EVALUATION OF WORK ENGAGEMENT AS A MEASURE OF

PSYCHOLOGICAL WELL-BEING FROM WORK MOTIVATION

by

MARK ANTONISON

A DISSERTATION

Submitted in partial fulfillment of the requirements

for the degree of Doctor of Philosophy
in
The Department of Industrial & Systems Engineering
and Engineering Management
to
The School of Graduate Studies
of
The University of Alabama in Huntsville

HUNTSVILLE, ALABAMA
2010

In presenting this dissertation in partial fulfillment of the requirements for a doctoral

degree from The University of Alabama in Huntsville, I agree that the Library of this
University shall make it freely available for inspection. I further agree that permission
for extensive copying for scholarly purposes may be granted by my advisor or, in his/her
absence, by the Chair of the Department or the Dean of the School of Graduate Studies.
It is also understood that due recognition shall be given to me and to The University of
Alabama in Huntsville in any scholarly use which may be made of any material in this
dissertation.
_________________________
(student signature)

___________
(date)

DISSERTATION APPROVAL FORM

Submitted by Mark Antonison in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Industrial & Systems Engineering and accepted on behalf of the
Faculty of the School of Graduate Studies by the dissertation committee.
We, the undersigned members of the Graduate Faculty of The University of Alabama in
Huntsville, certify that we have advised and/or supervised the candidate on the work
described in this dissertation. We further certify that we have reviewed the dissertation
manuscript and approve it in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Industrial & Systems Engineering.

_____________________________________Committee Chair
(Date)
_____________________________________

_____________________________________

_____________________________________

_____________________________________

_____________________________________ Department Chair

_____________________________________ College Dean

____________________________________ Graduate Dean

ABSTRACT
The School of Graduate Studies
The University of Alabama in Huntsville

Degree Doctor of Philosophy College/Dept. Engineering/Industrial & Systems

Engineering and Engineering Management
Name of Candidate Mark Antonison
Title Evaluation of Work Engagement as a Measure of Psychological Well-Being
from Work Motivation
Employee motivation is related to organizational improvements and employee
well-being; therefore, understanding motivation levels within the workforce is important
to achievement of organizational goals. Employee well-being metrics with sensitivity
beyond job satisfaction may provide managers with better feedback on employee work
motivation. Direct measurement of work motivation is difficult because it is a higher
order latent variable that must be inferred from measurements of other employee workrelated attitudes and behaviors. A relatively new construct called work engagement has
been developed to measure employee well-being at work. Work engagement is defined
by three highly correlated factors: vigor, dedication and absorption. This dissertation
evaluated using work engagement as a measure of work motivation. Goal theory and job
characteristics theory provided the framework for two research models. Antecedent
variables from the two motivation theories were treated as predictors of an employees
level of work engagement, including goal orientations, job goal attributes, occupational
self-efficacy, and job characteristics. The three factors of work engagement were treated
as dependent variables.

The models were tested using a cross-sectional field study approach. Research
model variables were operationalized in a web survey by combining measurement scales
of previously validated survey instruments from the literature. The web survey was
distributed to graduate engineering students with jobs and to all employees of an
engineering technical services company.
Results were analyzed to evaluate the strength and significance of correlations
and regression paths between variables. Structural equation modeling was employed to
evaluate goodness-of-fit between the theoretical research models and the sample data set.
The goal theory research model produced favorable results. The implication for
management is that work engagement can be used to measure work motivation in an
engineering technical services organization. The results show that goal theory, when
deployed in an organization that promotes healthy workplace practices, will have a
positive effect on employee psychological well-being from increased work motivation.
This effect can be reliably and validly measured by managers using the 9-item version of
the Utrecht Work Engagement Survey instrument.

Abstract Approval:

Committee Chair ____________________________________

Department Chair ____________________________________
Graduate Dean

____________________________________

ACKNOWLEDGMENTS
The work described in this dissertation would not have been possible without the
assistance of a number of people who deserve special mention. First, I would like to
thank Dr. Sampson Gholston and Dr. Anthony Morris for their guidance and support
throughout all stages of the work. Dr. Phillip Farrington provided me with technical
insights that challenged me in a positive manner to reach for my personal goals in this
academic program. Dr. Dawn Utley provided the original encouragement for me to start
my graduate program of study here at the University of Alabama in Huntsville.
Dr. Michael P.J. Benfield provided invaluable course instruction about the research
methods employed in this research study. Other faculty members of my department have
always been very helpful with comments and suggestions.
I would like to thank my wife, Anastasia, and my children, Alexander, Angelina
and Shinichi, who encouraged me to begin work on this degree and supported me
throughout my graduate program of study. This dissertation was truly a family affair.
Anastasia and Alexander, being fellow students enrolled with me in the same Industrial &
Systems Engineering Department here at UAH, were a constant source of inspiration.
Angelina and Shinichi, also fellow students enrolled in the UAH School of Liberal Arts
and School of Science, provided support and inspiration by their strong work ethic and
unwavering commitment to UAH.

TABLE OF CONTENTS
Page
List of Figures..............................................................................................................xii
List of Tables................................................................................................................xvii
Chapter
I. INTRODUCTION AND OBJECTIVES.................................................................1
......................................................................................................................................A.
Background..................................................................................................................1
......................................................................................................................................B.
Employee Well-being...................................................................................................3
......................................................................................................................................C.
Description of the Problem..........................................................................................4
......................................................................................................................................D.
Work Motivation..........................................................................................................4
......................................................................................................................................E.
Research Objective.......................................................................................................6
II. LITERATURE REVIEW.......................................................................................9
......................................................................................................................................A.
Introduction..................................................................................................................9
......................................................................................................................................B.
Psychological Well-being.............................................................................................10
......................................................................................................................................C.
Work Motivation..........................................................................................................11

D. Goal Theory of Work Motivation...............................................................13

...................................................................................................................1.
Goal Attributes..............................................................................................................17
...................................................................................................................2.
Self-Efficacy.................................................................................................................20
...................................................................................................................3.
Feedback and other Goal Moderators..........................................................................22
...................................................................................................................4.
Goal Orientation as a Measure of Values and Personality...........................................23
...................................................................................................................5.
Relevance of Goal Theory............................................................................................25

E. Job Characteristics Theory of Motivation..................................................26

...................................................................................................................1.
Core Job Characteristics...............................................................................................27
...................................................................................................................2.
Internal Work Motivation.............................................................................................29
...................................................................................................................3.
The Job Diagnostic Survey..........................................................................................30
F. Comparison of the Two Motivation Theories............................................32
G. Theory of Engagement...............................................................................33
H. Other Explanations of Engagement............................................................36

I. Review of Work Engagement Research.....................................................36

...................................................................................................................1.
First Empirical Study of Burnout and Work Engagement............................................38
2. An Empirical Study of Daily Fluctuation of Work Engagement..........39
3. Work Engagement Study with Respect to Service Climate..................40
4.
5.
6.
7.

Antecedents and Consequents of Work Engagement...........................41

Work Engagement and the Job Demands-Resources Model................42
Work Engagement Discriminant Analysis Study.................................44
Work Engagement in Japan..................................................................44

...................................................................................................................8.
Work Engagement Studies in the United States...........................................................45
...................................................................................................................9.
Assessment of Construct Validity from Published UWES Studies..............................46
J. Need for Further Research.........................................................................47
III. RESEARCH STATEMENT..................................................................................49
A. Introduction................................................................................................49
B. Research Issues..........................................................................................51
C. Description of Research Study...................................................................54
D. Research Hypotheses..................................................................................55
E. Significance of Research............................................................................55
IV. RESEARCH METHODOLOGY..........................................................................57
A. Research Approach.....................................................................................57
B. Two Work Motivation Research Models....................................................59
C. Description of Test Instruments.................................................................62
1. Utrecht Work Engagement Survey Instrument.....................................63
9

2. The Job Diagnostic Survey and Revised Job Diagnostic Survey.........64

3. Achievement Goal Orientation Survey Instrument..............................66
4. Job Goal Specificity and Job Goal Difficulty Measurement Scales...........68
5. Occupational Self-Efficacy Measurement Scale..................................69
D. Demographics Measurements....................................................................70
E. The Pilot Study...........................................................................................71
F. The Full Study............................................................................................73
G. Full Study Approach to Data Collection....................................................73
H. Assessment of Factor Purity through Validity Studies...........................................74
V. RESEARCH RESULTS..........................................................................................77
A. Data Collection Results..............................................................................77
B. Demographics Analysis..............................................................................78
C. Analysis of Survey Results Utilizing Descriptive Statistics......................83
D. Measurement Scale Reliability Analysis....................................................86
E. Correlation Coefficient Analysis of Latent Variables.................................88

......................................................................................................................................F.
Confirmatory Factor Analysis......................................................................................93
1. Factor Analysis Results for the Pilot Study..........................................95
2. Factor analysis Results for the Full Study............................................100
......................................................................................................................................G.
Sample Size Analysis...................................................................................................110
......................................................................................................................................H.
Regression Analysis MPS versus Work Engagement..................................................112

10

......................................................................................................................................I.
Structural Equation Modeling......................................................................................114
1. Structural Equation Modeling Methodology.........................................114
2. Goodness-of-Fit Measures for Model Evaluation.................................116
3. SEM Analysis of Work Engagement.....................................................119
4. SEM Analysis of Goal Theory Research Model...................................126
5. SEM Analysis of Job Characteristics Research Model.........................129
VI. CONCLUSIONS AND RECOMMENDATIONS................................................134
......................................................................................................................................A.
Principal Conclusion....................................................................................................134
1. Work Engagement Factor Structure......................................................135
2. Evaluation of Goal Theory Structural Equation Model Results............136
3. Evaluation of Job Characteristics Structural Model Results.................138
4. Correlation versus Causation................................................................139
B. Recommendations......................................................................................140
1. Goal Setting Practices...........................................................................141
2. Areas for Further Study.........................................................................142
APPENDIX A: INSTITUITIONAL REVIEW BOARD FORMS...............................143
APPENDIX B: INDIVIDUAL SURVEY INSTRUMENTS.........................................146
APPENDIX C: COMBINED WEB SURVEY INSTRUMENT...................................153
APPENDIX D: DESCRIPTIVE STATISTICS MEASUREMENT VARIABLES......171
APPENDIX E: SAMPLE SIZE ANALYSIS...............................................................178
APPENDIX F: REGRESSION ANALYSIS MPS vs. WORK ENGAGEMENT.......183

11

APPENDIX G: STRUCTURAL EQUATION MODELING ANALYSIS..................191

REFERENCES.............................................................................................................249

12

LIST OF FIGURES
Figure

Page

1.1 The Framework for a Healthy Workplace (Grawitch et al. 2006).......................2

2.1 Flow of Topics Covered in Literature Review.....................................................9
2.2 Graphical Depiction of Work Motivation (Latham and Pinder 2005).................12
2.3 Goal Theory Model of Work Motivation (Locke and Latham 2002)..................14
2.4 Four Dimensions of Achievement Goal Orientation (Baranik et al. 2007)..............24
2.5 Job Characteristics Theory (Hackman and Oldham 1975)..................................27
2.6 First Published Theory of Engagement (Kahn 1990)..........................................33
2.7 Engagement is a Continuum Variable (Kahn 1990, 1992)..................................34
2.8 Theory of Work Engagement (Schaufeli et al. 2002)..........................................37
2.9 Theoretical Dimensions of Work Engagement (Schaufeli et al. 2002)................37
2.10 Model of Daily Fluctuation of Work Engagement (Sonnentag 2002)...............40
2.11 Work Engagement as an Independent Variable (Salanova et al. 2005)..............41
2.12 Antecedents and Consequences of Work Engagement......................................42
(Koyuncu et al. 2006)
2.13 Job Demands-Resources Model of Work Engagement......................................43
(Schaufeli and Bakker 2004)
2.14 Second Job Demands-Resources Model of Work Engagement.........................43
(Bakker et al. 2006)
3.1 Attributes of Work Motivation versus Dimensions of Work Engagement...........51
3.2 Attributes of Job Characteristics Theory versus Work Engagement Theory............52
4.1 Proposed Goal Theory Research Model of Work Engagement...........................60
4.2 Proposed Job Characteristics Research Model of Work Engagement.................61
5.1 Confirmatory Factor Analysis Web Survey Instrument Pilot Study....................96
5.2 Confirmatory Factor Analysis Work Engagement Survey Instruments Pilot Study.........97
5.3 Confirmatory Factor Analysis Job Diagnostic Survey Instrument Pilot Study................98
13

5.4 Confirmatory Factor Analysis Goal Orientation Survey Instrument Pilot Study.............99
5.5 Confirmatory Factor Analysis Occupational Self-efficacy Scale Pilot Study..................100
5.6 Confirmatory Factor Analysis Web Survey Instrument Full Study.....................101
5.7 Confirmatory Factor Analysis Work Engagement Survey Instrument Full Study...........103
5.8 Confirmatory Factor Analysis Job Diagnostic Survey Instrument Full Study.................104
5.9 Confirmatory Factor Analysis Goal Orientation Survey Instrument Full Study..............105
5.10 Confirmatory Factor Analysis Goal Theory Antecedent Variables Full Study...............106
5.11 Confirmatory Factor Analysis Goal Theory Research Model Full Study.....................107
5.12 Confirmatory Factor Analysis Job Characteristics Research Model.................108
Full Study
5.13 Work Engagement Score vs. Motivating Potential Score Full Study................113
5.14 Structural Equation Modeling Process...............................................................115
5.15 Second Order Three-Factor 17-item UWES Model of Work Engagement............119
5.16 Second Order Three-Factor 9-item UWES Model of Work Engagement...............121
5.17 Final Structural Equation Goal Theory Research Model...................................127
5.18 First Six Factor Job Characteristics Structural Equation Model........................129
5.19 Final Five Factor Second Order Job Characteristics Three Factor..........................132
Second Order Work Engagement Model
B.1 UWES Work Engagement Survey Instrument (Schaufeli et al. 2006)................146
B.2 Job Goal Attributes Measurement Scales (Wright, Bradley E. 2004).................147
B.3 Occupational Self-efficacy Measurement Scale..................................................148
B.4 Achievement Goal Orientation Measurement Scale...........................................149
B.5 The Revised Job Diagnostic Survey....................................................................151
B.6 Two Related JDS Job Characteristics Measurement Scales................................152

14

D.1 Box Plot Work Engagement Measurement Variables Pilot Study Results..........171
D.2 Box Plot Work Engagement Measurement Variables Full Study Results...........171
D.3 Box Plot Job Characteristics Measurement Variables Pilot Study Results.........172
D.4 Box Plot Job Characteristics Measurement Variables Full Study Results..........172
D.5 Box Plot Self-efficacy Measurement Variables Pilot Study Results.....................173
D.6 Box Plot Self-efficacy Measurement Variables Full Study Results.......................173
D.7 Box Plot goal orientation Measurement Variables Pilot Study Results..............174
D.8 Box Plot Goal Orientation Measurement Variables Full Study Results.............174
D.9 Box Plot Goal Attributes Measurement Variables Full Study Results................175
E.1 Communalities for Combined Web Survey Instrument......................................180
E.2 Communalities for Goal Theory Research Model Variables...............................180
F.1 Work Engagement Score vs. Motivating Potential Score Pilot Study.................183
F.2 Results of Regression Analysis Work Engagement Score...................................184
vs. Motivating Potential Score Pilot Study
F.3 Work Engagement Score vs. Sum Job Characteristics Score..............................184
Pilot Study
F.4 Results of Regression Analysis Work Engagement Score vs. Sum.....................185
Job Characteristics Score Pilot Study
F.5 Work Engagement Score vs. Motivating Potential Score....................................186
Full Study
F.6 Results of Regression Analysis Work Engagement Score...................................187
vs. Motivating Potential Score Full Study
F.7 Work Engagement Score vs. Sum Job Characteristics Score....................................187
Full Study
F.8 Results of Regression Analysis Work Engagement Score vs. Sum.....................188
Job Characteristics Score Full Study
F.9 Four-in-one Residuals Plot for MPS Regression Analysis..................................189
15

Full Study
F.10 Four-in-one Residuals Plot for Sum Job Characteristics Score.........................189
Regression Analysis Full Study
F.11 Probability Plot for Regression Analysis Variables Full Study..........................190
G.1 Typical Structural Equation Measurement Models for Latent Variables............195
G.2 Single-Factor UWES 17-item Model of Work Engagement...............................201
G.3 First Order Three-Factor UWES 17-item Model of Work Engagement.................202
G.4 Second Order Three-Factor UWES 17-item Model of Work Engagement.............203
G.5 One-Factor UWES 9-item Model of Work Engagement....................................206
G.6 First Order Three-Factor UWES 9-item Model of Work Engagement...............207
G.7 Second Order Three-Factor 9-item UWES Model of Work Engagement..........208
G.8 One Factor Revised Job Diagnostic Survey Job Characteristics Model.............211
G.9 First Order Five-Factor Revised JDS Job Characteristics Model.......................212
G.10 Second Order Five Factor Revised JDS Job Characteristics Model.................213
G.11 First Order Seven Factor Job Characteristics Model........................................214
G.12 First Order Four-Factor Goal Orientation Model.............................................217
G.13 Second Order Four Factor Goal Orientation Model (invalid solution).............218
G.14 Three Factor First Order Model of Goal Theory Antecedents to Work............223
Motivation
G.15 Goal Theory First Order Three Factor Work Engagement Model....................227
G.16 Goal Theory Second Order Three Factor Work Engagement Model................228
G.17 Goal Theory One Factor Work Engagement Model.........................................229
G.18 First Six Factor Job Characteristics Structural Equation Model.......................234
G.19 Five Factor First Order Job Characteristics One Factor 9-item Work..............237
Engagement Model

16

G.20 Five Factor First Order Job Characteristics Three Factor First Order..............238
9-item Work Engagement Model
G.21 Five Factor First Order Job Characteristics Three Factor Second Order..........239
9-item Work Engagement Model
G.22 Five Factor Second Order Job Characteristics One Factor 9-item Work..........240
Engagement Model
G.23 Five Factor Second Order Job Characteristics Three Factor First Order..............241
9-item Work Engagement Model
G.24 Five Factor Second Order Job Characteristics Three Factor Second Order............242
9-item Work Engagement Model
G.25 Work Engagement Multi-Group Analysis Settings...........................................244
G.26 Short Version Three Factor Second Order Work Engagement..........................245
Structural Equation Model Multi-Group Analysis

17

LIST OF TABLES
Table
Page
4.1 Summary of Individual Test Instruments Selected from the Literature.................62
5.1 Combined Web Survey Instrument Data Collection Results.................................77
5.2 Gender, Full Study and Pilot Study........................................................................79
5.3 Ethnicity, Full Study and Pilot Study.....................................................................79
5.4 Level of Supervisory Duties, Full Study and Pilot Study......................................79
5.5 Level of Education, Full Study and Pilot Study.....................................................80
5.6 Current Job by Category, Full Study and Pilot Study............................................80
5.7 Number of Years Employed, Full Study and Pilot Study.......................................81
5.8 Years Employed at Current Job, Full Study and Pilot Study.....................................81
5.9 Statistics of Engineering Services Company Full Study Population.....................82
Compared to Full Study Sample Results
5.10 Comparison of Aggregate Latent Variable Measurement Results Pilot and Full Study.......85
5.11 Measurement Scale Reliability Estimates............................................................87
5.12 Correlations Between Work Engagement and Goal Theory Pilot Study...................89
5.13 Correlations Between Work Engagement and Job Characteristics Pilot Study............89
5.14 Correlations Between Work Engagement and Goal Theory Full Study.....................90
5.15 Correlations Between Work Engagement and Job Characteristics Full Study.............91
5.16 Correlations Between Goal Theory and Job Characteristics Full Study......................91
5.17 Results of Regression Analysis Work Engagement vs. Job Characteristics.............113
5.18 SEM Analysis Results for Full Version 17-item UWES Work Engagement............120
Survey Instrument Compared to Published SEM Results

18

5.19 SEM Analysis Results for Short Version 9-item UWES Work Engagement............122
Survey Instrument Compared to Published SEM Results
5.20 Tests for Invariance Work Engagement Multi-Group Analysis...........................123
5.21 Scale Items Used in the Two Structural Equation Research Models...................125
5.22 SEM Analysis Results for Final Goal Theory Research Model...........................126
5.23 Standardized Regression Path Coefficients for Goal Theory Research Model................128
5.24 SEM Analysis Results for Six Job Characteristics Research Models...........................133
C.1 Test Instrument Scale Items vs. Position in Web Survey Instrument.........................154
D.1 Descriptive Statistics 71 Scale Items Measured in Web Survey Instrument..............176
E.1 Summary of Minimum Sample Size Recommendations from Literature.............181
E.2 Summary of Statistics for Minimum Sample Size for Research Study.................182
G.1 Structural Equation Modeling Process Steps........................................................191
G.2 Goodness-of-Fit Results for Structural Equation Measurement Models..............196
G.3 Test for Equal Variances Between Work Engagement Factors..............................199
G.4 SEM Analysis Comparative Results for the Full 17-item UWES Work...............204
Engagement Survey Instrument
G.5 SEM Analysis Results for the Short Version 9-item UWES Work.......................209
Engagement Survey Instrument
G.6 SEM Analysis Comparative Results for the Revised Job Diagnostic Survey..............215
G.7 SEM Analysis Results for the Achievement Goal Orientation Survey.................219
G.8 SEM Analysis Results for the Goal Theory Antecedent Variables Model.................224
G.9 SEM Analysis Results for Goal Theory Models of Work Engagement................231
G.10 Regression Path Coefficients for Goal Theory Second Order.............................232
Three Factor Work Engagement Model
G.11 SEM Analysis Results for Job Characteristics Research Models........................243
G.12 Results Work Engagement Multi-Group Nested Models Analysis.....................246
19

G.13 Tests for Invariance Work Engagement Multi-Group Analysis..........................247

20

CHAPTER I

INTRODUCTION AND OBJECTIVES

A.

Background
In today's competitive business environment, a technical organization depends

more than ever on the efficient functioning of its most valuable resource, its employees.
Engineering management professionals can benefit from an increased understanding of
the underlying psychological factors that influence the attitudes and behavior of their
technical work force. In recognition of this fact, a holistic approach to the understanding
of employee attitudes is emerging in occupational health psychology (OHP). This
holistic approach extends beyond simple measurement of job satisfaction and job
performance, moving toward an integrated strategy of organizational behavior known as
healthy workplace practices. The concept of healthy workplace practices has evolved
initially from a narrow focus on the financial bottom-line to more recently include health
and fitness programs for employees. Corporate sponsored health initiatives were
previously concerned with avoidance of unhealthy practices. Current best practices are
now focused on the optimization of employee psychological health (American
Psychological Association 2008).

The healthy workplace can be defined as any organization that maximizes the
integration of employee goals for his or her well-being along with company objectives
for profitability and productivity (Grawitch et al. 2006). Empirical links have been
demonstrated between healthy workplace practices, employee well-being, and positive
organizational outcomes. A framework for relating these three concepts and several key
attributes is depicted in Figure 1.1 (Grawitch et al. 2006). This model includes a
representative summary of the components of the overall framework of a healthy
workplace.

Employee Well-being
Physical/Mental Health
Stress (reduction)
Motivation
Commitment

Healthy Workplace
Practices

Job Satisfaction

Work-Life Balance

Morale / Climate

Health & Safety

Recognition
Organizational Improvements

Employee Involvement

Competitive Advantage

Employee Growth &

Performance/ Productivity

Development

Absenteeism / Turnover
Accident / Injury Rates
Product / Service Quality
Customer Service / Satisfaction

Figure 1.1 The Framework for a Healthy Workplace (Grawitch et al. 2006)

The concept of a healthy workplace highlights the modern organization's need to

broaden the range of desired business outcomes to include multiple measures of
employee well-being considered important to performance measurement. The measures
listed in the model can provide management guidance for development of a more
comprehensive set of indicators for employee and organizational health.
It should be noted that motivation and job satisfaction are treated as separate and
distinct measures of employee well-being in the model. While many companies regularly
survey their employees to determine employee job satisfaction levels, there is much less
evidence of agreement in both business and academic literature how to measure
employee motivation levels. Identification of a reliable performance metric for a
workers motivation level that is separate from job satisfaction would provide managers
with two complementary options for employee well-being indicators.

B.

Employee Well-being
Measures of employee well-being can be considered indicators of future

employee performance. Grant et al. (2007) defines employee well-being in terms of the
overall quality of an employees experience and functioning at work. They state that wellbeing is defined by its three core dimensions: psychological, physical, and social. The
psychological dimension is concerned with an individuals subjective experience and
functioning. The physical dimension consists of bodily health and functioning. The
social dimension of well-being includes the relational experiences between individuals
(Grant et al. 2007).

C.

Description of the Problem

With respect to measures of employee well-being, there is a need to broaden the

range of performance measures important to an organization. Keeping employees

working to their full potential is the ultimate goal of management efforts to maintain and
increase employee motivation in their work. Employee motivation factors are related to
other factors that contribute to the accomplishment of organizational improvement goals.
A simple metric with sensitivities beyond that of job satisfaction that provides managers
with more direct feedback on employee motivation is needed. A psychological test that
can reliably measure the strength of the relationship between workplace practices and
employee work motivation would be very useful to managers concerned with the
occupational health of their workforce.

D.

Work Motivation
Work motivation is an essential aspect of organizational behavior. It is the driving

force within employees which impels them to action, putting forth effort on behalf of
their organization. The resulting level of effort is measured in terms of job performance.
Managers need to know what motivates their employees as they try to maximize the job
performance of their workforce to achieve the aims of the organization. Work motivation
is considered a latent psychological variable. This means it cannot be measured directly
but must be inferred from the measurement of other more directly observable attitudinal
or behavioral variables.
In order to better understand the attitudes and behaviors of employees,
organizations use surveys to obtain feedback from the workforce. These surveys are

typically self-report measures of the worker's own perceptions and attitudes about their
work environment. Historically, asking employees to complete a job satisfaction survey
is a standard approach used by management to collect data. Through evaluation of job
satisfaction survey results, managers try to make inferences about employee motivation.
When organizations follow this approach to measure employee behavior, its management
assumes that they can expect higher performance from a workforce that has a higher level
of job satisfaction. Restated, the expectation of these managers is that increased job
satisfaction results in improved job performance.
Unfortunately, many managers today still do not know that empirical studies
conducted as far back as the 1960's and 1970's refuted that traditional viewpoint with
convincing evidence that the reverse causal path was actually more likely, that increasing
job performance influences increased job satisfaction (Lawler and Porter 1967). In the
decades that followed, additional studies were subsequently published with conflicting
results concerning the size of any actual effect of this theorized causal path on both
performance and satisfaction. In an effort to put the academic debate and its various
conflicting research claims into proper perspective, an authoritative meta-analysis was
finally conducted by Iffaldano and Muchinsky (1985) which had a damping effect on
further research of the subject. Their conclusions pointed to a very weak correlation of
0.17 between job satisfaction and job performance. Since that study was published, no
better psychological construct has been promoted in the academic literature, and today
the measurement of work motivation using a job satisfaction survey continues to be
accepted practice for both industry and academia (Mayer et al. 2007).

