T2 + e2 Y2 = U2 + E2
AX = e2 - e1 Ay = 52 - 61

Again the
tithin ti

cone 1 ati

thee aga
a Spuric
\ “
cuffeitr
tions wl

Spuriou

model 1

as hef

 

 

 

69

2(62 - €1)(€2 - 81)
r = _____——______——————
AxAy n
OAXOAY

_ ZGZEZ ‘ 28281 - 28182 + 28151
nOAXOAy

:he fact that errors of measurement are not correlated between or

 

time periods means that every term is zero. Thus the dynamic

ition is:
t

rAxAy = 0
‘ain it has been shown that simple unreliability does 223 predict
ous correlation for the correlation between change scores for
nt variables. The one result in the impact and dynamic correla—
hich might be an artifact produced by simple unreliability is the

5 negative regression found in the diagonal of the impact matrix.

A close examination of the quantitative predictions made by this

or the cross—lagged correlations reveals one other failing. If
x1 = T + e1
yZ = U + 52

re, then

m + elltU + e2)

 

rX1>'2 =
n
“X1OY2

ZTU + ZelU + ZTEZ + 26152

 

n
Ox1°>'2

ZTU
nOx1°Y2

_ °TU
Ox1°>'1

= rx1Y1

 

 

hat is,
sinilar t

This is r

retest 0
observed

Inns thi

the que:
variati
instmm
18 the:
from or
hility'
eluded
depend
"Weigh
tuatit
PUI‘po:
Clefin
the f
Weigh
in a
site

had

 

70

t is, the cross—lagged correlations are predicted to be not just
‘lar to the static correlations, but they are predicted to be equal.
5 is not true in the data for either firm.

A side effect of this result is the fact that if the test-
est correlation is equal to the reliability of the scale, then the
erved reliability of the change score for each variable would be zero.

this model fails by predicting too many zero reliabilities.

The Transient Factor Model

 

There are a number of sources of variation in the responses to
questionnaire by the managers of both Firms A and B. Some of that
iation can be attributed to the unreliability in the measuring
trument. Another portion of that variation is change in true scores.
there another component? Consider the change in a person's weight
n one week to another. Little of the variation is due to "unrelia—
Lty" in the sense of "error of measurement.” Can it then be con—
led that all the variation is due to a change in true score? That
:nds on what sort of concept of "weight" is used as a reference. If
ght” is defined as instantaneous mass, then by definition all fluc—
ions in weight represent changes in true score. However for such
oses as the study of obesity, heart disease, body types, etc., this

ition would be pointless and misleading. To illustrate, consider
fact that during a summer game a football player may show ”apparent”
t 1055 of 25 or 30 pounds; i.e., a ”water loss" which will vanish
matter of hours. Clearly the relevant concept of "weight” in this
tion is a hypothetical "true" weight which would have been obtained

he person been weighed under "standard” conditions.

 

 

 

S(
denation:
"instabil
be obtain
:easure a
he corn
the new.

variable
of equiv

Simple e

”coeffi
ject to
actual}
instnn
0f tial
hhile
toeffi
is ove
stab).
no)
A ten
tar),
his
bang
day

the

71

Some time ago (1960) Cronbach noted the distinction between the
ions in observed scores produced by transient factors (i.e.,
bility”) and errors of measurement. The usual reliability would
,ained by administering two different instruments (designed to
re attitudes regarding the MBO system) at the same point in time.
)rrelation between these alternate forms would indicate how well
:asurement on form A agrees with the measurement of the same
)les using form B. Cronbach calls this correlation the "coefficient
ivalence." The difference in scores between these forms is the

error of measurement discussed in the previous model.

Cronbach also defined a second reliability coefficient, the
Licient of stability," to assess the reliability of measures sub—
.0 transient conditions. The ”coefficient of stability" is
.ly a test—retest correlation obtained by administering the same
ment at two points in time that are far enough apart that one set
nsient factors is replaced by another set of transient factors,
close enough in time that there is no change in true score. This
:ient of stability indicates how stable a particular measurement
r time. That is, the "true score" for this coefficient is the
attribute of a manager (e.g., general attitude towards work or
ich will be found in both administrations of the questionnaire.
rary condition such as a quarrel with the boss over some momen—
iget problem, feeling a financial pinch regarding a recent
dinvestment, a temporary lull in sales, etc., may cause a
to respond, on all the items of some scale, higher or lower one
he would the next. The effect of such events is a lowering of

retest reliability or "coefficient of stability.” The effect

of transifi
apposite.

md hence

bilities
equivaler
estimate
trmsien‘
that ext
coeffici
is an ox
comm

the ch31

Cross
tions

U10 8

 

both

dict

 

 

72

ansient factors on the coefficient of equivalence would be the
ite. They would enter into the two measurements at the same time
ence would Spuriously inflate the coefficient of equivalence.

The coefficient used in this study to estimate scale relia—
ies is coefficient alpha (Cronbach, 1951). Coefficient alpha is
alent to an "alternate form" reliability coefficient, i.e., an

te of the coefficient of equivalence. If there are significant
'ent factors in the MBO questionnaire, then coefficient alpha is to
extent an overestimate of the relevant dynamic reliability
icient, i.e., the coefficient of stability. If coefficient alpha
overestimate, then the reliabilities of the change scores will be
;pondingly inflated. This casts doubt upon the reliabilities of
lange scores found in the replication that were not already zero.

Unfortunately in the present study the time interval between
mments is eighteen months. Thus there is no way to differentiate
rn real change and instability. That is, there is no means to
, the effects of ”instability” upon the reliability measures given
.wo administrations of the questionnaire. However, this can be
f there are measurements made at thrge_times. As will be shown in
xt chapter, there is such data and it does suggest pronounced
ent factors.

Mathematically the effect of instability on static correlations,
lagged correlations, dynamic correlations, and impact correla—
is indistinguishable from the effect of error of measurement. The
together in a single "error” term. Thus a model which contains
ror of measurement and instability will make the same major pre—

5 as the model built on simple unreliability alone. However, the

 

 

l

 

existence
etuivalen
ing the e

the like.

factor:
native t
the date
579 neg;
positiw

ing bot

deviati
actual]
50196 d:
feel 6‘
Sonal

Wort

Candy

 

73

istence of significant transient factors makes the coefficient of
uivalence (the usual ”reliability") useless for statistically eliminat—
g the effect of "error" from the data by correcting for attenuation or

e like.

The "Mood" Model

To this point, the effects of unreliability and transient
ctors and the assumption of no real change can be considered an alter-
tive to the real change models to explain some of the results found in
.e data. However these no change models have not been able to explain
e negative off—diagonals in the impact correlation matrix or the
sitive dynamic correlations. The "mood" model will succeed in predict-
g both.

The "mood" model is an attempt to identify a component of the
viation of the observed responses from the true score that would
tually be common to the entire set of responses to the questionnaire.
me days the manager will feel especially good and some days he will
31 especially bad. This could be the result of personal health, per—
1al problems, a conflict situation, and so on. All of these are
idom and transient. However, if a manager's mood contributes signifi—
tly to his measured attitude on one of the MBO dimensions, then it
1 also contribute to the others. That is, if the manager's mood is a
tor in his questionnaire responses, then it is a factor that is

on to all variables in the MBO system. Therefore consider a model

Ch assumes (1) that there is no change in the true score of the
ager between administrations and (2) that the change observed is in

t a function of the "mood" of the manager at that point in time.

 

 

the no

Thet

betau

def“

 

 

 

74
Mathematically, the mood model for two variables is

.= +-+.
x1 T hl e1

andyi=U+hi+€i

observed score for variable x

where: Xi =
yi = observed score for variable y
T = true score for variable x
hi = mood variable at time i
ei = error of measurement for variable x at time i
gi = error of measurement for variable y at time i

Since the mood variable is not correlated across administrations,

the cross—lagged correlation formula is very simple. Thus,

T h U + h +
rx y 2( + 1 + e1)( 2 52)
1 2 noxloyz

ll

Z(T + h1)(U + hz)
nOX1°Y2

ZTU + Zhlu + ZThz + Zhlh2

 

nOx1°Y2
ZTU
nO'XlOyz

= GTU

 

oxoy
The time subscripts have been dropped from x and y in the last equation

becaus = = = = 0
e 0X1 OX2 OX and Cyl OYZ y'

The static correlations for the mood model are similarly

ierived

 

 

 

 

ills sf
in the

Sputio

lags)

the c}

fact

1501;

mode

The;

75

2(T + hl + e1)(U + hl + 81)

 

r
le
l nOX1°Y1

_ ZTU + Ehlz

 

n
OXlOyl

ZTU + 2h12

nOXlOY1 noxloyl

 

> OTU

 

rx1Y1

lus since mood affects both x and y, mood produces a Spurious increase
1 the static correlation between them. Furthermore, it will produce
Juriously high correlations for both administrations of the data.
Since mood increases the static correlations but not the cross—
%5, it can greatly and falsely inflate the estimated reliability of
.e Change scores for the system variables. That is, in the mood model

1x2 = the coefficient of stability. Coefficient alpha as used in

2
2—2 which is inflated. This reflects the

15 study is rXlxl = rxlxz + 02
x

Ct that mood is a transient factor for any one variable considered in
olation.
For the diagonals of the impact correlation matrix, the mood
161 Operates as follows:
x1 = T + h1 + el
x2 = T + h2 + e2
so Ax = h2 — hl + e2 ~ el = Ah + Ae

trefore

 

 

which f

Hence ‘

ahdt

Whit
the

bill

76

_ 2(T1 + hl + 81) (Ah + Ae)
x,Ax _

 

noonx

_ ZTlAh + zTAe + ZhlAh + ZelAh + ZelAe
n

 

oonx
Several assumptions must be made. The errors of measurement are

defined to be uncorrelated with any other variable. Similarly, if mood

is transient and random, it is also uncorrelated with any other variable.

In particular, if mood is a transient variable, then mood at time 1 will

not be correlated with mood at time 2. By using these assumptions,

—Zh12 — zelz = —02h - 026 = _ 02h + 02

r =

e
XAX

noxOAx OXUAX OXOAX
which is the negative diagonal impact correlation found in the data.
Now, for the off—diagonal impact correlations,
Y2 = U + h2 + 52
Y1 = U + hl + 81

so Ay = Ah + As

2(T + hl + el)(Ah +AE)

 

noony

ZTAh + ZTAe + zhlAh + ZhlAE + ZelAh + 261A:

 

nOXOAy
and by USing the same assumptions as before,

_ "Ehlz — _Ozh

XAy HOXOAY CXOAY

 

- - - ' , Wh ? Because
which 15 negative but smaller than the rx,Ax correlation y

the off-diagonals do not have the error term which accounts for unrelia—

)ility or the regression effect. Thus a model has been found whlch

 

 

accounts
is, the

fact th:

FOHOh

land

them

this

the (

altd‘

test

77

:counts for the off—diagonal negative correlation as artifacts. That
, the mood model predicts negative impact correlations despite the
tct that it assumes ng_real change between administrations.

What does the mood model predict for the dynamic correlations?

>r the correlation between change in variable x and change in variable

Ax = h2 - h1 + e2 — el = Ah + Ae
and

Ay = h2 — hl + 52 — 61 = Ah + As
so that

= 2(Ah + Ae)(Ah +Ae)

r
AXA-
Y nOXOAy

_ zAhZ + ZAhAc + ZAhAe + ZAeAe

 

n
OAXOAY

)llowing the previously established assumptions,

2

XAhZ 0 h

rAxAy = fi-'—"—’ ---—
OAxOAy OAxOAy
is is a positive correlation. Thus the fact that mood affects both
and y not only produces a spuriously high static correlation between
em, it produces a spurious dynamic correlation as well.
Thus the mood model predicts every result in the data. Since
is model fits the data and assumes no real change in the MBO variables,

e dynamic correlations and impact correlations may simply be artlfacts

d the reliability of the change scores could be spurious.

Summar
Two alternative models have been derived which explain the

' tant
sults in the data, i.e., the pattern of correlations, the cons

 

variance
diets re
researcl
change :
general

model a

assmrs
that a]
lhethe:

data g.

 

 

78

ariance, and the poor change score reliabilities. The mood model pre-
icts results which are congruent with both the Chesser analysis and the
esearch reported here. The mood model assumes that there is no real
hange in the MBO system during the two time periods. However the
eneral factor model also explains the data. And the general factor
odel assumes real change in the system during the two time period

tudy. Thus, the dilemma is posed. One model which fits the data
ssumes real change while the other model which fits the data assumes
hat all the observed change is spurious. Thus one cannot determine
hether or not there was real change in the managers' attitudes from the

ata gathered in two time measurements.

 

 

dmn*
from th
there 1
third a
an indi

longitt

over 2
each ‘

Would

Eacl

CHAPTER 5

TEST OF THE REAL CHANGE HYPOTHESIS

Is there a direct empirical test of the hypothesis of real
hange in the MBO system? This chapter will first show that if data
rom three administrations of the MBO questionnaire are available, then
here is indeed such a test. The test will then be applied to the
bird administration data from Firm A. The result of the test will be
n indication of which mathematical model best fits the data of the

ongitudinal study of Firm A.

Transient Factors and Unreliability

 

Assume that there is no real change in an MBO system variable
ler a three—year period (in this study the three administrations were
ich 18 months apart). Then the three successive scores for variable x

>uld be characterized by

x1 = T + e1
x2 = T + e2
x3:T+e3

where:
T = unchanging true score for variable x

e- = aggregate of transient factors and .
unreliability for variable x at time 1

ch of the static correlations for the three time periods is the same,

e.,

- 2(T + ei)(T + ej)
rfixj‘ Hg nOXiOxj

 

 

isthec

311m
in t1
Woul

latl

80

2
= ET + zTej + zeiT + Zeie.

 

J
“0 .o _
x1 xJ
2
‘7—
0 x
since 02Xi = 02x. = 02T + 026. Stated differently,

r13 = r12 = r23
is the coefficient of stability as Cronbach defined it.

The most important feature of this model is that r13 = r12 and
r13 = r23. To see the importance of this, suppose that there had been
no unreliability or transient factors in the observed variable. Then
r12 would be low to the extent that there was a large amount of real
Change from time 1 to time 2 and r23 would be small to the extent that
there was a large amount of change from time 2 to time 3. But if there
is considerable change from time 1 to time 2 and more change from time 2
to time 3, then there is greater change from time 1 to time 3 than from
time 1 to time 2 and r13 would be expected to be less than r12. The
fact that r13 is not smaller than r12 in this model is a direct reflec—
tion of the assumption that the observed change from time 1 to time 2
was not real change but only apparent change-

What are the cross—lagged panel correlations in this model?
3Uppose there was perfect measurement for two variables, x and y. Then,
in this no change model, T would be the true score for x1 and x2 and U
vould be the true score for variable yl and Y2- The cross-lagged corre-
lations would be

rxlyz = rTy2 = rTU

: =r
rY1X2 r”X2 UT

 

 

that is,
the cross

relation

the cros
tions, l

ships at

81

And Since rUT = rTU ,

rle2 = rTU = rUT = rY1X2
rat is, the cross-lagged correlations are equal to each other. In fact,
he cross-lagged correlations in this model satisfy an even stronger

elation. Since
rxlyl = I‘Tyl = rTU
rsz2 = rTYz = rTU

e cross-lagged correlations are actually equal to the static correla—

ons, rxly2 = erYl' For three time periods these symmetrical relation-
ips are:

T1 = T2 = T3 = T and U1 = U2 = U3 = U
115,

rxly3 = rx1U3 = rxlu = rTu

ry1X3 = rY1T3 = rYiT = rUT = rTU = rxiys
rxly3 = rTU = rxlyz

rY1X3 = rUT = rY1X2

rx2y3 = rx2U3 = rsz = rTU = ry2X3
rx2y3 = rTU = rxzyl = rX1Y2 = rX1Y3

it is, for all i and j

r = r . . = T . = r U = rTU
ice the cross-lags are all equal. But actually this formula h01d5 for
= j as well as i # j, and so the static correlations are also all
1&1. Therefore, if there is perfect measurement, the cross—lags are

.equal to each other (i.e., they are symmetrical) and are equal to

>static correlations.

 

"error 0

the stat

That '1:
Each 0
other

Simple

PNdil
study
Perio

model

when

art

82

If the variables in this model are less than perfect, i.e., have

"error of measurement" of the simple kind, then for

Xi = T + 81 and ya = U * Ej

the static correlation and the cross—lags are given by
th + ei)(U + a.)
I‘ , . = -—____J__
xlyJ n0 0
Xi Yj
= ETU + zeiU + ETEj + Zeiej
n
Ox.0 .
l YJ
ZTU
n
oxioyj
_ OTu

O _O _
XIX]

OTU
oxoy

 

That is, each rx_y_
1 J

each other. Thus the fact that the cross-lags are all equal to each

is equal to the same constant and hence equal to

Other and are equal to the static correlations is not affected by adding

Simple error of measurement or "error" due to transient factors.
However, for the mood model, the cross-lagged correlations are

Predicted to be less than the static correlations for a two time period

StUdY (r ). But what about a three time

X1Y2 = rx2Y1 < rX1Y1 = rX2>’2

Period study? If there is no change in true score over time, the mOOd
model can be written

—. . .=T+h.+e-
Xi—Tl‘l'hl’f‘el l 1

Vi = Ui + hi + ei = U + hi + 5i

a are defined as the values of T, U, h, e, and e

vhere Ti, u. h. ei’ i

1, 1’

- - ' bles
it time i. After eliminating the terms involv1ng uncorrelated varia ,

 

 

For cro

WRONG

ndth

That i

That
Slum

the j

tin

alt:

83

 

T + h. + . U h. .
rx. . = Z( 1 e1)( + J + SJ)
1yJ nOX 0x m
i j”
noxoy
since = and = .
oxi ox Oyj 0y

For cross—lags, i and j refer to two different times so that the mood

components, hi and hj’ are uncorrelated. Thus for cross-lags

r : ZTU
xin noxoy

 

OTU

OX0),

if i g j

 

and the cross-lags are all equal to each other.
However for static correlations, i and j refer to the same time.
That is, for correlations, j = i, so that

_ ZTU + Zhihi

xiyi ncxcy
OTU * 02h

OxOy

2
= OTU +70 h
oxoy oxoy

 

 

That is, the static correlations are always equal to each other, but are

SPuriOUSly higher than the cross-lag correlations by the amount given in
the second term of the last equation.
In summary, the mood model predicts that
r13 = r12 = T23 < r11 = r22 = r33'
hese relationships will be used to test the data for three administra-

3i0h$ to the Firm A managers for the presence of real change in their

mtitudes.

 

lo sin
by the
Furthe
assume
in the

comb

then

T0(

3C0

Th1

84

Real or Cumulative Change

In this section real change is assumed in the observed variable,
.e., a change in the true score. To simplify the discussion, first
onsider only true scores. Let the change in true scores be given by

AT = e

o simplify the discussion, assume that the variable T is not affected
y the other variables in the system; i.e., consider a univariate model.
urther assume that the change in T is not related to T itself; i.e.,
ssume that the change in T is a simple accumulation of random events
n the person's environment. After this over-simplified model is
ompleted, the effect of alternate assumptions will be considered.

From the equation AT = 5 these specific equations can be derived

AT = T2 — T1 = 51

AT=T3—T2=€2

T = T1 + (T2 - T1) = T1 + e1
T3 2 T2 + (T3 — T2)
T = (Tl + £1) + 62
T3 = T1 + 61 + 52

> contrast this model with the previous model of no change, "The" true
:ore is identified as the value of the true score at time 1. Then

T1 = T

T2 = T + 61

T3 = T + 81 + 82
e difference is that in the no real Change model, the new value was

- and
tained by erasing the old e (aggregate 0f tran51ent factor

_ . n e" from
reliability) and replacing it by a new 6: 1-6" the Chang

 

115:3?”

tine l t
time 2 t
is intrc
Don‘s II

cuiul at

them

test-j

 

 

85
time 1 to time 2 was mathematically obliterated before the "change" from
time 2 to time 3 was introduced. In the real change model, once the 51
is introduced, it stays in the equation. This mathematical fact corres-
ponds to the verbal statement that "real change should be cumulative or
cumulated over time."
What then are the correlations? Since xi = Ti’
_ 2T1
noTloT2

T
2
rx1X2
DOTloTZ
ole
0102

 

where 0T1 is replaced by Oi' If we also simplify the notation for the

test—retest correlations from rxlxz t0 r12, then
01
r = _—
12
02
2T T
Similarly, rX2x3 = 5"ELT2'
0T20T3

= 2T2(T2 + 62)
I'ICITZOT3

02T2

0203

 

and hence

Q
N

|

r23 =

Q
04

 

how an

SUIpri

That 1

0T thl

time

These

but c

his

on

Thu

 

the

 

86
Finally,

ETiTs

 

I‘
XIXS
nOT10T3

ZT1(T1 + 61 + 62)
n
6H@%
2
0 T1
0103
and hence
_ 01
r1 —

3 a

How are these correlations related to one another? The key relation is

surprising, but simple:

r r -01 02-Ol-r
12 ~ 23 O2 ' O3 03 13

That is,
r13 = r12 ' r23

or the time 1, time 3 correlation (r13) is the product of the time 1,
time 2 correlation (r12) and the time 2, time 3 correlation (r23).
These correlations would be fractions (i.e., 0 < r < l) in any context
bUt one in which there is no change. Since r23 < 1, this means that

r13 = r12 r23 < r12 - 1 = r12
Thus if there is real change in the system, r13 < r12?’ That is,
r13 = T12 only when there is no change between time 2 and time 3.",
Similarly, since r12 < 1,

r13 = r12 r23 < 1 - r23 = r23
< r23; the time 1, time 3 correlation is less

”11.15, 1‘13 < r12 and T13

than either the time 1, time 2 correlation or the time 2, time 3 corre-

lation. This corresponds precisely to the verbal statement: "If there

 

 

is teal
3, then

tineSt

in the
change
tion is

:onside

there:

produc

the ‘

87
real change from time 1 to time 2 and real change from time 2 to time
then there is more change during the total interval from time 1 to
w 3 than there is in either subinterval."