A more recent analysis of the job satisfaction and job performance relationship
incorporated improved meta-analysis techniques (Judge et al. 2001). The authors
concluded that the expected correlation between the two measures may be closer to 0.30,
a value which organizational behavior researchers usually consider the lower limit for
noteworthy correlation between two behavioral variables (Muchinsky 2006). This study
also presented a qualitative analysis including a discussion of seven possible causal path
models relating job satisfaction and job performance. Their conclusion was that an
integrated model which included both path directions and up to seventeen other mediator
and moderator variables provided the best explanation of the relationship between these
two psychological constructs. While this integrated job satisfaction model is interesting
from an academic perspective, managers of organizations would benefit from the
development of a simpler and more practical explanation of how measurement of
employee job satisfaction relates to performance. There are still more fundamental issues
concerning the lack of utility of job satisfaction as an attitudinal measure of motivation.
Porter and Lawler (2000) state that job satisfaction levels do not tell management how
well employees are performing as much as they are an indicator of how well the company
is rewarding employees. Other psychological constructs with greater precision and
validity are needed to explain the level of an individual employees work motivation.

E.

Research Objective
The objective of this research is to determine if work motivation can be

adequately represented by an existing self-report measure of psychological well-being

that has recently emerged in occupational health psychology called work engagement.

Work engagement theory stresses the assumption of optimal functioning at work in terms
of personal well-being, which is a distinguishing characteristic of the eudaimonic
viewpoint of psychological well-being. Three highly correlated dimensions are used to
define the attributes of work engagement: absorption, dedication, and vigor. These
three dimensions are defined with terminology similar to descriptions of employee
behaviors displayed by a highly motivated workforce. The objective of this research is to
determine if there is a sufficiently strong correlation between work engagement and work
motivation to warrant the use of work engagement by managers as a criterion variable for
the measurement of work motivation within their organization.
In order to evaluate the utility of work engagement as a criterion measure of work
motivation, its relationship to factors defined in one or more relevant work motivation
theories should be examined. Two theories of work motivation, both supported by an
extensive body of empirical data presented in the academic literature, have been
identified as suitable candidates for this task: the goal theory of work motivation (Locke
and Latham 2002), and the job characteristics theory of motivation (Hackman and
Oldham 1975). The first work related motivation theory, goal theory, is considered
relevant to this study because it defines the mediating variables of goal setting to be the
three components of action: direction, effort, and persistence. Mediating variables are
psychological factors defined and used to explain the linkage mechanism theorized to
exist between a cause variable and its related effect variable. Similarities in the definition
of these components of action to the dimensions of work engagement will be evaluated in
this study.

To strengthen any case that may be made either for or against the proposed use of
work engagement as a measure of work motivation, a second work related motivation
theory, the job characteristics theory of motivation (Hackman and Oldham 1975). Job
characteristics theory has its own substantial body of supporting knowledge, and it is
cited as one of the theoretical foundations of the earliest empirical study of engagement
presented in the literature (Kahn 1990). Job characteristics theory is considered relevant
because it provides an alternate perspective which may generate additional information as
a second independent examination of relationships between work engagement and work
motivation. These two work related motivation theories each incorporate several
different and distinct independent variables that can be readily influenced through
managements application of healthy workplace practices. It is an objective of this
research study to evaluate the premise that managers can tailor their organizational
improvement plans to leverage organizational factors measured by goal setting and job
characteristics variables in a manner intended to increase their employees work
motivation level, and thus have a positive effect on employee well-being as measured by
their level of work engagement.

CHAPTER II

LITERATURE REVIEW

A. Introduction
A thorough review of engagement, goal theory, and job characteristics theory was
conducted of the literature of both occupational health psychology and business
management. The results of the review provided a theoretical and empirical roadmap for
this research study. The salient features of two management theories of motivation and
their relationship to the theory of engagement was examined. The flow diagram shown in
Figure 2.1 provides an overview of the sequence of topics presented in the literature
review.

Figure 2.1 Flow of Topics Covered in Literature Review

The connections that were found between the various psychological constructs
presented here inspired the objectives of this research study. These psychological
constructs are abstract theoretical variables developed to explain the phenomena of
interest. The constructs are used to explain higher level concepts such as work
motivation, and lower level psychological factors used to describe specific attitudes or
behaviors of individuals. Employee well-being is defined, followed by the concept of
work motivation. Two separate work motivation theories are reviewed to identify their
theoretical relationships to well-being. The various constructs that comprise these
two motivation theories are examined in detail because they have several similarities and
theoretical connections to the theory of engagement and are therefore sources of
potentially useful predictor variables for measurement of work engagement within a
technical workforce. The literature on work engagement is reviewed to better understand
the lessons learned from earlier work engagement research studies. The current status of
research on work engagement as a measure of well-being is also discussed.

B. Psychological Well-being
A benefit of healthy workplace practices is the recognition of the human element
of the organization. The untapped potential of the workforce is related to the overall level
of employee well-being. Management practices designed to produce positive
organizational outcomes are interconnected to employee well-being. Management tactics
and strategies exist that can elicit increased levels of organizational performance and
competitiveness by drawing from the increased potential of human resources and energy
made available through higher levels of employee well-being within the organization.

10

There are differing theoretical perspectives within the study of psychological

well-being. The Hedonic approach is based on an individuals achievement of happiness
through the pursuit of pleasure and avoidance of pain. Job satisfaction can be considered
a measure of hedonic well-being (Waterman 1993). A second theoretical approach to
well-being is based on the Aristotelian viewpoint of Eudaimonia that focuses on meaning,
the potentialities of an individual and their self-realization, defining well-being as the
degree to which a person is fully functioning (Ryan and Deci 2001). Ryan and Deci
(2001) conclude that motivation can be considered a measure related to eudaimonic wellbeing. The eudaimonic approach to psychological well-being provides the philosophical
underpinning for the theoretical framework of this research study.

C. Work Motivation
The term motivation is derived from the Latin word for movement movere.
Atkinson (1964, page 11) defined motivation as the contemporary influence on
direction, vigor, and persistence of action. Motivation affects three aspects of an
employees actions at work including: direction, or what a person does (behavioral
choice); intensity, or how hard a person works (amount of effort); and duration, whether a
person changes or maintains the initially chosen behavior (persistence of action) (Latham
and Pinder 2005). The concept of work motivation is concerned with a set of energetic
forces that impel action, originating both within as well as beyond an individual to
initiate and determine work-related behavior (Latham and Pinder 2005). This definition
of work motivation provides a suitable framework for motivation in this research study.
Figure 2.2 provides a graphical depiction of this definition of work motivation.

11

Figure 2.2 Graphical Depiction of Work Motivation (Latham and Pinder 2005)

Work motivation has also been described as a psychological process that is the
result of an individuals interaction with their organizational environment (Latham and
Pinder 2005). Organizational context is important in any holistic understanding of work
motivation. The employee work environment fit has been evaluated from multiple
perspectives in the literature. While providing useful insight into factors that influence an
employees motivation level, a limiting factor to evaluations of employee-environment fit
is that the variables of interest are usually treated as stable rather than dynamic states.
Since employees also affect and help create their work environment, treating specific
aspects of environment as independent variables may oversimplify the dynamic nature of
work motivation (Latham and Pinder 2005). An employees daily state of well-being will
influence the daily direction, level and duration of their work behaviors. Understanding
an employees well-being is therefore useful because it helps managers to better
understand the dynamic nature of an employees level of work motivation.
12

A review of motivation literature was conducted with the purpose of identifying

one or more specific theories that appeared to have the closest fit with the objective of
this research which was to evaluate the concept of work engagement as a measure of
work motivation. A recent review of research on motivation by Eccles and Wigfield
(2002) helped to clarify which theories of motivation might be best suited for this
research study. The review categorized four types of motivation theories: theories
focused on expectancy, theories focused on the reasons for engagement, theories
integrating expectancy and value constructs, and theories integrating motivation and
cognition (Eccles and Wigfield 2002). Their review highlighted goal theory as the most
suitable candidate for an explanation of work motivation focused on the reasons for
engagement of employees in a work context. Goal theory was therefore selected as a first
theory of motivation for this research study.

D. Goal Theory of Work Motivation

Goal theory is based on the premise that conscious goals affect an individuals
actions. A goal is defined as the object or aim of action, and there is usually a specified
or perceived time limit associated with accomplishment of the goal. Research studies on
work motivation have consistently shown that high performance is not always the result
of sheer effort or persistence. It is also the result of cognitive understanding of the task,
as well as the strategy or plan necessary for completing the task (Latham 2007).
According to Latham (2007), goals provide a simple and reasonable motivational
explanation for observed variation in employee productivity and performance that is not
attributable to differences in either an individuals ability or in the work situation itself.

13

Some employees perform better than others simply because they have different
performance goals (Locke and Latham 1990). Locke and Lathams goal theory model of
work motivation is depicted in Figure 2.3.

Figure 2.3 Goal Theory Model of Work Motivation (Locke and Latham 2002)
The causal assumptions of this theoretical model are supported by over
five hundred research studies spanning 35 years (Locke and Latham 2002). The model is
comprised of several antecedent variables which are believed to precede and influence
the level of motivation indicators as measured by the goal and efficacy mechanisms in the
model (work related behaviors), which determine work performance and outcomes that in
turn influence job satisfaction. The effect which goal and efficacy mechanisms have on
performance and outcomes is moderated by other goal moderator factors both within and

14

beyond an individual. It should be noted that there is no single explicit variable specified
in the model that is labeled work motivation. The goal theory of work motivation is
treated as a higher order concept that is described by specification of antecedent
(predecessor), mediating (explanatory), and consequent (outcome) latent variables in a
nomological network. It is the goal and efficacy mechanisms (mediators) in Figure 2.3
that appear to represent the core characteristics of work motivation in goal theory. In
their theoretical model, Locke and Latham (2002) attempt to explain the effect goals have
on employee actions in four ways. First, goals provide direction to an employees
attention. Second, goals energize effort. Third, goals influence persistence in a positive
way. Fourth, goals influence cognitive effort to obtain knowledge and strategies that will
assist them in the accomplishment of a goal.
The goal theory model of work motivation depicts three categories of
psychological constructs as antecedents of goal and efficacy mechanisms: values, goal
attributes, and self-efficacy. These constructs are believed to precede and influence the
goal and efficacy mediator mechanisms in the model, which include the
three components of action that describe the motivational aspect of employee work
related behavior. As shown in the model in Figure 2.3, values and personality have the
most distal influence on goal choices, goal and efficacy mechanisms, and therefore on
performance and outcomes. The effect that values and personality traits have on goal
setting choices, which influence work related behavior, is less direct and therefore weaker
compared to the more direct influence that goal difficulty, goal specificity and selfefficacy will have on the goal and efficacy mechanisms that represent work related
behavior in the model. In actual work settings, it has been found that assigned work

15

goals will exert an overriding proximal influence on employee behavior through the
attributes of goal specificity, goal difficulty and the employees level of self-efficacy
toward the work goal (Latham 2007). In this model the goal and efficacy mechanisms
(mediator variables) are direction, effort, persistence and task strategies. They are used to
describe the salient aspects of motivated employee behavior. It should be noted that these
mediator variables are the three components of action in the definition of work
motivation presented earlier in the literature review. Mediator variables are included in
theoretical models that contain explicit cause and effect paths to explain the causal
relationship between either predictor and criterion or independent and dependent
variables (Barron and Kenny 1986). The goal and efficacy mechanisms variables are
labeled as mediators by Locke and Latham (2002) to explain the causal linkage between
the goal theory antecedents of values, personality, goal attributes, and self-efficacy to the
goal theory consequents of performance and outcomes, including task success and
rewards, and job satisfaction.
Additional goal moderating factors are included in the model that influence the
goal theory consequents of performance and outcomes, even though moderator variables
are not theorized to explain the models primary cause and effect path relationships. The
level of effect produced by an employees motivated behavior on performance and
outcomes is moderated either positively or negatively by other factors categorized as goal
moderators. The moderator variables include the amount of organizational feedback
received by the employee, employee commitment to the goal, an employees abilities
available to accomplish task strategies for goal attainment, and the complexity of the task
itself.

16

The goal theory model further shows the goal theory consequents, performance
and outcomes, as the causal variables influencing job satisfaction. Other organizational
factors however, act as moderator or mediator variables that can either positively or
negatively affect the level of employee job satisfaction. As this model illustrates, high
levels of job satisfaction do not necessarily indicate the presence of high levels of work
motivation (Porter and Lawler 2000).

1. Goal Attributes
Goal attributes are considered the most influential factors on goal setting behavior
that results in positive performance and outcomes because of their proximal relationship
to the goal and efficacy mediating mechanisms that describe work motivation. To
summarize a key finding of Locke and Lathams (2002) goal setting research, they state
that specific, difficult goals consistently lead to better performance than specific easy
goals or general goals such as do your best, or no goals (Locke and Latham 2002).
Several meta-analyses have estimated the effect sizes (Pearson r correlations) of the
variables goal specificity and goal difficulty, reporting values ranging from 0.42 to 0.80
(Locke and Latham 2002). The size of the effect that difficulty and specificity can have
on goal performance is relatively large. Why then do employees not always meet their
assigned work performance goals? It appears that the manner in which these
two attributes of goal difficulty and goal specificity are treated by management is critical
to the successful definition of a clear and unambiguous goal in a work assignment
context. Stated another way, when employees are simply tasked by their manager to do
their best, they often dont. The primary reason appears to be that since do your best

17

goals have no clearly defined external reference, employees end up defining them in a
unique manner peculiar to their individual circumstances and temperament. This
tendency allows a wide range of performance levels in an organization, which research
indicates is not the case when work goals are more clearly specified.
Specific work goals help focus the attention of employees, letting them know
precisely what they are required to do in their assigned tasks. Setting clear goals tends to
help employees direct their efforts more effectively by clarifying for them the
relationships between effort, performance and rewards within their organization (Steers
and Porter 1974). Goal specificity also reduces employee variation in performance, by
reducing the ambiguity about what level of performance the employee is expected to
attain (Locke and Latham 2002). The specificity of an employees assigned job goals is
best defined in terms that are consistent with explicit organizational goals. Selfreferenced goals are not as effective as externally-referenced goals with respect to
attainment of high levels of employee performance (Locke and Latham 2002).
Locke and Latham (2002) also found that there is a positive linear relationship
between goal difficulty and performance. A problem with the empirical research about
goal difficulty is that the construct has not been consistently defined in the literature (Lee
and Bobko 1992). Locke defined this construct as the probability that a goal can be
reached (Lee, Locke and Latham 1989). Empirical studies by Locke and Latham (2002)
focused on objective goal difficulty, an approach usually taken in a controlled laboratory
setting where goal difficulty is measured in quantitative terms usually defined by
objective performance norms. In uncontrolled field studies, the goal difficulty construct
that is actually measured is the test subjects subjective or perceptual definition of

18

difficulty. The underlying philosophy is that perception of goal difficulty plays a role in
an individuals cognitive decision to expend effort toward a particular goal (Lee and
Bobko 1992). Subjective goal difficulty is defined by Lee and Bobko (1992) as a
measure of the challenge of a goal to that individual. This conceptualization of goal
difficulty will yield lower effect sizes because it is a messy operationalization of the
construct that will likely be confounded with other unknown variables (Lee and Bobko
1992). How difficult the goal is perceived by an individual may be confounded with an
individuals self-efficacy and their perceived individual abilities (Lee and Bobko 1992).
Goal difficulty has two potentially competing effects on motivation. Job goal
difficulty can enhance the employees motivational force to act because larger, more
difficult goals increase the gap between current performance and desired performance,
requiring greater effort to attain the positive self-evaluation which occurs when the more
difficult goal is attained (Bandura 1986). Alternatively, when an individual makes a
subjective assessment of goal difficulty from an expectancy sense, hard goals produce
low expectancy while easy goals produce high expectancy. Low expectancies produced
by employee perceptions of very difficult goals can offset the motivational aspect of the
discrepancy-creation and discrepancy-reduction process postulated by social cognitive
theory (Lee and Bobko 1992). Goals should be difficult yet allow the expectancy of their
attainment to be high enough to generate employee acceptance and buy-in toward the
goal. This aspect of goal difficulty highlights the significance of self-efficacy as an
antecedent variable affecting motivation and goal attainment behavior.

19

2. Self-Efficacy
The objective of a usable theory of work motivation is to identify and explain the
relationships between conscious work performance goals and the level of ensuing task
performance, rather than just discuss vague intentions or desires. Social cognitive theory
(SCT) was first developed by Albert Bandura (1986) to provide a better understanding of
the complexity of human behavior in the workplace, and offer ideas for more effective
behavioral management of the human resources within a modern organization. The
principles of SCT are essentially the same as goal theory, with the primary focus being
the cognitive constructs of self-regulation mechanisms and human agency embodied in
the concept of self-efficacy. A primary tenet of social cognitive theory (Bandura 1986) is
used to explain the underlying self-regulation behavioral mechanism at work when an
individual sets a goal. Goals by themselves do not motivate a person. It is the cognitive
discrepancy created within individuals when they compare current performance to
desired performance that motivates their subsequent behavior to reduce or eliminate the
discrepancy. Self-dissatisfaction with the current state causes individuals to self-regulate
their actions in ways that lead to a more positive self-evaluation of their performance
(Bandura 1986).
The second concept embodied in self-efficacy is that of human agency, which
refers to a persons capacity (or freedom) to act in the world by engaging in social
structures (Bandura 1986). The task-specific concept of self-efficacy as first presented in
the psychology literature by Bandura (1977) defines self-efficacy as the conviction that
one can successfully execute a given behavior required to produce certain outcomes.
Self-efficacy influences the degree of effort put forth by an individual, and their

20

persistence in the face of obstacles to achievement of their desired outcomes. A high

level of self-efficacy in a person promotes greater effort by that individual and more
persistence. Conversely, a person with low self-efficacy is more likely to withdraw effort
prematurely and fail at their assigned work tasks. Bandura (1977) explains that selfefficacy refers to expectancies in an individuals self-perceived ability to put into action a
strategy that meets situational demands. Self-efficacy can influence both goal attributes
and goal mediator mechanisms, especially the task strategies developed by employees in
order to achieve a work goal (Latham 2007). The perceived gap between the current state
and the desired outcome will only drive performance if an individual sees the
performance goal as achievable and therefore worthy of effort.
Occupational self-efficacy is a more generalized abstraction of the concept of selfefficacy than the task-specific assessment of self-efficacy developed by Bandura (1977).
Occupational self-efficacy is considered to be a stable long-term measure of an
individuals level of work-related self-efficacy (Schyns and Collani 2002). This is a
newly developed work domain-specific conceptualization of the generalized self-efficacy
psychological construct. The construct allows comparison of employees in different
organizational contexts with respect to their level of self-efficacy. The original nine item
scale showed reasonable correlations with related personality constructs of generalized
self-efficacy, self-esteem, and locus of control, and with organizational factors including
task demands, job satisfaction, and commitment. A shortened six scale item version of
the occupational self-efficacy construct has been validated in five countries (Rigotti et al.
2008). The level of abstraction of the occupational self-efficacy construct matches the
level of abstraction of engagement as specified by work engagement.

21

3. Feedback and other Goal Moderators

Several goal moderator variables included in the goal theory model of work
motivation shown in Figure 2.3 are theorized to be able to affect the direction and
strength of the path relationship between goal and efficacy mechanisms and performance
and outcomes. One of the most important moderator variables listed in the model is
feedback. Goal setting is a discrepancy creating process. Motivation requires a feedforward control loop that produces a discrepancy or gap between the present state and the
desired future state. Action is then taken to reduce this discrepancy, or as Bandura (1986)
explains, an equilibrating reduction in the discrepancy occurs through an individuals
chosen behavioral strategy that is then adjusted by a feedback loop. Feedback moderates
goal setting behavior. It helps enhance the goal attainment strategy. Without feedback an
individual will have no idea if progress has been made toward attaining the desired goal.
Goal setting needs feedback to help an individual achieve higher performance levels
(Latham 2007).
The three other goal moderator variables listed in the goal theory model of work
motivation include goal commitment, ability and task complexity. These three variables
are more difficult to measure correctly as generalized state-like variables and did not
appear in the literature with validated measurement scales that reflected long term
attitudinal or behavioral properties of employee motivation. These three variables are not
included in this research study because their task specific nature does not match the level
of abstraction of the operationalized versions of the other latent variables in this study or
of engagement as operationalized by work engagement.

22

4. Goal Orientation as a Measure of Values and Personality

The discussion on goal setting has so far focused primarily on the attributes of
goal difficulty, goal specificity and self-efficacy which are task-specific and proximal
state-like psychological constructs. Goal setting behavior also has personality-specific
aspects. Values and personality are included in the theoretical model of work motivation
because they influence goal choices, goal-setting behavior, and therefore work
motivation. An individuals goal orientation provides insight about the influence of their
values and personality on their goal choice. Achievement goal orientation was identified
in the literature as a conceptualization of the effects that an individuals values and
personality have on goal choices, and it was determined to be a suitable measurement
variable in this research study for values and personality (VandeWalle 1997).
Goal orientation research attempts to predict and explain not only the tasks people
choose, but how they will behave when choosing their task strategy for executing the
chosen task. Goal orientation is therefore a dispositional determinant of behavior, while
goal attributes are situational determinants of behavior (Seijts et al. 2004). The research
on achievement goal orientation started in the educational domain, and current research
has extended the theory to a work domain (Baranik et al. 2007). Individuals have
different purposes or goals for engaging in achievement behavior, and these different
orientations toward goals can be structured in a theoretical framework that explains how
individuals approach and react in an achievement goal setting. People who are concerned
with developing their competence or mastering a task set mastery goals. Performance
goals are set by people who are concerned with demonstrating their competence relative

23

to others. Each category of goals can also be defined in terms of an individual either
trying to approach favorable judgments and demonstration of competence, while others
try to avoid unfavorable judgments and demonstrating incompetence compared to
others. These two orientations of approach and avoidance can be exhibited with respect
to either mastery or performance goals, hence the 2 x 2 dimensional framework of the
achievement goal orientation construct depicted in Figure 2.4 (Baranik et al. 2007).

Figure 2.4 Four Dimensions of Achievement Goal Orientation (Baranik et al. 2007)
Research on goal orientation suggests that approach-oriented goals may foster
more adaptive behaviors and are indicative of positive thinking, while avoidance-oriented
goals may foster maladaptive behaviors, which can lead to negative thinking. It should
be noted that most goals can typically be framed in either an approach or avoidance
viewpoint. Empirical evidence suggests there are psychological well-being benefits to
approach goals versus avoidance goals (Coats and Alpert 1996).
Management support for incorporation of business strategy within an organization
that focuses on increased employee mastery through learning goals can enhance the
positive goal orientation behavior of their workers (Locke and Latham 2002). Learning
goals focus the employee on developing mastery and increased understanding of their
work assignments. Learning goals are associated with higher levels of performance in
24

complex tasks. Learning goals can promote adaptive behavioral strategies, while
performance goals may sometimes actually promote maladaptive behavioral task
strategies. A specific high learning goal is best in jobs that involve complex task
strategies, because the learning goal increases the probability that an effective goal
attainment process will be identified, mastered and implemented (Latham 2007).

5. Relevance of Goal Theory

The goal theory of work motivation has been examined and two connections
between work motivation, well-being and goal theory are considered relevant to the
research objective of evaluating the correlations between work motivation and work
engagement. First, goal theory explains work motivation using the three components of
action, direction, intensity and duration, to describe the cause and effect relationship
between goal antecedents and performance consequents. These three components of
action appear similar in definition to the dimensions of work engagement. Second, goal
theory is concerned with attainment of goals through task strategies that maximize
employee performance, increasing the level of employee functionality toward goal
attainment. This aspect of goal theory appears to be aligned with the eudaimonic
viewpoint of psychological well-being which equates development of an individuals full
potential to higher levels of psychological well-being. The following components of the
goal theory model of work motivation were included in this research study: the two goal
attributes represented by the measures of job goal difficulty and job goal specificity,
occupational self-efficacy and achievement goal orientation. The goal moderator variable
feedback was also included because it was repeatedly referenced in the literature.

25

Goal theory is however only one of several theories of motivation which attempt
to incorporate employee well-being into their theoretical framework. The incorporation
of a second theory of motivation into this research study will provide an alternate
approach to achieve the research objective of evaluating the correlations between work
motivation and work engagement.

E. Job Characteristics Theory of Motivation

Job characteristics theory emerged in the 1970s as managers looked for new
ways to boost employee productivity. Work redesign through job enrichment provided
the impetus for research into job characteristics because there was insufficient data at the
time about the relative effectiveness of the new management initiatives (Hackman
1980). There was insufficient knowledge of how jobs affect people and how to measure
and thereby understand what happens when jobs are changed. The purpose of work
redesign is to increase organizational productivity by improving the quality of work
experiences within an organization, thereby increasing internal motivation of employees
to improve their job performance (Oldham 1976). This was to be accomplished by
alteration of specific characteristics of jobs, incorporating greater responsibility and
autonomy into a set of more meaningful tasks that a worker could identify as a whole
work package. The goal was to promote development of self-managing workgroups
(Hackman 1980). At the time work redesign, which was using a behavioral approach,
was viewed as a departure from the classic scientific management approach of
simplification, standardization and specialization of jobs. It was apparent to many
managers that their traditional management methods often had unintended dysfunctional

26

consequences. It should be noted that job characteristics theory was just one of several
approaches to work redesign. Another competing approach at the time was Herzbergs
two-factor theory of hygiene and motivators in the workplace. The common theme to
these theories of work redesign was that motivation is improved if the worker is
appropriately matched to his job.

1. Core Job Characteristics

The theory of job characteristics is one aspect of work design theory developed in
the 1970s that is still considered relevant for use in organizational research today. Its
primary premise is that internal motivation of employees can be influenced by core job
characteristics. The model depicted in Figure 2.5 includes the five core characteristics of
jobs which Hackman and Oldham (1975) empirically determined can be designed in a
manner that enhances internal work motivation.

Figure 2.5 Job Characteristics Theory (Hackman and Oldham 1975)

27

Internal work motivation refers to the degree to which an individual experiences

positive internal feelings when performing effectively on the job. The content of an
individuals job is one of the critical determinants of their internal motivation level. By
improving or enriching the characteristics of a job, the level of internal work motivation
should be increased in many situations (Oldham 1976). The validity of job
characteristics theory was tested through a series of research studies using a survey
instrument called the Job Diagnostic Survey or JDS (Hackman and Oldham 1975). The
five core job characteristics of the Job Diagnostic Survey are described below:
1) Task significance: the extent to which the job has a substantial and perceivable
impact upon others in the immediate organization or external environment
(Hackman and Oldham 1975, pg. 161).
2) Task identity: the extent to which the job requires completion of a whole and
identifiable piece of work; doing a job from start to end with a visible outcome
(Hackman and Oldham 1975, pg. 161).
3) Skill variety: the extent to which the job requires the worker to perform different
activities calling for different skills and abilities (Hackman and Oldham 1975,
pg. 161).
4) Autonomy: the extent to which the job gives the worker freedom, independence,
and discretion in scheduling work and determining procedures (Hackman and
Oldham 1975, pg. 162).
5) Feedback from the job itself: the extent to which the worker, in carrying out the
activities required by the job, receives information about the effectiveness of his
efforts (Hackman and Oldham 1975, pg. 162).
These five core job characteristics are treated as independent variables in the job
characteristics model. The effect of these five independent variables on the dependent
variables of personal and work outcomes is mediated by three psychological states which
are described below:

28

1) Experienced meaningfulness of the work refers to the degree to which the

employee experiences the job as generally meaningful, valuable and worthwhile.
Skill variety, task significance and task identity are the antecedents of
experienced meaningfulness (Hackman and Oldham 1975).
2) Experienced responsibility for outcomes of the work is defined as the degree to
which the employee feels personally accountable and responsible for the results
of the work that the employee does. Autonomy is the antecedent of experienced
meaningfulness (Hackman and Oldham 1975).
3) Knowledge of the actual results of the work activities is the degree to which the
employee knows and understands, on a continuous basis, how effectively they are
performing their job. Feedback is the antecedent of knowledge of results
(Hackman and Oldham 1975)
2. Internal Work Motivation
The conceptual definition of internal work motivation incorporated into the job
characteristics theory is not exactly the same concept of motivation that is the basis of
goal theory. Job characteristics theory describes internal motivation as an internal
psychological process that is most closely related to the concept of intrinsic motivation.
Intrinsic motivation refers to the degree to which an employee experiences positive
internal feelings when they are working effectively on the job. Intrinsic motivation
results from the perceived personal rewards inherent in a task or activity itself; in the case
of work, the enjoyment that comes from performance of job duties and assignments
(Ryan and Deci 2000). Intrinsic motivation is related to a sense of competence and selfdetermination, having a keen interest in and obtaining high levels of personal satisfaction
from work related tasks and work performance (Ryan and Deci 2000). It is usually
evaluated within the context of its dichotomous relationship with extrinsic motivation.
Extrinsic motivators will come from influences and factors outside of a persons
psychological processes. Money or other rewards are a typical source of extrinsic
motivation. Coercion or the threats of punishment are other extrinsic motivators. It has
29

been observed that extrinsic motivators over time will tend to diminish levels of intrinsic
motivation (Ryan and Deci 2000).