Does this conclusion depend on the simplifying assumptions made
the particular model above? First, remove the assumption that the
nge in T is independent of its initial value. A more general assump-
n is that the change in T is in part a linear function of T; i.e.,
sider the regression equation

AT=0LT+e
re: a = the regression coefficient of change on initial score. This
iuces
T2=T1+AT=T1+0LT1+€1
=(1+a)Tl+gl
T3=T2+AT=T2+0LT2+€2
=(1+a)T2+€2
= (1+0L)[(1+0t)Ti+51]+€2
= (1+a)2T1+ (1+0‘)€1+62
. can be arranged for contrast by identifying ”the” true score with
true score at time I. Then,
T1 = T
T2 = (l + a) T + 51
T3=(1+a)2T+(1+00 61“62
6 are two critical features of these equations. First, in compari—
to the simplified real change model, it is shown that T is fiTSt
iplied by (l + a) and then by (l + alz- In the case Of 5331

BSSion to the mean,” the constant would be negative and SO (1 + at)

 

 

 

would b

true SC

Tnat is
decree!
me o
the to

n goes

111th

WOU1(

ms
for
The
tial
. ft‘

she

88

uld be less than 1. For fractions which act as multipliers of the

ue score,

1 > (1 + 0.) > (1 + a)2 > . .
at is, if a is negative, then the multiplier of the true score

creases as a function of the time interval involved. Thus, the influ—
:e of the initial score on later measurements steadily decreases and

3 correlation between T1 and later measurements Tn would go to zero as
goes to infinity.

If there is no real regression to the mean, then a is positive.
a > 0, then (1 + a) > 1 and hence

1 < (1 + a) < (1 + a)2 < .
this case the influence of the initial score T on later measurements
dd decrease, but not to zero.

The second and most important feature of these equations for
sent purposes is the fact that 81 continues to appear in the equation
‘TZ' This again reflects the fact that real change will cumulate.

multiplication by the constant (1 + a) only means that the ”cumula-
n" is not a simple additive process. What about the correlations?
er some routine but tedious algebra, the test-retest correlations are
wn to be
(1 + a) ole
OTICTZ
(1 + a) C1sz
r23 = 0'1"on3
(1 + “)2 O2T1

0T1°T3

The ”pr

m this

That

i The 1
Univ
ml

real

prc

m

 

 

89
a "product rule” that held for the simplified model can now be tested
this more general model,

(1 + a) 02 (1 + a) 02
1 2
0102 0203

 

r12 r23 =

(1 + d)2 012 022

0102203

(1 + a)? 012

0103

= 1‘13

r13 = r12 - r23
product rule still holds! Thus, the product rule still holds for a /
variate model in which the measurement of the variable is perfect and
1 Change is cumulative. Again the immediate implication is that for
1 change,
r13 < r12 and r13 < r23.

A continuous variable whose test—retest correlations satisfy the
duct rule is called a linear Markov process. A test-retest correla—
1 matrix for one variable that satisfies the product rule is called
lttman Simplex (Guttman, 1954)-

Now suppose that the assumption of perfect measurement is

)ped. That is,

re e is the sum of the random component due to unreliability and the

' we have
10m component produced by tran51ent factors. For true scores

product rule

I = T = rT T
T1T3 T1T2 2 3

 

obtain

attent

mere

COITE

ties

and]

Int

90

What about the observed variables? Actually these are quickly
led from the classic reliability formula for "correction for

lation."

I‘xy = rXT rTU rUy

T is the true score for variable x and U is the true score for
)le y. If the right reliability coefficients are used, then the
lations rKT and rUy can be expressed in terms of the reliabili-
Df x and y. In the present context this means that the coefficient

ability is used to write

rxT = ‘/ rxx and rUy = ‘/ ryy

en CG

rxy = M rxx rm“ ryy

e present case, this yields correlations in the form
15‘13‘2 = er1x1 rT1T2 V rXZXZ
rx2x3 = V I‘XZXZ TTZTS V rx3x3
rxlxs = V rxlxl rT1T3 V I‘x3x3

tution into the product rule yields,
r = r r T I‘x x I'x x rTzT V IX X
X2 XZX3 xlxl T1 2 2 2 2 2 3 3 3
/ r I‘ r
rxlxl rxzxz TlTZ TZTS X3X3
r r (I 1'
I.szz V x1X1 TlTS x3x3

rx2x2 ' rXIXS

 

 

 

Thus th

i.e., L

Sou Si

para
dras

ohse

91

s the product r is not equal to r

. unless r

X1x2 rxzxs X1X3 X2X2 = 1;

., unless the measurement is perfect. Instead, we have

1‘x x - rx2x3
rx1x3 = l 2

erX2
since
rx x l, 1 l
2 2 --—-
rxzxz

it is shown that

rX1X3 rxlxz rX2X3
thermore this inequality does not yield the hypothesized inequali-

; between r and

X1X3 rxlXZ or Does the hypothesized relation

rx2x3‘

1? Or to phrase it negatively, can there be a combination of

imeters in this model for which the reliabilities are changing so

stically over time that the rank order of the correlations between

2rved scores is inverted from the rank order of the correlations

:he true scores? There are combinations which eliminate one

[uality or the other. However if the variance of observed scores

‘eases over time, then r13 r12 (but not necessarily r13 r23). If
variance is increasing from time to time, then r13 r23 (but not
ssarily r13 r12). Of major importance to the present study is the
Where the variance stays the same across time. In this case

rxlxl = rXZXZ = rX3X3 = constant

hence

TXIX3 = ‘/rx1x1 IT1T3 V XSXE
= ‘/rx1x1 rT1T2 TTZTS V rx3x3

 

 

 

If the:

ad in

If th

3C1

d0!

 

92

‘lr

= r . r X3X3

V rxlxl Tsz \er2X2 T2T3 V/;;1:__
2 2

 

re is real change, then r 3 < l and so r <

T2T X1X3 IXIXZ

similar fashion

_VrX1X1

r
x2X2

= I‘ " I‘
T1T2 X2X3

re is real change, then IT T2 < l and so
1

 

r " r r " r

rxlxs < rxsz
f the variance is not changing, then both inequalities hold.

The one simplifying assumption in this model which has not been
d is the assumption that a particular variable is not influenced

other variables in the system.

If there is some (possibly unobserved) variable that does inter-
usally over time with the observed variable, then the product rule
at hold and the inequalities might or might not stand up depend-
‘the nature of the interaction. A general discussion of this case
Pnd the scope of the present paper. However, in the case where
hiances of the causally interacting variables do not change from

time and where the correlations are also constant, there is a
t theorem. Under these conditions, let ri. = the correlation

J

the value of T at time i and time j. Then

 

 

 

In par

Or not

oatic

r13 (

cméi

IESI

hr:

the

93
rin —-)0 as n9”

trticular this means that eventually rin is less than r12. Whether
t the model predicts an immediate decrease (r13 < r12) is not

at the present time, though for most sets of possible mathe-

al parameters it would be true. That is, the predictions,

r12 and r13 < r23, have been shown to be plausible under these
tions but they have not been proved. Thus if the data showed, say,
r12, it would not absolutely rule out real change in a multi-

te interactive system.

The Test for Real Change
How then can the hypothesis of real change be tested given the
retest correlations for three measurements on a given variable?
the strong test is applied: Does the product rule hold? If
r13 = 1"12 ° 1‘23
:here is strong evidence that

1. the coefficient of equivalence is 1.00;

2. all observed change is real;

3. that part of the real change which is not attributable to
differences in initial value is attributable to nonrecurrent
random factors; i.e., the observed variable can be studied
in isolation.

If the variable is not perfectly measured, the product rule will

ld for the observed test-retest correlations (i.e., for the
d variables). However, if the coefficient of stability is known,
e test—retest correlations should be corrected for attenuation.

ulting correlations are the estimated correlations between true

 

 

 

SCONS

ad t}
negaz
varia

KEYS

then

posi
vari
ore
r13
str

obs

 

94

)res and those correlations should be tested for the product rule.

Suppose that the product rule does not hold. If r13 < r12 r23

1 this is not due to sampling error, then there must be strong
:ative interaction between the observed variable and some other
iable. Thus there is real change but its nature depends in critical
3 on some other variable that may not have been observed.
On the other hand, suppose that r13 > r12 r23. The question

n becomes: how much bigger? In particular, is r13 so much bigger
n r12 r23 that r13 > r12 or r13 > r23? Suppose that both statements
true, i.e., r13 > r12 and r13 > r23. Then there should be a strong
itive interaction between the observed variable and some other
iable. A second test for this is that there should be a sizable in—
ase in the variance of the observed variable over time. Thus if

> r12 and r13 > r23, then there is real change but its nature is ”I
angly determined by some other variable which may not have been

rved.

The cases for which r12 r23 < r13 < r12 and/or r12 r23 < r13 <
are more ambiguous. There is real change in the observed variable,
unless the coefficient of stability is known there is no way to
53 the relative contributions of instability and possible inter—
ng outside variables.

Finally there is the case r13 = r12 = r23. Here the strong
Umptive hypothesis must be ”no change." If this is true, then

= r23 = r13 is the coefficient of stability for that variable. If

coefficient of stability equals the coefficient of equivalence, then kg,

appropriate model is simple unreliability; both transient factors

ood can be ruled out. If the coefficient of equivalence is larger

 

 

 

chant

affect

COMI‘

obser

ROI?

ohse‘

ad

fact

trat

95

the coefficient of stability, then the observed variable is

:ted by transient factors. Whether or not mood makes a significant

vibution in this case cannot be tested in the data for any one lk*"
,r

Ned variable but requires the cross-lag correlations for two or

variables.

Thus as was noted at the beginning of the chapter, the third
‘vation is critical to distinguish between real and apparent change
hence critical for the differentiation of the mood and general
tr models. The next section will present data from a third adminis-

on of the questionnaire to Firm A.

The Firm A Third Administration Data

The MBO study questionnaire (Appendix A) was administered in
ary, 1972 to the managers of Firm A who had participated in the
and second administrations. Seventy—three (73) questionnaires
sent and fifty—three (53) replies were received. Of the twenty ‘
ers who did not reSpond, it was learned that fifteen managers had
rated, retired, or deceased in the eighteen months' time period {
an the second and third administrations. These responses were
i according to the seven scale research model developed by Chesser
\ppendix B). Then the seven scale results for the three adminis—
)ns were correlated and became the basis for testing whether or not
was real change occurring in the MBO system.

The means and standard deviations of each of the seven scales at
woint in time are shown in Table 5-1. Unlike the total sample for
_ or the sample for Firm B, the subsample from Firm A shows a

L1 pattern of increasing means over time. Two of the variables,

96

able 5—1.-— Means and Standard Deviations for First, Second, and
Third Administrations--Seven Scale Research Model--
Firm A Managers (n=53)

 

Means Standard Deviations
3 Description Time 1 Time 2 Time 3 Time 1 Time 2 Time 3

 

Superior-Subordinate

Relationship 3.17 3.22 3.25 .36 .35 .38
Goal Clarity and

Relevance 2.77 2.97 3.01** .61 .48** .47**
Orientation Toward MBO 3.09 3.36 3.26 .86 .96 1.01
Performance—Reward

Association 3.76 3.70 3.67 .87 .65** .75

Subordinate's

Influence Over Goals 2.73 2.98 2.93 1.19 .83 .93
Satisfaction With Job 3.13 3.12 3.35 .96 .78 .87
Success in Attaining

Goals 2.78 3.09 3.19* 1.28 .35 .69**

 

ignificant difference between time i value and time 1 at .05 level.

ignificant difference between time i value and time 1 at .01 level.

 

 

Goal

the

the

gene

Whit

tin

Goa

 

 

 

97

a1 Clarity and Perceived Success, show a significant increase over the
ree year period from first administration to third administration.
Using a test for the difference in variances (see note below),
e variances (the standard deviations squared from Table 5—1) show a
neral pattern of decreasing magnitude. There are two of the variables
ich show a significant (p < .01) decrease in variance from time 1 to
e 3. They are Goal Clarity and Relevance and Success in Attaining
Is. Also, variable 4--Performance—Reward Association—-has a signifi—
t decrease from time 1 to time 2. These significant differences
gest real change within these system variables.
Table 5-2 presents the correlation matrices for the three time
riod data. These correlations were calculated for the responses to
= questionnaire by the S3 managers from Firm A who were identified in
, three administrations. Change scores are not represented in the
rix.
The product rule test for real change requires that the time 1,
e 3 correlation (r13) for an MBO variable be less than either the
e 1, time 2 correlation (r12) or the time 2, time 3 correlation (53).
ires 5-1 and 5-2 are used to test these predictions. In Figure 5—1,
time 1, time 3 correlations are plotted against the time 1, time 2
relations. The variables-~Subordinate's Influence and Perceived Suc-
—-were included in the graph even though they are not used to
mate the general factor. Two variables——Goal Clarity and Relevance

Performance—Reward Association, respectively——do not meet the

 

The author is indebted to Professor John Edward Hunter, Michigan

State University, for the derivation of this test.

 

 

 

98

 

em. n

KN. u

”.H mo odd; ucmofimacwfim

Ho>oH Ho.

Ho>oH mo.

| \r .

 

 

 

 

wmooosm po>floohod

new spa: :OHuomMmHumw

monosfimcH m.opwcflweonom

coauwfiuomm< vhmzom-oo:mfiaomhom

om: pnmzoe COHumpcoHHo

oocm>oaom wed quHwHu Hmoo

”$598323 om opmcfluho smnaoflomsm
:OHH Haemoo oHLdHHm>

 

 

 

m.oEHH :H mucHom own

I|||||||kBthiIIblllillllllllllll!

 

 

 

HN Va 5
ON MH 0
ma NH m
wH HH v
NH OH m
0H m N
mm m H

 

mm.N.H QEHHU

 

Honfioz oHQmHHm>

 

 

 

:u um mofinmflhm> Eoumxm noosuoo mmHew:oHsuHo« ):.

 

 

 

aoH em 0H m. m A RH. eH o. H v en m- m m m mH- o om .Hm mm
em aoH oH «H n. Hm. H. o- 0 mm 0. NH DH an OH mH m- V. cm .m on
0H 0H oeH DH mm mH. A- Hm. N 3 mm m. mH mm mm a e- mm mm ”me mo
m. «H oH OOH vm em mm: H». e .Hn mm oH mm no 0H mH a H- He ma om
m n. mm em cOH em m. 0H. Hm Hm nN ow Km ow H» mm oH mH mm mm an
n Hm: mH' vm om ooH n NH. aH WH vm we v mu H. m m: NH m a. on
«H. H. 5. mm- m. m 00H as m e m o- m- 0 0H 0. 0H s m. a. m
«H a. Hm- Hm. 0H. NH. oe oaH HH an- A. mH. 0H- NH: n- VH- wH o- a. HH NH-
0. o a e Hm OH 0 HH aoH H nH a. o Hm mm AH A m- m. “NH mm
5 mm o H» Hm NH e an. H aaH cm on oH mm NH mN VH em Hm mm mm
v a. mm mm mm em m a. nH mm ooH on on om NH mm we mm vH .mm mm
.vn NH nu ow om me an mH. u- on on ooH m H. H n v- m- mm m mm
m. OH mH mw mm v m. 0H. 0 OH on m ooH no we «0 He mm so am e
m on mm mm om n- 0 NH. H» mm om H- no ooH om me am am Hm .Hm om
m OH on 0H Hm H- 0H m. mm RH NH H ve mm aaH H» m mm mm an me
n mH s we mm m a. ¢H. NH mm mw n ea we Hm ooH mm mm mm mH o-
mH. N. e. m OH on 0H 9H A wH ev v- Hv om n mm omH 0H mN mm v
o «- mm H. 0H NH a o. n- «N mm m- Nm nm mm mm 0H ooH am an mm
on on am He mm m m. a. m. an ¢H ma ov H» mm mm mm om

Hm m me we mm e. H- OH NH we on m cm am 5» mH mm Hm

no mm mm on an an m NH. am mm mm mm 0 cm me o- v mm Mm me ooH
Hm VH a ow MH 0 9H NH m nH HH v NH aH q 0H 0 m mu m H

upcoMoq

J

Hm
vH
N

.nmv-I
HN

lm‘nmflkfl‘r‘m'“n'n
.4 HH HH ‘4"

gmvnn
H

,xaak<z-m amymgaomm

 

 

 

99

5—1.-- Scatterplot for Time l,Time 3 versus Time l,Time 2

 

e
Cross-Legged Correlations for the MBO Variables which
Estimate the General Factor - Firm A Managers (n = 53).

l,Time3

-Lagged

lations

 

 

 

Time 1, Time 2 Cross-Legged
Correlations

 

e Description

Superior-Subordinate Relationship
Goal Clarity and Relevance
Orientation toward MBO
Performance-Reward Association
Satisfaction with Job

Subordinate's Influence
Perceived Success

variablesare not included in the estimate of the general
r.

 

1qu

the

sug

 

 

100

rement for real change. The other variables do show a real change
rrelation. However, in Figure 5—2, all of the M80 variables meet
riteria for real change. Thus, the test—retest correlations
at that in each variable separately there has been real change.
Is there real change in the general factor? There are several
_n which the three time period data can be used to answer this
.on. First, the MBO variables which make up the general factor as
tted in Chapter Two have demonstrated real change when considered
Ltely. However this could mean change in the Specific factors
it change in the general factor. The test for the general factor
test of r13 < r12, r23 for the cross-lagged correlations. This v'
s displayed in Figure 5—3. In this figure, the time 1, time 2
and time 2, time 3 cross—lagged correlations (r23) have been
8d for each of the MBO variables. These averages are then used
independent variable in the scatterplot and the time 1, time 3
ation (r13) for each variable is the dependent value. This plot
equivocal results. For the variables with low average time 1,
and time 2, time 3 correlations there is an indication of real
[i.e., r13 < (r12 + r23)/2]. However, the variables with the
1 average correlations are approximately equal to the time 1,
correlations which suggest no change in these variables.
A second test for real change in the general factor utilizes an
d static and cross-lagged correlation for those variables which .,
the general factor within each of the three administrations.
, by averaging the off-diagonal static correlations among the
riables which represent the general factor, an average static

tion can be determined for each time period. Similarly, by

 

101

gure 5-2.—— Scatterplot for Time l,Time 3 versus Time 2,Time 3
Cross-Lagged Correlations for the M80 Variables which
Estimate the General Factor—-Firm A Managers (n=53)

 

 

 

 

.7
.6 5"
Time 3
t'o
1 ns +3
.4
4
+ 1
+
.3
6
z++
.2
o?
.1
(96
0.0 L ‘ 4 L I
o 0 1 2 .3 4 5 6 7

Time 2, Time 3 Cross-Legged
Correlations

 

Description

Superior-Subordinate Relationship
Goal Clarity and Relevance
Orientation toward MBO
Performance-Reward Association
Satisfaction with Job

Subordinate's Influence
Perceived Success

riables are not included in the estimate of the general factor.

 

 

 

102

-3.-— Scatterplot for Time l,Time 3 versus the Averaged Time 1,

Time 2 and Time 2,Time 3 Off—Diagonal Cross-Legged
Correlations for the MBO Variables which Estimate the
General Factor-~Firm A Managers (n=53)

 

/,
/
/
.6 _ 1,17
9/
//
. '5 ' ,/
me 3 /
gged /
'Lons .h .
. //
+ / +
+ /<
.3 _ / ’
/++

 

.2 .3 .h .5 .6 .7

Averaged Time l,Time 2 and Time 2,Time 3 Off—Diagonal
Cross-Lagged Correlations

 

 

 

 

 

103

Lging the off—diagonal cross-lagged correlation for these same
.bles within a particular cross-lagged matrix, an estimated cross-
d correlation can be detenmined. The estimated static and cross—
d correlations for the general factor are found in Table 5—3.
In the table it is shown that the averaged time 1, time 3 corre—
n (513) for the general factor is less than either the averaged
1, time 2 correlation ($12) or the averaged time 2, time 3 correla—
(353). That is,
T13 < T12 M 3:23
is the requirement for real change in the general factor. But
is actually the same evidence as shown in Figure 5-3, and that
2 showed that this test is equivocal. One—half of the component
>les of the general factor show real change and one—half of them
The averaged static correlations increase in magnitude as a
on of time; that is, ihl < T32 < r33. Since the static correla-
s a function of the ratio of the relative strength of the general
, there are two ways in which the static correlation might ‘
se-—one, by an increase in the variance of the general factor, and \
y a decrease in the variance of the specific factor. This poses
eresting question: "Does the variance of the general factor go
does the variance of the specific factor go down?” The answer to
s a test of the differences in the variance of the general factor.
ately the general factor is not actually observed. Therefore a
test cannot be performed. However, since five of the MBO vari—
re highly correlated with the general factor, each can act as an
or of the general factor. Thus the sum of the five indicators

give a reasonable estimate of it. Since the variables differ

 

 

 

 

 

Table 5-3.-- Matrix of Averaged Static and Cross Lagged
Correlations for the General Factor During

the First, Second, and Third Administrations—-

Firm A Managers (n=53)

 

Variable .1
1. General Factor — time 1 .21
2. General Factor - time 2 .23
3. General Factor - time 3 .16

IN

.23
.30

.25

[m

.16
.25

.37

 

Significant value of r:
.05 level = .27

.01 level = .34

 

 

 

 

105

ly in their reliability and their correlation with the general

r, the variables were summed at the level of items rather than as
ard scores. That is, the general factor was estimated by pooling
tems from the five highly saturated scales into one large test.

est was then scored for all three administrations. It should be

that this estimated general factor is actually a variable in its
'ght and has its own specific factor. That specific factor is a
ed sum of the Specific factors of the five MBO variables and

gh it is relatively smaller for the estimated general factor, it
ot be absent. Table 5-4 contains the means and standard devia—
for the estimated general factor at each of the three points in

There is no significant difference in the variance of the general

over time. This implies that the increasing static correlations
: variables that load highly on the general factor may be due to a
me in the variance of the specific factors. At present there is ”
' to test this.