3. The Job Diagnostic Survey

The perceived erosion of American industrial competiveness in the 1970s relative
to the emerging competitive threats of Japanese and German industry provided the
rationale for work redesign programs and development of the Job Diagnostic Survey.
The survey instrument was designed for use both in diagnosis of jobs prior to their
redesign and to assess effects of redesigned jobs on the people who did them. It is meant
to assess jobs as perceived by the individual employee level of analysis. Individual
responses would be aggregated to attain a picture of a particular job or work group within
the organization being surveyed. The instrument is made up of two sections. The first
section asks respondents to assess how much of each job characteristic they perceive their
job to have, and the second section then asks the employee to indicate how accurate a
number of statements are relative to the characteristics of their job. The complete job
diagnostic survey had separate sections for job dimensions, psychological states, affective
responses to ones job, and measures to ascertain the employees growth need strength,
considered by Hackman and Oldham (1975) to be a moderator variable which could
either increase or decrease the effect of the core job characteristics on internal motivation.
The developers of the job diagnostic survey (Hackman and Oldham 1975)
believed it is possible to generate a summary score that reflects the overall motivating
potential of a job in terms of certain core job dimensions or characteristics. The
Motivating Potential Score (MPS) is to be calculated using equation (2.1):

30

Because of the multiplicative relationship in the formula, if any core dimension is

low, then the resulting MPS for a given job will be low. A job that is high in MPS does
not affect all individuals in the same way. People who value feelings of accomplishment
and personal growth should respond more positively to a job with high MPS. People who
do not value personal growth and accomplishment may find such a job causes them
anxiety and can be uncomfortable. Therefore individual growth need strength is included
in the job characteristics theory as a moderator variable of work performance outcomes.
Research studies today continue to examine the relevance of the JDS as a
diagnostic tool for work design (Buys et al. 2007). Managers can use the JDS to evaluate
core job characteristics for work redesign and job enrichment strategies by measuring the
content of jobs with the JDS. Core job characteristics can be positively influenced by
application of management best practices, and the content of individual and group work
roles optimized for employee well-being. One must be aware of the impact of
managerial decisions and seek to optimize well-being by carefully balancing
organizational situation-specific trade-offs. Empirical evidence suggests a high degree of
congruence between the work environment and employees generally results in higher
performance and employee well-being (Grant et al. 2007).

F. Comparison of the two Motivation Theories

31

There appears to be some structural similarities between the goal theory model and
the job characteristics model. Goal attributes and self-efficacy are treated as goal theory
antecedents, and are analogous to the five core job characteristics which are antecedents in
the job characteristics model. The consequents of job characteristics theory, personal and
work outcomes, appear analogous to the goal theory consequents, performance and
outcomes. The three psychological states in job characteristics theory are considered
mediator variables, explaining the cause and effect relationship between antecedent and
consequents in job characteristics theory, which is analogous to the use of the goal and
efficacy mechanisms as mediator variables in goal theory. The construct of employee
growth needs strength is treated as a moderator variable in job characteristics theory, and
may be considered analogous to goal commitment and abilities moderator variables in goal
theory. One important difference is that while feedback is a moderator variable in goal
theory, in job characteristics theory feedback is treated as an antecedent variable. Another
difference is that the job characteristics theory of motivation can be effectively represented
with the use of a single validated survey instrument, the Job Diagnostic Survey, while no
single validated survey instrument was found in the literature that represented the goal
theory of work motivation as depicted in Figure 2.3. Several separate validated survey
instruments were identified that measure the goal theory constructs chosen for inclusion
into this research study. The specific constructs from the two motivation theories
incorporated into this research study have relevance to one or more aspects of the theory of
engagement which is presented next in this literature review.

G. Theory of Engagement

32

The word engagement can have different meanings depending on the context of its
use. The Websters online dictionary (YourDictonary.com 2009) defines the state of being
engaged as the act of sharing in the activities of a group. It further explains that to be
engaged is to be involved in activity, or to be occupied, busy, greatly interested, or
committed. The main similarities between definitions of engagement and motivation appear
to be the concepts of action and involvement. Motivation may be viewed as the behavioral
factor that impels a person to act, while engagement may be a measure of the level of
intensity of action produced by motivation. If true, then the level of engagement displayed
by an individual may be a useful indicator of that individuals motivational state.
William Kahn published the first conceptual model of personal engagement in the
academic literature in 1990. According to his theory, momentary psychological conditions
affect the level of personal presence, or expression of a persons preferred self in task
behaviors. A person either presents (engages) or withdraws (disengages) their self in a
given role (Kahn 1990). The construct is summarized in Figure 2.6.

Model of PERSONAL ENGAGEMENT (Kahn, 1990)

Psychological Conditions
Needed for Engagement:
Meaningfulness
Safety
Availability

Multiple levels of influences:

Individual
Interpersonal
Group
Inter-group
Organizational

Three Dimensions
of Engagement:
Cognitive
Emotional
Physical

Figure 2.6 First Published Theory of Engagement (Kahn 1990)

There are three dimensions of personal engagement: cognitive, emotional, and

physical. Kahn (1990) also defines the antecedents of personal engagement. Certain
33

psychological conditions are needed for engagement: meaningfulness, safety, and availability.
There also exist multiple levels of influences affecting personal engagement: individual,
interpersonal, group, inter-group, and organizational. A fundamental component of the
theory is that the level of personal engagement in different work roles will vary on a
continuum and engagement can be measured with a psychometric scale as shown in
Figure 2.7 (Kahn 1990, 1992).

Engagement and Disengagement Continuum

(Kahn, William.A., 1990, 1992)
Role B

Role A

Disengagement
(withdrawal of self)

Engage ment
(presentation of self)

Figure 2.7 Engagement is a Continuum Variable (Kahn 1990, 1992)

Kahns (1990, 1992) research was an exploratory effort, and he chose a qualitative
approach to data collection using the personal interview method. He conducted a series
of interviews while immersed inside several different organizational settings. This
approach enabled the collection of a rich set of behavioral observation data that was
subsequently analyzed and used to lay the theoretical groundwork for the definition of the
construct of personal engagement. A review of the engagement literature found that
Kahns model of personal engagement has been empirically tested in one published study
(May et al. 2004) which produced tentative support for the theory, with the caveat that

34

the three-dimensional structure of work engagement as operationalized in the study was

inconclusive (May et al. 2004).
Additional organizational behavior studies published after Kahns (1990, 1992)
seminal work on the subject have considered engagement as an independent variable
(Salanova et al. 2005), a dependent variable (Schaufeli and Bakker 2004), and even a
moderator variable (Sonnentag 2003). Engagement is important to study because it is
linked to positive individual work related outcomes. Engagement is related to good
health and positive work affect. Researchers have explored the dynamics of engagement
from different viewpoints, including multiple levels of engagement (Salanova et al.
2005), the relationship of engagement in work and family roles (Rothbard 2001),
engagement in individual roles verses team roles (Bakker et al. 2006), and the variation in
engagement levels over day intervals (Sonnentag 2003). Theoretical models for different
versions of engagement have been proposed in the academic literature, including personal
engagement (Kahn 1990), work engagement (Schaufeli et al. 2002), employee
engagement (Harter et al. 2002), (Saks 2006), and self-engagement (Britt et al. 2005).
Engagement is typically conceptualized as a distinct construct that has been empirically
shown to be separate but related to other constructs including flow, job involvement, and
organizational commitment. Engagement also has an opposite construct first identified as
disengagement by Kahn (1990) and later on in the literature as burnout (Maslach et al.
2001). Burnout is a condition related to a reduced level of effectiveness and chronic
health problems in employees. Burnout can be induced by stress within the workplace, or
by a lack of job resources (Schaufeli et al. 2002).
H. Other Explanations of Engagement

35

The term engagement has been used more loosely by a variety of management
consultants that have published articles in either non-academic or for-profit management
consultant newsletter forums (Marcum 1999) (Haudan 2002). These articles typically
describe a variety of sometimes conflicting approaches to promote their unique blend
of business management and organizational transformation best practices, productivity
enhancement and employee well-being programs. What is missing from these articles on
engagement is a clear grounding in established management theory. No explicitly
defined methodology following established scientific experimental design practices and
no psychometric properties of their purported survey instruments are usually published.
For engagement to be viewed as a distinct and clearly defined psychological construct
additional research with theoretical rigor and empirical support is needed to clarify a
more precise meaning of engagement as a measure of organizational behavior that is
useful to engineering management.

I. Review of Work Engagement Research

One particular research path has emerged in the occupational health psychology
literature as the result of a group of research psychologists extending their empirical
studies about job burnout into a new research path focused on work engagement. Their
theory of work engagement defines the construct as a positive, fulfilling work-related
state of mind; a long-term persistent affective-cognitive state (Schaufeli et al. 2002). The
salient features of work engagement theory are summarized in Figure 2.8.

36

Figure 2.8 Theory of Work Engagement (Schaufeli et al. 2002)

Work engagement theory stresses the assumption of optimal functioning at work
in terms of personal well-being, which is a distinguishing characteristic of the
eudaimonic viewpoint of psychological well-being. The primary characteristics of
engagement are the presence of energy and attachment to work. Schaufeli et al. (2002),
the researchers who developed the construct state that work engagement is a function of
the job resources that fuel work motivation. Work engagement mediates the link between
work-life factors and work outcomes. Three highly correlated yet distinct dimensions
are used to define work engagement: absorption, dedication, and vigor. Their
characteristics are summarized in Figure 2.9.

Figure 2.9 Theoretical Dimensions of Work Engagement (Schaufeli et al. 2002)

37

These three dimensions have been operationalized in the Utrecht Work

Engagement Survey instrument (UWES), an individual self-report questionnaire with the
title: Work and Well-being Survey. The UWES is now used in dozens of countries
around the world. Thirteen language versions are available and an international data-base
exists that currently includes engagement records of over 20,000 employees (Shimazu
et al. 2008). The sections that follow present summaries of several relevant
organizational behavior studies which have considered work engagement in a variety of
research contexts. The final section on previously published work engagement research
summarizes the only two U.S. based research studies found in the literature review search
on work engagement.

1.

First Empirical Study of Burnout and Work Engagement

This first published study on psychometric properties of the Utrecht Work

Engagement Survey instrument (Schaufeli et al. 2002) described the validation and
refinement process of the instrument from its initial 24-item survey format which
contained nine items measuring vigor, eight items measuring dedication, and seven items
measuring absorption. Reliability analysis was used to reduce the original version of the
survey to its present 17-item version, and also maximize internal consistency.
Two separate samples were taken: 314 undergraduate students from a Spanish university,
and 619 employees from twelve public and private Spanish organizations. The authors
studied the relationship between burnout and work engagement. These two latent
variable constructs were defined by two different sets of three reflective factors. The
hypothesized latent variable structures were confirmed simultaneously in two successive

38

confirmatory factor analyses. The study results found that the two constructs of work
engagement and burnout are negatively correlated to each other, and the inter-relations
for the three scales were all positive within each construct, demonstrating consistency as
measures that accurately reflect the theoretical variables as defined in their research
study. They concluded that alleviation of workplace stressors is a priority to reduce
employee burnout. Providing adequate resources can mitigate the effect of workplace
stressors and increase the level of work engagement (Schaufeli et al. 2002). Structural
equation analysis from the first study indicated the 3-factor model of engagement with
vigor, dedication and absorption modeled as separate but highly correlated factors fit well
to the data of both samples.

2.

An Empirical Study of Daily Fluctuation of Work Engagement

Sonnentag (2003) was concerned with measuring day-to-day levels of work

engagement. The UWES scale of work engagement was incorporated into the study as
shown in Figure 2.10. The premise is that engagement at work expends energy and
resources, requiring a period of recovery-leisure time in non-work activities. A sample
size of n = 147 subjects in one organization was tested daily for five days. Daily fluctuations
in engagement level were observed. Day-level Work Engagement was found to be positively
correlated (Pearson r correlation = 0.31) to Day-Level of Recovery and also positively
correlated (Pearson r correlation = 0.48) to Day-Level Personal Initiative. This study used a
16-item version of the UWES survey instrument and changed the 7-point response scale
numbering scheme from 0 (never) to 6 (always), used in the original instrument, to a 1(never)
to 7 (every day) response scale. A principal component analysis conducted with the data did

39

not result in a clear factor solution. The researchers used an overall one factor scale for
engagement that produced a high reliability Cronbachs alpha of 0.91. This study measured
daily fluctuations in level of work engagement, a significantly shorter time frame than stated
in the theory for work engagement where it is defined as a measure of a long-term persistent
affective-cognitive state (Schaufeli et al. 2002).

Control Variables
De mogr aphic Variables
Gender

Solid Lines denote hypothesized effects.

Dotted lines denote effects of control variables.
Day-Le vel
Recovery

Age
Tenure
Work place Characteristics
Method control

Day-Le vel
Work Engage ment

Time control
Situational constraints
Time pressure
Trait Work Engagement

Day-Level Personal Initiati ve

Day-Level Pursuit of Learning

Vigor, De dication, Absorption Model

Trait Personal Ini tiative

Trait Personal Learning

Recovery, Work Engagement, and Pro-active Behavior: a study

of day- level variation in Work Engagement (Sonnentag, 2003).

Figure 2.10 Model of Daily Fluctuation of Work Engagement (Sonnentag, 2003)

3. Work Engagement Study with Respect to Service Climate

Another study using the UWES looked at service climate and its mediating effect
on employee performance and customer loyalty (Salanova et al. 2005). Work
engagement was treated as an independent variable in this study. As shown in
Figure 2.11, this theory has the two variables, level of resources and work engagement,
predicting service climate. Service climate then predicts employee performance which
predicts customer loyalty. Structural equation modeling was used to analyze results of
40

the empirical study. A sample size of 342 front line service unit employees in 85 work
units was tested with the UWES engagement scale and other scales to measure service
climate and resource levels. A separate customer survey was used to measure customer
loyalty and customer perception of employee performance. The three factor structure of
work engagement did not fit the data well at first, and the researchers modified the
absorption scale and then also dropped items 1 and 3 from the vigor scale. This revised
model met analysis criteria for 3 factors.

Re porting by EMPLO YEES

Training

Autonomy

Technology

Re porting by CUS TOMERS

Service
Climate

Service
Climate

Organizational
Resources
H2

Dedication

Employee
Performance
H3

H1
H4

Work
Engage ment

Vigor

Employ ee
Performance

Absorp tion

Customer
Loyalty

Customer
Loyalty

Linking Organizational Resources and work engagement to e mployee Performance

and Customer Loyalty: the mediation of Service Climate (Salanova et. al., 2005).

Figure 2.11 Work Engagement as an Independent Variable (Salanova et al. 2005)

4. Antecedents and Consequents of Work Engagement
A further independent study by Koyuncu et al. (2006) continued to build upon the
UWES model of work engagement, studying possible antecedents and consequences of
work engagement. A sample population of 286 women professionals in a Turkish bank
was surveyed. The measures of the study included personal demographics and work
situation, organizational life experiences, work outcomes including career satisfaction

41

and intent to quit, the three factor UWES work engagement scale, psychological wellbeing, and emotional exhaustion, as shown in Figure 2.12. While this study used the
17-item UWES instrument, the researchers changed the 7-point response scale numbering
scheme of the original instrument from 0 (never) to 6 (always) to a 5-point response scale
numbering scheme from 1 (strongly agree) to 5 (strongly disagree).

Antecedents

Consequences

Organizational
Life Experiences:
Workload
Control
Rewa rd and Recognition
Co mmunity
Fairness
Person-to-Organization
Value-fit

Work Outcomes:
Work Engage ment:
Vigor
Dedicat ion
Absorption

Career Sat isfaction

Intent to quit

Psychological
Well-being
Emotional
Exhaustion

Work
Engagement
Research
Model
(Koyuncu
al. 2006)
Engage
ment Research
Model
(Koyuncu
et. al. et
, 2006).

Figure 2.12 Antecedents and Consequences of Work Engagement (Koyuncu et al. 2006)

The researchers performed a hierarchical regression analysis of the results. A

positive correlation was found between engagement and work outcomes. They reported
Cronbachs alpha reliability measures of 0.92 to 0.90 for the three factors of the work
engagement instrument. No validity analysis results were reported for this study.

5. Work Engagement and the Job Demands-Resources Model

A study conducted by Schaufeli and Bakker (2004) focused on how the work
engagement construct fit into the context of the job-demands job resources model. The
study examined if this particular model helped to better explain the two constructs of
burnout and engagement. Two theoretical models from this study are shown in
42

Figures 2.13 and 2.14. Additional studies conducted by the researchers replicated this
effort, varying areas of research focus and studying different sample populations (Bakker
et al. 2006) (Hakanen et al. 2008). The researchers used structural equation modeling
methods in their analyses and concluded in all studies that their models exhibited a good
fit to the experimental data.

Job Demands

Job Resources

Burnout

Health
Proble ms

Engage ment

Turnove r
Intention

Research Model of Predictors and Consequents of Burnout and Engage ment

(Schaufeli and Bakker, 2004)

Figure 2.13 Job Demands-Resources Model of Work Engagement (Schaufeli and Bakker 2004)

Mental
Emotional

+
Job Demands

Strain

Physical

Etc .

Organizational
Outcomes

Support
Autonomy

Job Resources

Motivation

Fee dback

The job demands-resources model

(Bakker et. al., 2006)

Etc .

Figure 2.14 Second Job Demands-Resources Model of Work Engagement (Bakker et al. 2006)
6. Work Engagement Discriminant Analysis Study

43

This study by Hallberg and Schaufeli (2006) investigated whether work engagement
as measured by the UWES could be empirically separated from the similar constructs of job
involvement and organizational commitment. Discriminant validity was tested through
inspection of latent inter-correlations between the constructs, confirmatory factor analysis,
and patterns of correlations with other constructs. They concluded that the constructs of
engagement, job involvement and organizational commitment can be clearly identified and
measured. All three constructs refer to positive attachment to work so some variance will be
shared, but without too much overlap to result in redundancy. The constructs displayed
different associations with health issues, job characteristics and turnover intentions. The
latent inter-correlations between the constructs ranged from between 0.35 and 0.46,
indicating between 12 percent and 21percent of shared variance. The authors state that this
supports the assumption that the constructs are related but do not overlap to the extent where
they become redundant psychological constructs.

7. Work Engagement in Japan

In order to study and apply the concept of work engagement in Japan, the translation
of the UWES into kanji and a research study applying the resulting UWES-J has been
reported in the literature (Shimazu et al. 2008). The purpose of this study was to validate the
Japanese version of the Utrecht Work Engagement Scale (UWES-J). Employees from
three independent samples completed the questionnaire (n = 2,334). Their analysis of the
17-item survey instrument was unable to reproduce the theoretical three factor structure of
work engagement. Further confirmatory factor analyses of the 9-item shortened version of
the survey instrument did recover a three factor structure, using the multiple-group method,

44

maximum likelihood estimation and Varimax rotation methods. Their comparison of a

three factor model with a one factor model concluded that a one-factor model which assumes
that all nine engagement items load on one single factor fitted the data better than the original
three factor model. The one-factor structure was found to be invariant across all
three samples. Scale internal consistency was sufficiently high (Cronbachs alpha = 0.92).

8. Work Engagement Studies in the United States.

At the time of this research study an extensive search in the literature found only
two other research studies on work engagement in the United States. Both studies are
dissertations which document their primary aim to confirm factor validity of the UWES
instrument in the United States while comparing it to other psychological constructs of
interest. The first reported U.S. study is a dissertation published by J. R. Halbesleben
(2003) at the University of Oklahoma. His research focused on further investigation of
the relationships between burnout and engagement and their correlates through analysis
of several studies using the UWES and other survey instruments. The work engagement
study was focused on confirmatory factor analysis of the UWES-17 survey instrument.
Three samples were collected, including 405 introductory psychology students with jobs
along with two other small samples. The results of the structural equation models
indicate that the 17-item three factor solution generated the best goodness-of-fit results
relative to alternative structural equation models. It is noteworthy that the correlations
reported between the three factors were significantly lower than other published results in
the literature (Halbesleben 2003).

45

The second research study in the United States to use the UWES instrument was
the dissertation published by Claura P. Louison (2007). The UWES survey instrument
was labeled employee engagement. This study compared work engagement and job
involvement as predictor variables to three criterion variables: job satisfaction,
organizational commitment and turnover intentions. A non-random snowball sampling
procedure was used and the size of the final sample analyzed was 232 full-time
employees located throughout the United States. Confirmatory factor analysis using
structural equation modeling methods yielded poor goodness-of-fit values for both the
three factor and one factor 17-item version of the UWES. Hierarchical linear regression
analysis was employed to evaluate the relationships between predictor and criterion
variables and inconclusive results were reported.

9. Assessment of Construct Validity from Published UWES Studies

A reasonable body of published statistical information related to the validity and
reliability of the UWES instrument was found in the literature. The instrument appears to
have demonstrated an acceptable level of reliability to researchers that have utilized work
engagement in a variety of applications as part of their research. The instrument was
used in a different organizational contexts and countries, and the results obtained were
consistent between studies. The correlation matrices of the study results indicate that the
inter-correlations of the scale items are highly correlated and positive, while the crosscorrelations between different measures are negative. There has however been some
disagreement about the precise factor structure of the work engagement construct.

46

One study could not identify the three factors of work engagement in their principal
component analysis (Sonnentag 2003). Another study modified the measurement scales
(Salanova et al. 2005). The Japanese study found the 9-item short version of the survey
instrument produced a better fit in confirmatory factor analysis than the 17-item version
of the survey instrument (Shimazu et al. 2008). The UWES instrument appears to be still
undergoing additional validation studies by more than one researcher, so there may be
additional information available in the future that could add clarity to this assessment.
More important to the relevance of this research study, no other research study appears to
have been published yet that explicitly examines the relationships between work
engagement and work motivation theories.

J. Need for Further Research

Explanations offered by both researchers in the academic literature and
management consultants in business journals concerning the construct of
engagement are not yet clear and consistent in their explanations of how or when
an employee is either engaged or disengaged. Supporting theory varies as well as
recommendations how the construct of engagement should be measured. There is
a need to bring greater clarity to the current descriptions of engagement in the
workplace through the continued application of accepted psychometric analysis
methods to assess the validity and reliability of test instruments that purport to
measure the most promising engagement concepts. Work engagement is one such
construct that has been operationalized in the Utrecht Work Engagement Survey
instrument. It has been made available to researchers in the public domain and

47

has been independently evaluated in academic peer-reviewed literature through

published psychometric results of studies that have been conducted in more than
one dozen countries throughout the world. The number of studies conducted in
the United States to date is still limited. Further research is required to assess the
equivalence of the published psychometric properties of the construct within
United States technical services organizations. Furthermore, it is important that
the usefulness of the work engagement construct as a measure of employee
motivation be considered, since psychological constructs with greater precision
and validity than job satisfaction by itself are needed by managers to better
explain the level of employee work motivation within their organizations.

48

CHAPTER III

RESEARCH STATEMENT

A. Introduction
The underlying premise of this research study is that managers desire a better
understanding of employee motivation within their organization in order to energize their
workforce more effectively, channeling that energy toward the accomplishment of
organizational goals and objectives. A variety of existing motivation theories attempt to
explain how motivation occurs in employees from very different perspectives. No single
theory of motivation appears to be dominant in todays management best practices. It
would therefore be advantageous to managers if some practical measurement of work
motivation emerges from current occupational health research as a standard or criterion
for the measurement of an employees level of work motivation. This motivation
measure should have validity and invariance across a diverse range of organizational
contexts and management practices.
This research study begins by considering work motivation as a component of
psychological well-being. The level of work motivation in an employee is believed to be
positively correlated with the occupational psychological health and well-being of that
employee. The antecedents of work motivation are latent factors related to healthy

49

workplace practices deployed by management to enable the work force to achieve

organizational improvement objectives and goals. Another way to view these healthy
workplace factors is to consider at least some of them to be useful predictor variables of
an employees level of work motivation. A manager that can consistently implement the
right mix of healthy workplace practices within their organization will be more likely to
maintain a highly motivated workforce. One issue is how to determine which
measurements of work related factors are the most suitable predictors of an employees
work motivation related component of psychological well-being. When operationalized
in reliable and valid attitudinal and behavioral questionnaires, the selected motivational
well-being factors should provide adequate predictive capability that managers will want
to use to them to measure how healthy their work environment is. In the ideal case, the
psychologically healthy work environment impels employees to achieve increased levels
of performance within their organization. An objective of this research study is to
evaluate the relationships between several organizational factors theorized in the
literature to be antecedents or predictors of work motivation. Evaluation of the strength
of correlations between these predictor factors and work motivation can provide evidence
concerning which set of predictor factors is the most useful to a manager attempting to
predict work motivation of employees.
A second issue is what psychological factor or factors are the most suitable
candidate for measurement of the resultant effects on employees produced by these
predictor variables of work motivation from psychological well-being. These factors can
be labeled criterion factors, because they are believed to be suitable measures of the key
characteristics or traits of a highly motivated employee. Work engagement is theorized to

50

be a criterion variable for work motivation in employees, measuring their level of energy
while on the job and the strength of their attachment to work, characteristic traits of a
motivated employee. The strength of the measured relationships between predictor and
criterion variables in this research will help determine if work engagement is a
motivation related construct with significant enough validity and utility to warrant its
usage by managers as an organizational performance metric of employee work
motivation from psychological well-being.

B. Research Issues
The organizational factors believed to motivate an employee to high levels of
performance on the job are intrinsic to psychological health and well-being in the
workplace. One such psychological construct that shows promise as a work motivation
criterion variable with these desirable characteristics is work engagement. While work
engagement is purported to be a measure of employee work and well-being, is work
engagement also a valid and useful measure of the motivational component of
psychological well-being? Are work engagement dimensions measuring the effects of
work motivation factors through employee actions on the job? The attributes of work
motivation and the dimensions of work engagement are compared in Figure 3.1.