Another test of change in the general factor is to compare the
correlations for the general factor with the cross—lagged corre—
s for that factor. That is, from the table, it is observed that
d $52 are greater than $12‘ Also, r22 and $33 are greater than
This is the pattern of relationships expected in the mood model
the transient factor spuriously inflated the static correlations.
Lf there is no change in the general factor, then the static

Ltion may be spuriously high due to a transient factor such as

A third test for real change is to directly calculate the test- /

correlations for the estimated general factor. To do this, the

106

able 5-4.-— Means and Standard Deviations for the General Factor
and Changes in the General Factor During the
First, Second, and Third Administrations—-

Firm A (n=53)

 

 

Standard

:ale Description Means Deviations
1. General Factor — Time 1 3.096 0.35
2. General Factor — Time 2 3.193 0.34
3. General Factor — Time 3 3.224 0.39
4. Change in General Factor 0.096 0.34

(Time 2 - Time 1)

0.031 0.31

5. Change in General Factor
(Time 3 — Time 2)

 

 

107

sponses from the fifty-three Firm A managers to the five scales which
timate the general factor were scored as one molar factor. In addi—
on, change scores (time 3 - time 2 and time 2 - time 1) were calcu—
ted. These scores were then used to calculate test-retest and impact
rrelations. The matrix for these correlations are found in Table 4-5.
e test—retest correlations for the estimated general factor when
rrected/for attenuation provide a test of the product rule. That is,

= test-retest correlation between time 1 and time 2

r
GlGZ for the estimated general factor

erG3 = test—retest correlation between time 2 and time 3
for the estimated general factor

rG G = test-retest correlation between time 1 and time 3
1 3 for the estimated general factor

ai = coefficient alpha internal reliability for the esti-
mated general factor at time i (i = l, 2, 3)
an, the product rule test is:
1‘6ng, 10162 rGZG3

ng the data from Table 4-5

 

. .51

 

indeed
.45 é .43
h satisfies the test for the univariate model.
Could the differences in the test-retest correlations be due to

ce? One test for this would be to examine the differences in the

 

 

 

108

>le 5-5.—- General Factor Correlation Matrix for First, Second,

and Third Administrations-—Firm A (n=53)

 

General Factor - Time 1
General Factor - Time 2
General Factor - Time 3

Change in General Factor
(Time 2 — Time 1)

Change in General Factor
(Time 3 - Time 2)

l
1.00
.51
.40

—.53

-.04

 

Z.
.51
1.00
.66

.46

-.26

2
.40
.66
.23

.56

lb

.46

.23

1.00

-.21

{U1

.04
.26
.56

.21

.00

 

m V

 

 

109

ee test—retest correlations of the estimated general factor for
tistical significance (see note below).

The strategy for the test is to perform a 2 test on the sum of
differences between the test-retest correlations of the three time
Lods. The standard error for the sum of these differences is

O = 6 (1—E) s
n-l

where: 5 = average correlation for the three test-retest
correlations

n = number of managers in the sample

1 for this study

0. M: 6__(-_4§)_=.23
53-1 52

the z score is calculated as

_ (rGle ‘ rGlG3) + (r6263 " rclc3

 

G

= (.51 — .40) + (.66 — .40)

 

z = 1.61
h in using a one-tailed test is on the border for being significant
he .05 level (2 @ .05 = 1.64). Interpreted this means that by a
istical test the matrix of test—retest correlations of the esti-
d general factor is not "flat"; i.e., there is a significant differ-

between the test—retest correlations. This is support for the real

   
    
 

e hypothesis at the .05 level.

The author is indebted to Professor John Edward Hunter of

Michigan State University for the derivation of this test.

 

 

110

Conclusions

The results of these tests of the real change hypothesis cannot
interpreted without qualification. In the three time period study
ere is evidence, though not decisive, that there is real change in the
served variables of the MBO system for a subsample of the Firm A
nagers.
The means and standard deviations for the seven scales of the
m1A research model do suggest real change for several of the
riables. Also, the analysis of the cross-lagged correlations for the
) variables produced evidence of real change. When five of the seven
riables of the MBO system are grouped in order to provide an estimate
the general factor, the corrected test—retest correlations do support
a product rule test for real change. Also, the average static corre-
Lions for the variables that correlate highly with the general factor
:rease as a function of time, which suggests that more of the variance
explained by the general factor across time. If this increase in
tic correlations were due to an increase in the absolute strength of
general factor, then it would mean a self-facilitory growth law for
general factor. This in turn would mean that the impact correlation
ld be zero or positive and the variance of the general factor was
easing.
0n the other hand, it was assumed in Chapter Three that the real
ge in the general factor was regression to the mean. This was used
xplain the constant variance shown in the data. The data for the
mated general factor does demonstrate regression to the mean and a

tant variance across the three time periods. Why then do the static

 

 

 

 

111
correlations increase across time? One explanation would be to assume
that there is a decrease in the variance of the Specific factors over
time. At present, there is no obvious hypothesis as to why this might
)6 true.
A second qualification refers to the difference between the
average static correlation and the average cross-lagged correlation.

[he fact that the static correlations are larger than the cross—lagged

 

:orrelations could also be due to a Spurious inflation of the static
:orrelation due to the mood variable.

It is important to note that the subsample of fifty-three Firm A
lanagers is deviant from the total sample of seventy—three Firm A
mnagers in the two time period study. That is, the static correlations,
5 well as the dynamic correlations, for the subsample during time 1
md time 2 are different from those same correlations in the total
lample. If this difference were due to the abnormally bad static corre—
.ations for the subsample at time 1, then most of the evidence for real
hange could be attributed to sampling error.

The data does suggest that there was real change in the general
actor for the questionnaire used in this study. However, the test was
eak. The qualifications placed on the conclusions of the test for real
hange can be removed only with the availability_9f_additional data to

rovide a more precise analysis.

 

 

 

CHAPTER 6

SUMMARY AND IMPLICATIONS OF THIS RESEARCH

This research has addressed a number of methodological issues in
1e assessment of attitude in a longitudinal study. These issues center
ton the assessment of real change in the variables of a Management By
'jectives (MBO) system.

Data were collected at three points in time in one organization
irm A) and at two points in another (Firm B). During the replication
'the Chesser study and the subsequent revision of that research model
veral contradictions to the basic assumption of real change were
nsidered and evaluated. The analysis and explanation of these contra-
ctions defined the direction of the research reported here. The
ecific research objective for this project has been to determine
ether the observed changes in attitude for managers who participated

the MBO programs of two large organizations were real or only

parent.

Several mathematical models have been derived to explain the

1 or apparent change in attitudes in that data. One model, which did
the data, assumed real change and represented a general factor of

t real change. This model with its various assumptions did explain
data in the two time period study. At the same time another model,
"mood" model, which assumed there was no real change in the atti—

es of the managers, also explained this same data. It was concluded

t two administrations of the MBO study questionnaire were not

ficient to test either model.

 

 

 

 

 

 

The third administration of the questionnaire to the Firm A
agers provided a test which resulted in the conclusion that real
mge was observed in the MBO system during the three time period
dy.« Then, the model for real change in the general factor, as
ived in Chapter Two, was used as a basis for interpretation of the
a from this longitudinal study.

A significant finding from this test for real change in mana-

ial attitudes was the existence of a general factor which underlies

 

MBO system in both organization samples. The content of this factor
ars to be comprised of at least five of the seven variables
eloped from the MBO study questionnaire that are highly correlated
seem to be appropriate estimates of the general factor. These five
iables are
1. Superior-Subordinate Relationship
2. Goal Clarity and Relevance
3. Orientation Toward MBO
4. Performance—Reward Association
5. Job Satisfaction
The General Factor model assumes that the one general, or molar,
or accounts for all of the observed correlations between the
ables of the MBO system. Another component of the observed scores,
"specific" factor, is assumed to be the residual or that portion of
of the true scores that remains after the general factor is par-
led out. This model also assumes that the Specific factor is not
elated with any of the other specific factors within one administra-
or between administrations. Thus the specific factor may not be a

meaningful one, although it is an important component in the model.

 

 

 

114

he general factor model does predict the three time period data quite
ell and does meet the product rule requirements for real change.

There are several possible interpretations of the general
actor. One is that it is some representation of the manager's satisfac—
ion with life or life—style. Another might be the manager's attitude
award work and the environment in which that work is accomplished. A
lird possibility might be designated as the manager's attitude toward
1e mastery of the job or task assigned. The primary reason that the
:finition cannot be more precisely defined is because the boundaries of
1e MBO system are not precisely defined.

A common denominator for all three of these interpretations for

 

.e general factor is that they are manifestations of attitudes that

11 affect all managers and may be in part outside the identified

stem under study. Also it is assumed that they would account for all

e correlations in the system. Obviously there are other possibilities,

d future research will be directed at the discovery and analysis of

se and other potential general factors.

A second finding which resulted from the consideration of the no
1 change models was the potential for some ”mood" variable to be
sent and influencing the managers of the study. Mood as used in the
el was the consideration of a transient factor or a temporary psycho—
ical condition that acted as an error of measurement in the longi-
inal study of observed scores for the managers. The mood variable
rates as a significant component in the static, dynamic, and impact
relations for the data. Since mood was assumed to be transient and
dom and therefore not correlated between administrations, it does not

e an influence on the cross-lagged correlations.

 

115

Could the "Hawthorne Effect” be the mood variable? No, as the

1
Hawthorne studies indicated, the effect would be a constant for all

 

anagers and would be evident only in the changing means for the observed
cores. Since the means in the variables for this study did not change,
he traditional interpretation of the Hawthorne effect would not satisfy
he requirements for the mood variable.

A second possibility which might define the mood variable is the
situation at wor .” The Litwin and Stringer (1968) study of the
ffects of organization climate (technology, leadership style, rules,

olicies, organization structure, etc.) upon aroused motivation of
1

 

hnagers provides some indication of the determinants of a managerial
hod. These "climate" variables must be transient conditions to be con—
idered as candidates for the mood variable. Also, unless these condi-
ions are individualized, their effects will not cause a change in the
ariance or the correlations for the variables of the system. Thus,
rganization climate is not a candidate for the mood variable.

Although the data from the third administration of the MBO study
estionnaire supported the test for real change and the general factor
del, it is important to note that the general factor model does not
1e the possibility or presence of the mood variable. The two models
3 compatible. If it is assumed that the error term in the general
:tor model is an aggregate of mood and unreliability, the mood com-
lent and its influence in the model can be used in conjunction with
> real change in the general factor to interpret the data. Further
earch will be required to fully test the compatibility of these

els.

”gt—w

 

w

116

Implications of This Research

 

One of the most important contributions of the research reported
ere is the relevance it has for further research into and the actual

ractice of an organization change program such as MBO.

Implications for Practice

It is obvious that the assessment of real change in attitudes of
anagers participating in change programs can have big payoffs for
rganization developers. Real change in attitudes is a difficult
lenomenon to measure. Once measured, it is even more difficult to
iterpret effectively. One conclusion from this project which is very
1portant to organization change Specialists is that at least three time‘/
:riods are required to arrive at a conclusion that change has occurred.

Another significant finding for organization change or develop-
nt programs is that a variable outside the system under study may be
e real source of change within a particular system. For example, the
neral factor, however defined, is the real change component. A change
ogram that does not attempt to identify, or recognize, the potential
such a variable may not achieve the results desired. The same is
ue for a "mood" variable. Future research will attempt to more
early define the relevant variables that are correlated with present
3 variables. This will give more consideration to the identification
the system boundaries.

The fact that the no change hypothesis could not be unequivo-
)ly rejected for the three time period Study in Firm A deserves some
xsideration. For the Firm A managers, an MBO program was implemented

:ween time 1 and time 2. Also, between time 2 and time 3, Firm A

 

 

 

117

mdertook a major reorganization. Yet, the real change in attitudes
bserved in the MBO system was very small. Thus, the resistance to
mange by the Firm A managers must be considerable. The resistance to a
mange in attitude may be represented by these variables, mood and the
eneral factor, which were not included in the original orientation of
his longitudinal study. Organization development programs should
ecognize the existence of factors such as these and direct some atten—

Hon toward the real change in these variables.

Implications for Further Research
‘ This research project is not an end itself, but a means to a
Ftter understanding of changes in the attitudes of managers participat-
ng in a program such as MBO. The results of this project suggest
ther hypotheses to be tested in the continuation of this longitudinal
tudy.

One of these future hypotheses will consider the question: "What
re the boundaries of the MBO system?" This question is based upon the
Vidence that the causal factors for change in both the general factor
d the mood model were not explicitly defined as entities in the
iginal research model. This being the case, future research is
quired to define the existence of the general factor and the mood
Ctor and to ascertain whether they are in part or in total a part of
3 explicit system. That is, are home, family, or community (political,
rvice, and social organizations) a part of the system or just the work
:anization and MBO? Until the boundaries of the system are defined,

general factor cannot be comprehended. If the questionnaire used

e were incremented to assess the managers' commitment to these areas

 

 

 

118
of interest, the important parts of the boundaries of the system could
be identified.
Future research must take into consideration some deficiencies
in the instrument. Revisions would include

1. Some attempt to develop a set of items to measure mood
directly.

2. Items which assess how MBO assists a manager in working
with his subordinates.

3. Items to augment those scales which have poor internal
scale reliability.

Firm B has granted permission to administer the questionnaire
for a third time in the Spring of 1973. The third administration data
;hould provide an important test of the real change hypothesis. Not
)nly will the sample be larger than that for the Firm A managers, the
’irm B managers could be pooled with the Firm A managers for a better
est of the three time period data.

In summary, the research reported here is an important step in
etter understanding the impact of organization development and change
rograms and, in particular, Management By Objectives, upon the atti—
ldeS of managers. The development of the mathematical models of
:titude change has explained a large amount of empirical data. Most
mortant, a theoretical framework for further longitudinal studies has

en provided.

 

 

 

 

APPENDICES

 

 

Note:

1.

APPENDIX A

MBO STUDY QUESTIONNAIRE

Items 1—47 common to administrations at time 1 and
time 2 for both Firm A and Firm B.

Items 48-55 administered at time 1 and time 2 to
Firm B only.

Management By Objectives Study

answer the following questions as truthfully as you can.
ionswer questions in o truthful and careful manner.
two academic researchers conducting this study.

The success of this study depends on your willing-
Your responses will be held in the strictest confidence
The company will receive only summary data concerning this

raring questions having to do with the Management by Objectives, assume the question is referring to lost
MBO effort unless the question specifically states this is not the case.

at, in your. opinion, was the level of difficulty Extremely difficult
o the objectives set for your position? Quito diff'cult
I

Moderately difficult
Not too difficult

—_____...

 

 

 

 

 

Easy

at, in your opinion, was the level of difficult Quite difficult
o the personal development objectives set or - -

you? Moderately difficult

Not too difficult

Easy

No personal development objective

what Extent did the objectives set for you un- ‘- To a very great degree

der MBO reflect the most serious and press-

. T 1d
ing needs of your department and the company? 00 grea egree

To a moderate degree
To a minor degree
Did not focus on any real needs of department or company

What degree did the personal development ob- To a very great degree
jectives set for you reflect your personal de- Too great degree
7 .

Velopment needs ' To a moderate degree
To a minor degree

Did not focus on real deficiencies

w often were you given feedback on your pro- ‘ VOTY f’OQWMIY
grass on your objectives ? Frequently
Occasionally

Rarely -.
Nnvm

To (A Very umut (lemma
l u (I great degree ____
To a moderate degree ,__
To (1 rmrmr dam".
Hot at all Liuull‘].bluwd _

What extent were your objectivm. clearly stu-
ted with respect to results expecto

119

 

 

120

hot extent was the relative importance of
>ur various objectives pointed out to you?
I .

To a very great degree
To a great degree

To a moderate degree
To a miner degree

No clues given as to the relative importance of performance goals

wt extent do you feel you control the means
reaching your objective ?-

often were you given feedback on your pro-
iss on your personal development objectives

'hat extent do you feel you had too many
iectlves ?

much emphasis did your boss put on attain-
your personal development objectives ?

did the amount of effort you put into your
last your compare to that of previous years ?

lo relations with your boss at the present
e compare to your relations With him dur-
previous years ?

uccessful were you in attaining the
actives set for you under MBO ?

To a very great degree

To a great degree

To a moderate degree

To a minor degree

Do not control means of reaching goals

Very frequently
Frequently
Occasionally
Rarely

Never

To a very great degree

To a great degree

To a moderate degree

To a minor degree

Not given too many performance goals

A very strong emphasis
A strong emphasis

A moderate emphasis

A minor emphasis.

No emphasis at all

Very much greater
_ Much greater
Somewhat greater
A little less

A great deal less

Our relationship is much improved

Our relationship is moderately improved
No change

Our relationship is somewhat worse
Our relationship is much worse

Performance was much higher

Performance was a little higher than the goals 5 f
e

Performance was about equal to the goals 5 t
e

Performance was a little less th
Performance was much less th

'you in attainipu

uccessful were . _
objectives

tonal development
your ?

improwment was much
Improvement was a little higl
improvement was about
lmpmVoment was a little
improvoment Was much

an lite goals set
an the goals set

higher than goals set
ier than the goals set
equal to the goals set
loss than the goals 3
loss than the goals 5::

 

llll ill

llll

NW NW Ill/l Hill I

 

121

lb ectlves have been set for you for 3979. under
00. How does the level of ehiectivu com-
pare with the level of these goals la st year ?

his year new personal development obiectives
have been set for on under the MBO system.
How does the dizficulty of these obiectives
compare to those of last year ?

ho had the most influence on setting the
ebiectives for you ?

to amount of change associated with my
job is:

0 number of contacts w/persons Much more frequent than contacts w/persons inside my dept.
outside my Dept. are: More frequent than contacts w/persons inside my dept.
Equal in frequency to the contacts w/persons inside my dept.

Less frequent than contacts w/parsons inside my dept,

Much less frequent than contacts w/persons inside my dept.

w much of an interest do you think the com-
pany has in the M80 system ?

in! much of an interest do you think your boss
tea in the MBO system ?

Very much higher ._._..__.

Much higher ______

A little higher __...__t ._
Alum! tine sumo
A little lower
Much lower

Much more difficult
A little more difficult
About the some _________
A little loll difficult m
Much less difficult

My boss had much more influence than l _—
My boss had somewhat more influence than i _______

My boss and .l had about equal influence __
i had somewhat more influence than my boss

i had much more influence than my boss __..__._

Much more than most other jobs at my level
More than most other tabs at my level
Equal to most other jobs at my level

Less than most other iobs at my level
Much less than most other jobs at my level

A great deal of interest

A moderate amount of interest
Some interest

Very little interest

No interest

A great deal of interest
A moderate amount of interest

Some interest
Very little interest

No interest \
N

ch statement best describes H° '°'°lY makes suggestiOns ,0 me
he manner in which your boss He gives me some ideas but i could use muCh more he! \

' f rmin our . p.
win! you m per 0 9 Y Sanguines my boss helps me how to plan to m h N

ob? , lective and sometimes he doesn't, ac 0" °b‘
Generally when l encounter a serious obstacl -\

will suggest ways of overcoming it, o my boss
Generally when a serious obst '

OClO Ol’lsosl Id; \

with my boss and we revise the

iective.

360:: it
5"“05” and the ob.

 

 

lilcli statement lwst (lesi'rllu-s thc prom-”i
difficulty your boss has in measuring

'I
you! ’ir't lur'imltu'rt 7

rich statement best describes the concern
of your boss for your career ?

iich statement best describes the lKind of
feedback you generally get from your
boss about your performance ?

122

My work is too complex to express in terms of
standards of performance

My boss is barely able to determine if I have done
a good io

Sometimes my host. knows enouqh about the work
l rlu in "who u iwlunrnr-ut ulmut my 'wrlm-
mum n, uriirinltmnh lm‘ iluann't

l have some measures of performance in practic-
ally every area of responsibility.

l have verifiable worl< objectives: l mean, at the
date agreed upon, my boss can tell readily
how close l'vecome toaccomplishing my goal.

My boss feels this is my responsibility, not his.

He might discuss career plan with me but views
this outside his responsibility.

He will discuss my long term career objectives
with me if l push him to do so.

We have agreed on specific things l need to do
or my self-improvement.

My boss is interested in my development and views
setting workobiectives as part of this process.

l'm luclry if l get any hint from higher manage.
ment on how well l'm doing my io .

There are too manytimes when l really don't know
what my boss expects 0 me.

The only real feedback about my performance
comes through official channels.

lgetsome specificfeedback about my performance
but l need more.

Much of the information l get about my performance
is objective and not just subiective and this

helps.
w often does your bossask your opinion when Almost always
1 problem comes up that involves your work? Most fth 9'
0 e me
Sometimes
Rarely
what extent you do feel you can influence the de- To a
. a g V . yer
ISions of your boss regarding tnmgs about which y great degree
'ou are concerned ? To a great degree
To a moderate degree
To a minor degree
Not at all
our opinion, how capable a manager is your boss ? E
xtremely Co 5
P0 le
Quite Capable
COPflhle
Not tog COPOble
N°l ccPt'lble

good is your boss in dealing with people ?

n all, how satisfied are you with your 5055 ?

Very effective

Quite effective
Moderately effective
Not too effective

lmffecfin

nVery satisfied
F . cmte satisfied
cirly well satisfi d
l' e

A .ittle dissetis“
Very rlir “led

“soiitlh I

 

a“

 

\

 

 

123

onsidering your skills and the effort you put into
the iob, how satisfied are you with the pay ?

you had a chance to get a much better paying iob
working for another company in this area, how
would you feel about changing ?

your opinion, to what extent will your actual iob
performance affect your future salary increases ?

your opinion, to what extent will your actual job
performance now effect your future promotions ?

general, how much time did your boss devote to
the M30 system during 1970?

o

:i had the most influence on setting personal de-
velopment obiectives for you 7

‘ your boss indicate any priorities for your per-
sonal development objectives?

i well do you like the M80 system .7

general, how applicable do you think the MBO
system is to your io ?

I helpful has the MBO system been to you in
”trimming the duties of your job?