Figure 3.1 Attributes of Work Motivation versus Dimensions of Work Engagement

51

Similarities in the components of the two definitions point to the possibility of

convergence between the two constructs. Will a study that measures several work
motivation predictor variables along with work engagement dimensions find evidence
that the factors theorized to be predictors of work motivation are indeed positively
correlated to the three dimensions of work engagement? Is work engagement a useful
criterion for measuring the motivational aspect of employee work-related behavior?
Answers to these research questions will be developed by evaluating the relationships
between work engagement and the antecedents of work motivation as defined by the goal
theory model of work motivation (Locke and Latham 2002). Goal theory is however
only one of several management theories available to managers seeking a better
understanding of employee attitudes and behaviors in the workplace. Upon closer
examination of the several theories of engagement, one also finds a recurring theme that
job characteristics appear to be important factors in the prediction of employee work
engagement and possibly work motivation. In fact, the description of the antecedents of
work engagement (Schaufeli et al. 2002) appears to be very similar to key elements of the
job characteristics theory defined by Hackman and Oldham (1976) shown in Figure 3.2.

Figure 3.2 Attributes of Job Characteristics Theory versus Work Engagement Theory
52

The theory of job characteristics is also mentioned in the original theory of

personal engagement (Kahn 1990). Can a group of several key job characteristics
variables in the Job Diagnostic Survey which Hackman and Oldham (1975)
recommended for management use to predict the motivating potential score of a job also
predict the level of work engagement in an individual? Is there a cause and effect
relationship between the motivating potential score of a job and the level of work
engagement in an individual? Answers to these research questions can be developed by
evaluating the relationships between the core job characteristics as defined and measured
by the Job Diagnostic Survey and the three dimensions of work engagement.
Estimation of correlation coefficient values and regression path coefficients
between predictor and criterion variables from the research study results may provide
insight about the utility of the work engagement construct as a criterion measure of
employee work motivation. The effect which predictor variables are observed to have
on criterion variables may be evaluated strictly through correlation studies, but there is no
guarantee that a strong correlation will provide proof that the predictor variables actually
cause the observed relationship between predictor and criterion variables. A more
powerful case will be made to managers if evidence of a cause and effect relationship can
somehow be demonstrated between these predictor and criterion variables of work
motivation in employees. Evidence that supports the theorized cause and effect
relationships between predictor and criterion variables may be demonstrated through use
of an analysis methodology known as structural equation modeling.

53

C. Description of Research Study

The organizational context for this research study is engineering technical
services organizations in the United States of America. All employees in an organization
are included in the survey sample frame. Engineers, scientists, and engineering
technicians constitute the majority of the employees sampled in this research. The
purpose of this research study was to answer three questions. The first question was
whether or not an employees level of engagement at work could be measured by the
Utrecht Work Engagement Survey (UWES) instrument with sufficient reliability and
validity for it to be useful to engineering managers. The second question was whether or
not several latent variables from the goal theory of work motivation, which are factors
that can be influenced by engineering managers, could predict an employees level of
work engagement as measured by the UWES. The third question was whether or not
several latent variables from the job characteristics theory of motivation, which are
factors that can be influenced by engineering managers, could predict an employees
level of work engagement as measured by the UWES. The research plan included data
collection using a web survey instrument that combined criterion measures of
psychological well-being as defined by work engagement with two groups of predictor
factors as defined in the two different motivation theories. Data was also collected
regarding demographic characteristics of the sample. A psychometric analysis of the
survey results included correlation analysis, confirmatory factor analysis and structural
equation modeling analyses that were used to compare the goodness-of-fit of the sample
data results to theorized causal relationships between predictor and criterion variables.

54

D. Research Hypotheses
The three questions that this research study has attempted to answer have been
formulated as research hypotheses presented as follows:

1) Ho: The scale items of the Utrecht Work Engagement Survey neither reliably nor
validly measure work engagement in an engineering technical services firm.
Ha: The scale items of the Utrecht Work Engagement Survey reliably and validly
measure work engagement in an engineering technical services firm.
2) Ho: Goal theory latent variables do not predict an employees level of work
engagement in an engineering technical services firm.
Ha: Goal theory latent variables do predict an employees level of work
engagement in an engineering technical services firm.
3) Ho: Job characteristics theory latent variables do not predict an employees
level of work engagement in an engineering technical services firm.
Ha: Job characteristics theory latent variables do predict an employees level
of work engagement in an engineering technical services firm.
The three hypotheses will be evaluated by the statistical goodness-of-fit results
obtained from structural equation modeling analyses of the UWES instrument and two
structural equation research models that will be presented in the next chapter. A detailed
explanation of the methodology employed in this research study is also presented in the next
chapter.
E. Significance of Research
This research study produced a set of measurement data that was analyzed using
several statistical methodologies to evaluate the latent variable factor structure of work
engagement. A comparative study between these results and previous studies on work
engagement determined the level of invariance of the construct between sample
55

populations of workers in technical occupations that have been previously studied in

Europe and Japan with engineers and technicians in the United States of America.
Structural equation modeling in this research examined the causal paths between
variables defined in two currently accepted work motivation theories, goal theory and job
characteristics theory. An assessment of the construct validity of work engagement as a
criterion measure of work motivation and psychological well-being was performed by
developing two structural equation models that were used to evaluate causal path
relationships between work engagement and two different sets of predictor variables
chosen from their respective work motivation theories. This analysis could lay the
groundwork for future research on the utility of work engagement as a criterion measure
of the motivational component of psychological well-being.

56

CHAPTER IV

RESEARCH METHODOLOGY

A.

Research Approach
This research study attempted to evaluate relationships between several

psychological constructs identified in the literature as predictors of work motivation and

the construct of work engagement. After completing a review of the relevant literature on
work motivation theory and theory of engagement, an observation was made that several
core elements of the theory of work engagement appear very similar to key elements of
both the goal theory of work motivation and the theory of job characteristics. Early on it
was determined that the variables chosen for this study should be consistent with the
framework of the two different motivation theories, i.e., Goal Theory and Job
Characteristics theory. Two research models were proposed as independent assessments
of the relationships between work motivation predictor variables and work engagement.
The specific variables chosen for inclusion into the two research models were consistent
with the nomological frameworks of each motivation theory. These latent variables were
combined in logical groups along with the three work engagement latent variables of the
UWES survey instrument in the web survey instrument developed for this research.

57

The theoretical research population selected for this study consists of all
employees of engineering technical services organizations in the United States of
America. A typical population subject could be an engineer, scientist, engineering
technician, computer programmer, or even an administrative employee of the technical
services organization. A majority of employees are expected to have technical jobs. A
pilot study was first conducted for the purpose of testing the research methodology and
also enabled preliminary analysis of the latent variables using the pilot study sample data
set. This approach facilitated identification of any potential issues which might need to
be addressed prior to deployment of the research methodology in a full study sample
population. The research methodology followed the sequence of activities presented
below:

1.
2.
3.
4.

Conceptualization of latent variables into two research models.

Selection of test instruments with measurement scales for the latent variables.
Combination of measurement scales in a web survey instrument.
Data collection from sample populations through use of email, URL hyperlink
and internet web survey technology.

5.
6.
7.
8.
9.

Demographics analysis.
Analysis of survey results utilizing descriptive statistics.
Measurement scale reliability analysis.
Correlation coefficients analysis of latent variables.
Confirmatory factor analyses of individual and combined survey instruments.

10. Develop structural equation models to evaluate cause and effect relationships
between latent variables in the two research models.
The pilot study produced the first sample data set that was evaluated following the
sequence of analysis steps described above. Each step produced additional information
which combined together provided the basis for formulating preliminary answers to the

58

three research questions. More important, analysis of the pilot study results identified a
conceptual weakness in the originally proposed goal theory research model that was
addressed by adding two important goal attributes variables, job goal difficulty and job
goal specificity, to the full study goal theory research model and a revised web survey
instrument. The full study was then implemented and its sample data set was evaluated
using the same methodology described above. At each step during evaluation of the full
study results, a comparative analysis between pilot study and full study results was also
conducted.

B.

Two Work Motivation Research Models

This study developed and evaluated two separate research models. The first

research model adapts the framework of the goal-setting model of work motivation to
evaluate relationships between goal setting theory predictor variables and work
engagement criterion variables. Specification of causality paths between research
variables will be guided by the assumptions in Locke and Lathams (2002) goal theory
model. These assumptions are that values and personality will have a distal influence on
the goal setting attributes of goal specificity and goal difficulty. Goal specificity and goal
difficulty have a proximal influence on the goal and efficacy mediating mechanisms of
direction, effort and persistence, which influence performance and outcomes. Selfefficacy will have a proximal influence on goal attributes and direction, effort and
persistence. Goal moderators such as feedback moderate the influence that direction,
effort and persistence have on performance and outcomes. Predictor variables included
achievement goal orientation, occupational self-efficacy, two job goal attributes, and

59

two measures of the goal theory moderator variable feedback, all acting as antecedents of
an employees work motivation state. Work engagement and its three dimensions were
treated as the criterion variables for measurement of work motivation in the full study
goal theory research model shown in Figure 4.1.

Figure 4.1 Proposed Goal Theory Research Model of Work Engagement

The second research model evaluated relationships between several core job
characteristics acting as predictor variables of work engagement, and work engagement is
included as the criterion variable measuring the resulting level of an employees work
motivation. Shown in Figure 4.2, the predictor variables include the five core job
characteristics measured by the revised job diagnostic survey plus two additional related
job characteristics from the original job diagnostic survey. Work engagement and its
three dimensions, treated again as the criterion variables in this second research model,
are treated as psychological state variables related to personal and work outcomes.

60

Figure 4.2 Proposed Job Characteristics Research Model of Work Engagement

An important issue in the selection of variables for both models was the matching
of specificity levels, or level of abstraction, for all the variables in this research study.
The theory of work engagement defines the construct as a positive, fulfilling work-related
state of mind, a long-term persistent affective-cognitive state (Schaufeli et al. 2002). The
predictor variable measurement scales in the survey instruments selected from the
literature are all considered to be measures of long-term individual employee behaviors
or attitudes.
C.

Description of Test Instruments

Predictor variables were measured using several separate measurement scales

found in the literature review and merged together in a single web survey instrument.
The individual survey instruments containing the measurement scales used in this

61

research were selected based on their direct applicability to the two theories of motivation
selected for this research study. In addition, the survey instruments selected met the
requirement of matching the level of abstraction of the work engagement survey
instrument. All survey instruments chosen for this study met have been validated in
previous research studies and have demonstrated acceptable reliability. Table 4.1
provides a summary list of the measurement scales selected for use in this research.
Table 4.1 Summary of Individual Test Instruments Selected from the Literature

The scales were reproduced intact as presented in the literature for the purpose of
comparing the psychometric properties of the sample results to published reliability
ratings and factor analysis results for each of the chosen measurement scales. Each set of
predictor variables was organized into a separate section with its own short introduction
also reproduced intact from the original published version of the instrument. The pilot
study instrument consisted of 15 measurement scales and a total of 62 survey questions.
The results of the pilot study provided data that confirmed the reliability of all but one of
the 15 measurement scales. As previously mentioned, the pilot study results also
identified the need to add two additional goal theory predictor variables to measure the
62

two goal attributes constructs of goal specificity and goal difficulty, yielding a full study
survey instrument with a total of 17 scales and 71 survey questions. Each individual
survey instrument and its list of survey questions are presented in Appendix B. The final
web survey instrument developed for the full study is presented in Appendix C. Table C.1
provides a map between test instruments selected from the literature, their scale items and
the position of each scale item in the web survey instrument by identified by survey
section and question number.

1. Utrecht Work Engagement Survey Instrument

The 17-item Utrecht Work Engagement Scale (UWES), a work and well-being
self-report questionnaire measures three dimensions of well-being that a worker feels
while at work: 6 items measuring vigor, 5 items measuring dedication, and 6 items
measuring absorption. Questions are to be answered using a seven point Likert-type
response scale that measures how often the employee has experienced the feelings at
work, from 0 (never) to 6 (every day). More recently, a 9-item short version of the
survey has been validated by the developers of the UWES, with three scale items per
dimension. The three dimensions are highly correlated with respect to each other, yet
empirical studies have shown they here is sufficient discriminant validity to warrant the
three factor structure. The first reported measurement scale reliabilities were Cronbachs
alpha of 0.79 for vigor, 0.89 for dedication, and 0.72 for absorption (Schaufeli et al.
2002). The scale reliabilities reported for the short version of the measurement scales
were a Cronbachs alpha of 0.77 for vigor, 0.85 for dedication, and 0.78 for absorption
(Schaufeli et al. 2006).

63

2. The Job Diagnostic Survey and Revised Job Diagnostic Survey

In order to evaluate work engagement within the framework of job characteristics
theory, a well-documented survey instrument was identified in the motivation research
literature, the Job Diagnostic Survey developed by Hackman and Oldham (1975). They
intended for the instrument to be used as a diagnostic tool by managers interested in
evaluating the motivational potential score, or MPS, of a given work situation, in order to
develop a strategy for job enrichment through redesign of the work. The JDS measures
the perceptions of employees about core job characteristics in their work environment.
The first study to be published about the job diagnostic Survey (JDS) by Hackman and
Oldham (1975) listed internal consistency reliabilities ranging from 0.59 to 0.71. They
also reported that the items that compose each scale showed adequate discriminant
validity. Ratings of the job characteristics by different sample groups and outside
observers showed moderate convergence of most dimensions. Correlations among the
JDS scales are moderately positively inter-correlated. Job dimensions are positively
related to measures of personal and work outcomes. Analysis of variances of the JDS
scales between various jobs showed differences that were statistically significant for all
scores. Some scales are more sensitive to between-job differences than others. This
finding has been replicated many times in different work situations by researchers that
incorporated the JDS into organizational behavior studies. The JDS continues to be a
relevant and empirically well-documented survey instrument. Different job situations can
therefore be evaluated and compared using the JDS scores.

64

The Job Diagnostic Survey was one of several job characteristics survey
instruments found in the literature. A second instrument titled the Job Content
Questionnaire (Karasek et al. 1998) was also developed in the 1970s. Its purpose was to
make comparative assessments of several psychological constructs which however were
not directly related to job design characteristics. The instrument was dropped from
further consideration early in the literature review phase of this research.
A third instrument, the Job Characteristics Inventory (JCI) was developed around
the same time as the JDS in an attempt to offer researchers and managers an alternative to
the JDS (Sims and Szilagyi 1976) (Sims et al. 1976). The Job Characteristics Inventory
was developed from the same foundational research used to develop the JDS. There were
six job characteristics in the final version of the JCI, including Variety, Autonomy,
Feedback, Friendship, Task Identity and Dealing with Others (Sims et al. 1976). This
competing survey instrument did not find widespread acceptance and its research stream
appears to have dried up by the late 1970s with no empirical studies found in the
literature from the mid-eighties to the present. The Job diagnostic Survey continued to be
used in research and improved upon during the later eighties up to as recently as 2007,
and the JDS was therefore selected for use in this research study.
In follow-on empirical studies, researchers using the original job diagnostic
survey did not always recover a true five-factor structure. The revised job diagnostic
survey was developed ten years later in response to factor purity problems identified with
reverse score scale items (Idaszak and Drasgow 1987). The reverse score items were
reworded so that all scale items are positively scored. The revised job diagnostic survey
continues to be utilized, with the most recent study found in the literature confirming its

65

factor validity (Buys et al. 2007). This study reported the following Cronbachs alpha
scale reliabilities: 0.74 for skill variety, 0.67 for task identity, 0.70 for task significance,
0.72 for autonomy, and 0.79 for feedback from job itself (Buys et al. 2007). The revised
version of the Job Diagnostic Survey was selected for incorporation into this research
study. There are five factors in the JDS with three scale items each. A seven point Likert
type response scale was used. The questions are arranged into two separate sections
allocating one question from each factor in section one and the other two questions per
factor in the second section arranged in randomized order. Two additional measurement
scales in the original version of the JDS have been added to this study because they were
considered related factors of interest. Feedback from agents attempts to measure
feedback from coworkers and supervisors, and Dealing with others attempts to measure
the interconnectedness of a particular job with other jobs in the organization (Hackman
and Oldham 1975). Reported scale reliabilities were 0.78 for feedback from agents and
0.59 for dealing with others (Hackman and Oldham 1975).The scale item wordings of
these two related factors appear to have face validity and provide a more complete
understanding of employee perceptions of feedback on the job.
3. Achievement Goal Orientation Survey Instrument
Objectives of incorporating goal orientation into the research effort were two fold:
first, does an individuals goal orientation have any significant correlation to their level of
work engagement; and second, the full study considered the possibility of interactions
between goal orientation variables and goal attributes variables. Previous research in this
area has indicated significant interaction effects can exist between mastery approach
goals when manipulated to appear more difficult, predicting higher performance
66

outcomes (Senko and Harackiewicz 2005). The 2x2 framework for achievement goal
orientation in a work domain survey instrument measured how individuals approach and
react to achievement goals (Baranik et al. 2007). The survey instrument developed by
Baranik et al. (2007) is an extension of earlier foundational work on goal orientation
measures, and therefore was the logical choice for this study because it represents the
latest and most developed theory of goal orientation in a work domain to be found in the
literature. Their achievement goal orientation variables were used to measure differences
in dispositional trait-like characteristics of individuals that are theorized to be predictors
of an employees goal choice. People who are concerned with developing their
competence or mastering a task typically set mastery goals. People who are concerned
with demonstrating their competence relative to others will be oriented to set
performance goals. Each category of goals can also be defined in terms of an individual
trying to approach favorable judgments and demonstration of competence, while others
instead try to avoid unfavorable judgments and demonstrating incompetence compared to
others (Coats et al. 1996). The two orientations of approach and avoidance can be
exhibited with respect to either mastery or performance goals, hence the 2x2 dimensional
framework of the survey instrument (Baranik et al. 2007). A seven point Likert type
scale was used in the instrument. Two studies were conducted by Baranik (2007) in order
to develop a sufficiently reliable mastery-avoidance scale, confirm the reliabilities of the
other three scales, and perform a confirmatory factor analysis of the four-factor structure
of the final 18-item survey instrument. Cronbachs alpha was reported as 0.89 for
mastery approach, 0.74 for mastery-avoidance, 0.88 for performance approach, and 0.77
for performance-avoidance. The first study consisted of 341 introductory psychology

67

students at a mid-sized southeastern university who met the criteria of having held a job
or currently working at a job. The second study consisted of 307 students that met the
same criteria.

4. Job Goal Specificity and Job Goal Difficulty Measurement Scales

Two goal attributes variables, job goal specificity and job goal difficulty, were
added to the full study research model to provide a more complete assessment of goal
setting antecedents. The measurement scales for goal specificity and goal difficulty were
used to capture proximal assigned work goal attributes. Goal attributes are important
antecedents of work motivation. Measurement scales for three different versions of the
two goal attributes were found in the literature. The earliest published set of goal
attributes measurement scales (Steers 1974) that were identified and evaluated as
potential candidate scales for this research were found to have been incorporated into
two more recently developed goal attributes measurement scales (Wright 2004) (Fang et al.
2004). Of the two newer sets of goal attributes measurement scales (Fang et al. 2004)
focused on assigned goals while the set of second measurement scales used a more
generalized wording of job goals in each scale item (Wright 2004). The usage of the
term assigned was considered to be potentially confusing or restrictive to survey
respondents, so the two goal attributes scales by Wright (2004) were selected for this
research. These two measurement scales were developed and validated in a public sector
organization setting. Cronbachs alpha measurement scale reliabilities were reported as
0.74 for job goal specificity, and 0.85 for job goal difficulty (Wright 2004). Two different
ranges for the Likert-type response scales were used by Wright, a 1 (strongly disagree) to

68

6 (strongly agree) agreement scale, and a 0 (never) to 4 (always) frequency scale. Scale
items within each job goal attribute scale incorporated both of these response scales. The
scores of items within each scale could therefore not be added or averaged together
directly in their raw format. Individual item response values were therefore standardized
to an equivalent score ranging from 0.0 to 1.0, and the standardized scores were used for
all subsequent evaluations and analyses.

5. Occupational Self-Efficacy Measurement Scale

Birgit Schyns and Gernot von Collani (2002) developed a survey instrument that
is intended to be a reliable one-dimensional measure of the psychological construct of
occupational self-efficacy. Occupational self-efficacy is considered to be a stable longterm measure of an individuals level of work-related self-efficacy (Schyns et al. 2002).
Other self-efficacy measurement scales were also considered. Task specific measures of
self-efficacy (Bandura 1982) were deemed unsuitable for this research study because
their short-term level of measurement specificity was not well matched to the long-term
attitudinal and behavioral measurement frame of the work engagement construct. A
generalized self-efficacy scale (Chen et al. 2001) was identified that was better matched
to the specificity level of work engagement. It was discarded because the generalized
nature of its scale items was deemed significantly less useful than the occupational selfefficacy scale (Schyns et al. 2002). The developers of the occupational self-efficacy
construct have published the results of three studies that assessed the reliability and
validity of the instrument as a self-efficacy construct for organizational behavior research
(Schyns and Collani 2002). The first study conducted by Schyns and Collani included

69

153 participants that were acquaintances of psychology students who recruited them for
the study based on the criteria that the subject had to have had job experience or was
currently employed. The sample included a broad range of professions and educational
levels. A second study consisted of 326 participants that were selected on the condition
that they currently worked in a hierarchically low position (no subordinates). A
third study consisted of 100 participants of a cross-section of blue-collar workers. The
results of the three studies validated the survey instrument and indicated it was
sufficiently reliable to be used in further research studies that seek to evaluate selfefficacy in its generalized work-related form. A six point Likert type response scale was
used for all six positively scored scale items. The scale has been validated in five
countries (Rigotti et al. 2008). The six item version of the measurement scale was used in
this research study. They reported a scale reliability of Cronbachs alpha of 0.85 in
Belgium and 0.90 in Great Britain (Rigotti et al. 2008).

D.

Demographics Measurements
The final section of the combined survey instrument contained a list of

seven demographics related questions which include gender, ethnicity, education level,
supervisory duties level, job description by selecting from a list of job categories, the
total number of years employed in their profession, and the number of years at their
current job. These questions were not required fields, thereby allowing a respondent to
complete the survey but remain demographically anonymous if they chose to for privacy
reasons. The primary aim of this research was to treat the sample population as a single
unit of analysis at the organizational group level. These demographics may be useful in a

70

follow-on analysis effort of demographic subsets if the organizational level analysis

results warrant such additional effort.

E.

The Pilot Study

The theoretical research population selected for this research consists of all

employees of engineering technical services organizations in the United States of

America. A pilot study was first conducted by collecting data from a convenience sample
believed to be roughly equivalent to a subset of the theoretical population: engineering
graduate students enrolled in the industrial & systems engineering department of a local
university either currently employed or who had recently held a full-time job. These
engineering graduate students were from a variety of organizations including private
companies, government research and development organizations, or staff members of the
university. The data collection methodology selected for the pilot study was to use
internet web browser technology, email, and a web-based survey development software
application for survey creation, distribution, and data collection. All individual survey
instrument scale items and response scales were reproduced in their original published
format as closely as was possible. Each individual instrument was organized into a
separate section with a short introduction.
The engineering graduate students were briefly informed of the purpose of the
pilot study during class and asked to voluntarily complete the web survey on their own
time by accessing the instrument via an internet URL hyperlink. The pilot study
produced 60 completed surveys, with the majority of respondents being engineers.
Three other job categories represented in the pilot study sample provided some of the job

71

description diversity expected in the theoretical research population. The results of the
pilot study were deemed favorable enough to warrant expansion of the research effort to a
full study sample population. Prior to deployment of the web survey instrument in a
larger full study population, preliminary analysis of the pilot study results identified a
deficiency in the initial set of goal theory predictor variables. Goal orientation variables
had very weak correlations to other goal theory variables and the three dimensions of
work engagement. This result prompted inclusion of two additional goal theory variables
identified during the literature review but not included in the pilot study. The addition of
two job goal attributes variables made the goal theory research model significantly more
robust and better aligned with the nomological network of the goal theory model of work
motivation (Locke and Latham 2002).
Work engagement had been placed first in the sequence of fifteen measurement
scales in the pilot study web survey instrument, and preliminary factor analysis of the
pilot sample data set caused concern that the work engagement survey instrument might
have suffered from its starting position in the layout design of the pilot study web survey.
Research on large multi-item survey instruments has shown that the first 25 survey items
tend to be of lower measurement quality than questions from 26 to 100, with a significant
drop-off in measurement quality beyond the 100th item (Andrews 1984). To compensate
for this possibility, the full study web survey instrument was modified to have
four separate versions with work engagement, job characteristics, goal attributes/selfefficacy, and goal orientation each placed first in a version of the web survey instrument.
The four versions were then randomly administered to one fourth of the full study organization.

72

F.

The Full Study

The next opportunity for deployment of the web survey instrument was provided

by a local engineering technical services company. This engineering services company,

consisting of 877 employees at the time of the study, is a privately owned business entity
that provides engineering and technical services through multiple technical services
support contracts with the U.S. government. The company manages an entirely on-site
support contractor workforce located at a geographically diverse collection of
U.S. government laboratory facilities engaged in either research and development or test
and evaluation for the U.S. Army, Navy, Air Force and NASA. The relatively small
number of administrative function employees located at corporate headquarters also
participated in the full study.

G.

Full Study Approach to Data Collection

The data collection methodology selected for the full study was to again use

internet web browser technology, email, and a web-based survey. Access to the
electronic survey by employees in the full study organization was provided through an
email that contained a secure web address hyperlink to enable private and anonymous
web access via personal computer to the electronic survey. A single introductory email
letter was sent out to all members of the sampled organization. No follow-up reminder
emails were sent. The single email explained the purpose of the survey and requested
their voluntary participation by asking them to open the embedded web address hyperlink
which provided a secure connection to the electronic survey located on the web services
company server computer. The email and survey in web format as it appeared on a

73

respondents computer screen is presented in Appendix C. Respondents were required to

complete all questions on each page before being able to proceed to the next page of the
survey. The survey could be partially completed and the respondent was allowed to reuse
to the web link to return to their own partially completed survey at a later date for
completion of their survey. Only one completed survey was permitted per each unique
personal computer IP address, thus providing a measure of confidence that only
one survey response set per employee was obtained. The IP addresses were tracked
privately by the web survey services company but were not included in the data results,
thereby providing complete anonymity to the survey respondent. The web survey
company managed the secure and confidential collection of all survey data and stored the
results. The survey results were downloaded for analysis in a comma separated file
format that was compatible with spreadsheet software applications.

H.