Very satisfied
Quite satisfied
Fairly well ultisflnll
A little rliqurtinliml
Very illneutiefiml

i would strongly prefer to stay here
lwould somewhat prefer to stay here
l would have 0 hard time deciding

l would somewhat prefer to change

l would strongly prefer to change to the other company

To a very great degree
To a great degree

To a moderate degree

To a minor degree

lt will not affect it at all

To a very great degree

To 0 great degree

To a moderate degree

To 0 minor degree

They will not be related at all

A great deal of time
Quite a bit of time

A moderate amount of time
A small amount of time
Very little time

My boss had much more influence than I

My boss had somewhat more influence than l
My boss and I had equal influence

l had somewhat more influence than my boss

l had much more influence than my boss

Ye S
No

I like it very much

I like it pretty well

l like it in some ways but not in others
l don't like it very much

l don't like it at all

Very applicable

. Quite applicable
Foirly opplicoble

Not too applicable

Not at 0” applicable

Very helpful
Quite helpful
Fairly helpful
Not too helpful
Not at all helpful

 

__...———

l

l

 

124 . .
r interesting is the work in your present iob? Extremely interesting
Quite interesting
Fairly interesting
Neither interesting nor uninteresting
Not at all interesting
:h of the statements best describes Received only praise with no criticism
to amount of praise you received Received mostly praise with iust a little criticism
om your boss about your performance Received about an equal amount of praise and criticism
5‘ year ? Received mostly criticism with iust a little praise
Received only criticism with no praise
concerned do you feel your boss would be if you Very concerned
Iiled to achieve the objectives established for Quite concerned

. . . . 9
our lab to 0 Significant degree . Somewhat concerned

Just slightly concerned
Not at all concerned

t kind of criticism would you receive from your Extremely severe criticism
055 if you failed to achieve the obiectives es-

Q ite s ver crit‘c'
Iblished for your job to a significant degree ? U 8 e ' ism

Somewhat severe criticism
Mild criticism
No criticism at all

important is it for you to know what your boss Extremely important

7 n l
ants you to do . Qurte important

Somewhat important
Slightly important
Not at all important

important is it for you to have definile P°llCle5 Extremely important
id procedures to help you in Performing your job ? Quite important
Somewhat important
Slightly important
Not at all important

your boss establish priorities for your perform- Yes
we sea 5 ? No
1 your performance goals were estab- I felt I had more than a 90% chance of attainment
shed, what did you feel about the pro- I felt l had about a 75% chance of attainment

:bility of their attainment? I felt I had about a 50% chance of attainment
. i felt i had about a 25% chance of attainment
I felt l had less than 0 "1% chance of attainment

 

llll

l

‘ l

Ill/l II If!!! Mill Hill

 

 

125
N satisfied are you with the present amount of in- VOIFY 30'lele
Fluence you have on the decisions of your boss Qurte satisfied
that relate to your work? Fairly well satisfied
A little dissatisfied
Very dissatisfied

v important is it to you that you do a better iob Extremely important
han other people who have or had your iob ? , Quite impoflom
Somewhat important

Slightly Important

Not at all Important

'our opinion, to what extent will effort increases To a very great degree
in your part load to increases in the level of your To 0 9,00, degree
a f n ?
per 0mm co To a moderate degree

To a minor degree

They will not be related at all

what extent do you experience a feeling of per- To a very great degree
:rrlleatiz:omplishmelnt and satisfgction in fully Tea great degree
' p 9 your 9°° assignmen s ' To a moderate degree
‘ To a minor degree

No feeling of personal accomplishment and satisfaction

in your present situation in life, rank Opportunity to use one's skill
he following items in order of their. Opportunity to experience a sense of accomplishment
mpertance, 1 thru 7, considering l to Salary
be most important and 7 the least Recognition in current iob
mportant. Promotions
Pleasant co-worhers

Job Stability

en your present_situation in life, how important - Extremely important
re uture promotions to you . Quite important

Somewhat important
Slightly important
Not at all important

Met—0"!

ase make additional comments about MBO strengths or weaknesses. suggeStions

' changes and/or improvements will be particularly helpfu1,

_.___________

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX B

CHESSER SEVEN SCALE RESEARCH MODEL

Note: These scales were developed by Chesser using data
from Firm A managers (Chesser, 1971, pp. 9-15).

 

 

 

 

APPENDIX B

This Appendix lists the items in each of the seven scales of the
ioral Model and its corresponding coefficient of internal relia-

y (r11) and change score reliability (rdd).

Aperior—Subordinate Relationship

r11 = .96 rdd = .94

The scale is composed of the following items, keyed numerically
: sample questionnaire in Appendix A:

5. How often were you given feedback in your progress on
your performance goals?

8. To what extent do you feel you control the means of
reaching your performance goals?

9. How often were you given feedback on your progress on
your self-improvement goals?

11. How much emphasis did your boss put on attaining your
self-improvement goals?

21. How much of an interest do you think the company has
in the CPA program?

22. How much of an interest do you think your boss has in
the work planning and review program?

23. Which statement best describes the manner in which
your boss helps you in performing your job?

25. Which statement best describes the concern of your boss
for your career?

26. Which statement best describes the kind of feedback you
generally get from your boss about your performance?

27. How often does your boss ask your opinion when a
problem comes up that involves your work?

28. To what extent do you feel that you can influence the
decisions of your boss regarding things about which
you are concerned?

126

 

29.
30.
31.

36.

43.

44.

45.

127

In your opinion, how capable a manager is your boss?
How good is your boss in dealing with people?

All in all, how satisfied are you with your boss?

In general, how much time did your boss devote to the
CPA program?

Which of the statements best describes the amount of
praise you received from your boss about your
performance last year?

How concerned do you feel your boss would be if you
failed to achieve the goals established for your job
to a significant degree?

What kind of criticism would you receive from your boss
if you failed to achieve the goals established for your
job to a significant degree?

lal Clarity and Relevance

1”11

Th

= .90

e

rdd = .87

scale is composed of the following eight items:

What, in your opinion, was the level of difficulty of
the performance goals set for you?

What, in your opinion, was the level of difficulty of
the self-improvement goals set for you?

To what extent did the performance goals set for you
under the program reflect the most serious and pressing
needs of your department and the company?

To what extent did the self-improvement goals set for
you reflect your personal development needs?

To what extent were your performance goals clearly
stated with respect to results expected?

To what extent was the relative importance of your
various performance goals pointed out to you?

Which statement best describes the present difficulty
your boss has in measuring your performance?

Did your boss indicate any priorities for your self—
improvement goals?

 

fizz-(as?

 

 

 

 

 

128

rientation Toward MBO

11

= .80

rdd = .50

The scale consists of the following three items:

39.

40.

41.

How well do you like the CPA program?

In general, how applicable do you think the CPA
program is to your job?

How helpful has the CPA program been to you in
performing the duties of your job?

erformance-Reward Association

r11

= .84 r = .68

dd

The four items in the scale are:

34.

35.

46.

47.

In your Opinion, to what extent will your actual job
performance affect your future salary increases?

In your opinion, to what extent will your actual job
performance affect your future promotions?

How important is it for you to know what your boss
wants you to do?

How important is it for you to have definite policies
and procedures to help you in performing your job?

)ordinate Influence Over Goals

1‘11

= .75 rdd = .50

The scale is composed of two items:

18.

37.

Who had the most influence in setting the performance
goals for you?

Who had the most influence in setting self—improvement
goals for you?

 

 

129

Satisfaction With Job

 

r11 = .58 I‘dd = .35

32. Considering your skills and the effort you put into the
job, how satisfied are you with your pay?

33. If you had a chance to get a much better paying job
‘ working for another company in this area, how would you
feel about changing?

Perceived Success

= .6 = . 0
I11 5 rdd 3

14. How successful were you in attaining the performance
goals set for you under the overall Performance
Appraisal Program?

15. How successful were you in attaining the self-
improvement goals set for you last year?

APPENDIX C

OTHER FINDINGS IN THE REPLICATION OF THE
CHESSER MBO STUDY

Note: The data from two administrations of the MBO
questionnaires for Firm B managers (n=ll7) were
used in this replication study.

 

APPENDIX C
OTHER FINDINGS IN THE REPLICATION OF THE
CHESSER MBO STUDY
This appendix describes the step—by-step replication of the

:sser research using data from Firm B. The appendix begins with a
cription of the fourteen scale model of the MBO behavioral system.

5 section includes a detail analysis of the inter—scale correlations,
1e reliabilities, scale means, and standard deviations for that

e1. The next section presents a similar analysis for the seven scale
e1. The third section describes the use of dynamic and cross—lagged
el correlations for the assessment of causal relationships and to

1d an effects diagram of the MBO system.

Fourteen Scale Model

 

The 55—item, Likert type questionnaire was administered to
agers of Firm B at two points in time. The first administration in
:h 1970 produced 600 completed questionnaires and the second adminis-
:ion in August 1971 yielded 548. These responses were multiple group
:or analyzed using the system of correlational analysis programs
Led "PACKAGE" developed by Hunter and Cohen (Hunter and Cohen, 1971).
Table C-l shows the structure of the fourteen scale research
:1 originally developed by Chesser (see Chesser, p. 10). The scales

resent the following variables in the MBO system.

130

 

 

 

 

 

 

131

Table C—1.-- Fourteen Scale Model

 

 

 

Questionnaire
ale Description Item Numbers
1. Use of goal oriented methods 5,9,11,21,22,36
2. Satisfaction with boss 29,30,31
3. Self—improvement goal clarity 2,4,38
4. Performance goal clarity 3,6,7,24
S. Orientation toward MBO 39,40,41
6. Boss concern with failure 44,45
7. Boss SUpportiveness 23,25,26,43
8. Influence over boss 8,27,28
9. Need for policy 46,47
0. Association between performance and rewards 34,35
1. Influence over goals 37,18
2. Performance goal difficulty 1,10
3. Satisfaction with job 32,33
1. Success in attaining goals 14,15
2f: Chesser, p. 10.

 

.-, "/

 

 

 

 

132

Scale
Number Description
1. Use of goal-oriented methods
2. Satisfaction with boss
3. Self-improvement goal clarity
4. Performance goal clarity
5. Orientation toward MBO
6. Boss concern with failure
7. Boss supportiveness
8. Influence over boss
9. Need for policy

10. Association between performance and reward

11. Influence over goals

12. Performance goal difficulty

13. Satisfaction with job

14. Success in attaining goals
11 of these variables are perceptions of the manager or a reflection of
is attitude toward the goal oriented system.

Table C—2 presents the inter-scale correlations for the first and

econd administrations of the questionnaire to both Firm A and Finn B.

he pattern of correlations between the two administrations is very

 

imilar. To determine any statistically significant difference between
he Time 1 and Time 2 correlations, both are transformed to 2 values (a
ischer r to z transformation). The standard error of the difference

etween the two correlations is obtained by:

 

where: nl = number of subjects in Time 1 sample
n = number of subjects in Time 2 sample
(McNemar, p. 190)
ten the ratio of (21—22) to its standard error (021.22) is calculated to
ltain a 2 value. Mathematically, this is:
z = (z

l—zz)/Ozl—z2

 

133

 

 

 

 

 

 

 

 

 

NN. ON. NN. OH. ON. ON. OO. OO. NH. OO. OH. HN. NH. OH. OH. OH. ON. ON. OH. OH. OH
OO. NO. OO. NH. NO. OO. ON. OH. OH. NO. HO. OH. NO. NO. OO. OH. OO. NO. OO. OH. OH
OO. OH. NN. OH. OO. NN. NO. OO. OO. OO. OO. OO. NH. OO. NH.- NO.- OO. ON. NO. HO. NH
OO. OO. OO.- OO. HN. OO. OO. NO. ON. NN. OO. OH. HO.- OH. OH. OO. ON. HO. OH. HN. HH
OO. OO. NH. ON. OO. OO. OO. NH. OO. OO. OO. ON. OO. NO. OO. ON. OO. OO. OO. NN. OH
NO. NO. OO. OH. OO. OO. NO. OO. ON. OO. ON. OH. ON. ON. NH. ON. OO. NO. ON. ON. O
NO. OO. OO. OH. OO. OO. NO. NO. ON. NO. ON. HN. OO. OO. OO. NO. OO. OO. OO. OO. O
NO. OO. OO. HN. OO. OO. HO. OO. OO. OO. OH. ON. HO. OO. ON. NO. NN. ON. OO. OO. N
OO. NO. OH. OH. ON. ON. OO. ON. HO. OO. OO. OH. HO. HO. NO. OO. OO. HN. ON. OO. O
OHezoe ammuOHOOHHO+ OO. OO. HO. HO. OO. NO. HN. ON. ON. HN. ON. NN. OO. OO. OO. OO. O
NHHHOHO HOOO OOOOEHOOHOO. OO. NO. NO. OO. OO. OO. ON. OO. OO. NO. HO. HO. O

NHHHOHO HOOO ecoEo>OHOeH-OHOO. ON. ON. OH. OO. OO. OO. HN. OO. O

HOHHoOOO OOH: OOHHOOOOHHOOO OO. NO. HO. HO. N

OOOOHoz OOHOOHNO HOOO Oo OOO. H

O O m
NO HO N< HO NO HO NO H< NO HO NO H< NO HO N< H< NO HO N< H< m“
OHCONUMMMoou :ofiumHOHHou paw moEmz ofinmflhm> m.

 

 

ammg :oHpmHumacHEoz ocoovo

 

133(a)

 

 

 

 

 

 

 

 

OH. wo.n No. mo. NH. mo. No. Ho.a wN. ON. mH. mH. OH. mN. NN. HN. ON. Hm. ON. Ho.- OH

mO. Om. Om. OO. NH. HN. oH. ON. OO. OO. Hm. mo. Om. mO. OO. OH. mN. mN. mH. NH. mH

oH. mo. 00. O0. Ho.- mH. mo. No. mH. No. om.- mo.. 00. mo. mH.u oH.: OO. ON. Ho. mo. NH

OH. OH. NH.| moi ON. NH. 0H. Oo. OH... OH... Ho... emf mo. wH. mo. me. am. om. Ho. 00.: HH
phmzomIMmmMKHoHHomO HN. Nm. HH. OH. mO. OO. Om. ON. Nm. Nm. mm. wN. om. mO. oH. HN. oH
NOHHom How Huoez.1 we. HH. oo. mo.- NN. mN. Ho. NH. MO. Nm. oH. NN. m

mane: Ho>o oocosHmcHO ww. ON. mo. NO. Hm. NO. ON. mH. w

mmoco>HuHommdm HOHNOQSmO MO. om. OH. mm. N

OHSHDH cqu :Hoocou NOHHomsmO o

m

O

m

N

H

0H m m N o H

mm Hm N< H< Nm Hm N< H< Nm Hm N< H< Nm Hm N< H< Nm Hm N< H< W”
T.

9

s

mHCOHonmoou :OHumHoHHou pzm moswz OHanHm>

 

 

Hc>oH HO. MOO. u Ho>oH OO. -- O EHHO
Ho>OH HO. MON. u Ho>oH OO. -- < EHHO

"H mo mosHm>

HOSE HOOHO

 

mmooosw vo>HoohomO

133(b)

coHOOOOOHHOO OOO.

NH.: mm.u om.u NH.|
mo. OH. OH.| HH.

OOHOOHOOHO HOOO.

OH.- OO.- OH.- OO.-
HO.- OH. HH. NH.
NN. HN. OO. OH.L

Hmoo Ho>o mucouscHO

OH
MH

NH

MVLDOWOOOI

HN

 

mH

NH

 

 

 

mm Hm N< H<

NO HO NO HO

 

 

 

WHCOHUHHH®OJ :34434DHH3J 1:10 ODHHa:

)4334a‘3-

SI q? "_[lBA

nus—h?”

 

 

 

134

if the z value is found to be significant, this indicates that the

:orrelations are different. Table C-3 shows the significant r value

.ifferences for Firm A and Firm B.

 

 

 

Table C—3.-- Significant Differences Between Correlations
for Firm A and Firm B
(Tl—T2)
Firm N1 N Oz 2 Significance Significant
2 1- 2 Level Difference

A 128 119 .119 .05 .233

.01 .277
B 600 548 .044 .05 .086

.01 .119

 

 

 

 

 

 

le fourteen scale research model shows a very similar correlational
lttern across the samples. The scale reliabilities for the fourteen
:ales are exhibited in Table C—4. These reliabilities are the average
’the coefficient alpha reliability measures produced by the PACKAGE
'Ogram for the Time 1 and Time 2 administrations. Coefficient alpha is
measure of the expected correlation of one test with an alternative

rm of the test containing an equal number of items (Nunnally, p. 196).
esser collapsed the fourteen scales down to seven in order to improve
6 internal and change score reliabilities. The resulting seven scale

del is discussed in the following section.

 

 

 

 

135

Table C—4.--Comparison Between Firm A and Firm B Reliabilities of
Fourteen Scales

 

Internal Change Score
Scale Description Reliability Reliability
(rxx) (r d d)

 

Firm Aa Firm Bb Firm Aa Firm BC

 

 

 

1. Use of goal-oriented methods .95 .87 .90 .74
2. Satisfaction with boss .90 .91 .88 .85
3. Self-improvement goal clarity .60 .48 .54 .12
4. Performance goal clarity .68 .69 .53 .45
5. Orientation toward MBO .80 .86 .50 .59
6. Boss concern with failure .57 .62 .35 .24
7. Boss supportiveness .78 .72 .70 .55
8. Influence over boss .60 .64 .29 .38
9. Need for policy .48 .59 .31 .09
10. Association between performance
and reward .84 .77 .68 .52
11. Influence over goals .75 .69 .50 .52
12. Performance goal difficulty .44 .38 .44 .00
13. Satisfaction with job .58 .58 .35 .12
14. Success in attaining goals .65 .54 .30 .06

 

aChesser, p. 44.

bInternal reliability for Firm B is the averaged standard score
coefficient alphas for fourteen scales from data samples of 600 managers
in first administration and 548 managers in second administration.

cCalculated by Equation 10.25, McNemar, p. 157.

136

Table C-5 contains the means and standard deviations for the
fourteen scale model. These statistics are based on the reSponses from

117 Firm B managers identified in both the Time 1 and Time 2 samples.

Seven Scale Model

The results of the condensation of the fourteen scale model into
a seven scale model are diSplayed in Table C-6. Appendix B contains a
description of the scales and the items which make up each scale.

One scale, Superior—Subordinate Relationship, contains eighteen
items. Another, Goal Clarity and Relevance, combines those scales which
previously measured both performance goals and self—development goals.
The scale, Performance—Reward Association, consolidated responses by the
managers to subscales representing the perceived relationship between
dollar rewards and performance plus a response to the subscale represent-
ing the individual's need for rules and policies.

Since the upper limit of scale combinations was defined by the
number of items in the questionnaire, some change in the character of
the scales was inevitable. The strategy for scale development was three—
fold. First, by ”external" analysis, if scales or items do cluster
together, they will demonstrate a similar pattern of correlations with
other items not in the scale. Also as an external indicator, if items
in a scale are similar they should not have opposite signs for their
correlation coefficients with a third item. Second, as an "internal"
analysis measure, the scales as augmented should have items which are
highly related to one another. In other words, they should have an
acceptable measure of internal reliability for a standard such as

Coefficient alpha. Third, the scales as condensed should possess a

 

 

 

137

)le C-5.-— Scale Means and Standard Deviations of Fourteen Scale Model
For Firm B Managers at First and Second Administration of
the Questionnaire

 

 

Mean§_ Standard Deviations
as First Second First Second
Administration Administration Administration Administration

3.28 3.31 0.81 0.72
3.95 4.04 0.98 0.88
2.74 2.78 0.51 0.57
3.16 3.21 0.57 0.47
3.73 3.51 0.89 0.92
3.39 3.33 0.79 0.76
2.54 2.49 0.78 0.74
3.48 3.51 0.64 0.65
4.29 4.16 0.68 0.76
3.54 3.35 1.01 1.01
2.64 2.80 0.90 0.90
2.51 2.58 0.63 0.65
3.30 3.34 0.89 0.88
3.10 2.98 0.77 0.79

 

 

 

138

Table C-6.-—Seven Scale Research Model

 

 

 

Questionnaire
Scales Item Numbers

1. Superior-subordinate relationship S,8,9,11,21,22,
23,25,26,27,28,
29,30,31,36,43,
44,45

2. Goal clarity and relevance 1,2,3,4,6,7,24,
38

3. Orientation toward MBO 39,40,41

4. Performance-reward association 34,35,46,47

5. Subordinate influence over goals 18,37

6. Satisfaction with job 32,33

7. Success in attaining goals 14,15

 

Reference: Chesser, 1971, p. 11.

 

 

 

 

 

139
similarity of content among the various items within that scale. The
responses from Firm B have been scored using the seven scales developed
for Firm A data. The inter-scale correlation coefficients are shown in
Table C-7. The criteria for testing significant differences between
Time 1 and Time 2 correlations are the same as those used in the fourteen

scale model.

DeveIOpment of Effects Diagram

 

In this section, the data from Firm B has been used to calculate
the dynamic correlations and cross—lagged panel correlations between the
scales of the seven variable model. From these correlations, an infer—
ence of causal relationships was made and an effects diagram was
developed. This same methodology and presentation is used for the pooled

data base for both organizations. A comparison of the effects diagrams

 

developed from the Firm A and Firm B data will close this section.

Dynamic and Cross—Lagged Panel Correlations

Dynamic correlations are Pearson Product Moment correlations
between the change scores for the variables in the model. The responses
for the variables have been scored at two points in time (actually
eighteen months apart). Then differences in these scale scores (Time 2
score - Time 1 score) were determined. By using dynamic correlations,
the statistical association between these changes on pairs of variables
within the model can be analyzed. The correlations between change scores
(dynamic correlations between raw change scores) may contain variance
that is due to managers in the Firm B sample having different initial

scores. To eliminate the variance in these difference scores that is

 

 

 

140

 

 

 

NN. ON. ON. ON. ON. ON. OO.- OO. NN. ON. NN. NN. OOooosw Oo>Hooeoa

OO. NO. HO. OO. OO. NO. OH. HN. OO. OO. OO. HO. OoO
:qu :oHuommOHumm

ao. co. mo.- mH. NN. wN. No. mo.4 mH. HN. wo. No. mHmow Ho>o ooze
stmcH opwcHwHoasm

HO. OO. NH. HH. NO. OO. ON. OH. NO. HO. OO. NN. :oHeOOooOOO
phm3omuoocmEHOMNom

OO.H OO.H OO.H OO.H om. O0. HN. mm. mv. MO. OO. mO. Om:
whmzoe :oHuOHCOHHo

OO.H OO.H OO.H OO.H ON. NN. NN. OO. eoqa>OHOO
One OOHHOHO HOOO

OO.H OO.H OO.H OO.H OHOOOOHHOHoO
oumchHonsmlgoHNmmsm

r—t

 

COHumHhomoo
OHOOHHO>

 

NO HO NO H< NO HO NO H< NO HO N< H<

 

 

 

atqetxaA

 

Hva coHumapchHEp< vacuum wen HHmV :OHHOHHOHCHEO< HONHO
a EHOL a:m amid :DOJnHOnO:OES< 3:39D0 3:6 OHCH :34;3+.u+:4:i..