Assessment of Factor Purity Through Validity Studies

In order to answer the research questions of this study, an analytical assessment of

the factor structure contained in the sample data results was performed. The validity of
the constructs was evaluated through the application of both confirmatory factor analysis
and structural equation modeling analysis techniques. Construct validity refers to the
degree to which inferences can be legitimately made from the operationalization of the
theoretical constructs as they have been defined in their respective theories and
operationalized in a survey instrument with scales designed to measure the theoretical
components of the constructs (Andrews 1984). Validity studies help to determine how
well the survey instrument scales actually measure the concepts of the goal theory of

74

work motivation, job characteristics theory and work engagement. Convergent and
discriminant validity are both considered sub categories of construct validity. The
two concepts work together. If convergent and discriminant validity can be demonstrated
then evidence for construct validity has been demonstrated (Andrews 1984).
For convergent validity, measures of constructs that theoretically should be related
to each other should actually be observed to be related to each other by correspondence
or convergence between similar constructs. Discriminant validity refers to the principle
that the indicators for different constructs should not be so highly correlated as to lead
one to conclude that they measure the same thing. The researcher must conduct a
statistical test to determine whether two constructs differ enough that they can be
discriminated from each other. Factor analysis has been used to conclude that constructs
are different if their sets of indicator variables load most heavily on different factors in a
confirmatory factor analysis. For discriminant validity there must be evidence that the
measures that should not be related are in reality not related. The relationship between
measures of different constructs should be low to be able to discriminate between
different constructs. Cross-construct correlations should be much lower than convergent
correlations. It is important to look for patterns in the correlation matrix of survey results
to support convergent and discriminant validity (Trochim 2006).
An ideal comprehensive factor validity study would include usage of the multitrait-multi-method matrix approach (MTMM), but this would require multiple
measurement methods be used to measure each of several concepts. With respect to the
chosen goal theory constructs, the job characteristics constructs and work engagement
research, the empirical studies to date have typically been limited to one method, the use

75

of a one-time self-report questionnaire administered to a cross-sectional sample of a

population. No studies in the engagement literature have reported the use of the MTMM
method. Individual studies have sometimes reported factor analytic assessments of
construct validity using confirmatory factor analysis, but in most of the studies, structural
equation modeling (SEM) was used to evaluate construct validity.
Empirical studies of the discriminant validity of the goal theory attributes
measurement scales, the Revised Job Diagnostic Survey and the work engagement
UWES measurement scales are reported in the literature. Most of the previously
published research studies related to the selected survey instruments typically included
data results from large size samples (i.e., greater than 500 responses). These large sample
sizes helped to build a strong argument that the measurement scales and their related
survey instruments are a valid and reliable measurement of their respective psychological
constructs. This research will add to the body of literature with respect to the construct
validity of these constructs when their measurement scales are deployed within a United
States engineering technical services organization context.

76

CHAPTER V

RESEARCH RESULTS

A.

Data Collection Results

Between the two studies conducted for this research, a total of 420 completed surveys

were obtained for subsequent data analysis, 60 surveys from the pilot study and 360 surveys
from the full study. The results are summarized in Table 5.1. The two sample sets were
analyzed separately. A comparative discussion between pilot and full study results is also
included. The response rate of 41% for the full study is considered acceptable because
response rates discussed in the literature for electronic web based survey methods are
typically in the 25% 35% response rate range (Kaplowitz et al. 2004).

Table 5.1 Combined Web Survey Instrument Data Collection Results

77

The pilot survey data was collected within a one month time period. The set of
results that was obtained from the full study of the engineering services company took
five months to collect because separate divisions of the company were surveyed
sequentially in time at the request of the organizations management, due in part to
waiting periods for authorization from local management to proceed with survey
distribution and also due to a hurricane catastrophe which affected three of seven work
sites of the company at the beginning of the full study survey distribution process. A
decision was made to defer sampling those divisions affected by the hurricane for
three and one half months to allow any adverse effects the hurricane had on
organizational facilities to be mitigated before surveying those employees. The final set
of usable survey responses consisted of nearly equivalent quantities of completed surveys
for each of the four versions of the full study survey instrument that had been distributed.
A review of descriptive statistics measures including means and standard deviations
indicated there were no discernable differences observed between the different versions
of the survey instruments or the different divisions of the company. The technical
services organization surveys were therefore combined into one data set, n = 360,
representing the technical services organization in subsequent analyses.

B.

Demographics Analysis
A summary of responses to demographics questions contained in the survey are

provided in Tables 5.2 through 5.8. The demographic data for the respondents includes
gender, ethnicity, job description, supervisory responsibilities level, education level, and
years of employment. The results are provided separately for both pilot study (n = 60)

78

and full study (n = 360) to allow comparison of the two samples to each other. The
two separate study populations are very similar in gender and ethnicity. The majority of
respondents were male, white/Caucasian employees as shown in Tables 5.2 and 5.3.

Table 5.2 Gender, Full Study and Pilot Study

Table 5.3 Ethnicity, Full Study and Pilot Study

As shown in Table 5.4, approximately 75% of survey respondents from both the
pilot study and full study had either lower level or no supervisory responsibilities.
Table 5.4 Level of Supervisory Duties, Full Study and Pilot Study

79

The engineering graduate students in the pilot study sample had a higher level of
advanced college degree education than the full study sample population of the
engineering services company as shown in Table 5.5. The education level differences are
logical outcomes from the nature of the full study sample population. A large percentage
of the full study engineering services company personnel classified themselves as
technicians in Table 5.6.

Table 5.5 Level of Education, Full Study and Pilot Study

Table 5.6 Current Job by Category, Full Study and Pilot Study

80

A significant difference between the two sample populations was that the pilot
study engineering graduate students were younger, with fewer years of professional
employment and fewer years at their current job than the employees of the engineering
services company. The results are presented in Tables 5.7 and 5.8.

Table 5.7 Number of Years Employed, Full Study and Pilot Study

Table 5.8 Years Employed at Current Job, Full Study and Pilot Study

81

Gender in the full study sample population was determined as best as was
possible by examination of employee first name. Age and education for the full study
sample population were obtained from published statistics on the company website for
the same year the study was conducted. The demographic information about the
company that was available is presented Table 5.9 for comparison to the full study sample
results.

Table 5.9 Statistics of Engineering Services Company Full Study Population Compared
to Full Study Sample Results

A comparison of the full study demographic results for gender, education and age
to the demographic data of the engineering services company indicates the full study
sample data set is a reasonable equivalent representation of the company sample
population.

C.

Analysis of Survey Results Utilizing Descriptive Statistics

82

The two studies conducted for this research produced sets of raw data in the form
of response scores for each measurement variable represented by the web survey
questions (scale items). Surveys with incomplete sets of response scores were discarded.
The raw scores were first reviewed visually using scatter plots and box plots to identify
unusual patterns and check for any obvious discrepancies in the raw data sets. Aside
from the presence of some outliers, the patterns in the raw data appeared acceptable with
varying amounts of dispersion and non-normality in the distributions. Comparison of the
box plots for each study immediately implied that differences exist between the pilot and
full study results. The pilot study results produced a consistent trend of lower median
response score values relative to the full study results.
Descriptive statistics were computed next for all measurement variables including
the mean, standard error, standard deviation, and two measures of non-normality,
skewness and kurtosis. A complete listing of descriptive statistics and plots for pilot and
full study results is presented in Appendix D. Comparison of means and standard
deviations for each measurement variable again indicates that differences exist between
the two sample populations. Skewness and kurtosis were also computed because they
provide an assessment of how much the assumption of normally distributed measurement
data has been violated. Two thirds of the seventy-one measurement variables exhibited
significant non-zero values for both kurtosis and negative skewness. The potential
impact of non-normally distributed measurement results must be reviewed during
subsequent factor analysis and structural equation modeling which are based on the
maximum likelihood method for estimating factor coefficients (Byrne 2001).

83

The next step was to compute raw scores of the latent variables for each survey
respondent. The latent variables in this research are reflective measures, represented by
calculating a mean score for each of the seventeen measurement scales set of survey
questions. The individual survey respondent latent variable mean scores were also
aggregated into overall mean scores for each latent variable for both study samples. The
two sets of aggregate latent variable mean scores were analyzed using a 2 sample t-test to
evaluate statistical differences in mean scores between the pilot and full studies. The
results are presented in Table 5.10 for both pilot and full study, including standard error
of the mean scores, standard deviations, medians, skewness and kurtosis. Thirteen of the
seventeen latent variables produced t-statistic values with magnitudes larger than 2,
indicating those aggregate mean score results are from statistically different populations.
The aggregate mean scores show that both study sample populations responded
positively to all questions about their perceived level of work engagement, indicating a
majority of respondents feel they have a relatively high level of work engagement. The
full study results show overall higher levels of engagement than the pilot study results.
Questions about attitudes towards key job characteristics and job goal attributes also
produced consistently positive responses, well above the midpoint of each response scale.
The aggregate scores for occupational self-efficacy were also well above the mid-point of
the self-efficacy response scale. A key result of the goal orientation measures is that both
study populations appear to be strongly oriented toward approach goals while the scores
are significantly below the midpoint in the respondents orientation toward avoidance
goals.

84

Table 5.10 Comparison of Aggregate Latent Variable Measurement Results Pilot and Full Study

D.

Measurement Scale Reliability Analysis

85

A measurement scale reliability analysis was performed next to determine the

extent to which the items within each of the seventeen measurement scales in the
combined survey instrument are related to each other. It is a necessary but not sufficient
criteria for determination of construct validity. Cronbach's alpha measures how well a set
of scale items measure a single one-dimensional latent variable. Cronbachs alpha is
considered a coefficient of reliability and provides an overall index of the repeatability or
average internal consistency of the scale as a whole. Since the measurement scales were
all validated through prior empirical studies, a Cronbachs alpha of 0.7 or greater was
considered a minimum acceptable level of reliability for this research study (Lance et al.
2006). Intraclass correlation coefficients (ICC) were used to compute inter-rater
reliability estimates to evaluate how well questions are organized into groups of items in
a measurement scale. The ICC value for a measurement scale describes how strongly
items in the same group resemble each other. ICC values greater than 0.3 infer the
measurement scale has acceptable reliability as a group measure (MacLennan 1993). The
results of the measurement scale reliability analysis are presented in Table 5.11, with the
results for the pilot study and full study each listed separately. The estimated values for
Cronbachs alpha ranged from 0.649 to 0.914. The estimated ICC values for each
measurement scale ranged from 0.293 to 0.714. The full study results exceeded the
minimum thresholds for acceptable measurement scale reliability except for the Dealing
with Others job characteristic measurement scale, which produced a Cronbachs alpha of
0.649. Dealing with others was therefore excluded from use in the job characteristics
research model.
Table 5.11 Measurement Scale Reliability Estimates

86

E.

Correlation Coefficient Analysis of Latent Variables

87

Correlation coefficient analysis was conducted to test whether linear relationships

between pairs of latent variables were statistically significant, produced the same algebraic
sign predicted by theory, and to assess the effect size of the linear relationships between pairs
of latent variables. Pearson r correlation coefficients were calculated. The results are
presented in two tables of correlation coefficients from the pilot study with a discussion of
their meaning to the overall research, followed by three tables of correlation coefficients from
the full study results. Correlations are the top value and its related P-value is the lower value.
The coefficients shown in Table 5.12 include the factors of work engagement and the seven
goal theory latent variables included in the pilot study. The coefficients shown in Table 5.13
include the factors of work engagement and the seven job characteristics latent variables.
P-values greater than 0.05 infer those correlation coefficient estimates were not statistically
significant. The mastery avoidance, performance approach, performance avoidance and
dealing with others latent variables did not have statistically significant correlation
coefficients with one or more other variables. Correlation coefficients in some cases were
also less than 0.3 which is considered too low to be of interest for an effect size. These pilot
study results were considered problematic because there was little to no correlation between
the two sets of predictor variables and the work engagement criterion variables. The pilot
study results imply that values and personality traits as represented by goal orientation have a
weak influence at best on an employees level of work motivation when other more proximal
factors are such as feedback are present in the work environment. More important, the pilot
study left out the two most important goal theory factors that influence the level of work
motivation in an employee: job goal specificity and job goal difficulty.
Table 5.12 Correlations Between Work Engagement and Goal Theory Pilot Study

88

Table 5.13 Correlations Between Work Engagement and Job Characteristics Pilot Study

89

The most important conclusion that emerged from analysis of the pilot study data
results was the need to incorporate measurement scales for the two goal setting attributes
of job goal difficulty and job goal specificity into this research before proceeding with
distribution of the survey in a full study. Appropriate measurement scales had already
been identified during the literature review, and the two goal attributes scales were added
into the web survey instrument grouped next to the measurement scale for occupational
self-efficacy. The full study correlation coefficients results shown in Table 5.14 include the
factors of work engagement and the nine goal theory latent variables. The correlation
coefficients shown in Table 5.15 include the factors of work engagement and the seven job
characteristics latent variables. The correlation coefficients shown in Table 5.16 include the
predictor factors for goal theory versus job characteristics predictor variables.
Table 5.14 Correlations Between Work Engagement and Goal Theory Full Study

90

Table 5.15 Correlations Between Work Engagement and Job Characteristics Full Study

Table 5.16 Correlations Between Goal Theory and Job Characteristics Full Study

91

The results of the correlation analysis were consistent with both underlying theory
and other researchers published results. Patterns that showed up in the pilot study results
can be seen again in the full study results. All the predictor variables and criterion
variables were positively correlated to each other, as predicted by their respective theory
with the exception of the four achievement goal orientation variables. The three work
engagement factors were very highly correlated in both study samples, as the literature on
work engagement predicts they should be. With respect to the three goal theory
antecedent latent variables of job goal difficulty, job goal specificity and occupational
self-efficacy, the two goal attributes have relatively low correlation to each other which
infers that they are distinct constructs, while occupational self-efficacy was more highly
correlated with specificity rather than difficulty. This result may mean that a specific job
goal increases a persons occupational self-efficacy to a greater extent because they have
a better understanding of the details of the goal. A difficult job goal may be expected to
increase a persons occupational self-efficacy to a lesser extent because the more difficult
the job goal is, the more uncertainty person is in their belief that they can achieve the job
goal. This result is consistent with self-efficacy and goal theory. Autonomy correlated
fairly high with several other job characteristic factors as did feedback from job itself. As
a result, the correlations between the five job characteristics variables did not exhibit as
clear and distinct a factor structure as predicted by job characteristics theory. The
correlation coefficient results between dealing with others and the other job characteristic
variables are evidence that the construct should not be considered a key job characteristic.
The two feedback variables are sufficiently distinct to justify them being treated as
separate constructs, and feedback from agents has a strong enough correlation to job goal

92

specificity and dedication to warrant its inclusion in the goal theory structural equation
model. Two other job characteristics, skill variety and task identity also produced large
enough correlations with goal attributes to be considered potential goal moderator
variables, but they were not included in the goal theory structural equation model due to
model complexity considerations.
The correlation analysis yielded results for achievement goal orientation variables
that were substantially different than the job goal attributes or job characteristics. A clear
understanding of the goal orientation of the survey respondents in this study is
complicated due to the number of non-significant correlation terms which have to be
treated with caution. Goal orientation theory predicts the approach orientations should be
negatively correlated with avoidance orientations. It is not as clear how mastery and
performance orientations should be correlated. The results did not provide clear support
for the approach versus avoidance relationship, and the mastery versus performance
relationship is also not clear. It appears from the results of this study that goal orientation
is a complex distal personality trait that does not have nearly as strong a relationship to
work engagement that the proximal job goal attributes of job goal difficulty and job goal
specificity variables appear to exhibit in the full study results.

F.

Confirmatory Factor Analysis

Confirmatory factor analyses (CFA), also called common factor analysis, were

conducted to generate and then evaluate recovered latent factor structures within the
sample data sets of both pilot and full studies. The purpose of this analysis was to test for
factor purity within the sample data sets with respect to the theorized factor structures

93

presented in the literature for the individual survey instruments and their measurement
scales. CFA assumes that there are latent factors that have a causal influence on the scale
item scores. The method analyzes the data set correlation matrix and assumes linear
relationships among scale item scores and extracts factors that are uncorrelated (OBrien
2007). The factor analytic technique employed in this research is based on the maximum
likelihood method of common factor extraction, which analyzes only the common variance
among the scale items. The common variance is that portion of total variance shared
among a set of items that can be explained by common factors (OBrien 2007). The
unexplained portion of variance in the data set is labeled unique variance and is comprised
of variance specific to an individual scale item combined with random measurement error
(OBrien 2007). The results of the CFA are a set of extracted factors with eigenvalues
greater than one. The latent factor structure implied in the sample data set can be checked
by comparing the number of factors extracted to the theorized number of factors, and
results evaluated to see how well the recovered factor loadings align to their respective
measurement scales. Another check is to look for cross-loading of individual scale items,
where they produce loading values greater than 0.4 on more than one extracted factor
(Ximnez 2006).
The procedure that was applied in this analysis was to first generate a factor
solution with any loadings less than 0.2 automatically suppressed. The next step was to
review the factor solution and manually suppress any cross-loading values less than 0.4
produced by a scale item on a recovered factor that appeared to not be loaded on its own
theoretical factor. The determination of which theoretical factor was being represented by a
given extracted factor was evident by the observed patterns of all loadings for each

94

measurement scale in a factor solution. CFA solutions also sometimes produced a few
scale item loadings that were less than 0.4 on its own theorized factor. Those loadings were
kept in the solution presented here even though they indicate that particular scale item
made a relatively weak contribution to the common variance of its theorized factor and is
therefore a candidate for deletion from its respective measurement scale.

1.

Factor Analysis Results for the Pilot Study

The first step in the factor analysis of the pilot study was to simply load all the

raw scale item scores from the fourteen measurement scales in the web survey instrument
together into a factor analysis data set, and evaluate the resulting factor solution which is
shown in Figure 5.1. The fourteen extracted common factors recovered from the pilot
study data set with eigenvalues greater than 1 represent a cumulative 80.39 % of total
explained variance in the pilot sample data set. These fourteen recovered factors do not
completely align themselves with the theorized factor structure. Fourteen out of
seventeen work engagement scale items loaded together on Factor 1 which is interpreted
as a generalized one dimensional work engagement factor. Three work engagement scale
items from the Vigor scale loaded independently on their own factor along with one other
absorption scale item which generated a strong cross-loading with those three separate
vigor items. Another factor extracted with combined loadings was Factor 2 where the
self-efficacy scale items combined with mastery approach scale items. That result was
unexpected and cannot be readily explained. Other factors appeared to show up aligned
with the theoretical structure of the measurement scales, albeit with a significant number

95

of scale items generating cross-loading values greater than 0.4. The small sample size of
n = 60 prevents an analyst from drawing of any strong conclusions based on these results.

Figure 5.1 Confirmatory Factor Analysis Web Survey Instrument Pilot Study
The next step in the factor analysis was to evaluate subsets of the pilot study
measurement scales based on their individual survey instruments selected from the
96

literature. This included the Utrecht Work Engagement Survey instrument, the Revised
Job Diagnostic Survey instrument, the Achievement Goal orientation survey instrument
and the occupational self-efficacy scale. The results of these four factor analyses are
presented in Figures 5.2 through 5.5. The work engagement survey instrument factor
analysis shows two factor solutions. The first solution shown in Figure 5.2 is for the full
17- item instrument (Schaufeli et al. 2002) and the second is for the 9-item short version
of the work engagement survey instrument (Schaufeli et al. 2006

Figure 5.2 Confirmatory Factor Analysis Work Engagement Survey Instruments Pilot Study
Four common factors were recovered from the seventeen scale items, representing
a cumulative 69.8% of total explained variance. There are a significant number of crossloadings among the four factors recovered from the pilot study sample data set. The
vigor scale again splits up into two factors, producing a four factor work engagement
solution instead of the theorized three factor solution. The factor analysis results for the
97

9-item short version work engagement survey instrument is also presented in Figure 5.2,
and it shows a single recovered factor that appears to represent a generalized one-dimensional
version of the work engagement construct. That single factor represents a cumulative
56.0% of total explained variance in the pilot sample data set.
The pilot study sample data set for six job characteristics theory measurement scales
was analyzed next. The results are presented in Figure 5.3. This factor analysis solution
extracted four common factors with eigenvalues greater than 1, representing a cumulative
74.1% of total explained variance. Scale items from three of the six measurement scales
loaded on Factor 1, and several scale items produced cross-loading values greater than
0.4. It is interesting to note that the two feedback measurement scales loaded separately
without any cross-loading between their common factors.

Figure 5.3 Confirmatory Factor Analysis Job Diagnostic Survey Instrument Pilot Study
98

The next survey instrument to be analyzed with the pilot study sample data set
was Achievement Goal Orientation. The factor analysis solution is presented in
Figure 5.4, and the five extracted factors with eigenvalues greater than 1 represent a
cumulative 72.2% of total explained variance. Four of the factors appear to match their
respective measurement scales with cross-loading, while the Factor 5 appears to be a
spurious factor generated from the cross-loading by two mastery approach scale items.

Figure 5.4 Confirmatory Factor Analysis Goal Orientation Survey Instrument Pilot Study
The final instrument evaluated using the pilot study sample data set was the
occupational self-efficacy scale. The results are presented in Figure 5.5 and show the

99

one extracted factor with an eigenvalue greater than 1 representing a cumulative 60.5% of
total explained variance.

Figure 5.5 Confirmatory Factor Analysis Occupational Self-efficacy Scale Pilot Study
After completing the factor analyses of the pilot study, the results were reviewed
with caution due to the small sample, n = 60. Factor purity in the recovered common
factors did not match theory in most cases. A clear and unambiguous assessment of the
pilot study factor analysis, especially for work engagement was considered desirable
before proceeding with a full study. The author of the work engagement survey
instrument, Dr. Schaufeli, was contacted and presented with the factor analysis results.
His recommendation concerning the factor solution for work engagement was to proceed
with the full study and collect a much larger sample size (i.e., greater than n = 500),
which would enable later factor analyses to recover a factor structure that hopefully
showed better agreement with their respective theoretical latent variables.

2. Factor Analysis Results for the Full Study

A set of confirmatory factor analyses was conducted for the full study sample data
set following the same procedures and guidelines used to evaluate the pilot study sample
data set. The first step was to again load all the raw scale item scores from the now

100

sixteen measurement scales in the full study web survey instrument together and evaluate
the resulting factor solution which is shown in Figure 5.6.

101

Figure 5.6 Confirmatory Factor Analysis Web Survey Instrument Full Study
102

The factor solution presented in Figure 5.6 shows fourteen extracted factors with
eigenvalues greater than 1, representing a cumulative 67.4% of total explained variance.
The 17 scale items for work engagement loaded on Factor 1. This loading structure
appears to represent a generalized one-dimensional work engagement factor. The other
factors which were also recovered from the full study sample data set appear to align
themselves with their respective measurement scale, except the skill variety and job goal
difficulty scale items all loaded on Factor 2. There were also two cross-loading values
greater than 0.4, one loading on Factor 1 and the other on Factor 2. The scale item job
goal specificity 3r which is one of the three reverse worded scale items included in the
web survey instrument failed to load even weakly on its theoretical factor and the scale
items only significant cross-loading value of 0.363 generated in the solution with the
feedback from job itself factor was not deleted to avoid leaving any row of the solution
blank. Three scale items generated weak loadings under 0.4 within their theoretical
factor group that were also left in the factor solution results presented in Figure 5.6.
The next step in the factor analysis was to evaluate subsets of the full study
measurement scales based on their individual survey instruments selected from the
literature. This included the Utrecht Work Engagement Survey instrument, the Job
Diagnostic Survey instrument, the Achievement Goal orientation survey instrument and
the occupational self-efficacy scale which is combined with the two job goal attributes
measurement scales to form a logical grouping of goal theory antecedent latent variables.
The results of these four factor analyses are presented in Figures 5.7 through 5.10. The
work engagement survey instrument factor analysis shows two factor solutions. The
first solution is for the full 17- item instrument (Schaufeli et al. 2002) and the second is

103

for the 9-item short version of the survey instrument (Schaufeli et al. 2006). The results
are shown in Figure 5.7. Three common factors with eigenvalues greater than 1 were
recovered from the seventeen scale items, representing a cumulative 64.3% of total
explained variance. The 9-item short version of the work engagement survey instrument
produced a one factor solution with an eigenvalue greater than 1, and that common factor
represents a cumulative 60.6% of total explained variance. The three recovered factors in
the 17-item solution do not however align themselves well with the three theoretical
factors of work engagement. Ten of the seventeen scale items load on Factor 1, and three
vigor scale items, vigor 4, 5 and 6, produced their own common Factor 3, a pattern
similar to the results of the pilot study factor analysis. The solution also shows three
scale items generating cross-loading values above 0.4.

Figure 5.7 Confirmatory Factor Analysis Work Engagement Survey Instrument Full Study

104

It appears from this analysis result that the 17-item version of the work engagement
survey instrument may indeed require a larger sample size before a stable three factor
structure properly aligned with theory emerges from a common factor analysis solution.
The full study sample data set for six related Job Diagnostic Survey and job
characteristics theory measurement scales was analyzed next. The results are presented
in Figure 5.8. This factor analysis solution extracted five common factors with
eigenvalues greater than 1, representing a cumulative 70.7% of total explained variance.
Scale items from the task identity and the feedback from job itself measurement scales
loaded together on Factor 1. The two feedback measurement scales loaded separately
without any cross-loading between their common factors. Scale item skill variety 1
loaded weakly on its factor, but the value of 0.352 was left in the solution.

Figure 5.8 Confirmatory Factor Analysis Job Diagnostic Survey Instrument Full Study
105

The next survey instrument to be analyzed with the full study sample data set was
Achievement Goal Orientation. The factor analysis solution is presented in Figure 5.9,
and the four extracted factors with eigenvalues greater than 1 represent a cumulative
58.2% of total explained variance. The four extracted factors appear to match their
respective measurement scales as predicted by theory.

Figure 5.9 Confirmatory Factor Analysis Goal Orientation Survey Instrument Full Study
The remaining three measurement scales in the full study web survey instrument
were three goal theory antecedent variables: job goal specificity, job goal difficulty and
occupational self-efficacy. They were grouped together and a confirmatory factor
analysis was conducted to determine their latent factor structure in the full study sample
data set. The resulting factor solution is presented in Figure 5.10. The three factors
106

extracted with eigenvalues greater than 1 represent a cumulative 68.3% of total explained
variance. The three factor structure recovered from the full study sample data set
matches the theoretical factors. The scale item job goal specificity3r generated a weak
loading of 0.309, indicating it is not contributing significantly to its common factor.

Figure 5.10 Confirmatory Factor Analysis Goal Theory Antecedent Variables Full Study
A final step in the common factor analysis of the full study sample data was to
evaluate the two sets of scale items which were selected for use in the goal theory
research model and job characteristics research model. The common factor results for the
thirty scale items selected for use in the goal theory model are shown in Figure 5.11.

107

Figure 5.11 Confirmatory Factor Analysis Goal Theory Research Model Full Study

The six factors extracted with eigenvalues greater than 1 represent a cumulative
64.9% of total explained variance. The factor structure recovered from the full study
sample data set matches the theoretical factors with work engagement producing a single
strong common factor. The scale item job goal specificity3r generated a loading less than
0.2 on its own theoretical factor and no cross-loading greater than 0.4 on any other factor.

108

The final common factor analysis was for the set of scale items selected for use in
the job characteristics research model. The factor solution is shown in Figure 5.12. The
six factors extracted with eigenvalues greater than 1 represent a cumulative 68.4% of
total explained variance. The factor structure recovered from the full study sample data
set matches the theoretical factors with work engagement again producing a single strong
common factor. The scale item skill variety 1 did not generate a significant loading on its
theoretical factor.

Figure 5.12 Confirmatory Factor Analysis Job Characteristics Research Model Full Study
109

The confirmatory factor analysis results for the full study sample data set
indicates that the theoretical latent factors as represented by their measurement scales and
individual scale items were recovered in some but not all cases when evaluated as a
single group in the web survey. Better results were obtained for the groups based on their
test instruments from the literature. The two research model factor analyses generated
factor structures deemed acceptable for further analysis in structural equation models.
Some scale items appeared to be candidates for removal from their measurement scale
due to weak factor loadings. The full study sample size of n = 360 enabled a clearer set
of latent factors to be recovered from the full study sample. A larger sample size would
appear to improve chances for recovery of a factor structure that matches theory and
thereby mitigate the desire to revise measurement scales through elimination of problem
scale items. An additional factor analysis effort was proposed that would simply combine
the two sample data sets to produce a single larger sample size data set. This factor
analysis approach was not implemented. Subsequent review of the relevant literature on
this topic provided strong clarification that any common factor analysis that combines
two samples into one common factor analysis data set is actually invalid and should not
be performed. The following excerpt is reproduced from Comrey et al. (2000) to clarify
this issue of simple combination of multiple sample groups:

Multiple-group CFA must be conducted on covariance matrices and never on

correlation matrices, because correlations represent standardized measures of
association and are thus expected to differ across groups (because groups may differ
in mean level and variance on a latent factor) even when factor pattern matrices (that
are based on analyses of covariances) are invariant (Comrey et al. 2000, pg. 293).

110

The proper methodology for confirmatory factor analysis of multiple groups of

data sets is therefore the use of structural equation modeling which is based on analysis
of covariance structures. The structural equation models developed in the final step of
this research utilized the multi-group method in the AMOS Version17 SEM software
when it was considered appropriate to do so. The multi-group analysis method evaluates
each sample group of data separately and produces a unique set of solution parameters for
each group. The single set of goodness-of-fit measures generated at the end of the SEM
analysis solution output file take into account the fits generated by individual group
sample data sets. Multi-group analysis has been employed in this research to first
develop models unconstrained between groups. In addition, multi-group analysis was
used to test for invariance of work engagement across two sample groups.

G.

Sample Size Analysis

Analysis of sample size adequacy was necessary to determine the effectiveness of this

research study in capturing the characteristics of the general theoretical population structure
from which the sample was obtained. To have sufficient confidence in the factor solutions
generated from the sample population, there is general agreement that the larger the sample size
the more likely the sample factor solution will adequately represent the theoretical population
because measurement error will be reduced and the factor structure will be more stable. A
detailed sample size analysis was only possible after the confirmatory factor analysis solutions
provided results for measurement scale item communalities. Details of these communalities
results are presented in Appendix E.

111

A variety of recommendations have been presented in the psychometric testing

literature concerning minimum required sample size, usually focused on the ratio of number of
survey responses to number of variables (scale items, or survey questions). Conflicting
recommendations from a minimum ratio of 5-to-1 to as much as 20-to-1 responses per variable
have been proposed. A more accurate assessment of a minimum sample size is related to the
number of variables, number of factors, number of variables per factor (over-determination of
the factor), and the size of the factor solutions communalities (MacCallum et al. 1999).
Estimation of the communality of a variable is important in factor analysis because the
communality of a variable is the portion of the variance of that variable that is accounted for by
the common factors (MacCallum et al. 1999).
A theoretical treatment of the sample size issue in recovery of factor structures from a
sample data set is presented in the psychometric testing literature which defines a measure, K,
the coefficient of congruence that attempts to quantify the level of agreement between a sample
covariance matrix and a true population covariance matrix (Tucker et al. 1969). Tuckers
recommendations state that a value of K = 1 indicates perfect agreement, values of K = 0.98 to
1 are considered excellent, and values of K = 0.92 to 0.98 are good. Values less than 0.92 are
considered borderline. The results of Monte Carlo simulations used to determine adequate
minimum sample sizes for a given K value have been published (Mundfrom et al. 2005).
Details of the sample size requirements versus range of communalities and the ratio p/f , the
number of scale items p to the number of theoretical factors f, are presented in Appendix E.
When comparison is made between the published recommendations and the full study
web survey instrument which has a p/f ratio = 4.2, the full study sample size of n = 360 appears
to meet the recommended minimum sample size requirements for a K value of 0.92. This

112

implies there should be good agreement between the sample recovered factor structure and
the theoretical factor structure. The goal theory research model with p = 30 variables and f = 6
theoretical factors can be more directly compared with the published recommendations with its
p/f ratio = 5 and number of extracted factors F = 6. The full study sample data set, n = 360,
appears to meet all recommendations for minimum sample size at a value of K = 0.98, which
indicates there should be excellent agreement between the goal theory research models
recovered factor structure and theoretical factor structure. The full study sample size of n = 360
appears to meet the recommended minimum sample size requirements.

H.

Regression Analysis MPS versus Work Engagement

The structural equation modeling process is the primary analysis method used to

evaluate the hypothesis that job characteristics are meaningful predictors of an employees
work engagement level. There is one issue related to job characteristics theory which cannot
be examined by a structural equation model, the motivating potential score (MPS) of a job
presented in the job characteristics theory of Hackman and Oldham (1975). They had
proposed that managers use the Job Diagnostic Survey results to compute a motivating
potential score, or MPS, for the job or work environment being evaluated. Their formula for
calculating an MPS was presented earlier and was designed so that if any one factor was
found to be low, the resulting MPS would also be low. Further research studies presented
evidence that using a simple summation of survey scores from the Revised Job Diagnostic
Survey was sufficient and less confusing (Buys et al. 2007). Regression analyses were
performed to evaluate linear relationships between a simple summation score of the 17-item
work engagement results for each survey respondent versus their MPS score and their

113

summation score of the five job characteristics used in the MPS. Details of the regression
analyses are presented in Appendix F. A summary of the results is presented in Table 5.17,
and a representative scatter plot for the full study sample population is shown in Figure 5.13.

Table 5.17 Results of Regression Analysis Work Engagement vs. Job Characteristics

The results of the regression analyses for the pilot study sample data set appeared
promising for consideration of either an MPS score or a summation score of job
characteristics to be useful to a manager as simple predictor variables for work engagement.
The adjusted R-sq(adj) values generated in the pilot study analyses explain nearly 50 percent
of the variance in the pilot study data set.

Figure 5.13 Work Engagement Score vs. Motivating Potential Score Full Study
114

The results of the full study regression analyses present a weak case for either the
MPS score or the summation score of job characteristics as a predictor measure of the level
of work engagement in an organization. Figure 5.13 illustrates a potential problem with the
use of an MPS score to predict work engagement. Several data points with high standardized
residuals imply there are employees with very high levels of work engagement who view
their jobs very unfavorably. Other outliers indicate there are employees who view their jobs
favorably but have low engagement scores relative to other employees. A simple metric like
the MPS has limitations and may not capture all factors influencing work engagement.

I. Structural Equation Modeling

The final analysis step in this research was to develop structural equation models
from the full study survey results for the purpose of evaluating the two research models
that were proposed in the research statement. The structural equation models are used to
answer the research questions about work engagement and its theorized relationship to
work motivation. Details of the SEM analysis are presented in Appendix G.

1. Structural Equation Modeling Methodology

SEM is a statistical analysis method that combines regression analysis, path
analysis and factor analysis. SEM is also described as the analysis of covariance
structures because the covariance structure in the sample data set is compared with a
theoretical covariance structure defined by the structural equation model, testing a
proposed set of causal paths between measurement variables and their theoretical latent
variable and between theoretical latent variables in the model.

115

Among its strengths is the ability of SEM to model constructs as latent variables.
Error terms are also included in the model for each measurement variable (the individual
survey instrument questions are the measurement variables) in addition to error terms for
each latent variable. This allows the model to explicitly capture the unreliability of
measurements in the model, which in theory allows the structural relations between latent
variables to be accurately estimated.
Structural equation modeling was performed using AMOS 17.0 software
(Arbuckle 1995). In SEM, the qualitative causal assumptions of a proposed theoretical
model are represented as either correlation paths, or as regression paths. The cause and
effect assumptions of these path lines are tested for their significance and validity (Byrne
2001) (Arbuckle 1995). Structural equation models do assume that causation implies
correlation. The SEM method assumes that the structure of associations observed in an
empirical data set exist because there is some underlying causal process that is imposing
the observed structure between the latent or underlying variables which have been
measured by the researcher. The SEM process in this research followed the sequence of
analysis activity shown in Figure 5.14.

Figure 5.14 Structural Equation Modeling Process

116

A checklist procedure used to evaluate the SEM solution output files, developed
directly from the evaluation methods recommended by Byrne (2001), is presented in
Appendix G. The modeling process started with development of measurement models
for each theoretical latent variable representing the measurement scales contained in the
survey instrument. The measurement models were then combined into first and second
order structural models to specify the causal relationships between the latent variables of
individual survey instruments. The third step was to develop structural models of the
goal theory research model. The fourth step was to develop structural models of the job
characteristics research model. Several alternate configurations of each research model
were constructed and evaluated. A final step was multi-group analysis of the work
engagement 9-item instrument to test for invariance between pilot and full study samples.

2. Goodness- of-Fit Measures for Model Evaluation

Structural equation models are typically evaluated by their goodness of fit
measures. The goodness-of -fit measures generated in the solution results of an SEM
analysis are used to evaluate differences between the sample data set covariance matrix
and the structural equation models theoretical covariance matrix structures (Byrne 2001)
(Arbuckle 1995).
Any one goodness-of-fit statistic by itself is not considered sufficient for
determination of model adequacy. The reporting of too many fit statistics is also not
recommended. A minimum recommended set of model fit statistics has been presented
for the structural equation models, the minimized Chi-squared statistic for a converged
solution CMIN, model degrees of freedom DF, the comparative fit index CFI, and the

117

Root Mean Square Error of Approximation RMSEA. A large chi-squared statistic is

indicative of poor model fit. A ratio of CMIN/DF that is approximately 2 or less is
considered a good model fit (Byrne 2001). In this study the criteria for an adequate
model fit as measured by CMIN/DF will be less than 3 and for a good model fit 2 or less.
Models with CMIN/DF greater than 3 will be rejected because of poor fit between sample
data set and theoretical model covariance matrices. However CMIN is sensitive to
sample size, and the larger the sample size the larger the estimated chi-square statistic.
This becomes an issue because almost all structural models with medium to large sample
sizes will produce large chi-squared estimates. CMIN should therefore not be the sole
criteria for assessment of model fit. For this reason several additional goodness-of-fit
indices are used to assess model fit. A second recommended statistic is the Comparative
Fit Index, CFI (Byrne 2001). The CFI, which is least affected by sample size, ranges
from 0 to 1. A cutoff value of 0.90 or greater was originally recommended as the cutoff
value for adequate model fit. More recent recommendations suggest that a cutoff value
closer to 0.95 or greater indicates a well-fitting model (Hu and Bentler 1999). In this
study the criteria for an adequate model fit as measured by CFI will be a value of 0.90 or
greater and for a good model fit 0.95 or greater. Models that produce a CFI value less
than 0.90 will be rejected because of poor fit between sample data set and theoretical
model covariance matrices. Another recommended statistic is RMSEA, the Root Mean
Square Error of Approximation, also relatively unaffected by sample size, which ranges
in value from 0 to 1(Byrne 2001). An RMSEA value of 0.05 or less indicates a good fit,
and values of RMSEA from 0.05 to 0.08 indicate an adequate fitting model (Hu and
Bentler 1999). RMSEA is a fit measure that includes a 90% confidence interval in its

118

calculation. The low 90% value should be lower than 0.05 and the high 90% value
should be less than 0.08. In this study the criteria for an adequate model fit as measured
by RMSEA will be values less than 0.08 and for a good model fit 0.05 or less. Models
that produce an RMSEA value greater than 0.08 will be rejected because of poor fit
between sample data set and theoretical model covariance matrices. Strict interpretations
of some cutoff criteria may however cause a researcher to introduce excessive type I error
in the assessment of structural equation models, rejecting models when they are actually
acceptable representations of the theory being tested (Marsh et al. 2004). In this study a
model is accepted as correctly specified and considered to have adequate fit between
sample data set and theoretical model covariance matrices if it meets the combined
criteria of CMIN/DF less than 3, CFI greater than 0.90 and RMSEA less than 0.08.
Evaluation of goodness-of-fit results for structural equation models is the primary
method used in this study to evaluate a models validity and also rank it with respect to its
alternative structural equation models. The goodness-of-fit measures found in published
studies for the survey instruments selected from the literature have been incorporated into
the SEM results tables of each survey instrument structural model, allowing direct
comparisons to be made between the published structural models and their equivalent
models in this study. The next three sections present the SEM models and their results
which provide the most direct evidence for answers to the three research hypotheses
about the validity of work engagement and the evaluation of two motivation theories as
predictors of work engagement. A research null hypothesis is rejected and its alternate
hypothesis accepted if sufficient evidence is presented in the form of adequate goodnessof fit measures results for its respective structural equation model.

119

3. SEM Analysis of Work Engagement

Structural equation models of work engagement were developed first to evaluate
the validity of work engagement as measured by the UWES survey instrument.
Three structural models were developed and tested for the full 17-item version of the
survey instrument. The SEM analysis proceeded with evaluation of a one factor first
order model, moving next to evaluation of a three factor first order model, followed by
the evaluation of a three factor second order structural equation model. The 17-item
work engagement structural model that produced the best goodness-of-fit measures was
the three factor second order model presented in Figure 5.15.

Figure 5.15 Second Order Three-Factor 17-item UWES Model of Work Engagement
120

Table 5.18 SEM Analysis Results for Full Version 17-item UWES Work Engagement
Survey Instrument Compared to Published SEM Results

The final 17-item UWES SEM model goodness-of-fit measures are listed in
Figure 5.15 and the results are compared to the first published SEM results for work
engagement in Table 5.18 (Schaufeli et al 2002). While the results compare somewhat
favorably with previously published data, they do not meet the goodness-of-fit cutoff
criteria chosen for this research study. CMIN/DF was greater than 3 and CFI was less
than 0.90. The RMSEA value of 0.07 is near the upper limit of an adequate fitting
structural model. The results of the full 17-item structural equation model did not
provide strong enough evidence to reject the null hypothesis that the 17-item UWES
survey instrument does not validly measure work engagement in the sample population.
The next step was development of structural equation models for the 9-item version
of the UWES work engagement survey instrument, again testing three alternative
structural models. The model shown in Figure 5.16, representing a second order
three factor structural equation model, produced the best goodness-of-fit indexes for the

121

9-item short version UWES work engagement survey instrument (Schaufeli et al. 2006).
The 9-item UWES structural models were all superior in goodness-of-fit results to their
respective full 17-item versions of the UWES work engagement survey instrument.

Figure 5.16 Second Order Three-Factor 9-item UWES Model of Work Engagement
The goodness-of-fit results for the three 9-item short version UWES structural
models are shown in Table 5.19 and compared to published results from other studies in
the literature. The comparative fit index for the 9-item instrument, CFI = 0.96, appears to
support validity of the three-factor model of work engagement with a better fit to the
sample data than the 9-item one-factor model of work engagement, CFI = 0.93. The
three factor model also generated lower RMSEA and minimized chi-square statistic
values than the one factor model, all indications that the three factor model of work
engagement is a better theoretical representation of the construct than the one factor
122

model of work engagement. The combined goodness-of-fit results for the 9-item UWES
structural equation models provide strong enough evidence to reject the null hypothesis
that the UWES survey instrument does not validly measure work engagement in the
sample population. These results compare favorably to published results of the original
study for the 9-item UWES survey instrument (Schaufeli et al. 2006). There are
similarities in results between this research study conducted in the United States and the
published studies from Europe and Japan, for both one factor and three factor models.

Table 5.19 SEM Analysis Results for Short Version 9-item UWES Work Engagement
Survey Instrument Compared to Published SEM Results

Up to this point in the SEM process, multi-group analysis results have been reported
for an unconstrained model only. An unconstrained multi-group analysis allows all model

123

parameters including path regression weights, measurement error terms, latent variable residual
terms and variances to vary between groups. The next step was to conduct a multi-group
analysis with a series of nested models. The nested models progressively test for invariance of
subsets of model parameters across groups. Five models were developed, starting with the
unconstrained model, and each nested model progressively assumed more constraints between
groups, testing for a greater level of invariance between groups. Changes in chi-squared
statistics between each nested model are used to determine a P-value to test the hypothesis of
invariance between groups for each subset of estimated model parameters. The results of these
tests of invariance between groups are presented in Table 5.20.

Table 5.20 Tests for Invariance Work Engagement Multi-Group Analysis

The results show that invariance has been first demonstrated between the pilot and
full study groups for the measurement weights which means the regression path weights
124

between each measurement variable and its respective latent 1st order variable, Vigor,
Dedication, and Absorption. Next invariance between groups has been demonstrated for
structural weights which are the regression path weights between each 1st order latent
variable and the 2nd variable, Work Engagement. Thirdly, the structural residuals for each
first order latent variable have then been demonstrated to be invariant between groups.
Finally, invariance between measurement residuals has not been demonstrated. This
means that the measurement error terms for each scale item varied across the two groups.
The results obtained from the various structural equation modeling analyses of
work engagement, combined with prior common factor analyses of test instruments and
correlation analyses of latent variables, provided sufficient evidence to support the
decision to proceed with development of two structural equation research models
utilizing a reduced set of scale items from the original sample data set of
seventy-one scale items. A summary of the scale items incorporated into each research
model is presented in Table 5.21. The goal theory structural equation research model was
developed and evaluated first, followed by development and evaluation of the job
characteristics structural equation research model. Multiple versions of each research
model were evaluated in order to better understand the validity of different combinations
and directions of the causal path relationships included in each model.

125

Table 5.21 Scale Items Used in the Two Structural Equation Research Models

4. SEM Analysis of Goal Theory Research Model

126

The next analysis step was evaluation of the goal theory research model. The goal
theory SEM process started with a structural model of three goal theory antecedent
variables: job goal difficulty, job goal specificity, and occupational self-efficacy. The two
feedback latent variables from job characteristics theory were added to represent
feedback goal moderator variables. These five constructs represented the predictor
variables in the goal theory research model. The three factors of work engagement were
added to the structural model as criterion variables, with path arrows pointing from goal
theory predictor variables to work engagement criterion variables. The 9-item short
version of the UWES survey instrument represented work engagement in the goal theory
research model because its confirmatory factor SEM solution showed a significantly
better fit over the full 17-item UWES model of work engagement. The final
twenty-eight scale item structural equation goal theory research model shown in
Figure 5.17 includes specification of work engagement as a three factor second order
construct. The goodness-of-fit results for the model including CMIN/DF, CFI and
RMSEA presented in Table 5.22 indicate there is adequate fit between the structural
model and the sample data set. These SEM results provide strong enough evidence to
reject the null research hypothesis and accept the alternate research hypothesis that
antecedent variables from the goal theory of work motivation validly predict work
engagement in the sample population.

Table 5.22 SEM Analysis Results for Final Goal Theory Research Model

127

Figure 5.17 Final Structural Equation Goal Theory Research Model

128

Standardized regression weights from the final goal theory research model are
shown in Table 5.23 for regression paths between latent variables in the SEM model
presented in Figure 5.17. Estimates for three of the thirteen standardized regression
coefficients were below 0.30, but were left in the model because goal theory assumes
those paths should be theoretically significant, and their presence did improve model fit
slightly. The strongest path loadings were between the variable feedback from the job
itself to the variables of job goal difficulty and job goal specificity. Job goal difficulty
produced a stronger path loading than job goal specificity to work engagement.
Occupational self-efficacy produced a weak path loading to work engagement and
moderate sized path loadings to both goal attributes. The goal theory research model
results provide evidence that the three dimensions of work engagement, Vigor,
Dedication and Absorption may provide engineering managers with a practical way to
measure the long term effects within individual employees of the three goal theory
mediator variables of direction, effort and persistence.

Table 5.23 Standardized Regression Path Coefficients for Goal Theory Research Model

129

5. SEM Analysis of Job Characteristics Research Model

The development of the structural equation job characteristics research model
started with the five core job characteristics constructs in the Revised Job Diagnostic
Survey including autonomy, feedback from job itself, skill variety, task significance and
task identity, plus the related job characteristic feedback from agents was also added to
the model. The 9-item short version of work engagement was added to the model. This
first job characteristics structural equation model is presented in Figure 5.18.

Figure 5.18 First Six Factor Job Characteristics Structural Equation Model
130

The latent variable dealing with others was excluded due to its non-significant
correlation to work engagement variables. The six job characteristics constructs
represented the predictor variables with path arrows pointing to the second order work
engagement criterion variable, explicitly modeling the hypothesized cause and effect
relationships underlying this proposed theory of motivation. The SEM results for this
first job characteristics research model were very unsatisfactory. The goodness-of-fit
results for this first job characteristics research model are also shown in Figure 5.18. The
value of the CMIN/DF measure is above the cutoff criteria of 3 and the value of the CFI
measure is significantly lower than the cutoff criteria of 0.90. These results were
unsatisfactory and a clear indication of model misspecification. A review of the models
solution set using the SEM checklist described in Appendix G identified several
problems. The regression weights for path lines between the job characteristics predictor
variables and the work engagement criterion variable were statistically non-significant
with all p-values > 0.05, and their respective standard error estimates were also very high.
The matrix of standardized residual covariance terms contained numerous high values
exceeding 2.58, an indication of covariance problems. Feedback from agents scale items
generated the highest values with other latent variable scale items in the solution set
standardized residuals covariance matrix, indicating it should be eliminated from the
model. The next series of models were therefore reduced to the five core job
characteristics of the Revised Job Diagnostic Survey instrument, eliminating feedback
from agents from further use in the job characteristics research models.
The resulting structural models were now defined by a total of eight measurement
scales representing five job characteristics predictor variables and three work engagement

131

criterion variables, for a combined total of twenty-four scale items in the structural
equation job characteristics research models. In order to examine the relationships
between the various job characteristics and work engagement latent variables, a
2 x 3 matrix of structural models was developed. Three structural equation models were
developed that modeled job characteristics as first order predictor variables directly
connected by causal path lines, first to a single 9-item criterion factor for work
engagement, followed by a second model that included the three factors of work
engagement modeled separately as first order criterion variables. In the third model work
engagement was modeled as a second order construct with its three dimensions as
first order variables. A second set of three models were then developed that modeled
meaningfulness as a second order predictor construct for the three job characteristics
variables skill variety , task identity and task significance. This model follows the
nomological network originally presented in job characteristics theory (Hackman and
Oldham 1975). Autonomy and feedback from the job itself were left as first order
predictor variables. The structural equation model shown in Figure 5.19 shows how the
second order variable meaningfulness was incorporated into the second set of job
characteristics research models. This additional complexity actually produced better
fitting solutions than the first order job characteristics factor structural equation models.
The goodness-of-fit measures for this final job characteristics research model are also
included in Figure 5.19. The results of all SEM analyses were still not favorable and did
not provide support for the predictive validity of the job characteristics research model as
specified. A recurring pattern of poor goodness-of-fit results and numerous modeling
misspecification errors emerged from the SEM analysis with only slight variations in the

132

numerical results for each of the job characteristics structural models that were
developed. The null research hypothesis that the job characteristics research model does
not validly predict work engagement in the sample population could not be rejected.

Figure 5.19 Final Five Factor Second Order Job Characteristics

Three Factor Second Order Work Engagement Model

133

The goodness-of fit indexes for all six models are summarized in Table 5.24. The
conclusion reached was that misspecification issues in the job characteristics research
structural models were severe. Potential model improvements suggested by modification
indices generated by each SEM solution did not produce better fitting structural job
characteristics structural models. Rearrangement of regression paths also did not
improve model fit sufficiently to meet this studys pre-determined cutoff criteria for
acceptable fit between structural equation models and the sample data set. The
conclusion reached is that the theorized first order causal paths drawn directly between
job characteristics latent variables and work engagement are not valid as specified. The
implication is that there may be other unspecified factors in the work environment that
act as mediators between the job characteristics predictor factors and work engagement.
Table 5.24 SEM Analysis Results for Six Job Characteristics Research Models

134

CHAPTER VI

CONCLUSIONS AND RECOMMENDATIONS

A.

Principal Conclusion
This research study has evaluated the work engagement construct from a new

viewpoint that had not been previously published in the literature by explicitly
considering the utility of the construct as a measure of work motivation. The results of
this study help to advance the understanding of work engagement as a measure of
psychological well-being by proposing goal theory as a completely new viewpoint for
describing the antecedents to work engagement. This study has provided empirical
evidence that the construct of work engagement is useful as a valid criterion
measurement indicator of an individuals work motivation. The results of this study can
help organizations expand both their understanding of healthy workplace practices, and
show how appropriate application of goal theory principles can enhance employee wellbeing through increased levels of work motivation. Work engagement can be used by an
organization as a measure of employee well-being from work motivation. Demonstration
of invariance between the two sample groups in this research adds further validity to the
comparative use of work engagement across different organizations. The three research
questions that prompted this study are answered in the sections that follow.

135

1. Work Engagement Factor Structure

The confirmatory factor analysis results using structural equation modeling
support the conclusion that work engagement is best defined by the three factor structure
of vigor, dedication and absorption, as measured by the 9-item short version of the
Utrecht Work Engagement Survey (UWES) instrument (Schaufeli et al. 2006). The
results for the full seventeen item measurement scale the UWES consistently produced a
mixed set of factor loadings between the 17 scale items and the structural equation model
for the 17-item version of work engagement did not meet several goodness-of-fit cutoff
criteria, even though the results of this analysis compared favorably to results published
in the literature. A larger sample size may enable a better understanding of the factor
structure of the 17-item work engagement survey instrument. The short version of the
survey instrument dispenses with items four, five and six, providing greater clarity to the
definition of the vigor construct. The same approach is followed by the reduction in
number of scale items for both the dedication and absorption scales. This more compact
definition of the three factor concept of work engagement was found to produce a good
fit with work motivation in the structural equation goal theory research model developed
and evaluated in this study.
There have been other research studies that have published results which also did
not completely agree with the theoretical factor structure of the 17-item UWES. The
initial confirmatory factor analysis in this research study did produce a factor structure
that contained a significant amount of cross loading of scale items between factors when
using the data from the full seventeen item survey instrument. A simple explanation
could be that the high correlation which exists between the three work engagement

136

factors may induce a significant amount overlap in a survey respondents interpretation of

the meaning of each individual question item. When responses to highly correlated
questions are combined with a small sample population, work context factors may exert
an influence on answers that show up as higher than desired levels of unexplained
variance. With small sample size data sets, a factor analysis solution may produce a
sample-specific recovered factor structure with scale items that exhibit a wide range of
communalities (0.2 to 0.8), and this small sample factor structure may not match theory
or the empirical factor structures recovered from much larger sample size data sets.
These issues may be mitigated by use of more compact measurement scales that have
questions with clearly distinct and focused meanings consistent with their assigned
theoretical factor. That appears to be one of the reasons why the 9-item shortened version
of the UWES was developed by Schaufeli (2006). Unfortunately, the more compact short
version survey instrument, with less scale items per factor, increases the probability that a
single-factor solution may be recovered for work engagements set of three still highly
correlated factors. A single strong common factor should still be a useful measure of
work engagement to an engineering manager.