 

 

 

 

 

 

 

 

HH. u Ho>OH HO. HOO. u Hm>OH OO. -- O SOHO
OO. u Ho>OH HO. MON. n HO>OH OO. -- < SOHO

VH mo mmsfig ufiwowwflcmfim

 

OO.H OO.H OO.H HO. OO. OO. OH.- OH.» OO.- NO.- OH.O ON. OH. OO. OO. OOOoosm OO>HO0~OO .N
OO.H

OO.H OO.H OO.H OO.H Ho... oH. wof mo. om. mm. Nm. OO. now
52.. :oOpowwOUmw

\O

OO.H OO.H ooé OO.H NN. MS. 3... oo. mHmow 996 ooco
555 ouwcfiiogm

LI“)

140 (a)

OO.H OO.H OO.H OO.H coUOdfloomm/O.
Easemlmoqmahomwom

<1“

om:
HOBO 309. co Um ucmfiho

m'

moqm>®me
OOO NOHNOHO Hmou .N

OHOOOOHOOHOm
oumcflwaogmuhofl 25m .

 

H

 

 

 

cosmic woo
0 NHHM
NO HO NO H< NO HO NO H< NO HO N< HO NO HO N< H< HO . >

 

 

atqepleA

 

 

 

 

 

 

 

 

 

141

predictable from the time 1 scores (initial scores), these dynamic corre-
lations can be "corrected" (Vroom, p. 64). To do this, a measure of the
difference between the observed change of a variable and the difference
that would be expected of a variable with the same initial score on the
variable is calculated. This method requires the computation of the
second-order partial correlation between difference scores with initial
scores on both variables held constant (rAxAy .xy)' Table C—8 shows the
matrix of raw and "corrected" correlation coefficients between change
scores for Firm A. Table C—9, page 143, is a presentation of similar data
for Firm B. Table C—lO is a matrix of the differences between dynamic
(corrected) correlations for the samples. Assuming that the dynamic

(corrected) correlations (rA X),) are equivalent to 2 values (by r to

XAY .
z transformation) a test of the differences between these coefficients
can ascertain whether the organization affects the relationship between
the variables. For both organizations at the .05 and .01 level, there
is not a statistically significant difference between dynamic correla—
tions for any of the relevant variables.

The logic of the cross—lagged panel correlation analysis is that
there is a time lag that occurs when one variable causes another. Pelz
and Andrews call this "causal priority" rather than causality (Pelz and
Andrews, p. 836). If one variable is to have a causal relationship with
another, the essential ingredient in that relationship is that it is
asymmetrical. That is, if variable x causes y, then the present state of
X should be more strongly associated with y's future state than with y's
present state (rxly2 > rlez). Table C-11 and Table C-12 display the
inter-scale correlation coefficients between the first and second

administrations of the questionnaire to Firms A and B.

 

 

 

 

 

142

 

 

.mo .9 .HNmH Anommeao

:oHoOmmooo eeuomnuoo age No boonHom .umsHm wopnomoa ON mmnoom omewgu
.p . .

Househomwm

om. u Hm>mH Ho.
MN. n Hm>wH mo.

"a wo modam> quuwmwcmflm

ZNH fi®®3H®£ COHfldHQHHOU ®£Pm

 

 

 

 

 

 

 

 

 

 

 

 

coNumNuomm<

oocm>ofiom

mflgchNuwHom

 

 

 

 

 

o.H o.H No. OH.- co. No.- OO.- oH.- ON. HN. OO.- co. No. OH.- mmbdem Oe>fimunem
o.H o.H oo.: 00.: ON. Om. ON. MN. ma. HH. om. vm. now flaw: cofipumeHumw
o.H o.H OO.- Ho.- OO.- No. HH. mo. mH. OH. mHmou Nm>o mace
usHmcH m.oumcHwHonzw

o.H o.H HH. mo. ma. NH. om. mN.
vgmzomlmocwgfiomawd
o.H o.H ma. mo. Hm. ON. Omz mhmzoe coaumucowho

o.H o.H mN. om.
Ea NOHSHO Haoo

o.H o.H
opmcflwhoaszHOHaomsm

_
O o m O m N H oHanHw>

 

 

 

 

 

 

 

 

< EHdm HO% WOHOOW OMfiNSU EOOZHQM WHC®HOH%M®OU COMHNHOHHOU UOHUQHHOU Ufid 3wm MO XHHHMZ I

umpcmfloflwmooo :ofiumHonhou macaw mmcmcu wouoongou mcw 3mm

 

|.mnu ofinmh

 

 

 

 

 

143

.u=0HUHmmmoo popumhaoo

ecu ND uoonHom .umHHm Umupomou OH mmaoom owcwnu

 

 

ON. u Ho>mH Ho.
wH. n fiw>0H mo.

”H Mo W®5Hm> HGNUfiWMCMflW

BNH E®m3H®£ COHHNHQHHOU 03PM

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mmp:0Honmoou COHpmHmhpou whoom owqmgu

fimuUOHHOU Ufiw 3mm

 

O.H o.H No.1 mH.- oH. HH. 0N. OH. OH. No.u oH. mo.4 mN. Ho.: mmmuusm uo>HeoHoa .N
O.H O.H Oo. OO. OO. ON. NH. NH. OH. OH. OO. OO. now aqu :oHuommOHqu .O

o.H o.H HH. OH. OO.- Ho.- NH. NN. oH. ON. mHmou He>o mono
:sHmcH O.mumcHeHon:m .m

O.H O.H ON. ON. ON. HH. OO. NO. :oHuaHUOOOO
vhdzomnoocmahowhod .O
O.H O.H ON. OH. ON. NH. oOz Oaazoe :oHuOOOOHHo .O

O.H o.H NO. wO. mosm>mem
Oew NOHOOHO HOOO .N

O.H O.H demcoHumHoO
muwcHwHondwunoNhomsm .H

N O O O O N H OHOOHNO>

 

m EHHm How moHoom omcmzu :omzpom mu:OHonwoou cowumHoauoo vmuoonaou van 3mm mo xwhumz 1:.

ago eHnwh

 

 

 

 

 

 

144

Table c-10.-— Matrix of Differences Between Dynamic (Corrected)
Correlations of Variables of Seven Scale Model
For Firm A and Firm B

 

Scales

 

 

 

1
2 .22
3 .06 .08

 

Scales 4 .15 .08 .13

 

 

 

 

 

Note: 1) The correlation values are assumed to be equal to 2
values.

2) The standard error of correlational differences = .215.

3) Significant difference at .05 level = .42 and at
.01 level = .50.

 

 

 

 

 

 

145

Table C—ll.—- Inter-Scale Correlation Coefficients Between First and
Second Administrations of the Questionnaire for Firm A

Second Administration

 

 

First Administration 1 2 3 4 5 6 7
l. Superior—Subordinate

Relationship .463 .01 .26 .26 .14 .35 .32
2. Goal Clarity and

Relevance .23 .23 .36 .13 .04 .17 -.06
3. Orientation Toward MBO .39 .09 .51 .19 .10 .29 .08
4. Performance-Reward

Association .20 .00 .03 .42 .23 .23 .12
5. Subordinate's Influence —.10 —.ll .05 .02 .17 .08 -.O7
6. Satisfaction With Job .05 —.05 .02 .33 .07 .48 -.01
7. Perceived Success .31 .07 .20 .08 .19 .19 .37

 

3The diagonal entries are correlations between a variable
measured at time 1 and time 2. Off-diagonal entries are correlations
between a variable measured at time 1 and a second variable measured

at time 2.
Significant value of r:

.05 level = .23
.01 level = .30

Reference: Chesser, 1971, p. 67.

 

 

 

 

146

Table C-12.—- Inter-Scale Correlation Coefficients Between First and
Second Administrations of the Questionnaire for Firm B

Second Administration

 

 

First Administration 1 2 3 4 5 6 7
1. Superior-Subordinate

Relationship .45:11 .22 .38 .30 — 13 .06 .25
2. Goal Clarity and

Relevance .31 .46 .24 .17 -.20 -.06 .13
3. Orientation Toward MBO .36 .39 .66 .32 —.04 .12 .24
4. Performance—Reward

Association .38 .31 .36 .60 .08 .22 .07
S. Subordinate's Influence .04 .07 .05 .01 .35 —.02 -.04
6. Satisfaction With Job .32 .16 .27 .37 .05 .52 .18
7. Perceived Success .20 .26 .12 .10 -.16 .06 .51

 

8The diagonal entries are correlations between a variable

nwaumdattflw].wdtmw2.

Off-diagonal entries are correlations

between a variable measured at time 1 and a second variable measured

at time 2.

Significant value of r:

.05 level
.01 level

.18
.24

 

 

147

Table C—lS and Table C—l4 are set up to display the difference
between pairs of cross-lagged panel correlations for all pairs of
variables in the seven scale model. These differences indicate the
degree of the asymmetrical relationship for any variable x at Time 1
with any other Time 1 variable y. A negative difference is interpreted
as the variable y being a better predictor of the future state of
variable x rather than the hypothesized temporal sequence.

For the managers of Firm A, there are three asymmetrical rela—
tionships between variables of the seven scale model that are significant
at the .10 level. For the managers of Firm B, there are two significant
asymmetrical relationships between system variables. A relationship
which is significant as indicated by the dynamic correlations and asym—
metrical as indicated by the cross—lagged panel correlations is found
between changes in Superior—Subordinate Relationship (variable 1) and
changes in Job Satisfaction (variable 6). It is interesting that for the
Firm A managers, changes in their relationship with their superior have
a causal relationship with changes in their satisfaction with the job.
Since this relationship is positive, after the Finm A manager perceives
an increase in the relationship with his boss, his satisfaction with his
job also increases. Although the relationship is still positive, the
opposite causal relationship is found for the managers of Firm B. For
the Firm B managers, an increase in job satisfaction causes an increase
in the relationship between the manager and his superior.

Having compared the means and variances for the scales across the
two organizations and finding them very similar, a decision was made to
pool the samples and average their respective correlation matrices (see

Table C—15 and Table C—16). The rationale for this pooling was that the

 

 

 

 

 

148

Table C-13.—- Matrix of Differences Between Cross-Lagged Panel
Correlations for Firm A

 

 

Second Administration

 

 

First Administration 1 2 3 4 5 6 7
Superior-Subordinate

Relationship -.22 -.13 .06 —.O4 .30 .01
Goal Clarity and

Relevance .27 .13 .07 .22 -.13
Orientation Toward MBO .16 -.15 .27 —.12
Performance—Reward

Association -.25 —.10 .04
Subordinate's Influence .15 .12
Satisfaction With Job —.20

Perceived Success

 

 

 

 

 

149

Table C—l4.-- Matrix of Differences Between Cross—Lagged Panel

Correlations for Finn B

 

 

Second Administration

 

 

First Administration 1 2 3 4 5 6 7
l. Superior—Subordinate

Relationship -.09 .02 —.O8 .17 .26 .OS
2. Goal Clarity and

Relevance —.15 -.14 -.27 .22 -.13
3. Orientation Toward MBO —.04 .Ol .15 .12
4. Performance-Reward

Association .07 .15 -.03
5. Subordinate's Influence .07 .12
6. Satisfaction With Job .12
7. Perceived Success

 

 

 

 

 

150

Table C-lS.-— Matrix of Dynamic (Corrected) Correlations Coefficients
for Pooled Firm A and Firm B Sample (Total n = 190)

 

Second Administration

 

First Administration 1 2 3 4 5 6 7

 

 

l. Superior—Subordinate
Relationship

2. Goal Clarity and
Relevance .36

3. Orientation Toward MBO .28 .19

4. Performance-Reward
Association .38 .19 .18

5. Subordinate's Influence .15 .14 -.O7 .03

6. Satisfaction With Job .35 .16 .20 .30 -.Ol

 

7. Perceived Success .16 .02 .18 .08 .08 .00

 

Significant values of r:
.05 level = .14

.01 level = .18

 

 

 

 

 

 

 

151

Table C—l6.—- Matrix of Inter-Scale Correlation Coefficients Between
First and Second Administrations of the Questionnaire
for the Pooled Sample

 

 

Second Administration

 

First Administration 1 2 3 4 S 6 7

 

 

l. Superior-Subordinate

Relationship .45 .12 .32 .28 —.14 .21 .28
2. Goal Clarity and

Relevance .27 .35 .30 .15 -.12 .06 .03
3. Orientation Toward MBO .38 .24 .59 .26 -.07 .21 .16

4. Performance-Reward
Association .29 .16 .19 .51 -.O7 .23 .10

5. Subordinate's Influence -.03 —.02 .00 .02 .25 .03 -.05
6. Satisfaction With Job .18 .05 .15 .35 -.Ol .50 .09

7. Perceived Success .25 .16 .16 .09 -.17 .13 .44

 

Significant values of r:
.05 level = .14

.01 level = .18

 

 

 

 

 

 

 

 

152

research model replicated within each firm and that the pooled sample
would improve the generalizability of the model. As the sample size
increased, the number of significant dynamic (corrected) correlations

also increased. Table C-ls indicates the fourteen pairs of variables in

 

the seven scale model that are significant. When the differences between
cross-lagged panel correlations for the pooled sample are calculated, it
is found that none of these differences is significant at the .10 level
(see Table C—l7).

Several conclusions are reached as the result of the analyses
presented in this section. First, the variables in the seven scale model
are similar across the samples. Second, the dynamic correlations indi-
cate that there are statistically significant associations between
changes in the variables in the model. Third, the inference of causal

priority at the .10 level is not possible at this point due to the

 

absence of significant differences between the cross-lagged panel corre—
lations. In order to develop the effects diagrams of the behavioral
system, a significance level for the asymmetrical relationships of .20

will be used.

Effects Diagrams
In the last few sections, the concept of changes of one variable
"having an effect on" changes in another variable has been used quite
often. To make this idea more precise, an effects diagram can be derived 'n.
for the three samples of data: Firm A, Firm B, and the Pooled Sample
(A + B). The process is as follows: It is assumed that the data are
reliable and subject to minimum sampling or random error. The variables

of the system are those of the seven scale model. For each significant

 

 

 

 

 

_i—71' _a._ 4-—---—-'w-—.~.---‘-n

153

Table C-l7.-— Matrix of Differences Between Cross-Lagged Panel

Correlations for the Pooled Sample

-‘
___'

Second Administration

 

 

First Administration 1 2 ' 3 4 s 6 7
l. Superior-SUbordinate

Relationship ~.1S -.06 -.Ol —.11 .03 .03
2. Goal Clarity and

Relevance .06 -.Ol -.10 .01 -.13
3. Orientation Toward MBO .07 -.O7 .06 .00
4. Performance-Reward

Association -.09 -.12 .01
S. SUbordinate's Influence .04 .12
6. Satisfaction With Job -.04
7. Perceived Success

 

 

 

 

 

 

 

154
dynamic (corrected) correlation in the system, a solid line is used to
connect the variables. When the cross—lagged panel correlations indi-
cate a causal relationship (at the .20 level or better), a unidirectional
arrow shows the direction of that relationship. For those cross-lagged
panel correlations which do not demonstrate an asymmetrical relationship,
a mutually reinforcing relationship is shown using arrowheads at both
ends of the solid line.

Figure C-l is the effects diagram showing the change relationships
in the MBO Behavioral System for Firm A and Figure C—2 is the effects
diagram for the managers of Firm B. Changes in Goal Clarity and Rele—
vance do not have the driver effect on the managers of Firm B as they did
for those of FirmA. Satisfaction with Job is the driver for the managers
of Firm B.

The effects diagram for the pooled sample (Figure C-3) demon-
strates the expected result of canceling all causal relationships and
having mutually reinforcing relationships between all the variables.

There are several possible explanations for the different
temporal sequences of causal relationships shown in the models. First,
the causal relationships shown in the effects diagrams of Firm A and Firm
B may be due to spurious correlations and not due to real changes. The
data as analyzed (i.e., constant means and standard deviations) and the
scale unreliability support this notion. Second, there may be actual
changes and causal relationships among the variables of the system but
the time Span of eighteen months may be too short or too long to assess
these causal relationships. Third, the variables of the system are con-
sistent across and within samples. However, the reliability of the
Change scores is too low to adequately discriminate between asymmetrical

relationships.

 

 

155

Figure C—1.-- Effects Diagram for Change Relationships in the

MBO Behavioral System — Firm A

(Reference:

Chesser, 1971, page 106)

 

 

OOO
spa: soaponNnaOOO

 

. 0mm “sheaves
GOmePQoNHo

 

 

.QH mmmqmso

 

O

 

 

 

OOHpmHoomm<

UHMSQMIoonOEHOMNem

ea momnmgo

EH monsmno

 

 

 

 

r

 

 

I“

_.__._____+“1

 

 

 

 

 

mHensOHpaHom
mpmnflwhonfimIOOHaoQSm
QH mmmmmno

 

 

 

 

 

moqm>oflom
One OOHHOHO HnoO

 

SH momsmso

__._____._______—.—.——.—._.._—~__

 

 

 

 

 

nHeow mmflcflnppd 4H
mmmoodm pm>flmonom
GO mowede

 

 

 

 

 

 

156

Figure C-2.-- Effects Diagram for Change Relationships in the

MBO Behavioral System - Firm B

 

TI

 

now anz
:OHuommeuwm
:H mowemnu

 

 

O

 

 

 

 

.:0HpmH00mm<

 

 

.oOz Bases
:OOHMHCoHHO
:H mowemnu

 

 

 

H

 

 

 

 

 

 

cameoMIoocmEHOMHom
:H momnmgo +

V

 

 

 

H /

 

mHnchHumHom
oumcwaonsmuHOHHomsm
ea mmmcmcu

 

 

 

 

mHmoo wchHmpp<
:H mmooodm wo>Hoo
when :H mommmcu

 

 

* .

 

. /

 

 

 

oocw>oHem

V O NOHsaHu Hmou

 

 

:H mowemnu

 

 

 

 

 

 

 

157

 

 

ples from

ge Relationships in the

MBO Behavioral System for Pooled Sam

Figure C—3.—- Effects Diagram of Chan

Firm A and Firm B

 

 

new :qu
:oHpomeHpmm
:H momcmzu

7!!!

OOO Osason

 

 

 

 

 

 

COHumpcoHHo
:H mowzwcu

 

 

 

 

 

 

 

 

 

COHumHoomm<
enmzomnoocmEHOMHod
. :H momcmnu

 

 

 

 

oocoszcH
m_ouw:anop:m
:H mowcmzu

mHzm:0HumHom
oumcHwHOQSmIHOHHomsm
:H momcmgu

 

 

 

wHwoo
wchHmuu< :H
wmooodw vo>Hoo
laud :H mowqmzu

 

 

 

     

oocm>oHem
O NOHNOHO Hmou
:H womqmgu

  

 

 

 

 

 

APPENDIX D

MBO QUESTIONNAIRE ITEMS, SCALE INTERCORRELATIONS, AND
LOADING MATRICES FOR THE REVISED FIFTEEN AND SEVEN SCALE
RESEARCH MODELS

 

Note: 1. MBO Questionnaire, Appendix A, contains .
all items referenced herein. L

2. Multiple group cluster analysis of 548
responses by Firm B managers to the
second administration of MBO Questionnaire
is reported herein.

 

 

 

 

 

APPENDIX D
MBO QUESTIONNAIRE ITEMS, SCALE INTERCORRELATIONS, AND
LOADING MATRICES FOR THE REVISED FIFTEEN
AND SEVEN SCALE RESEARCH MODELS

This appendix describes the scales produced by the multiple group
cluster analysis of the Firm B data. A fifteen scale model was initially
produced. In order to improve the internal scale reliabilities, the
scales were condensed and a seven scale model was developed. The

development of the fifteen scale model is presented first.

The Revised Fifteen Scale Research Model

For each of the scales in the fifteen scale model, a table is
presented which displays the stems of the items from the MBO study ques-
tionnaire for that scale in addition to the correlations between and
within the clusters. These correlations have been computed by the
PACKAGE routine for oblique multiple groups factor analysis with commu- ‘
nalities (Hunter and Cohen, 1971). By using communalities, the correla-
tion between an item and a scale in which it is not included is corrected
for attenuation. At the same time, the effect of unreliability within
the scale has been removed. The effect of unreliability for a particular
item has not been removed. The communalities which appear as diagonal
entries in the item correlation matrices represent the specific relia-

bilities of the item in the scale to which they belong.

158

 

 

159

Objective Feedback

Table D—l presents the three items which make up the objective
feedback scale, their intercorrelations and their correlations with the
other scales, corrected for attenuation. The diagonals indicate a
strong-weak gradient within the cluster. The first two items are con—
cerned with the manager's recall of the occasions on which he was given
information about his performance. There is the potential for some
confusion by the respondent if he includes personal development objec-
tives under the general heading of objectives as in item five. The
common denominator of these three items is the manager's feelings about

how his superior is evaluating his qualifications and development needs.

Interest in MBO

Table D-2 diSplays the three items which make up the interest in
MBO scale. Also included are their inter—item and inter-scale correla-
tions. The intercorrelations of the items form a rank one matrix and
have a strong-weak gradient from item 22 to item 21. These items are
also quite consistent in their correlations with other scales. The
content of these items is quite similar and seeks reSponses which concern
the genuine interest and involvement of the superiors and the organiza-

tion with MBO.

Goal Relevance

The eight items which make up the goal relevance scale are shown
in Table D-3. The first three items measure the manager's perception of
the congruence of his objectives with the needs of the organization.
These items seek to ascertain goal clarity as well as goal relevance.

The last five items concern the degree of understanding between the

 

 

Has—vane“ on

accoulrou mo 359—3! _ MM mm “M mHm

333232 Euros-3558:...— . an mv An OH».

«382m :8 N 3 3 OH...
Swuuamfluum 50.. av

NuHsunOHO Hnou ON an ON HHO

mHsou so OOOOOHOOH .N ON ON OHO

cm: vane—o... 833:3qu av : on mom

mmofizflommsw mmon ”no we we won

m3: 5:. 53332.5 nv om mm New

.822. you. vuoz ~n no on com

Ewan—D 35:25 an ov a. mom

onsHHna OOH: cuuucou mmOO On On NO OOO

353$“ _aou av 00 No new

OO: OH umosousH av ON NO NOO

OOJOOOOO O>HuuoHOo .OO ON no HOO

160

”9 3 on «N
3 Nm 3 On
On NO ON. NN

1.5-: «:3on EB

. . . oHaum

2.3:»
cm: 23 5 mm... Eunice 05 #55 no» ca 3335 5 mo 5:: no:

33.. 5.13s: 3: 2.» 3 333 O3; «so» 3H. 2.3 4...:- .6:
:33? 8: 2.3 5 n...— nng EPA .2qu so» av yaw—3.: 5 mo 5:.- so:
:emummhunua

on: :H wagon—i .Nom $64,439. 071.9

 

luau

H333:

ounuuomaou mo 35939:
EHuaHuomm< vunzomuoofifinomnam
nnouunw Hnou

538333 8H.
NOHsuHmeO Hsou

“Hmou :o 3533:

On: Bunch. COHunucoHuo
mno=o>muhona=w 3.3

30m 5:. :oHuummmHuam
.339. Mon nooz

Ewan: conga—OH

951.8..— .Hqu 5350 Owen
oocm>oHom Hmoo

0O: 5 30.32:

enema... 2.58.26

xfluwz wives v5 .-

H3332— .So» 95:33 :0 «Ha nmon .5: Be 33.3% 5:5 .5: HH
.30» .5 any—moan 59H .8 xouavoom 5:» so» one: game 30: m

HN.—onwan .50» :0 30.30.; .30» H5 xuaavoow :9...» so» Duo... curve :0: a

:oUmmuumon

 
  

enacted 2.33.30 :3. 33m --.H-n «Hank

. oHnuw

39.33.30 azuamngov

30330.30

36330.30 unwamoHocHov

 

 

161

manager and his superior regarding the evaluation of performance and the
priority of effort to be expended. The common elements in the entire set

of items for this scale are goal relevance and goal clarity.

Boss Concern with Failure

Table D-4 shows the two items that form the boss concern with
failure scale. The items are very consistent in their inter-item and
inter—scale correlations. The items show homogeneous content and rela-

tively strong reliability.

Influence Upward

The common element among the four items which make up this scale
is the influence of the manager upon his superior. Table D-5 shows that
the items are quite similar in content and have a consistent pattern of

correlations with other scales.

Need for Policy

Scale 506, Need for Policy, is shown in Table D—6. The two items
which form this scale are very compatible in content, internal consis-
tency, and external correlational pattern with other scales. This scale
measures the manager's need for structure or guidance as he participates

in the organization.

Satisfaction with Boss

The four items which make up the satisfaction with boss scale are
shown in Table D—7. This scale is a measure of the manager's satisfac-
tion with his boss as a boss. The inter-scale correlational pattern is

consistent and the inter—item correlation shows three strong items and

 

 

 

 

3:38... 3 3 3m

3:32—30 mo oooauyooay .. aw ow 2m
.83-«032 vyuxoxéofisyomyom. 3 on In
330.6 230.. mu cu mam

5333.33 3.. o om 2m
533:3 :8. .3 ow :m

«:8 8 35:9: - aw Sm

on: 32.6... sauna—Note .nN on mom
30:33:35 33 on so «on

30¢ 5; :Oyuuaumyuaw ed 0v son
.629. you. 332 an mm eon

Earn: oocogufi «a «n new

833. 5:. £850 38 2 2 3...
3:328. :39 .9 no new

.. on: 5 3.235 n». 2.. New

. _xoaneuoy o>yuuonao an. an “an

162

“on .~n .mv”
.mn mm .v.

a

Mayan: Magoo.— .Eo nooyuanoyyouyoucm 33w

 

«nosyuuoNao cognyyaqamo o>oaaua
on noun-u so» my :2. yoo» gym 2533 so» 3.8: 3333.3 pays

«mosyuooyno vogmayaauno
2623 3 33¢ so» «a on 3.6: anon yzox yooy soy ov coy—yous”. 3o:

Sflmnumoo

833. 5:. 58:8 9.8 33.225 22.1. 39:.

me

:03

3.33. a... o an o. 3 ow 3 «N 2m

388.58 «o 355.5 3 2 .3 .5 vn an 2. an . my...
83333 232735583." - om m~ .3 an .N on an . z...
383m :8 .N o a... 3 3 a 3m om 2...

:oypuaMmyumm non ny o
Nuyauymwya yaou Ny .w
myaou co ”anonymcy am an
on: eyesoy :oyumucoyyo an my

my nu my Nu ~n yam
n oy an ac vm an .yym
y 3 a. N. v 3. Sm
N on Nu «n N» on «on

«monotpyuyoaosm mmom - mu on vn on my nv om mom
33 5y: :oyuumwmyumm om am ow mm ow o." Nv ow Now

Nuyyoa you vomz

mm «a mm N« m.“ ha NN eon

a
235 853.2: .3 o 3 3 2 am 3 o. mom
cum—«mum saw—z Eoucou mmom ON on on on ”N on an on , vow
3:230: anon y» N» on av mv wm co my mom
8: am umoyoucn on mm an on an em mm no Nom

sneeze". 238.25 a on «n on I. 3 3 I. 8m

.y c» m ow .y cy my mm m.
on vy mu ~a - n ow am an
no oy my my oy nw ow «n my
.ou ~y my Ny yy «N y» an .w
.e« - my yy ow a. N~ om .
..y .n MN «m o. N~ «. Nn m
an oy o~ an N~ we we we a
o~ ow u» an o~ N» no no N

 

anyway—images Ea mooyuayvyyouyoufi 33m

530» 3538on ya?» you moyumyoyym £32336 «non yo?» v3
30:30.23

you ago—gov Honouyon yzo» you moyyyyomyn .98 3335 30: yoox 3a
Nyaox away 3 oyamaou

.35 on. lo: .8: .395 N52 ya so» you you soon 9:“: mo>yuuonao
«3558on yoo» «stowag-

5 was anon yoox 330$va «:3on of moflyomov 3.2. not?

N300: agencygo—u 22.8on yoox uuoaoy

:oN you you $533.30 3059333 133on on» one ooyuuu «as: o...
Nbamaou on» mo

30 E 9:3on .nsoyyom anon 05 uooymoy no» you you. mo>yyuonpo 3a
2.30 .98 muggy 3 «gonna say: 323.» xyyaoyo mgyuuomoo yoox eye:
«sex on

no. @353 mgyuoonno mooyyg yoo» .yo oucauyomay «$338 05 mm:

:oyumyyumon

8:329“ «moo .nom 23m 13.5 2%...

we

an

0H

3..

 

flown

 

 

 

 

1333: 8 3 a . 3 3m
3:309:00 .«o ouumuyoman mn um an .3 mam
5.33033 uym§mu3=ufiyowyom an on. on nm 3m

:oyuomMmyumm non vu vv Nn om Nam
33333 :8 w .n om N 2m

330 :0 3:0:wa vac vs N. Na.- 3.».
cm: 35.8. acyyooaoyyo an N.— ~.. 0 men
mmozgyuyonmam mmom nv on no mm wow
305 2:2 :eyuunumwumm tn no on at New
.320; you no»: mu .3 v to com
352.5 353:5 Nn «N. on an . mom
33?... 5y: EmuSu 30» mm o." my v« H 3m
85:63: Row on ow an n~ “ 25

am: :y umvyoucn mm on NV ON A Now
fiafiouy 233.25 B on Na 3 Sm

1.533: us ca 03

8:32:80 mo occur—om: _ 3 on my...
.83-«032133685583; N an «S
amouuam :00 a“ 0 2m

Eyuuaunyuum ooh.

baits :8

230 so 8:032: .3 an Sn

8: 225 832:35 nu 2 8m

163
NO

 

 

30:333me 33 a nu mom a» mw on mm n
32. 5:. 533335» «a om 3m mm Nn mm on an
3:2 you .52 . 3 3 8m .3 ma NN. on 2.
x vymsns 35qu a. my mom ww on on on mm
95:: 5y... F538 38 on N» com
35.67% 13w 0» an new
8: cm 393:: «w or. Sm . xyyua: wing... Ea S. . . 23m
fiaéofi 2.33.30 3 2. _8m
9 3 N.
. u. 9 3
«258.3.
. t yS» ufifiavy mo £303 05 3.3.69 :0» «new no» on 2,838 was: oh m
“3.5:: may—woo.— Ea 22:... 4 . 33m

«aye: .50» o» Saucy easy mmon .50» mo mcoynmuuv
of 3 gm: :ox conga—i mo «ESQ 05 5y: :ox 0y» voymmflmm .5: cm

. 3on you» 339:: 35
3% .50» may-Fora 5 nos . a; non—no Hogan a 5:: :omcymo yaox x3 macs .50» noon game 30: NN
30: on noyavauoym as $40.32* ova—Sana 93: 0» uu my ufiayomly :0: 5. human yaox mo mcommyuou 05 35:35 56 :0» «new so» on mN

. Nov 3 so» 35: «nos 50» uni. .55— 3 no» you «a an astound" to: 3 3 av:

Silva nouy

 

 

. . . . Buigoocoanwuu .mom uuuow nu.mvn manna.
.023. you. woo: .2; 33m i.e-o 3%...

 

 

Table D-7.-- Scale 507, Satisfaction With Boss

Item Description

29 How capable a manager is your boss?
30 How good is your boss in dealing with people?
31 How satisfied are you with your boss?

26 Which statement best describes the feedback you get from your boss
about your performance?

Scale Intercorrelations and Loading Matrix

2; 79 79 75 53
31 79 89 74 58
3o 75 79 65 44
25 53 5a 44 37

501 49 49 3? 58 Objective Feedback

502 58 5»? 48 60 Interest in MBO

503 46 40 34 5.43 Goal Relevance

504 44 32 32 42 Boss Concern With Failure

505 61 69 56 65 Influence Upward

506 29 24 24 20 Need For Policy

507 89 92 80 60 Satisfaction With Boss

508 71 81, 66 76 Boss Supportiveness

509 27 21. 21 2M Orientation Toward MBO

510 4 '9 ’3 8 Influence 0n Goals

511 18 19 14 16 Goal Difficulty

512 45 50 41 49 ’ Job Satisfaction

513 20 1b 16 1], Goal Success

514 36 36 32 40 1 Performance—Reward Association
515 P5 26 28 33 Importance Of Competence “
516 14 9 14 15 Residual

 

 

165

one weaker one. Item 26 is an indicator of how the manager may interpret
the actions of his boss based on his feedback and interaction with the

boss.

Boss Supportiveness

Table D—8 presents the four items which form the boss supportive-
ness cluster, their intercorrelations, and their correlations with the
other clusters. The pattern of intercorrelations is weakly rank one;
however it shows a consistent pattern of parallelism with other scales.
The content of these items measures the manager's perception of how

Supportive his boss is in the working environment.

Orientation Toward MBO

The three items of the orientation toward MBO scale (Table D-9)
concern the usefulness of MBO as a tool fOr helping the manager do his
job. The intercorrelations show a rank one matrix and a strong—weak
gradient. The pattern of the inter—scale correlations demonstrates that

these items hold together quite well.

Influence on Goals

Table D-lO presents the two items which compose the influence on
goals scale, their intercorrelations and correlations with other scales.
The items concern the objective setting process and seek to ascertain if

it is an authoritative or a participative process.

Goal Difficulty
The four items which constitute the goal difficulty scale measure

the challenge of the goals being sought by the manager. A response for

 

 

3.33“ V 2 5.3 2. 2m
oououomlou mo oofiflonan son .3 av m3
.83-«032 Euros-oucafiuowhon on n. vv In

166
2
fl
5
E
g.

-
N
um
um
-
.030... no". vooz an
Buzz: 3:9:qu n~ .

0.3:!— numz Eoocou 32. «N .mn mv vow

3:333. “woo . mm

Om: cm uwuhour: 0'

«05.30". 333.30 .3

an

to

no

anus: 55.3.. Eu 1 . l . 33m

:31» 8: B: 8.: :0» av So: to:

Sen «no» .«o
33.6 05 uiomuom 5 sex on soon :39? an: 05 no: 1650: to:

Sam .56» 3 mm noun? on: 05 :55 so» ou ofifiuflmnn so:
8:583“.

82 2039—. Smalucoto.v.8mfl 01.3”. 15000 Ouadh

 

:03

2:33: 3 S a. 3. 8m

85393 mo oufifionfl a." m~ «a a). m3
:ofiaaoomwé vuaxoméufiflowuua 3 - mm 3 «3
3326 :8 ma on a on 2m

.5333me new .3 a» nu 5n 2m

533.58 :8 v. 3 v o» :m

maaou :0 3.53%.: our 0 «a 0." 3m

Om: vumzoh .8335qu .n - 1 K ; mom
mmocoiukommam mmom on a. ~o an. mom

mmom 5E :oflummmflww an v» an co 3m

Dhfifimdn— ZHMZ EvUr—OU mmom NI 'N NH NV VOW
oucd>o~wz :00 an vn ma vm mom

8: 5 333.: 3 on 3 3 8m

xumfioi 258.25 .2 n» 3 on . 8m

“3.3a: Magoo..— vnm «5.5.2 .. canon

«.30» 33 uofiaoukom “cox Scan 309 Sex
arc vo>mouou 59A owing mo 2505 mufluumov amen uzoeouaum 52...

«mung 39.26.; «5.26 am:
5."... 28332 .Sox 3 9895.... so: «won .50» :5: 2332?" cu :0:

30m SPA Namesake.”
5 so» 2:0: 3.3 .50» cows: 5 3:55 may monakumov amen 52‘

200.30
59A how 309 .50» mo Emucou 05 monmuumou amen ucoeoumun some.

camouflage

mmoco>wuuonnaw noon .wom ammum Ilia: 3A2.

 

scum

 

 

 

167

this scale records a measure of the intrinsic worth of the manager's
performance. Although the scale as shown in Table D—ll is a weak rank
one matrix, the items show consistent internal and external patterns of

relationships.

Job Satisfaction

Table D-12 presents the two items which make up the job satisfac-
tion scale. The items are strongly related and consistent in their
behavior with other scales. A reSponse to this scale is an indication
of the satisfaction of the manager with his monetary reward for effort

expended and skill level possessed.

Goal Success

The three items of the goal success scale seek responses as to
how probable was success in attaining goals as well as how was success
perceived by the re5pondent. The item intercorrelations shown in Table
D—13 form a rank one matrix and a strong—weak gradient along the
diagonals. Additionally, the items have a similar pattern of correla—

tions with other scales.

Performance—Reward Association

Table D—l4 displays the two items which measure the perceived
relationship between effort expended by the manager and future salary
levels and promotions. These items are homogenous in content and hold

together very well internally and externally.

3.33. a {o 3 ...mm1 2...

oucuuomlou mo 35939.: 0
noun-«232 Bazomn3§ahomuom
mmuooam Goo

30m 5:. :oUumunUum ow s .
.028 .2: 302 3 N» a N can

25:5 353:5 .o c . v .o mow

953“... 5M... 58:3 3.3 ma ma .o.~ va «on

8 3:333. “new mm am on on new
«.10. On: :u umouofi: ow on Na mu Now

fiaéeoavfiuuunfl «n K 3 an Sm

 

1...; ".523 Ea . . 23m

.

name» an: no 035 3 059.3 32.30“? omofi mo xuuauflmg 05
23 noon no: .90» coup 25... 35,300.30 ago-332% Recon xoz

30:30.30 x5.- 03 v.2 no» doom :9» on

«Swine;
.50» how you 353035 «5 me .333qu no ~92 o5 nu: one.

So» how no» 333030
ufiaengov manor—om 05 we .3303“va we not: 05 3: “an:

833.525

Dazufiuuaagu .Zm diuu....s.:-n oz...—

 

 

EOUH

Hun—gnu”. nu ma. 3m
oocouonaou mo 359395 N." «a mam
3330034 uanozuooEEouuom Na

On: «95on :emumuzowko m o aom
mmoeoiukomnzm 33 N a wow
38 cum: :ofiuumwmwuwm « N 3m

.02: 8a 302 3 3 08
6.3.3: 3:335 VT. 5. V mom
93:: 5:. E856 33 - .3 3».

x0333 altars 2 on :5.

 

mm 2 3.
3. mm an
xwhum: «Sumo..— uzd . 4 . 3an

~50» how 353030 05 «£53m :o oozoawcw 309 05 no: on: 3

~50» now
mg?“ 030 goes—330v Ecomkom M533. .8 conga—n." “nos 05 on: 9.3 mm

newumiumoa

 

scum

28w :0 ouzoaafi .on 23m --.2-n flank

 

 

 

3.38m “mu 3 MN 3...

8a3§u 3 85.939: 3 .3 mu “3
833032 5330860538.": .3 nw 0 3m
3325 «com .ov .mn ms . 2m

Sflofluflaw 8.. 7 a 3 7 . 2m
533:3 :8 s 3. a. :m

2.8 8 8:33,: a. 2. 3. 2...

an: 7538. camuaucomho Na mm 4m 8m
nmocofiflommam 33 a." KN am new

«mom 5:. :owuoawmfiwm o a s 3m
3:8 3". «82 3 . a 8m

Edam: 35:35 an. an. mu mom

asnfin— 5%.: 53:00 mmom mu m." on van
3533”. 30w .3 s." an new

8: S. 3885 I 3 3 8m

xuaauoom 033030 mm aw nu . Sm

169

 

- -- - mn nu
. «N a~ «» “v.
an s» K. h.“

 

5.32.. 9533 23 “53383335 33m

undo» an:
333039 use-mougov "-3935 «5533 5 :95 82. Hammoousa .6:

3.85532. mo xuzmfinoun 05 «:23
noon :95 any :5: 62.3233 20: 33» oofiauomuon .30» :25

«on: ~35
:0» how «on 35.8030 05 «5532. 5 :9» £0: “33303 .5:

Suwmmuunua

383» .38..»3 :3»...sz$ 93.:

ma

av

v"

 

HOHH

 

 

guano”—
oocouomsu .«o 3593qu o ow... Mm
coauuwuomvd Easoznouqufiuomuom on N“

3326 :00 mm .N» 3.».

COA h N Mum
.HUHMmmuww no“. 00 00 Nam
xznumma 2.8 3 N :m
mamou :o 85:35 N no Sm

8: Ema; 833:3.5

oom
mmocgmflommsm mmom MM MM mom
«won :52 53333.3

now

.833 you vooz mm mm com

Essa: 352%: an an mom

Pusan“ fig F3280 mmom n." «a «on

3:338— :30 mm on , mom
on: 5 39:55 mm mm New
“.883“. $33.30 .on ‘ - 8m

WV Ni mm
2. rm... 3

 

3.5m: 33.va v5 203333335 Baum

pagan—ago 50% ”com :0»
333 .8: 393 many 5 non magma .332. a you 355 a was no» .3

«Ann
of :3: so» 93 vowmnmuam to: $3.30 Ea mid? .5; wcflovfimnou

535.3 moo

. 833333 n8. .2». 23m -125 «Ea...

mm

 

Eoun

 

 

 

 

 

170

Table D—l4.—- Scale 514, Performance—Reward Association

Item Description

34 To what extent will your actual job performance affect your future
salary increases? ‘

35 To what extent will your actual job performance now affect your
future promotions?

8 Do you feel you control the means of reaching your objectives?

Scale Intercorrelations and Loading Matrix

34 5" 61
.35 6; 64

501 3.5 39 Objective Feedback

502 44 36 Interest in MBO

503 43 48 Goal Relevance

504 31 27 Boss Concern With Failure
505 38 31 Influence Upward

506 18 16 Need For Policy

507 41 29 Satisfaction With Boss
508 46 38 Boss Supportiveness

509 32 40 Orientation Toward MBO
510 11 12 Influence 0n Goals

511 18 15 Goal Difficulty

512 49 29 Job Satisfaction

513 18 21 Goal Success

514 79 79 . Performance-Reward Association
515 44 4° ” Importance Of Competence
516 23 47, Residual

 

 

 

 

 

 

 

 

 

 

 

 

 

171

Importance of Competence

The three items which compose the importance of competence scale
(Table D—lS) measure the manager‘s need to achieve a certain level of
expertise or reward for successful job performance. The items form a
flat rank one matrix and have a consistent pattern of correlational

relationships with the other scales.

Residual

This scale (D—16) is composed of those items which did not meet

the criteria for inclusion in any of the above described scales.

The Revised Seven Scale Research Model

In an attempt to improve the internal scale reliabilities for
the research model, the fifteen scales were condensed into a seven scale
model. The cluster analyses of the scales for that model are presented

in the following sub—sections.

Importance of Goals

Table D—l7 presents the ten items which make up the Importance
of Goals scale. This macro scale has two scales which were defined in
the fifteen scale model as Goal Relevance (Scale 503) and Boss Concern
with Failure (Scale 504). The items of this scale demonstrate a consis-

tent pattern of correlational relationships with the other scales of the

 

system.