2. Evaluation of Goal Theory Structural Equation Model Results

A research question that was to be answered by measurement of work motivation
variables and work engagement variables was do the results of the study provide
evidence for convergent validity between the three aspects of work motivation and the
three dimensions of work engagement? The acceptable goodness-of-fit index results
presented in Table 5.22 for the final structural equation goal theory research model shown

137

in Figure 5.17 support an affirmative answer to the question. A second follow-on

question was can work engagement provide managers with a useful criterion for
measuring the motivational aspect of employee work-related behavior toward
achievement of organizational goals? The standardized regression path weights shown
in Figure 5.17 and Table 5.23 for the causal paths in the final structural equation goal
theory research model provide a positive indication of the relative strength and practical
significance of each path. The antecedent latent variables and causal paths connecting
those variables in the structural equation model are analogous to the causal paths between
constructs in the established goal theory model of work motivation as proposed by Locke
and Latham (2002). The three factors of work engagement are substituted in the goal
theory model as analogues to the goal theory mediator variables of direction, effort and
persistence. The results of the structural equation analysis support the model as correctly
specified. The results of this study provided sufficient evidence to reject the null research
hypothesis and accept the alternate research hypothesis that the antecedent variables from
the goal theory of work motivation do validly predict the level of work engagement in an
engineering technical services organization. These conclusions are however based on a
single study and a single method. Additional supporting evidence is needed using multimethod multi-trait methods in follow-on studies to substantiate the premise that these first
results are not an artifact of a single study, a single method or a single version of latent
variables. The results do show that this research path is promising enough to warrant an
additional research study.

138

3. Evaluation of Job Characteristics Structural Model Results

Evaluation of the results obtained from the structural equation job characteristics
research models were not as conclusive as the results of the goal theory research model.
Why did the various structural equation models not support casual path relationships with
work engagement? The answer may be two-fold. First, since job characteristics theory
does not define its internal motivation construct with the same attributes as the work
motivation construct, work engagement may not be an entirely appropriate measure of
an individuals internal motivation. Second, job characteristics are identified in goal
theory as moderator variables, not as mediator variables. The distinction may be
important to understanding the structural equation model results. Mediator variable are
constructs created in order to explain the casual relationships between two psychological
constructs. Moderator variables act as attenuators of that causal relationship. A
moderator variable, i.e., a job characteristic, may affect the direction or strength of a
causal relationship, but it may not by itself be able to adequately explain or account for
the existence of that causal relationship. An alternative explanation might be that the job
characteristics factors are simply antecedents to the unspecified mediator variables. The
one model that did include meaningfulness as a second order mediator variable produced
the best goodness-of-fit measures. The interpretation of job characteristics factors solely
as moderators of work motivation may help to explain why the job characteristics factors
did not generate statistically significant causal path regression weights in the job
characteristics research models. The two job characteristic variables that were also
included in the goal theory model, feedback from job itself and feedback from agents,
were placed in the model as moderator variables that had a direct regression path only to

139

the three antecedent variables of work motivation. Their effect on the work engagement
criterion variables was indirect. Treating job characteristics as mediator variables
themselves in a structural equation model may be a misspecification that the full study
sample data set could not support as a valid explanation of how an individuals level of
work engagement is best predicted. This is only one of many other possible explanations
why the job characteristics research models were determined by analysis to have
inadequate fit. There may have been too much error in the measurements for the
structural equation modeling process to produce satisfactory results. The confirmatory
factor analyses did not always recover a theoretical factor structure. The present study
should not be construed as a strong rejection of characteristics theory. Further research
would be required to confirm or reject the job characteristics structural model.

4. Correlation versus Causation

An aim of this organizational behavior research study was to contribute to
development of a more coherent understanding of which workplace practices under the
control of management are best harnessed to influence employee performance in a way
that enhances employee well-being and achieves organizational goals. Organizational
goals are achieved through the motivated actions of its employees. Many managers
attempt to influence employee performance in a positive way, but without accurate
insight into what factors actually directly motivate employees to act, the manager can
unwittingly be meddling with unrelated factors, which although they appear to be
correlated to performance, have no direct causal influence on it. Management which is
focused solely on improving job satisfaction may possibly expend much effort and

140

resources and still not produce significant or sustainable improvements to organizational

performance. Managers that embrace goal theory and understand how it can be used to
create a high performance organization will have gone beyond correlation and used
causation to effect positive improvements in organizational performance. That is why
understanding the causal forces affecting performance within the organization is so
important to a manager, even though the task may be daunting. Coupled with the
deployment of goal theory in the management processes of an organization, a metric that
can provide a valid and accurate assessment of the level of an employees work
motivation is important. Work engagement may very well be that motivation metric, if
the results of this research study can be replicated and further validated in future studies.

B.

Recommendations
The results of this research study, while promising, can at best be considered

sufficient rationale for continued research into goal theory mechanisms which influence
work motivation, work engagement and its relationship to work motivation, and
employee well-being. There is a need to replicate the results in another sample
population. The combined survey instrument used in this research study could be
reworked and streamlined into a more integrated and coherent survey instrument. There
is a need to revise the survey instrument to be more integrated and reduce its length, and
in the revision process attempt to eliminate some of the redundancy and overlap between
questions.

141

1. Goal Setting Practices

The results of this research study appear to be consistent with the extensive body
of knowledge for goal theory. Goal setting best practices within the organization should
be given priority attention by all levels of management. A formal goal setting process can
be an effective component of any employee performance review process, and an integral
component of employee growth and development initiatives. Practical opportunities to
apply goal setting theory applied within the workplace can be found from the highest to
the lowest levels of the organization, including executive management strategic planning
activities, management by objectives programs, balanced scorecard initiatives, to lean six
sigma projects undertaken by green belt certified employees working at the front lines of
organizational processes. The measurement of work, motivation, whether at the
executive level or the assembly line level, makes goal theory a prime candidate for
evaluating the merits of work engagement as a useful indicator of employee well-being.
Latham (2007) states that employee development programs can be implemented
which increase the level of employee self-efficacy. People with high self-efficacy set
higher goals, which are associated with higher performance. They are also better
equipped to respond to negative feedback by putting out more effort in a positive manner.
Ways to raise the level of employee self-efficacy include ensuring adequate training to
increase mastery and success experiences, coupled with persuasive communications that
express confidence employees can attain goals. Most important, leadership is needed at
all levels in an organization to inspire and cognitively stimulate employee behavior to act
in a manner congruent with organizational goals and business strategies (Latham 2007).
2. Areas for Further Study

142

A further research study should continue the construct validity process. A

discriminant analysis between work engagement and other motivation constructs could
incorporate multi-trait methods to evaluate construct validity. Another area for further
study could be discriminant analysis between work engagement and other well-being
constructs. The continued use of the full version work engagement survey instrument is
recommended to build a larger database of survey records to further test the full survey
instruments factor structure and instrument validity in other study populations
Another area for further research is to develop a compact goal theory web survey
instrument and continue testing goal theory variables as predictors of work engagement
in order to replicate results of this research. Adding the measurement of goal task
strategies to a follow-on study would incorporate the fourth mediating mechanism of
Locke and Lathams goal theory of work motivation.
Job characteristics should not be rejected outright as a predictor of work
engagement. Re-examination of the job characteristics research model assumptions is
necessary, considering other latent variables. Development of a revised and more
compact job characteristics web survey instrument is a potential follow on task, to
provide clarification of the mixed results produced in this research.

143

APPENDICES

Appendix A

INSTITUTIONAL REVIEW BOARD FORMS

143

144

145

Appendix B
INDIVIDUAL SURVEY INSTRUMENTS

Utrecht Work Engagement Survey

The 17 item Utrecht Work Engagement Scale (UWES), a work and well-being
self-report questionnaire measures three dimensions of well-being that a worker feels
while at work (Schaufeli et al. 2002). More recently, a 9-item short version of the survey
has been validated by the developers of the UWES. The UWES 17-item scale is shown
in Figure B.1. Questions marked with an SV comprise the shortened version of the
UWES (Schaufeli et al. 2006).

Figure B.1 UWES Work Engagement Survey Instrument (Schaufeli et al. 2006)
146

Job Goal Specificity and Job Goal Difficulty Attributes

The measurement scales for goal specificity and goal difficulty were used to
capture proximal assigned work goal attributes. Goal attributes are important antecedents
of work motivation. These variables were operationalized in two scales shown in
Figure B.2. The scales were developed and validated in a public sector organization
setting (Wright 2004). Two different Likert-type response scales were used by Wright, a
1 - 6 agreement scale, and a 0 - 4 frequency scale. In order to properly interpret the
results from these two measurement scales, the raw response scores have to be
standardized to values between 0 and 1 to account for the mixed use of two different
Likert response scale formats within each latent variable measurement scale.

Figure B.2 Job Goal Attributes Measurement Scales (Wright, Bradley E. 2004)
147

Occupational Self-Efficacy
Birgit Schyns and Gernot von Collani (2002) developed a survey instrument that
is intended to be a reliable one-dimensional measure of the psychological construct of
occupational self-efficacy. Occupational self-efficacy is considered to be a stable longterm measure of an individuals level of work-related self-efficacy (Schyns et al. 2002).
The developers of the occupational self-efficacy construct have published the results of
three studies that assessed the reliability and validity of the instrument as a self-efficacy
construct for organizational behavior research (Schyns and Collani 2002). The results of
the three studies validated the survey instrument and indicated it was sufficiently reliable
to be used in further research studies that seek to evaluate self-efficacy in its generalized
work-related form. A six point Likert type response scale was used for all six positively
scored scale items. The scale has been validated in five countries (Rigotti et al. 2008).
The six item version of the measurement scale used in this research is shown in
Figure B.3.

Figure B.3 Occupational Self-efficacy Measurement Scale

Achievement Goal Orientation in a Work Domain
148

The 2x2 framework for achievement goal orientation in a work domain survey
instrument measured how individuals approach and react to achievement goals (Baranik
et al. 2007). The achievement goal orientation variables were used to measure
differences in dispositional trait-like characteristics of individuals that are theorized to be
predictors of an employees goal choice. A seven point Likert type scale was used in the
instrument. The goal orientation measurement scales are shown in Figure B.4.

Figure B.4 Achievement Goal Orientation Measurement Scale

The Job Diagnostic Survey

149

A well-documented job characteristics survey instrument was identified in the

motivation research literature, the Job Diagnostic Survey developed by Hackman and
Oldham (1975). The revised job diagnostic survey was developed ten years later in
response to factor purity problems identified with reverse score scale items (Idaszak and
Drasgow 1987). The reverse score items were reworded so that all scale items are
positively scored. The revised job diagnostic survey continues to be utilized, with the
most recent study found in the literature confirming its factor validity (Buys et al. 2007).
The revised version of the Job Diagnostic Survey was selected for incorporation into this
research study. As shown in Figure B.5, there are five factors in the JDS with three scale
items each. A seven point Likert type response scale was used. Two additional
measurement scales in the original version of the JDS were added to this study because
they were considered related factors of interest. As shown in Figure B.6, Feedback from
Agents attempts to measure feedback from coworkers and supervisors, and Dealing
with Others attempts to measure the interconnectedness of a particular job with other
jobs in the organization (Hackman and Oldham 1975). The scale item wordings of these
two related factors appeared to have face validity and provide a more complete
understanding of employee perceptions of feedback on the job.

150

Survey section

Figure B.5 The Revised Job Diagnostic Survey

151

Survey section

Figure B.6 Two Related JDS Job Characteristics Measurement Scales

152

Appendix C

COMBINED WEB SURVEY INSTRUMENT

The text of an email is presented below that was sent to all employees of the full
study engineering services company with a listed email address, requesting their

participation in the survey.

Dear ERC employee, I am a team leader of an ERC test support group within the US
Army Redstone Technical Test Center located at Redstone Arsenal in Alabama. I need
your help with my final University of Alabama in Huntsville graduate school research
project. Could you please click on the secure web link listed below and complete a
research survey. I have received the permission and support of ERC management to
administer this survey to everyone within ERC. Your effort is greatly appreciated!

https://www.surveymonkey.com/s.aspx?sm=yVprghiwaftszRFuK_2fyUCQ_3d_3d

Sincerely,
Mark Antonison

Table C.1 Test Instrument Scale Items vs. Position in Web Survey Instrument
153

Table C.1 (Continued) Test Instrument Scale Items vs. Position in Web Survey Instrument
154

155

156

Appendix D
DESCRIPTIVE STATISTICS MEASUREMENT VARIABLES

Boxplot of work engagement measurement variables pilot study n = 60

6
5

Data

4
3
2
1
0
vi1 vi2 vi3 vi4 vi5 vi6 de1 de2 de3 de4 de5 ab1 ab2 ab3 ab4 ab5 ab6

Figure D.1 Box Plot Work Engagement Measurement Variables Pilot Study Results

Boxplot of work engagement measurement variables full study n = 360

6
5

Data

4
3
2
1
0
vi1 vi2 vi3 vi4 vi5 vi6 de1 de2 de3 de4 de5 ab1 ab2 ab3 ab4 ab5 ab6

Figure D.2 Box Plot Work Engagement Measurement Variables Full Study Results

157

Boxplot of job characteristics measurement variables pilot study n = 60

7
6

Data

5
4
3
2
1

Figure D.3 Box Plot Job Characteristics Measurement Variables Pilot Study Results

Boxplot of job characterisitics measurement variables full study n = 360

7
6

Data

5
4
3
2
1

Figure D.4 Box Plot Job Characteristics Measurement Variables Full Study Results

158

Boxplot of occupational self-efficacy measurement variables pilot study n = 60

6

Data

1
occseef1

occseef2

occseef3

occseef4

occseef5

occseef6

Figure D.5 Box Plot Self-efficacy Measurement Variables Pilot Study Results

Boxplot of occupational self-efficacy measurement variables full study n = 360

6

Data

1
occseef1

occseef2

occseef3

occseef4

occseef5

occseef6

Figure D.6 Box Plot Self-efficacy Measurement Variables Full Study Results

159

Boxplot of goal orientation measurement variables pilot study n = 60

7
6

Data

5
4
3
2
1

Figure D.7 Box Plot Goal Orientation Measurement Variables Pilot Study Results

Boxplot of goal orientation measurement variables full study n = 360

7
6

Data

5
4
3
2
1

Figure D.8 Box Plot Goal Orientation Measurement Variables Full Study Results

160

Boxplot of std job goal specificity and std job goal difficulty full study n = 360
1.0

Data

0.8
0.6
0.4
0.2
0.0

Figure D.9 Box Plot Goal Attributes Measurement Variables Full Study Results

161

Table D.1 Descriptive Statistics 71 Scale Items Measured in Survey Instrument

162

Table D.1 Descriptive Statistics 71 Scale Items Measured in Survey Instrument

(continued)

163

Appendix E
SAMPLE SIZE ANALYSIS

A variety of recommendations have been presented in the psychometric testing

literature concerning minimum required sample size, usually focused on the ratio of
number of survey responses to number of variables (scale items, or survey questions).
Conflicting recommendations from a minimum ratio of 5-to-1 to as much as 20-to-1
responses per variable have been proposed. A more accurate assessment of a minimum
sample size is related to the number of variables, number of factors, number of variables
per factor (over-determination of the factor), and the size of the factor solutions
communalities (MacCallum et al. 1999). Estimation of the communality of a variable is
important in factor analysis because the communality of a variable is the portion of the
variance of that variable that is accounted for by the common factors (MacCallum et al. 1999).
A theoretical treatment of the sample size issue in recovery of factor structures
from a sample data set is presented in the psychometric testing literature which defines a
measure, K, the coefficient of congruence that attempts to quantify the level of agreement
between a sample covariance matrix and a true population covariance matrix (Tucker et
al. 1969). Tuckers recommendations state that a value of K = 1 indicates perfect
agreement, values of K = 0.98 to 1 are considered excellent, and values of K = 0.92 to
0.98 are good. Values less than 0.92 are considered borderline. The results of Monte
Carlo simulations used to determine adequate minimum sample sizes for a given K value
have been published (Mundfrom et al. 2005). The two studies that have been cited both
conclude that the actual number of extracted factors F is less important than the ratio p/f
164

(the number of variables or scale items, p to the number of theoretical factors, f) and the
range of communalities for a specified value of K. The analysis results published by
Mundfrom et al. (2005) include results for number of extracted factors F = 3 to 6 in
one of their tables of recommendations. The two studies provide similar conclusions for
the minimum required sample size, n to a given cutoff value of K, the p/f ratio, and the
range of the pattern of communalities obtained from a confirmatory factor analysis.
Values for the patterns of sample factor solution communalities are grouped into three
range categories: high = 0.6 to 0.8, wide = 0.2 to 0.8, and low = 0.2 to 0.4, with higher
communalities allowing smaller minimum sample sizes and lower communalities
requiring larger minimum sample sizes. The factor solution extracted from the combined
survey instrument included p = 71 variables (71 total scale items) and f = 17 theoretical
factors, for a ratio of p/f = 4.2. As shown in Figure E.1, factor analysis produced a
wide range for communalities in 16 factors extracted with eigenvalues greater than 1.
The communalities extracted from the goal theory structural model with p= 30 variables
(30 total scale items) and f = 6, for a ratio of p/f = 5, were also a wide range of values
as shown in Figure E.2.

165

Figure E.1 Communalities for Combined Web Survey Instrument

Figure E.2 Communalities for Goal Theory Research Model Variables

Direct comparison of the full survey instrument recovered factor structure, with
F = 16 extracted factors, to published recommendations is not possible since the authors
only explicitly published results for Monte Carlo solutions with the number of extracted
factors, F up to eight. The study recommended incorporating smaller numbers of
theoretical factors (i.e., f = 6 to 8 maximum) in any given study so as to recover more
stable factor solutions and avoid non-convergence issues in confirmatory factor solutions.
166

A table summarizing the recommendations is presented in Table E.1 (MacCallum et al. 1999)
(Mundfrom et al. 2005).

Table E.1 Summary of Minimum Sample Size Recommendations from Literature

A summary of minimum sample size analysis parameters for the web survey
instrument and the goal theory research model is presented in Table E.2.

Table E.2 Summary of Statistics for Minimum Sample Size for Research Study

167

When comparison is made between the published recommendations and the web
survey instrument which has a p/f ratio = 4.2, the full study sample size of n=360 appears
to meet the recommended minimum sample size requirements for a K value of 0.92. This
implies there should be good agreement between the sample recovered factor structure
and the theoretical factor structure.
The goal theory structural model factor solution with p = 30 variables and
f = 6 theoretical factors can be more directly compared with the published
recommendations with its p/f ratio = 5 and number of extracted factors F = 6.
The full study sample data set, n = 360, appears to meet all recommendations for
minimum sample size at a value of K = 0.98, which indicates there should be excellent
agreement between the goal theory research model recovered factor structure and
theoretical factor structure.

168

Appendix F
REGRESSION ANALYSIS MPS VS. WORK ENGAGEMENT

The first two regression analyses evaluated the pilot study sample data set. The
first analysis compared the simple summation scores from the 17-item work engagement
survey calculated for each engineering graduate student with their respective MPS score
calculated from their responses to the of five core job characteristics latent variables: skill
variety, task significance, task identity, autonomy, and feedback from job itself. The
second regression analysis substituted the MPS score with a simple summation of the
response scores for the scale items that measure the five core job characteristics latent
variables in the MPS. The results are presented in Figures F.1 through F.4.

Figure F.1 Work Engagement Score vs. Motivating Potential Score Pilot Study

169

Figure F.2 Results of Regression Analysis Work Engagement Score vs. Motivating
Potential Score Pilot Study

Figure F.3 Work Engagement Score vs. Sum Job Characteristics Score Pilot Study

170

Figure F.4 Results of Regression Analysis Work Engagement Score vs. Sum Job
Characteristics Score Pilot Study
The results presented in Figures F.1 through F.4 from the pilot study sample data
set looked very promising for consideration of either an MPS score or a summation score
of job characteristics to be used by a manager as simple predictor variables for work
engagement. The adjusted R-sq(adj) values generated in the analysis explain nearly
50 percent of the variance in the pilot study data set. There were no observed severe
outliers in the data. One note of caution must be made: the pilot study data set was
obtained from a sample that does not actually represent any single distinct organizational
context. The pilot study data is a composite of many different organizations and work
contexts. These results should not be used to draw a conclusion about workplace
practices or job characteristics in any specific organizational context. They are more
likely to be generalized in nature with respect to the engineering occupation.
The next two regression analyses evaluated results from the full study sample data
set. The analyses compared the simple summation scores from the 17-item work
171

engagement survey calculated for each employee of the engineering services company
with their respective MPS score and with their simple summation of the response scores
for the scale items that measure the five core job characteristics latent variables in the
MPS. The results are presented in Figures F.5 through F.8.

Figure F.5 Work Engagement Score vs. Motivating Potential Score Full Study

172

Figure F.6 Results of Regression Analysis Work Engagement Score vs. Motivating
Potential Score Full Study

Figure F.7 Work Engagement Score vs. Sum Job Characteristics Score Full Study

173

Figure F.8 Results of Regression Analysis Work Engagement Score vs. Sum Job
Characteristics Score Full Study

The regression analysis results shown in Figures F.5 through F.8 for the full study
sample data set were less promising primary due to the presence of several data points
with high standardized residuals in the full study sample data set. Residual plots and a
normal distribution probability plot were generated to better understand the full study
regression analysis results. These plots are presented in Figures F.8 through F.11.

174

Residual Plots for 360sum uwes17 vs 360 MPS

Normal Probability Plot of the Residuals

Residuals Versus the Fitted Values

60

99.9

40

90

Residual

Percent

99

50
10

20
0
-20

1
0.1

-40

-20

0
20
Residual

40

40

Histogram of the Residuals

50

60
70
Fitted Value

80

Residuals Versus the Order of the Data

60
40
Residual

Frequency

60
45
30
15
0

20
0
-20

-30

-15

15
Residual

30

45

50

100 150 200 250

Observation Order

300

350

Figure F.9 Four-in-one Residuals Plot for MPS Regression Analysis Full Study
Residual Plots for 360sum uwes17 vs 360 sum rjds
Normal Probability Plot of the Residuals

Residuals Versus the Fitted Values

99

40

90

20

Residual

Percent

99.9

50
10

0
-20

-40

0.1

-40

-20

0
Residual

20

40

50

Histogram of the Residuals

80

90

Residuals Versus the Order of the Data

45

Residual

Frequency

70
Fitted Value

40

60

30

20
0
-20

15
0

60

-30

-20

-10

0
10
20
Residual

30

-40

40

50

100 150 200 250

Observation Order

300

350

Figure F.10 Four-in-one Residuals Plot for Sum Job Characteristics Score Regression
Analysis Full Study

175

Evaluation of the residuals plots indicates problems with the data sets violating
the normally distributed assumption of linear regression analysis and there also appears to
be several data points causing problems with the assumption that the variance of the
analysis residuals should be constant. Severe violations of these two assumptions by a
sample population data set would require additional analysis and data set transformations
that were not performed because of their limited impact on development of answers to the
research questions of this study. The Anderson Darling test results shown in the
probability plot in Figure F.11 for indicates the MPS score has severe non-normality in
the full study sample data set. This result may be partially due to the multiplicative
nature of the MPS score formula. The simple summation of job characteristics scores
variable may therefore be a better choice if a job characteristics scores data set is to be
evaluated using linear regression methods.

Figure F.11 Probability Plot for Regression Analysis Variables Full Study
Appendix G
176

STRUCTURAL EQUATION MODELING ANALYSIS

Structural equation modeling was performed using AMOS 17.0 software

(Arbuckle 1995). The modeling process started with development of measurement
models for each theoretical latent variable. The measurement models represent the
measurement scales contained in the survey instrument. The measurement models are
combined into structural models that specify the causal relationships between the latent
variables. The structural equation modeling process followed a four step process,
described in Table G.1.

Table G.1 Structural Equation Modeling Process Steps

The latent variable measurement scales were first modeled and evaluated. The
next step was to develop first and second order structural models of individual survey
instruments, as described by the theory supporting each instrument. The third step was to
develop structural models of the proposed goal theory research model. The fourth step
177

was to develop structural models of the job characteristics research model. Several
alternate configurations of each research model were constructed and evaluated.

Checklist Procedure Used to Evaluate SEM Solution Output Files

The following set of recommended SEM analysis checks was obtained from
Byrne (2001) and used throughout the SEM analysis process to evaluate the results in the
solution output files for each analysis run generated using the AMOS software.

1) Prior to analysis run, a check for model over-identification was performed to

ensure a unique converged solution is possible. If a model is under-identified,
meaning it has fewer degrees of freedom than the number of parameters to be
estimated in the analysis, the analysis run will be terminated without a unique
converged solution.
2) Analysis of the structural model should have produced a converged solution with
a positive definite covariance matrix for the sample data set, and there should not
be any correlation terms greater than one and no negative variances in the solution
output, which are problems that result from model misspecification issues or
anomalies embedded within the sample data set itself.
3) A check for statistical significance of all the estimated parameters including
variances and regression weights was performed. Critical ratios for each variable
can be interpreted as measures of statistical significance along with associated
probabilities. A P-value less than 0.05 was chosen as the cutoff level for
statistical significance of the results.
178

4) The algebraic signs of regression paths were reviewed to confirm they matched
their respective theoretical assumptions.
5) Evaluation of the size of standard errors was required, because very large or very
small standard errors are indications of problems with model specification or
significant measurement error in the sample data set.
6) Evaluation of the size of the terms in the standardized residuals covariance matrix
was important in order to properly assess the goodness-of-fit between the sample
data covariance matrix and the theoretical covariance matrix specified by the
model. The standardized residual covariance terms which are calculated for every
pair of scale items can be considered equivalent to Z- scores; acceptable values
should be less than +/-2.58 for an alpha criterion level of = 0.10.
7) Assessment of non-normality was checked. Severe non-normality violates the
assumptions of the maximum likelihood analysis method and may affect solution
results with poorly estimated regression coefficients, standard errors, and nonnormality will also affect estimated goodness-of-fit values.

8) Observations farthest from the solution centroid (Mahalanobis distance) were also
considered. Each row of data in the input file represents a set of answers to the
survey questions from an individual survey respondent, defined as an observation.
This check is equivalent to looking for extreme outliers in a response data set that
may exert an excessive influence on the covariance matrix solution because they
are not representative of the influence that the rest of the sample population has
on estimation of the solution covariance matrix.
9) The last step in evaluation of the solution results was to review the various
goodness-of-fit measures in the solution output file (Byrne 2001). All discussions
179

of goodness of fit variables and application of cutoff criteria followed the

guidance provided in the SEM literature by Byrne (2001), Hu and Bentler (1999),
and Marsh et al. (2004).
SEM Confirmatory Factor Analysis of Latent Variable Measurement Models

Measurement models were developed in AMOS for the seventeen latent variables
in this research study. The measurement models for Vigor and Autonomy latent variables
are provided in Figure G.1 as examples because their structure is typical for all the other
latent variable measurement models. Each scale item or question is an observation or
measured variable and is also called an indicator (i.e., vi1 through vi6 are questions 1
through 6 in the vigor scale of the work engagement survey instrument). A measurement
error term is included for each scale item measurement variable, with path regression
coefficients set to 1 because the path terms for measurement error variables are not
estimated. The scale item indicators are measuring aspects of a theoretical latent variable
that represents a psychological construct. The path arrows point from the latent variable
toward each scale item. This is a reflective modeling approach, meaning that each scale
item indicator variable is a reflection of its underlying latent variable.