Goal Setting Behavior
This scale consists of two subscales--Influence on Goals and Goal

Difficulty--which together total six items. In Table D-18 the existence

 

 

 

1:33“ an S an on. 3 3m

 

8.52.58 uo 8.5.35 W» en 3 3 on ms.
83.333 intoxioufifiomuom aw mw n n 3 «3
3325 "new mu ow Nu m an. 3m
833333 e2. 2. an a. Z. a . Sm
313:3 :8 a w 9. 3 R :m
225 :o 3533: w. n. :n v: a _ 3m
8: P332. 833.55 3 on v. v 3. 8m
unoceiflonnam anon Nu 3 o N on new
«mom 5:.— Emuuawfluuw w. on n on ma 3...
.02.; .3". e32 o 3 .3. T 3 8m
Essa: 3.53.2: a mn m or ca mom
932d..— 5: Eco—Eu moon nu o“. o... o.... an «on
353301 13o m« on on o nm new
n On: 5 30.3%: m nu v: a ma “MM
0> 00 .
1 answer: 2 8 s .3 ~. 0 3
3 ,n I 3 ma .3.
w «a a." a «a ~¢
cu w.“ .3 a" A 3
.3 o «a o o a.“
a m." an a o 3 ma
5.5:: 55.3.— ..Eu .2 . . 33m
«so» 3

2532-3.— 233 on- uifiomfl .5: .5325: 538a use» :36 mm
52.. £53.:— 50» 5 *3: e5 3 9:39:55 .8: 3

"uk- uaogaov AB 343:0 23¢qu 51. 303:3 .«o .385: 2:. 3
"ma pen .9. :3: “533033 ous—Eu .«a 252.. 2:. 2

has: “sagas
fin: Paulo”. no“ .59» 3.: van so» enouwo mo p.505 05 3v .5: Nu

a noun

 

1:32.. 63. «1.6411615 26....

 

Hosanna

3539500 no 3:339; NM NM mm mm
:ofiuumuomwi vuasomuounmauowuom ow av m~ :m
3.326 goo mm Aw on Mum
838333 “.2, - nu v NS
.333me :8 a m 3 :m
mamou :0 3:355 A Na a 3m
on: who}; :ofiumuzomuo 3. mm m." mom
wmoeo>3uomm=w mmom mw on v." mom
38 5:. 533333 3 .3 3 8m
3:: eon e82 ow Z on com
Berg: 353%: mm on. 3 men
2.5.?"— nuwz Emu—BU mem «n «N vN vow
UUE>OHO¢ HmOU ov «v mN mom
on: 5. owners: we 2 3 , 8m
firearm $23.30 em; on. 3. 8m
um ~n N» R
N» n... n» ~m
«n .3 an an

xwkuuz M5250..— Ufi accueu u Ommuw

pounce—gamma Row .50» 233950 iii 5 5303
.nmuum v:- ucoicmflmaouua stflom mo waioom n ooaomuoneno no» on

833335

woofiakowkon pen hack
5 330.35 3 33 so» ‘3 mom—woken.“ vacuum. Sm: «583 wan: oh

pooh kaox v5— o>az 0:: Shaun
.350 55 pom weapon a co so» K5 sex 3 um aw ufiuhomam Io:

mucouomaou mo 3.830%: than onaum 1:».an 03:.

N

m

so:

 

gaudnog o« o on a“ nu. 5.

83.382 23378.3!omu3 on 3.. o o. T .7 .
58:38 .2. u n N _.

353930 .«o oufiuuonau z «
Szuo 3:3: I. .2 S

35.5323. ounfivuofimuuoflonaw N." o
3 .3323 ufiuuom "moo ,nn 3 av 3 3» 3
q/ n-oo we oocauuoga_ om ow N» on “N MN

1

.3 Z S. a

 

3 2
3 a." nu n~ m ad
H .8 mm 3 S ... o
N a 3 S 2 .K o
a . a." a o a «a Nn
. . .3 N.— .. o an “N
5.53. «533 EH 1 . ‘ . «Gum

 

fine» :3 mo 035 3 gun-5o 3:50on 3.05 MO 5333:.
05 98v :0: Jon 505,25: 3330?? 29:233. dacoflom .62
«33.30.25 x55 03 can so» doom sex on

2.53me .50» you you 33533 we 31.3qu mo 33: 05 3: «on:

So» you you
325030 «55035.» ~25qu mo .333va «0 1.53 05 «a: use.

~45» you 335930 a...» 5:53 :o 3:2:pr 39.. 05 E:— of

.30» how 9.530030

ago-E3309 unsung gauge» :0 85:35 3.0... 05 EH: SE

555525

hof'ngoa ufluuwom =00 .~cm ouuow :63 3an

ha

u-t

 

53¢: E553 EB .52“ » 33w

«338030 3:333...» gumfia
3 33¢ so» .3 33 .30» scum 0380.— 50; 3:8. ammonia awn:

 

mv
33.33.30 connfifiumo
. 93!. Va 3 v3.1a :0» MW on 3:03 30» use» Hoow so» ov vofioucou 30x 3
amoiuuompo
Eon—9332.. 3:030.“ .50» you modwuowkm 331:: 3.5 .50» us an
"mag“ 3.85023 .50» new muwufloim #3333 309 .50» v3 mv
Tau» um:
cu mama—So .35 av 30: .8: .335 «53 you uom :35 .32 $533.30 3
woucmEouuom
.28.. «5.532: 5 an: mmon “so; 333..“qu 22 monflumuv anon fife. v~
V wxfiwmsou «Z»
we 32:. 9:393 .gowuom umoa 30:3 :0» how «on 333030 3a n
:53 5x0 333% cu uuommoa 5m: vegan 3.330 3:30.30 .50» one... o
«no» on
as» «.356.» 35,300.23 30.2.; .50» mo oozmuquEm 9:230." 05 mm: s
Ewumwuumon 33H

m~mcw .«o customs: .mOm buwum IIRTQ canny.

 

 

174

of a rank two matrix is very evident. However, when viewed collectively
the macro scale has a consistent pattern of correlations with the other
scales. The content of the influence subscale requires the re5pondent
to assess who had the most influence in setting both his performance and
personal development goals. For the goal difficulty subscale, the
manager's response indicates whether goal setting is accomplished only
as a technique for expanding or developing the manager's potential for

effectiveness.

Superior—Subordinate Relationship

The twelve items which comprise the superior—subordinate relation-
ship scale are shown in Table D-19. This scale has three subscales--
Boss Supportiveness - items 25, 23, 13, 43; Influence Upward - items 8,
28, 27, SO; and Satisfaction with Boss - items 29, 30, 31, 26. All
three subscales assess the manager's feelings about his interaction and

influence with his boss.

Utility of MBO

Table D-ZO contains the eleven items which form the utility of
MBO scale. It is actually composed of four subscales. The first sub-
scale (items 46, 47) is described as need for policy which could also be
called need for structure. The second subscale (items 39, 40, 31) is
concerned with a specific technique (MBO) which may satisfy the need for
structure or policy.

The next three items (21, 22, 26) appraise the "higher level"
interest of the boss and the organization in MBO as a useful tool. The

remaining three items (5, 9, 11) evaluate the amount of results

 

 

 

 

O
m
0
ts
s
0
IA
5
'
rs
N
'0
n
s
L
o
v
In
0
(\
N
v
v
Q
m

175
K
N
v
n
a
n
a
n
o
n
O
1
0
K
'n'
C
co
0
O
a
Q
N
6
v

 

xenon: «causes an. . . . ofimum

 

333320 unogngov

3:35A— usox 3:535 5 ”an 32. .50» 3c 33%..» 5:: .6:
$2530.30 .50» co nutmeg .50» so xuaéoow :05» so» one: game :0:

. «mgwuuumno ucoanofigov

,, “mgr—om so magmas use» co xuunvuou :22.» so» one: :oumo .8:
«5393 8: 23 3:; so» on :0: 3:

Son

use» we 332. 05 «55°qu 5 coop 53m? 8: 05 was H33»; :0:
Sam .50» 3 mm .3»? an: 2.5 x55 so» 0.. 38329? to:

we...“ use» mcmaomkom 5. so»
30: cu mgavouonn v5 mvmufiom ufifiwow PS: 3 an. a... «5:69: .5:
«3 3 so» 35: 30a .50» was: :05. 3 :9» you 5 ma ufiuuofifi. .6:

, 2.393

8: 05 5 no: >533 05 x35 :0» av amateur: 5 we cuss yo:
plasma» 8: oz» 5 an: 33 use» as?» .5» av 30.355 5 we 5:.- .6:
~22 5. Ivan? on: of 3 332.. 9.0» .3: one as? 5:.- 1o:

:3 “armamen—

82 mo 3:3: .8». 33m log 93.;

.00-1
H

3 3 3 a... 3 t. 1 2 8m
Mn “N kw mm mm on ow an mom

on av on .0? on ow, NN. «n 9 an nn an “em

.«a om «v an mv am a. an on mu 0. so on

on ma «5 as am cu m. “N on mm 0. an. an

v. vn av a“ we «v mm mm ow mm mm v. on

ma an an an av n. on mm om m~ 0. am mN

o. m ow av mm mm on ma a. o~ on ~v cm

H _ an n «v mv «m s. on on mm a” an on “N
ac av an on am am on «a on o~ «w on w~

h . an \N mw aw Aw ow - ma ow - a“ m~ a
. an A, om ow Hv an on m“ 0H NH A“ - my
. mm N» am ow om ad on Na Na v” «N aw ma
a. a: an 9. an mn «N a 5“ ma mm nv n~

.ao an vv om w. on mm aw mm mm m. n. VN

 

. xmuum: uzmvmoq can . . . 33m

wougowkom

as; «sand 30.. .50» 59G uom a?» guanvoow 05 34.333 anon 52.x

. , smmon .50» 5.“: :0» oh“ comwmmuum .10:
$302“ 5; 9:33 5. 300. 50» mm voom so:

«302 .50» m« “mania a oflnmmau :0:

gen .30» 3 333 was» 309 .30» mo 23m

$.33 5 25: sex cocoa—$5. we asses 05 51. sex can vomwmwumm 30x
Eye... .59»
330?: $5 .5393 a goon zodiac .5: xma 3.5 .50» 30v 5qu :0:
~32. use» we 20.33% 05. 3:03.32. :3 30% 3a :0» on

3:30.30

30» «5:32 we v.52: on» Hahucou 30x .03 so» ov £353 5:: 3.

«.30» “mg 3550.,“qu .50» 8.5%

32. use» Eu voice?" so» omdmkm .«o 2525 any 353va anon file.
mesogoum am:
:3: £5333 .50» cu ouch—.50 :0: mmon .5: 5“.) 25323 0v no:
Sch .5: ”5.3.5qu

5 :0» m3”:— mmon .50» :02: E... non—Ra 23 monwuumov 33 some.
«.3053 use» you mmon use» we Eon—.50 05 nonmuumuv amen 5.25

:owumwkumoo

.A .oua .1 AU

"the .mom 23w lid 033

Am

2:8

0
m

5N
mN

w

M

v

MN
mu

 

Baum

 

 

 

 

176

oriented feedback from the boss regarding performance to d
War
performan C6

and personal development goals.

Importance of Competence
Table D—21 presents the three items '
Wthh make .
u th
. . . _ p 15 Scale,
their inter—correlations, and their correlations with the th
0 er Scales
These items demonstrate a flat rank one correl ‘ .
ation matrix and
are
parallel across the other scales. All three have the idea of
Co"‘Petence

or mastery of the job in common.

Job Satisfaction

The two items which make up this scale are shown in Table 0-22
along with their intercorrelations and their correlations with other
scales. Both items evaluate the manager's satisfaction with his pay
compared to his input (skills and effort) and his next best alternative
job. The items are very similar in content and correlation with other

scales. They form a flat rank one matrix of intercorrelations

Performance—Reward Association

Table D-23 presents the five items which comprise this scale,
It is evident that there are two subscales in this macro scale. The
first three items (14, 49, 15) concern the manager's feeling of goal
accomplishment. The other two items (34, 35) assess the subject's per—

ception of the relationship between actual performance on the job and

future increases in pay and promotion opportunity.

Residual
Table D-24 displays the five residual items, their intercorrela-

tions and their correlations with other scales.

 

 

 

 

 

 

 

 

i.

177

 

1:33: o 3..

53:25: vantozuoogowuon Nu. nu
Swuoounmudm new 00 no

00:32:30 mo 3:359: nu wu

- 8: mo .325: on on.
3:25:33. dugguonawuuoiunaw an an
.3328 uefluow “moo o c-

230 mo 3:539: an on

. a. no

, Mr A.

53-: 5.533 E: . . a . 01:5

33.. .61 .49»: $5 5 non union .5»qu -

35955 521 ”cow .5»
you 855 a v2 so» u“ an

05 5d: :9» on: 36.3: .5... .9330 El 2:: .50» 25.33

533.539

538:3!" n2. .2; 38m in”... 3%...

”>2—

 

 

.d «.1 r

 

3.33: me am Mm
Swuamuomg vagomooofiauomuom on nv on
533333 new ww n." v
oueouomaou we oocmukomam on. mm mm
v 8: mo 32%: 3. an . nw
c .n 4 u h<wbaEuu mN AN n."
3.26:3 unmuuow Hwou a." a." on
326 we ouemukomam av on om
N» N» N»
mn nn nn
no nn nn
“June: «cw—vac... Ea . . - . 33m

Sandman—ad» v5 gonna:
go.“
Show.“ #050 55 new weapon a ow :9» «mg» no»

Eaummuumon

00.590930 we 85 9395

5 3395:..— 3 «.qu no» 3 newness: flowwo 3a.“: u:

agave—Emma 13m .30» ”53353 31$ 5
93an macaw—on we mimoow w 5093 so» on

«3553qu
083 3:: o._.

.30“ .52" we; 26: on:
3 «a mu. ufiuuonuw .5:

£3 23m lino Beak

Nm

Am

flown

3:33“ 2. 2 SN 2 N.” 8m
Smuuwuom: unugxuoofiakowuom vn ow on Nm no «on

 

 

 

 

3.38.. an an an on .3 . 8m Evading as. 3 a. m 3 7 03
83-383 Ears—55.8%: an an 3 u a" , 3» 853.58 we oufitonfl E. 3 3 an 3 mom
contents» can wwn an no 3L u can on: we .323: vv vv aw n~ ow won
85%?8 mo 85989: v.» v» 3 3 mu 3... @2332”... SSSSfimLSSEw n» 3 3 R 2 8m
8: mo .933: .3” n“ on . N NN vow hams—Eon mzmuuom flaou «a ow m -I on Now
3:23:13. ounfivkoimuuov—onam no m~ m on v“ mom mason mo oufifloan: mv «v 0» Z mu an
m .3323 «5:8 :8 ..T n. 3. 2 S New .
.l. “:00 .«o ous—Nahum: on on on o nu . 8m om ac a." aw m mm
. .3 em 0 on u 3.
‘ ad n - tn .3 ma 3 ca 0 nu MN on m—
m «a ~u a «w «v om ow NN KN An 3
cu Na 3 «a n 3. m 5 nn 5» 9H 1
.3 o «a o a 2
an an s a . ow Na
Suva: ”.533 v5 . . . 33w
«tun: «51.3 v:- . . - 011.5
38389:— 333
so» 093.: .6: 35.38qu nan aasuuw .59» :«z 23.8 «ma: 3. mm
. . ~50» muonmwuunw bmnmw
09 233089.:— aausw 9n- ufiekomlu so: .camuasumn ucumvnm .50» .536 mm 8.03m .50» uuomwa oufiEowkom new 12.59» .50» :«3 ucouxo pan: 3. vn
Sun 239.:— uaox 5.. x3: 05 mm Slump—35 .8: 3 two» um:

no.5” compo Scan—330v 1:6“qu «5:13: 5 no» one: «awmmuuuam :0: .2
"0.: 25-9—33 b. 033.8 min—on :3... 303:3 mo non—5.. 2:. on
2:353: .305 we xufimfinoun 23 39.?

 

 

u: no“ P 5:. 18:83. .356 mo “52.. 2c. 2 Sou :9. Eu :5: 60523.3 “no: 28» 855.33.. 5% 55. 3

«nu-9A 33> «cm: “3:5
-93 cu Pal—Bu no“ .50» 35 gm :0» 93%.: mo «So-.- 05 3.. .5: - so» now you $530.30 05 afifiauua :« :ox v32 163325 :0: 3
5335:: lo: Ewumfiumuo 503

1:33.. .8». 28m {42. 2.1... 533253. Enzozéufifiouuom .3m 23» lung ”32.

.4- ‘m ..n.' ~~ _..

APPENDIX E

OTHER FINDINGS IN THE FIRM B MBO STUDY

 

Note: The data from two administrations of
the MBO questionnaire to the managers
of Firm B were used in this analysis.

 

APPENDIX E

OTHER FINDINGS IN THE FIRM B MBO STUDY

This appendix presents the development of a revised research
model for the managerial attitude system using the re5ponses to two

administrations of the fifty-five item MBO study questionnaire by the

one hundred seventeen (117) managers of Firm B. To do this the first

section describes a fifteen variable model deve10ped from a multiple

group cluster analysis of the data. This section also contains a

description of the scale content, the inter-scale correlations, and the

change score reliabilities. The second section is a presentation of a

condensed model of these fifteen scales and representative statistics

about each of the new scales. The third section contains an effects

diagram for the condensed model for Firm B managers.

Fifteen Scale Model

The strategy employed for the analysis of the Firm B data was to

revise the original fourteen scale model developed by Chesser using the

additional data available (items 48—55) for Firm B. The first step was

to perform a fourteen scale multiple group analysis and carefully examine

the factor loadings of those items in the residual. The residual in this

case included all items not in the Chesser fourteen scale model plus the

new items available. After examination of the resultant clusters, a new

scale of three items which measured the importance of competence to the

manager was discovered. The scale consisted of the following items:

179

 

 

 

 

180

51. How important is it that you do a better job than other people
who have had your job?

52. To what extent will effort increases by you lead to increases
in job performance?

53. Do you experience a feeling of personal accomplishment,
satisfaction in fu11 completion of goals?

It also was evident that some of the items in the residual belonged in
established clusters. Several different combinations of items within and
between clusters were tried before a decision was made that further
manipulations would only blur important content considerations.

During these various studies the same three criteria were applied
to the analysis of the scales. They were:

1. Items within a scale should be internally consistent (that is,
they should correlate with one another).

2. Items within a scale should be "externally" consistent (that is,
they should have a similar pattern of correlations with other
scales).

3. The scales should have reasonably similar content.
Table E-l presents a description of the scales for the fifteen scale
model for the managers of Firm B. Table E-2 disPIays the means and
standard deviations for each scale in the model.

In the deve10pment of the original model, a minimum threshold for
internal reliability was set at .50 (r11 = .50). Using this same cri-
teria, the internal scale reliabilities shown in Table B-3 are all .50 or
greater with the exception of scale eleven--Goal Difficulty-—for the
first administration of the questionnaire.

The prdblem of low change score reliabilities for several of the
scales is present in the revised scales as it was in the Chesser
researCh. Scales 4, 6, ll, 12, 13, and 15 all have Change score relia-

bilities less than .25. The strategy for improving these low change

 

 

181

Table 5-1 .—— Fifteen Scale Research Model

 

g:

Scales

L

Questionnaire
Item Numbers

 

10.
11.
12.
13.
14.

15.

16.

Objective Feedback
Interest in MBO

Goal Relevance

Boss Concern with Failure
Influence Upward

Need for Policy
Satisfaction with Boss
Boss Supportiveness
Orientation Toward MBO
Influence on Goals

Goal Difficulty

JOb Satisfaction

Goal Success
Performance-Reward Association
Importance of Competence

Residual

9,5,11

36,22,21
7,6,3,4,24,16,33,48
44,45

8,28,27,50

46,47

 

29,31,30,26
25,23,13,43
40,41,39
37,13
2,1,10,17
32,33
14,49,15
34,35
51,52,53

12,19,20,42,61

 

 

 

182

Table E—2.-- Means and Standard Deviations for
Fifteen Scale Model

 

 

M3223 Standard Deviations

Scale Description Time 1 Time 2 Time 1 Time 2
501 Objective Feedback 2.89 3.01 0.79 0.77
502 Interest in MBO 3.82 3.77 .84 .72
503 Goal Relevance 3.00 3.04 .48 .43
504 Boss Concern With

Failure 3.39 3.33 .79 .76
505 Influence Upward 3.33 3.40 .71 .77
506 Need for Policy 4.29 4.16 .68 .76
507 Satisfaction With Boss 3.50 3.56 .59 .53
508 Boss Supportiveness 2.92 1.98 .52 .54
509 Orientation Toward MBO 3.73 3.51 .89 .92
510 Influence on Goals 2.25 2.36 .69 .67
511 Goal Difficulty 2.98 2.99 .43 .49
512 Job Satisfaction - 3.30 3.34 .89 .88
513 Goal Success 3.08 2.98 .74 .79
514 Performance-Reward

Association 3.64 3.47 .84 .82

515 Importance of
Competence 4.36 4.30 0.54 0.55

 

 

 

 

 

 

 

 

183

Table E-3.-- Internal Scale Reliabilities and Change Score
Reliabilities for Fifteen Scale Model

 

 

 

Time 1- Change
Scale Internal Reliability Time 2 Score
Correlation Reliability
Time 1 (r11) Time 2 (r22) r12 rdd
1 .83 .84 .53 .61
2 .72 .80 .47 .55
3 .70 .72 .39 .52
4 .56 .68 .50 .24
5 .75 .74 .33 .63
6 .59 .59 .55 .09
7 .88 .88 .34 .82
8 .67 .57 .25 .50
9 .85 .87 .66 .61
10 .69 .68 .35 .52
11 .37 .54 .41 .10
12 .56 .59 .52 .13
13 .58 .57 .52 .13
14 .77 .76 .56 .48

15 .49 .59 ~56 '00

 

 

 

184

score reliabilities was to condense these scales into seven macro scales.

Seven Scale Model
__________________

From the analysis of the reliability of the change scores for the
fifteen scale model, it was decided that further scale condensation was
necessary in order to improve the internal reliability of the scales.
Wherever it was feasible, recognizing content and correlational patterns,
the scales of the fifteen scale model were collapsed into macro measures.
The resulting improved seven scale research model is shown in Table E-4.

See Appendix D for the cluster analysis of these scales.

Development of Effects Diagram

In this section, the data from the Firm B managers will be used
to construct an effects diagram of the behavioral system. The method-
ology for this development will be the same as that used in Appendix C.

Table E-S is a matrix of raw and corrected correlation coeffi-
cients between change scores for the improved model. There are sixteen
pairs of variables in the model which possess a significant relationship
(p < .05). These will be indicated in the effects diagram by a solid
line connecting the related pair of variables. Table E-6 is a matrix of
differences between cross-lagged panel correlations for the total sample
of Firm B managers. These differences are calculated using the formula:

Ar = rxp’z ' Jt‘>'1X2

= the correlation between a variable x at Time 1

where: rx '
1Y2 with a variable y at Time 2

r = the correlation between a variable y at Time I
ylxz with a variable y at Time 2
Differences found to be significant are indicative of an asymmetrical or

. . . . <
Causal relationship. There are seven significant relationships (p .20)

 

 

 

 

 

 

Table E-4.—— Revised Seven Scale Research Model

 

 

 

Questionnaire
Scale Description Item Numbers
1. Importance of Goals 7,6,3,4,24,16
48,38,44,45
2. Goal Setting Behavior 37,18,2,1,10,17
3. Superior—Subordinate 25,23,13,43,28,27
Relationship 50,29,30,3l,26
4. Utility of MBO 36,22,21,46,47
40,41,39,9,5,11
5. Importance of Competence 51,52,53
6. Job Satisfaction 32,33
7. Performance-Reward 14,49,15,34,35

Association

 

 

186

 

 

 

cm.
ma.

M

Hw>®H HO.
H®>®H mo.

up we monam> unwoflmficme

 

 

 

_ _ - _ _
m m m m m m
O.Hm 0.“ mm. “ma. mm. “mm. He. "am. am. _em. Ho. "wo.- oe. "ma. schemaoomw<
_ n . n m m u m m phmzomnoocmshomnom .n
. . _
. . _ . _ . . _ . . _
m .o.H “ O.H 0H. mma. em. mam. Hm. mom oo moo ma mmH. coauumMmflumm now .0
_ . .
_ . . . _ .
m “ O.H “ O.H mm. "mm. NN. mmm. NH. "ma. AN. "NN. ouemumdsou
" u m m u m m mo oocmpnodsH .m
_ _
_ . _ . _ .
m u " O.H " O.H mm. mow. mm. "om. am. "em. on: mo spafifiue .4
. . . . . _
_ _ . . . _
m u u " O.H “ O.H co. "ma. HN. "AN. daemeoapaamm
" u n m m m m oumcfiunondmuHOfiuomsm .m
. _ _
. . . . . _
m u u n m O.H “ O.H om. “He. aofi>msom wcwuuom Hmou .N
. . . . . .
_ _ . _ _ _
m u n u u " O.H " O.H mfimou mo oocmuaomEH .H
p .r n n r H h
A e m e m N H manmbam>

 

 

 

 

 

 

 

 

mucowoflmmoou coapmaouhou macaw omcmco wouooyaou paw 3mm

 

 

 

Hope: pomw>om How monoom omnmcu :oozuom mucofioflmmoou :oflpmaohaou pouoonuou paw 3am mo xfiyumz nu.mum canoe

 

 

 

187

between the variables in the model. By combining the analysis of these
tables with that of the previous table which contained the significant
dynamic (corrected) correlations, the effects diagram for the total
sample of Firm B managers has been constructed (Figure E-l). There are
five of the seven asymmetrical cross—lagged panel correlations which are
identified for significant dynamic correlations.

Changes in satisfaction with the jOb is the variable which
assumes the role of the "driver" variable in this diagram. This variable
is a member of three of the five causal relationships that were found and
in each case is the causal factor in changes of the other variables. One
implication is that JOb Satisfaction is affected by forces outside the
system under study. Personal pressures, economic pressures, and pres-
sures from other relevant persons could all form a part or a total
explanation for the important influence of this variable. Another possi-
bility is the "halo" effect or the importance (as perceived by the
manager) of participating in a research program of this nature.

There are four variables which are mutually reinfbrcing and form
the "center" of the attitudinal system. These variables--Changes in
Superior-Subordinate Relationship, Performance-Reward Association,
Utility of MBO, and Importance of Competence-—a11 share a positive rela-
tionship which indicates that as any one of them increases, the others
will follow. The Importance of Goals scale is one of the factors which
directly influences this center core of the model. As the Importance of
Goals increases, the Superior-Subordinate relationship increases and its
positive relationship with the other core variable is influenced.

The Importance of Goals variable is itself influenced by another

variable, Changes in Goal Setting Behavior. For example, if Goal

 

 

 

188
in MBO BehaviOral System

-l.—-Revised Model of Change Relationships

Figure E

 

 

   

new spa:
:ofluomMmfipmm
ca mowcocu

 

om:

 

 
 

 

 

 

 

 

CC Ted wk<uu<
wwwzom
noocmshomaom
cw mom:w:o

mo swash»:
:H momqmgu

 

 

 

 

 

 

 

 

 

magmGOflpmHom
ouwcflvhonsmuHOMHoQSm
:H mowqmnu

 

 

 

 

 

»

i

 

 

 

 

mHmoo

 

mo oocmpaomEH
up womcmgu

 

 

 

mucouomaou
mo cosmpHomEH
ca mowcmnu

    

 

 

 

 

 

 

90fi>w5om
wcflupow Hmoo
cw mowcwnu

 

 

 

 

 

189

Setting Behavior (the goal setting process as perceived by the sUbor-
dinate) became more autocratic than consultative, this change would be
positively related to changes in the importance (relevance, clarity,
priority) of goals.

The effects diagram for the total sample suggests that as there
are changes in job satisfaction these changes bring about changes in the
core of the system. These changes in the core variables influence and
bring about changes in the goal setting process and the importance of

goals.

 

 

 

 

 

 

 

 

 

 

 

APPENDIX E

MODERATED CHANGE RELATIONSHIPS IN THE MBO SYSTEM

 

 

Note: 1. Firm B "High Cool" Managers (n = 58)

2. Firm B "Low Cool" Managers (n = 59)

 

 

 

 

 

 

 

APPENDIX F

MODERATED CHANGE RELATIONSHIPS IN THE MBO SYSTEM

The Ghiselli Self Description Inventory was administered to the
managers of both organizations concomitant with the first administration
of the MBO Study Questionnaire. A computer search was conducted to
determine if the Ghiselli dimensions moderated the change relationships
found in the seven scale model. Chesser found four highly correlated
dimensions which moderated these relationships. They were perceived
occupational level, initiative, self assurance, and intelligence
(Chesser, p. 80 ). Those managers who rated themselves high on these
dimensions are called the "high cool" managers while those who rated
themselves low on these dimensions are called "low cool” managers.

For the replication of the original research model deve10pment
using the Firm B data, these "moderators" were used to sort the total
sample of one hundred seventeen managers into two subgroups. These two
subgroups, the ”high cool" managers (n=58) and the ”low cool" managers
(n=59) are intended to be homogeneous with respect to the four Ghiselli
personality dimensions.

Table F—l and Table F—Z present the inter—scale correlation
coefficients or the cross—lagged panel correlations between the first and
second administrations of the questionnaire for the Firm B ”high cool"
and “low cool” managers, respectively. The two samples do have several
distinct differences when the dynamic (corrected) correlations are cal-

culated and analyzed (Table F—3 and Table F—4). There are ten of these

190

 

191

Table F-.1--- Inter-Scale Correlation Coefficients Between
First and.Second Administrations of the
Questionnaire - Firm B "High Cool" Managers (n=58)

 

Second Administration

 

First Administration 1 2 3 4 S 6 7

 

 

l. Superior-SUbordinate

Relationship .47a .19 .42 .35 —.15 .10 .32
2. Goal Clarity and

Relevance .42 .41 .35 .26 -.22 .09 .27
3. Orientation Toward MBO .60 .48 .66 .48 .05 .30 .35

4. Performance-Reward
Association .36 .35 .38 .46 .11 .21 .10

5. subordinate's Influence .12 .16 .00 .13 .49 —.12 -.04

 

6. Satisfaction With Job .29 .34 .32 .37 .04 .56 .26 ‘

7. Perceived Success .12 .23 .23 .08 -.24 -.02 .45

 

aThe diagonal entries are correlations between a variable.
measured at time 1 and time 2. Off-diagonal entries are correlations
between a variable measured at time 1 and a second variable measured

at time 2.

Significant value Of r;

.26
.33

.05 level
.01 level

 

 

 

 

192

Table F-Z .-- Inter-Scale Correlation Coefficients Between
First and Second Administrations of the
Questionnaire - Firm B "Low Cool" Managers (n=59)

 

Second Administration'

 

First Administration 1 2 3 4 S 6 7

 

 

l. Superior-SUbordinate

Relationship .403 .27 .35 .23 -.14 -.02 .18
2. Goal Clarity and

Relevance .29 .51 .15 .16 -.18 -.17 .03
3. Orientation Toward MBO .10 .27 .67 .19 -.13 -.06 .13
4. Performance-Reward

Association .38 .31 .37 .69 .04 .20 .05

 

5. Subordinate's Influence -.O7 -.02 -.09 -.12 .21 .08 -.04 1
6. Satisfaction With Job .30 .00 .25 .34 .03 .44 .09

7. Perceived Success .31 .29 .01 .11 -.08 .14 .57

 

aThe diagonal entries are correlations between a variable
measured at time 1 and time 2. Off—diagonal entries are correlations
between a variable measured at time 1 and a second variable measured

at time 2.
Significant value of r:

.26
.33

.05 level
.01 level

 

 

 

 

 

 

193

Table p—3 .-- Matrix of Dynamic (Corrected) Correlation
Coefficients for Firm B "High Cool" Managers

 

Second Administration

 

First Administration 1 2 3 4 5 6 7

 

1. Superior—Subordinate

Relationship 1.00
2. Goal Clarity and
Relevance .54 1.00
3. Orientation Toward MBO .29 .21 1.00

4. Performance-Reward
Association .46 .36 .32 1.00

5. Subordinate's Influence .13 .28 -.06 .18 1.00

6. Satisfaction With Job .43 .27 .17 .41 .04 1.00

 

7. Perceived Success .25 .07 .16 .19 .24 -.06 1.00

 

Significant value of r:

.05 level = .26
.01 level = .33

 

 

 

 

 

 

 

194

Table F-4 .-- Matrix of Dynamic (Corrected) Correlation
Coefficients for Firm B "Low Cool" Managers

Second Administration

 

First Administration 1 2 3 4 5 6 7

 

1. Superior-subordinate

Relationship 1.00
2. Goal Clarity and
Relevance .41 1.00
3. Orientation Toward MBO .24 .28 1.00

4. Performance-Reward
Association .42 .10 .14 1.00

5. subordinate's Influence .01 .00 -.15 -.05 1.00

6. Satisfaction With Job .18 .06 .08 .23 .05 1.00

 

7. Perceived Success .24 .12 .19 .25 -.05 -.13 1.00

 

.26
.33

Significant value of r: .05 level
.01 level

 

 

 

 

 

 

 

 

 

195

dynamic (corrected correlations) which are significant at the .05 level
for the high cool managers and there are seven for the low cool managers.
The methodology for the inference of causality is the same as that of
Appendices C and E. To locate the significant asymmetrical relationships
between the cross-lagged panel correlations, a matrix of differences is
set up for each sample (Table F-S and Table F—6). The most efficient
mechanism for discussing these differences is the effects diagram for

each of the samples.

Effects Diagram - ”High Cool" Managers

The relationships between the seven system variables for the
Firm B ”high cool" managers is shown in Figure F-l. For each of the
significant dynamic (corrected) correlations (at the .05 level) a solid
line is used to connect the two change variables in the diagram. A

dotted line shows that a dynamic (corrected) correlation is significant

 

at the .10 level or better. The inference of causal priority suggests
three asymmetrical relationships among the cross-lagged panel correla-
tions for the high cool managers at the .10 level or better. These are
shown in effects diagram as unidirectional arrows.

The relationship between changes in goal clarity and relevance
and changes in superior-subordinate relationship is different for the
high cool managers than for the total sample. For the sample of high
cool managers, increases in the clarity and relevance of goals leads to
increases in the relationship between the manager and his superior.

A comparison of the effects diagram developed here with the
effects diagram for the Firm A "high cool" managers (Chesser, p. 112)

shows even greater differences. The most obvious difference is that the

 

 

 

196

Table F—S .-- Matrix of Differences Between Cross—Lagged Panel
Correlations for Firm B "High Cool" Managers

 

Second Administration

 

First Administration 1 2 3 4 5 6 7

 

l. Superior—Subordinate

 

Relationship -.23 -.18 -.Ol —.27 -.19 .20
2. Goal Clarity and

Relevance -.13 -.09 -.38 -.25 .04
3. Orientation Toward MBO .10 .05 -.02 .12
4. PerfOImance-Reward

Association -.02 —.16 .02
5. subordinate's Influence -.16 .20
6. Satisfaction With Job .28

7. Perceived Success

 

Notes: 1) These differences are calculated as the correlation for a
First Administration Variable (X ) with a Second
Administration Variable (Y ) minus the correlation of the
First Administration Varia 1e (Y ) with the Second
Administration Variable (X2) or rx1Y2 — rx2Y1

2) r = .19
zz‘zl .20 level = .24
.10 level = .31
3) Significant Difference ( Ar) at .05 level = .37
.01 level = .44

 

 

 

 

197

Table F~6 .—- Matrix of Differences Between Cross-Lagged Panel
Correlations for Firm B "Low Cool" Managers

Second Administration

 

First Administration 1 2 3 4 5 6 7

 

l. Superior—sabordinate

 

Relationship —.02 .25 -.15 -.07 -.32 —.13
2. Goal Clarity and

Relevance -.12 -.15 -.l6 -.17 -.26
3. Orientation Toward MBO -.18 -.04 -.31 .12
4. Performance-Reward

Association .16 —.14 —.O6
5. Subordinate's Influence .05 .04
6. Satisfaction With Job -.05

7. Perceived Success

 

Notes: 1) These differences are calculated as the correlation for a
First Administration Variable (X ) with a Second
Administration Variable (Y2) minus the correlation of the
First Administration Variable (Y1) with the Second
Administration Variable (X2) or erYZ - rXZYl

2) r _ = .19
22 zl .20 level = .24
.10 level = .31
3) Significant Difference ( Ar) at .05 level = .37
.01 level = .44

 

 

 

 

.IIIHD‘

198

 

 

Figure F-l.-— Effects Diagram of Change Relationships in the MBO

 

 

 

now :uflz
:0flpoemmfipmm
:fl mowcmsu

 

 

a

 

‘

 

Om: 68m 309
:oflpmqufiHo
Cw mmwfimgu

i

 

 

 

 

‘

 

 

 

cowueAUOmm<
euezomnoocmEHOMAom
cw mowqmnu

 

 

 

mflnmcofipmaom
opmcfleaonsmsHOfiaomam.lu.
:H momnmcu

 

 

 

Behavioral System for Firm B "High Cool" Managers

 

 

a

 

 

ooao>oaom w

4| 5.330 Row [AI

 

Y

 

mfieou
mice»: 5
mmooosm eo>fioo
used :w mowcmno

 

.
_

. _

_
a

 

 

CH mownmnu

 

 

 

 

mfimoo ao>o

oucosfimcH
opmafieaondm
:H mowqmzo

 

 

 

—>—_ ‘_ 4 “E... ‘nlu‘- V... — - 41 _1_ — ~ - - . ‘__..—

 

199

number of significant relationships between system variables is much less
for the Firm A managers. A part of this difference is attributable to
the larger sample size for the Firm B managers. It should be noted that
two of the five significant relationships in the effects diagram for
Firm A are negative, while all the relationships for Firm B are positive.
Another difference between the Firm A and Firm B managers is
that of the "driver" variable. Clearly, for the Firm A managers,
changes in Goal Clarity and Relevance cause Changes in the Relationship
between Superior and subordinate. While this same causal relationship
is found in the sample of Firm B managers, the Changes in Goal Clarity
and Relevance variable is "driven" by the variables, Changes in Subor—
dinate Influence over Goals and Changes in Satisfaction with the JOb.

The net result of these causal differences is that the driver and output

 

 

variables have reversed themselves for the two samples of "high cool"

managers.

Effects Diagram - "Low Cool" Managers

In the original study, the effects diagram for the Firm A
managers who rated themselves as "low cool" on the Ghiselli dimensions
was the same as the effects diagram for the total sample of Firm A
managers. This is not the case for the Firm B "low cool" managers
(Figure F-2). There are three significant dynamic (corrected) correla-
tion coefficients at the .05 level with four additional coefficients
significant at the .10 level or better. To assess causal priorities, it
is required to search the matrix of differences between cross-lagged
panel correlations at a significance level of .20. Since only one

asymmetrical relationship is found at that level (variable 1 with

 

 

 

 

 

200

 

new spa:
:oMHQNMmflpmw
:H momcmgu

 

.
_
. _

_
a

       

   

om: phmzop
:ofipmueoflno
:fl momqozu

_

 

 

deflpmfioomm<
enezomxoocwenomaoa
:H momcmnu

 

 

 

 

magm:0fiuwfiom
opmcfieaonnmuhoflAoQSm
:M momqmcu

 

 

 

—

-2.-- Effects Diagram of Change Relationships in the MBO
Behavioral System for Firm B "Low Cool" Managers

Figure F

 

 

ouna>ofiom
w sunnaau seem
:a momcmnu

 

 

 

 

  

mHmoo

mega»? fl
mmooosm eo>fiou
chem :fi momnmzu

    

 

 

 

 

 

 

 

201

variable 3), it was determined that the relationships for the "low cool"
managers were mutually reinforcing.

The relationships established in the effects diagrams for the
managers of Firm B were found to be moderated by the Ghiselli dimensions—-
perceived occupational level, initiative, self assurance, and intelli-
gence. This finding replicates a similar finding for the managers of

Firm A.

7.!

 

 

 

 

 

BIBLIOGRAPHY

 

 

 

BIBLIOGRAPHY

Bentz, V. J. ”The Sears Longitudinal Study of Management Behavior."
Washington, D. C.: Paper presented at the American Psychological
Association Division 14 Symposium: Longitudinal Approaches to
Management Research, Sept. 6, 1971.

Carroll, 5., and H. Tosi. ”Goal Characteristics and Personality Factors
in a Management—by-Objectives Program." Administrative Science
Quarterly, 15, No. 3 (Sept. 1970), 295—305.

 

Chesser, R. J. ”MBO As a Behavioral System: A Focus on Change
Relationships and Inferences of Causality." East Lansing,
Michigan: Unpublished Doctoral Dissertation, Michigan State
University, 1971.

Cronbach, L. J. "Coefficient Alpha and the Internal Structure of Tests."
Psychometrika, 16 (1951), 297—334.

Cronbach, L. J. Essentials of Psychological Testing. New York: Harper
8 Row, 1960.

 

Ghiselli, E. E. "Self—Description Inventory." Berkeley, California:
Typewritten Xeroxed copy, University of California.

Guttman, L. ”A New Approach to Factor Analysis: The Radex." In P. F.
Lazarsfeld (Ed.), Mathematical Thinking in the Social Sciences.
Glencoe, Illinois: Free Press, 1954.

Hunter, J. E., and R. J. Chesser. "MODERATOR: A Family of Computer
Programs Designed to Show Simultaneously the Moderating Effects and
Nonlinear Relationships Among a Set of Variables.” East Lansing,
Michigan: Unpublished paper, Michigan State University, 1971.

Hunter, J. E., and S. H. Cohen. ”PACKAGE: A System of Computer
Routines for the Analysis of Correlational Data." Educational and
Psychological Measurement, 29 (1969), 697—700.

Hunter, J. E., and S. H. Cohen. ”Correcting for Unreliability in
Nonlinear Regression." Cleveland Ohio: Paper presented at Mid-
western Society for Multivariate Experimental Psychology, May 1972.

Ivancevich, J. M., J. H. Donnelly, and H. L. Lyon. "A Study of the.
Impact of Management By Objectives on Perceived Need Satisfaction.”
Personnel Psychology, 23 (1970), 139-151.

 

202

 

203

Kavanagh, M. J. "The Owens—Illinois Longitudinal Study of Manager
Performance." Washington, D. C.: Annual Meetings of the American
Psychological Association. Paper presented as part of symposium:
Longitudinal—Causational Approaches in Management Research:
Working Examples, September 1971.

Lawler, E. E. "A Correlational—Causal Analysis of the Relationship
Between Expectancy Attitudes and Job Performance." Journal of

Applied Psychology, 52, No. 6 (1968), 462-468.

Likert, R. New Patterns of Management. New York: McGraw—Hill Book
Company, Inc., 1961.

 

Litwin, G. H., and R. A. Stringer, Jr. Motivation and Organizational
Climate. Boston: Division of Research, Graduate School of
Business Administration, Harvard University, 1968.

 

Locke, E. A., and J. F. Bryan. ”Cognitive Aspects of Psychomotor
Performance: The Effect of Performance Goals on Level of
Performance.” Journal of Applied Psychology, 50 (1966), 286—291.

Mendleson, J. "Managerial Goal Setting: An Exploration Into Its
Meaning and Measurement.” East Lansing, Michigan: Unpublished
Doctoral Dissertation, Michigan State University, 1971.

Meyer, H., E. Kay, and J. R. French. "Split Roles in Performance
Appraisal." Harvard Business Review, 43, No. l (1965), 123-129.

McNemar, Q. Psychological Statistics. New York: John Wiley and Sons,
Inc., 1962.

Nunnally, J. C. Psychometric Theory. New York: McGraw—Hill Book
Company, Inc., 1967.

Pelz, D. C., and F. M. Andrews. ”Detecting Causal Priorities in Panel
Study Data.” American Sociological Review, 29 (1964), 836—848.

Raia, A., "A Second Look at Goals and Controls.” California Management
Review (Summer, 1966), 49—58.

Raia, A., "Goal Setting and Self Control.” Journal of Management
Studies, 2, No. l (1965), 34—53.

Spearman, C. ”General Intelligence Objectively Determined and Measured."

American Journal of Psychology, 15 (1904), 201-292.

Tosi, H., and 8. Carroll. ”Some Factors Affecting the Success of
'Management By Objectives.'” Journal of Management Studies, 7,
No. 2 (May 1970), 209—222.

Vroom, V. H. "A Comparison of Static and Dynamic Correlational Methods
in the Study of Organizations.” Organizational BehaVior and Human

Performance, 1 (1966), 55—70.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

’nICHIGAN STATE

    

(1111):, 3 1) mm

9

U

 

11m“