180

err6
err5
err4
err3
err2
err1

1
1
1
1
1
1

Measurement error terms

Regression
coefficient
err1
err2
err3
path terms

vi6
vi5
vi4
vi3

VIGOR

au1

au2

au3

vi2

AUTONOMY

vi1

Scale items the observed indicator

Latent Variable underlying
measurement variables
Theoretical psychological construct
Figure G.1 Typical Structural Equation Measurement Models for Latent Variables
There is one noticeable difference between the example measurement models for
vigor and autonomy. Measurement scales with only three scale items have zero degrees
of freedom and will not meet the requirements for an over-identified SEM model which
AMOS needs to calculate a unique default model solution. Model identification is
achieved by having more measurement data points than variables to be estimated.
To enable at least one degree of freedom for a unique solution in 3-scale item
measurement models, the latent variable has its factor variance set to 1 as shown in the
measurement model for autonomy. The evaluation of model identification has to be
performed during development of each structural equation model.
A summary of the results of this first phase of structural equation modeling
process is presented in Table G.2. The minimum recommended set of model fit statistics
has been presented for the measurement models, the minimized Chi-squared statistic for a

181

converged solution CMIN, model degrees of freedom DOF, the comparative fit index
CFI, and the Root Mean Square Error of Approximation RMSEA.

Table G.2 Goodness-of-Fit Results for Structural Equation Measurement Models

The results of the SEM analysis are consistent overall with the findings of the
earlier factor analysis. The vigor measurement scale has a high RMSEA most likely due
to the sample data sets recovered two factor solution identified in the earlier factor
analysis. A two factor measurement model was tested and it produced a RMSEA value of
0.05. A number of other measurement models also produced large RMSEA values.
Evaluation of regression path coefficients, standard errors, non-normality measures, and
standardized residuals covariance matrix terms did not yield any obvious problems that
182

could be addressed, so the SEM analysis process was continued to the next step.
Modification indices included in the results suggested the addition of several correlation
paths between individual measurement scale item error terms to improve RMSEA values,
but these paths were not added to the models at this stage in the confirmatory factor
analysis process.

First and Second Order Structural Equation Models of Survey Instruments

The next step in the structural equation modeling process was to combine the
individual measurement scale models and into first order models and second order
representing the three survey instruments of work engagement, job characteristics, and
goal orientation. The analysis results of this research study could be directly compared to
results of prior studies of the same instruments that were found published in the literature.
The aim of these comparative studies is to assess the similarities and differences if any
between the earlier studies and this current research effort
SEM Analysis of the Utrecht Work Engagement Survey
The Utrecht Work Engagement Survey instrument was evaluated first. Three
models were developed and tested for the full 17-item survey instrument. The first work
engagement structural equation model is a one factor model with the path arrows for all
17 scale items loading onto the single latent variable representing work engagement. The
graphical representation of the model is shown in Figure G.2. The second work
engagement structural equation model is a first order three factor model where path
arrows for the respective measurement scale items of the three separate factors load on
183

the three latent variables of vigor, dedication and absorption. Work engagement is
evaluated by incorporating into the first order model correlation paths between each
latent variable to estimate their relationships to each other. This model is presented in
Figure G.3. The third work engagement structural equation model is a second order
three factor model that explicitly modeled work engagement as a second order latent
variable. Regression paths point from work engagement to the three first order latent
variables of vigor, dedication, and absorption, because all three factors of work
engagement are specified as reflective measures of the work engagement construct. In
the second order model explicit factor variance terms are modeled for each of the three
first order latent variables and represent the unknown residual variance or disturbance in
the sample data set that is evaluated with the model. The variance for work engagement
is assumed to be fully accounted for by the three first order factor variance terms so its
own variance term is set to one and not estimated. This model is shown in Figure G.4.
Before the three factor second order model could be evaluated a check for
identification at the second order level revealed a just identified model. The three data points
are now the three first order factors so p = 3 and there are (3(3+1)/2 = 6 data points. The model
as first set up would require calculation of estimate three residual variances plus three
regression coefficients for the three second order path loading terms. To achieve an overidentified model at the second order level, a reduction in the number of estimated parameters of
at last one is required. One technique to achieve over-identification is to evaluate whether an
assumption of equal variances for first order factor terms is acceptable, and if so assign these
the variables the same residual variance term. Levenes test for equality of two variances was
performed in MINITAB and the results are shown in Table G.3 below.
184

Table G.3 Test for Equal Variances Between Work Engagement Factors

The results of Levenes test show the hypothesis that the variances of vigor and
absorption are equal could not be rejected, and the second order structural equation model
was therefore modified to have the variances for the vigor and absorption factors set
equal to the parameter var_a as shown in the SEM model. This provided the necessary
over-identification of the three factor second order structural model for work
engagement.
The models were evaluated using the multi-group method with n1 = 360 from the
full study engineering technical services company and n2 = 60 from the pilot study
graduate engineering students with jobs. The results of the structural equation model
evaluation for the full 17-item work engagement survey instrument are presented in
Table G.4, including comparisons with results from published studies. The fit statistics
for the one factor model are less acceptable than the fit statistics for the three factor
models, implying that the three first order latent variables may indeed constitute three
distinct factors supporting discriminant validity of the three factor theoretical model of
work engagement. The two three-factor models that were developed do not, however,
produce high enough goodness-of-fit statistics by themselves to support any strong
conclusions about the construct validity of work engagement as a three factor construct.
185

This weak conclusion about three factor construct validity is consistent with the mixed
results presented earlier in common factor analysis of the UWES17-item survey
instrument with both the combined full and pilot study sample data set, n1 = 360 and
n2 = 60. While comparison of published SEM results with this current research study
shows relatively close agreement between model fit statistics of similarly specified
structural equation models (Schaufeli et al. 2001), the results of the earlier common
factor analysis and the structural equation models presented in this analysis appear to
provide mixed results for a strict assessment of factor purity with the sample data set
obtained using of the UWES17-item survey instrument.

186

Figure G.2 Single-Factor UWES 17-item Model of Work Engagement

187

errvi1

errvi2

errvi3

vi1

errvi4

vi2

errvi5

vi3

errvi6

vi4

vi5

vi6

vigor

errde1

errde2

errde3

de1

errde4

de2

errde5

de3

de4

de5

dedication

errab1

errab2

1
ab1

1
ab2

errab3

errab4

ab3

ab4

errab5

errab6

1
ab5

1
ab6

absorption

Figure G.3 First Order Three-Factor UWES 17-item Model of Work Engagement

188

Figure G.4 Second Order Three-Factor UWES 17-item Model of Work Engagement

189

Table G.4 SEM Analysis Comparative Results for the Full 17-item UWES
Work Engagement Survey Instrument

The published version of the 9-item short version survey instrument for work
engagement was evaluated next (Schaufeli et al. 2006). Three structural models were
again developed and tested: a one factor first order model shown in Figure G.5, a
three factor first order model shown in Figure G.6, and a three factor second order model
shown in Figure G.7. The multi-group sample method was also used again. The
comparative goodness of fit results for the three models is shown in Table F.5. A review
of all final solution statistics found only one problem in the second order model.

190

Regression paths in all models were all significant and positive. Standard errors
were also reasonable in size with no extremely large or small values in all models. All
estimates of regression coefficients and variances were statistically significant (p<0.001),
except for the estimated variance of the residual disturbance term of the latent variable
dedication in the second order model, which was estimated to be non-significant with
p = 0.322 for group 1, n = 60, and p = 0.475 for group 2, n = 360. The standardized
residual covariance matrix terms in all models were within acceptable limits of +/- 2.58.
The fit indexes for the 9-item survey instrument appear to support the three factor
model of work engagement as a better fit to the sample data set than the one-factor model
of work engagement. This study replicates most of the results of the much larger size
study of Schaufeli et al. (2006), and the larger RMSEA value produced in this study
relative to the RMSEA value in their 2006 published SEM results may be attributed in
part to this studys smaller sample size. It is also interesting to note the similarities in
results appear fairly invariant between the different countries of the United States, Europe
and Japan for the one factor model, but the Japanese study (Shimazu et al. 2008) did not
show the same level of similarity with its three factor model solution.

191

Figure G.5 One-Factor UWES 9-item Model of Work Engagement

192

Figure G.6 First Order Three-Factor UWES 9-item Model of Work Engagement

193

Figure G.7 Second Order Three-Factor 9-item UWES Model of Work Engagement

194

Table G.5 SEM Analysis Results for the Short Version 9-item UWES
Work Engagement Survey Instrument

SEM Analysis of the Revised Job Diagnostic Survey

The second survey instrument to be evaluated with a structural equation model
was the revised job diagnostic survey which is comprised of five job characteristic factors
with each factor operationalized by its own 3-item measurement scale. The first model to
be evaluated was a one-dimensional single factor job characteristics model with all
15 scale items, shown in Figure G.8. The goodness-of-fit indexes were very low
indicating a very poor fit with the sample data set. This result supports discriminant
validity of the five job characteristics latent variables.

195

The second model was a five factor first order model with correlation paths between
all five factors, shown in Figure G.9. The fit of this model is significantly better than the one
factor model. These results also support the discriminant validity of the five job
characteristics factors. The five factor first order model was enhanced to include an
additional construct, the critical psychological state of experienced meaningfulness, inserted
into the model as second order factor related to the three job characteristics of task identity
skill variety and task significance. Correlation paths arrows were then added between
Meaningfulness, Autonomy and Feedback. This second order model is based on the job
characteristics model of Hackman and Oldham (1975), and it is shown in Figure G.10. The
fit indexes for this second order model results were nearly identical to the first order model.
These SEM results were compared with two other sets of published results for the revised job
diagnostic survey and there is close agreement with both earlier studies. The fit indexes meet
the guidelines for a moderately acceptable fit (CFI > 0.90 and RMSEA < 0.08) but fail to
meet the more stringent cutoff values for a very good model fit (CFI > 0.95 and RMSEA <
0.05). The standardized residual covariance matrix for both the first and second order model
results contained numerous terms with values greater than 2.58, considered an upper limit for
an adequate fit model. The standard errors for variances of all the latent variables were large
relative to the size of the variances themselves (25% to 35%). The standard errors for the
regression weights for each scale item to its latent variable were also high. The variance for
scale item skill variety item 3 was non-significant (p =0.489). These high standard errors are
an indication of model misspecification, consistent with the mixed results obtained in the
confirmatory factor analysis solutions.

196

As a final step in job characteristics SEM model evaluation, a seven factor first
order model was developed by adding the two related job characteristics of feedback
from agents and dealing with others that were associated with the original Job Diagnostic
Survey instrument. The final job characteristics structural equation model is presented in
Figure G.11.

1
au1
au2

1
1

au3

errau1
errau2
errau3

1
sv1
sv2

1
1

sv3

errsv1
errsv2
errsv3

1
ts1

job
characteristics

ts2

errts1

errts2

ts3

errts3

1
ti1

ti2

ti3

errti1
errti2
errti3

1
fbji1
fbji2
fbji3

errfbji1
errfbji2

errfji3

Figure G.8 One Factor Revised Job Diagnostic Survey Job Characteristics Model

197

errau1

au1

errau2

errau3

au2

au3

autonomy

errsv1

sv1

errsv2

errsv3

sv2

sv3

skill variety

errts1

ts1

errts2

errts3

ts2

ts3

task
significance

errti1

ti1

errti2

errti3

ti2

ti3

task identity

errfbji1

fbji1

errfbji2

errfji3

fbji2

fbji3

feedback
job itself

Figure G.9 First Order Five-Factor Revised JDS Job Characteristics Model

198

e10

e11

e12

au2

au3

au1
1

autonomy
er1

er2

er3

sv1

sv2

sv3

1
1

res1

skill variety

er4

er5

er6

ts2

ts3

ts1
1

res2

task
significance

meaningfulness

er7

er8

er9

ti1

ti2

ti3

res3

task identity
e13

e14

e15

fbji2

fbji3

fbji1
1

feedback
job itself

Figure G.10 Second Order Five Factor Revised JDS Job Characteristics Model

199

err1

err2

fbfa1

err3

fbfa2r

fbfa3

1
feedback agents

errau1

errau2

errau3

au2

au3

au1

autonomy

errsv1

sv1

errsv2

errsv3

sv2

sv3

skill variety

errts1

ts1

errts2

errts3

ts2

ts3

task significance

errti1

ti1

errti2

errti3

ti2

ti3

task identity

errfbji1

fbji1

errfbji2

errfji3

fbji2

fbji3

feedback job itself

err5

err4

dwo2

dwo1

err6

1
dwo3r

1
dwo

Figure G.11 First Order Seven Factor Job Characteristics Model

200

The two additional factors did not improve the model fit indexes, and the results
of this model contained all the problems listed above to a greater degree due to larger
model size. It appears that this seven factor model had too many large standardized
residual covariance terms to meet guidelines for good model fit. The specification of the
model may be acceptable but the sample data set contains a significant amount of
correlated error with respect to the twenty one job characteristics measurement variables
included in this study. Table G.6 summarizes the SEM analysis results for the revised Job
Diagnostic Survey Instrument and comparisons to published data from earlier studies.

Table G.6 SEM Analysis Comparative Results for the

Revised Job Diagnostic Survey

SEM Analysis of the Achievement Goal Orientation Survey

201

The third survey instrument that was incorporated into this research study was the
four factor achievement goal orientation in a work domain survey instrument. A
one factor goal orientation model was first developed with all eighteen scale items, but a
converged solution was not achievable due to a sample covariance matrix that was not
positive definite. This is evidence of discriminant validity between the four goal
orientation constructs. A four factor first order model was developed next and a
converged solution was achieved and the results are presented below. This model is
shown in Figure G.12. A four factor second order model was developed next and a
converged solution was achieved but the solution generated a negative variance for the
residual term for performance avoidance latent variable. This model is shown in
Figure G.13. Two additional second order models were attempted that are not shown. A
model with second order performance and mastery latent variables and then a model with
second order approach and avoidance latent variables both failed to achieve converged
solutions due to negative variances for the residual term for performance avoidance and
performance approach latent variables. These problems were not unexpected due to the
correlation analysis results for these constructs presented earlier. The second order
structural equation models of the goal orientation constructs based on the higher level
second order goal orientation, approach-avoidance and mastery-performance concepts
generated unsatisfactory and invalid results due to the unclear set of correlations within
the sample data set between goal orientation factors themselves at the higher order level
of analysis. A comparison of this studys valid four factor first order model results to the
earlier published results for this survey instrument is included in Table G.7. At this point
202

in the analysis process, a decision was made to discontinue further use of the goal
orientation sample data due to the analytical problems encountered in the results
documented in the correlation analysis and the structural equation models.

errmav1

errmav2

errmav3

errmav4

errmav5

errmav6

mav1

mav2

mav3

mav4

mav5

mav6

mastery
avoidance
errmap1

errmap2

errmap3

errmap4

map1

map2

map3

map4

mastery
approach

errpav1

errpav2

errpav3

errp av4

p av1

pav2

pav3

pav4

performance
avoidance

errpap1

errpap2

errpap3

errpap4

pap1

pap2

pap3

pap4

performance
approach

Figure G.12 First Order Four-Factor Goal Orientation Model

203

errmav1

errmav2

errmav3

errmav4

errmav5

errmav6

mav1

mav2

mav3

mav4

mav5

mav6

1
var1

mastery
avoidance

dist_mav

errmap1

errmap2

errmap3

errmap4

map1

map2

map3

map4

1
var1

mastery
approach

1
dist_map

goal orientation

errpav1

errpav2

errpav3

errpav4

pav1

pav2

pav3

pav4

1
var1

performance
avoidance

dist_pav

errpap1

errpap2

errpap3

errpap4

pap1

pap2

pap3

pap4

1
var1
dist_pap

performance
approach

Figure G.13 Second Order Four Factor Goal Orientation Model (invalid solution)

204

Table G.7 SEM Analysis Results for the Achievement Goal Orientation Survey

At this point in the structural equation modeling process, confirmatory factor

analysis in the strictest definition of its application was finished. The models that follow
are combinations of portions of operationalized measurement scales developed for
different theoretical research models. Prior to this study, the measurement scales of the
latent psychological constructs included in this research had not been operationalized in a
survey instrument which could empirically test the measurement scales together. Based
on the actual research processes that have been described by the original developers of
the several survey instruments chosen for this study, there is a legitimate role for
exploratory factor analysis methods to develop more comprehensive answers to the
research questions posed in this study. Each survey instrument has undergone a reduction
in the number of scale items from its original version based on the results of a series of
earlier studies conducted by the original survey instrument developer to enhance the
reliability and improve factor validity of the instrument. The modification of an existing
survey instrument for the purpose of testing a new or revised theoretical model is a
continuation of the process.
205

Confirmatory factor analyses in this study were followed by exploratory factor

analysis, wherein modifications to the model were evaluated for their merit. First, nonsignificant regression paths coefficients were considered sufficient reason to eliminate or
modify causal paths in a structural model that were not considered absolutely necessary
to explain the primary assumptions of the theoretical model. Second, scale items that
were generating high standardized residual covariance terms with other scale items were
evaluated and given serious consideration to their removal for the purpose of developing
a more parsimonious and better specified model.

SEM Analysis of Goal Setting Theory Antecedent Variables

The three latent variables that represent the antecedents to work motivation in
Locke and Lathams goal setting theoretical model of work motivation are goal difficulty,
goal specificity and self-efficacy. These three constructs were operationalized in this
study by incorporating measurement scales from two separate sources in the academic
literature. A three factor first order structural model was developed to evaluate the
constructs prior to developing the complete structural model of goal theory and work
engagement. The model is shown in Figure G.14. The results of the first SEM analysis
of these three constructs produced a solution with inadequate model fit indexes. Several
scale item regression paths did not have statistically significant regression weights, and
one job goal specificity path was negatively weighted, in violation of the assumptions of
goal theory. A review of the results found that a single survey response in the sample
data set was exerting a very strong influence on the SEM solution. The data point was
206

responsible for the solutions multivariate non-normality = 366, c.r. = 77, and its
Mahalanobis distance = 286 from the solution centroid. In all the previous structural
model solutions, maximum values for Mahalanobis distance = 93 to 36. A review of the
SEM literature determined that removal of this single outlier survey response from the
sample set should be considered as a legitimate action because of its extreme nature
compared with all other responses in the sample data set (Byrne, 2001). The recalculated
multivariate normality of the model after solving with the deleted response for n = 359
was now 192, c.r. = 42, and the largest Mahalanobis distance = 98. This solution still
produced mediocre goodness-of-fit indexes, but all regression paths were now
statistically significant and had the correct algebraic sign according to the assumptions of
goal theory. Standard error terms for variances and regression weights did not appear
unusual, but examination of the standardized residual covariance matrix identified five
terms with very high negative standardized residual covariance terms (> 4.0) related to
the scale item job goal difficulty item 5R. Job goal difficulty item 5R is a reverse worded
question that reads as follows: My job is easy. A re-evaluation of earlier analytical
results for the job goal difficulty measurement scale and the scale item itself, including
reliability analysis and earlier factor analysis did not identify any specific problems with
this scale item. The problem first surfaced in the structural equation modeling analysis.
The SEM analysis results from the studys sample data set may imply that survey
respondents in the sample appear to have had some amount of trouble answering this
question. The wording of this question was evaluated in the context of the wording of
other questions. Looking at the question retrospectively, there is no clear connection to
job goals in the wording. The question with its negative wording may actually have
207

caused an adverse attitudinal response in a significant enough portion of the sample

population, generating a high amount of unexplainable negative residual covariance with
other question responses in the structural equation model solution. A reduced three factor
first order model was developed by removing the problematic scale item job goal
difficulty item 5R. The results of this modeling process are summarized in Table G.8
below. The solution for this reduced three factor first order model showed modestly
improved goodness-of-fit indexes.

208

Figure G.14 Three Factor First Order Model of Goal Theory Antecedents to Work
Motivation

Table G.8 SEM Analysis Results for the Goal Theory Antecedent Variables Model
209

Structural Equation Modeling Analysis of Goal Theory Research Model

The next step in the SEM process was development of the goal theory model of
work motivation. The goal theory model was developed by starting with the structural
model of the three goal theory antecedent variables job goal difficulty, job goal
specificity, and occupational self-efficacy. The feedback from the job itself and feedback
from agents latent variables were added to the goal theory model to represent goal
moderator variables. These five constructs represent predictor variables in the goal
theory model. Work engagement was added to the model as the criterion variable with
path arrows pointing from predictor variables to the criterion variables, explicitly
modeling the hypothesized cause and effect relationships underlying the theory of work
motivation. Based on the overall analysis objective of maintaining as much model
parsimony as possible in this study, the 9-item short version of work engagement was
chosen as the most appropriate representation of work engagement in the goal theory

210

research model. The 9-item confirmatory factor SEM solution showed a significantly
better fit over the full 17-item model of work engagement.
Three goal theory models were developed. The three factor first order 9-item
model of work engagement was developed first. The three factor second order 9-item
model of work engagement was developed next. The one factor 9-item model of work
engagement was developed last. The initial iterations of each model still contained the
full five-item scale for job goal difficulty. The scale item job goal difficulty 5R had a
strong adverse effect on these first iteration model solutions, causing fit problems and
high standardized residual covariance terms with numerous other scale items in other
measurement scales. The decision in the previous SEM analysis to delete it from the job
difficulty scale was confirmed and all further model development of the goal theory
research model was based on a reduced four item job goal difficulty measurement scale.
The decision to exclude one extreme outlier survey response which had been identified
and deleted in the earlier goal theory antecedents structural model analysis was also
continued in this goal theory modeling process. The multivariate normality of the sample
data set n = 359 is 192, c.r. = 42, and the largest Mahalanobis distance = 98. In the first
iteration of the model, all possible causal paths were added between each of the
three antecedent variables of goal specificity, goal difficulty and self-efficacy to the
three factors of work engagement vigor, dedication and absorption to represent their
mediating effect on work motivation. All possible causal paths were added between the
two feedback variables and job goal specificity, job goal difficulty and occupational
self-efficacy to represent feedback as a moderating effect on the three antecedent
mediator variables. The first iteration solution for the goal theory model indicated that
211

the regression weights for several of these causal paths were non-significant (p > 0.05)
and those paths were therefore deleted from that model. A second iteration and third
iteration of model specification was performed. A review of the solution sets indicated
that a second scale item was causing problems, job goal specificity item 3R, causing fit
problems and high standardized residual covariance terms with numerous other scale
items in other measurement scales. The final iteration of the goal theory research model
contained twenty-eight scale items (two scale items deleted). It was determined that
further efforts would not improve the models. The final versions of the three models are
shown in Figures G.15 through G.17.

212

Figure G.15 Goal Theory First Order Three Factor Work Engagement Model
213

Figure G.16 Goal Theory Second Order Three Factor Work Engagement Model
214

Figure G.17 Goal Theory One Factor Work Engagement Model

215

To achieve identification of the second order three factor goal theory model of
work engagement, two constraints had to be incorporated into the model. At the
second order level in the model there were six latent variables with causal paths to the
second order work engagement latent variable, producing [6 (6+1) /2] = 21 available data
points. The model as it was initially specified required estimation of thirteen regression
paths and nine latent variable variances, or twenty-two unknowns, resulting in an underidentified model. The variances for vigor and absorption were again set equal to the
same parameter var_a, and the regression path from work engagement to vigor was set
to 1. This reduced the number of parameters to be estimated at the second order level to
twenty, providing one degree of freedom and an over-identified model which could
generate a unique solution.
In all the final versions of the three goal theory models, standard error estimates
appeared acceptable although several terms were higher relative to the majority of other
standards errors. All terms in the correlation matrix were all less than one, the covariance
matrix was positive definite. There was however a still a couple of standardized residual
covariance terms between pairs of scale items that were greater than +2.58. The impact
on these analyses is that these SEM models still contain some potential model
misspecification between those pairs of scale items, a result possibly related to the
method employed in this research study where the four distinct survey instruments that
were originally developed for different purposes by separate and unrelated researchers
were merged intact and without modification of any sort. The goodness-of fit measures
for each model are shown in Table G.9. The goal theory model based on specification of
work engagement as a three factor second order construct yielded the best fit. Several
216

indexes including, CFI and RMSEA, indicate there is reasonably adequate model fit to
the sample data set, supporting the hypothesis that work engagement is a potentially
useful construct for measurement of work motivation as defined by goal theory. This
result is considered significant for the first time deployment of a combined survey
instrument with a diverse composition of constructs and large number of scale items.

Table G.9 SEM Analysis Results for Goal Theory Models of Work Engagement

The standardized regression weights are shown in Table G.10 for the regression
paths between each latent variable in the optimal second order three factor goal theory
model of work engagement presented in Figure G.16.

Table G.10 Regression Path Coefficients for Goal Theory Second Order Three Factor
Work Engagement Model
217

Estimates for three of the thirteen standardized regression coefficients were below
0.3. These three regression paths were left in the model because goal theory assumes
those paths should be theoretically significant, and their presence did improve model fit
slightly over a model that did not include those three regression paths.

SEM Analysis of the Job Characteristics Research Model

The development of the job characteristics model of work motivation started with
the five core job characteristics constructs in the Revised Job Diagnostic Survey
including autonomy, feedback from job itself, skill variety, task significance and task
identity and the related job characteristic feedback from agents. Dealing with others was
excluded due to its low reliability and non-significant correlation to work engagement
variables. These six constructs represented the predictor variables with path arrows
pointing to the work engagement criterion variable, explicitly modeling the hypothesized
218

cause and effect relationships underlying this proposed theory of work motivation. The
9-item short version of work engagement was used to maintain model parsimony. This
first job characteristics structural equation model is presented in Figure 5.18.
The SEM results for this first job characteristics research model were very
unsatisfactory. The goodness-of-fit measures are shown in Figure 5.18 with the model.
The regression weights for path lines between the job characteristics latent variables and
work engagement, modeled as a second order variable, were statistically non-significant
with all p-values > 0.05, and standard error estimates were very high. Feedback from
agents scale items generated high values with other scale items in the solutions
standardized residuals covariance matrix, indicting it should be eliminated from the
model. The next series of models was reduced to the five core job characteristics,
eliminating feedback from agents. The resulting structural models were defined by a total
of eight measurement scales representing five predictor latent variables and
three criterion latent variables with a combined total of twenty-four scale items in the
structural model for job characteristics and work engagement.

219

Figure G.18 First Six Factor Job Characteristics Structural Equation Model

220

In order to further examine the relationships between the five latent variables of
job characteristics and work engagement, a 2 x 3 matrix of structural models was
developed. Three job characteristics models were developed for the first order version of
job characteristics, and three additional models developed for the second order version of
job characteristics, a total of six job characteristics structural models, merging each of the
models with the three factor first order model of work engagement, the three factor
second order model of work engagement and the one factor model of work engagement.
The series of SEM models is presented in Figures G.18 through G.23 below.
The combined sample set was applicable and incorporated into this analysis,
utilizing the multi-group SEM method, with n1 = 60 from the pilot study of engineering
graduate students with jobs and n2 = 360 from the full study of all employees in the
engineering technical services company. The multi-group method analyzes each group
separately at first, calculating estimates of parameters and evaluating the covariance
structure of each group separately, and then uses the results from each separate group to
calculate a single set of Fit indexes for the structural model. The multi-variate nonnormality of group n1 = 60 was moderate in value at 42.98, c.r. = 4.7, and the response
with largest Mahalanobis distance from the centroid of the n1 = 60 sample set was 42.07.
The values calculated for the group n2 = 360 were again very high, with multi-variate
non-normality = 208.0, c.r. = 55.9 and largest Mahalanobis distance = 98. There were no
clear outliers in this data, rather a steady gradient of values from the lowest to highest.
There was no analytical justification to remove any outlier sample data points from the
analysis.

221

The first job characteristics structural model that was tested was the RDJS
five factor job characteristics first order, UWES 9-item three factor first order model. The
initial solution generated serious model misspecification issues which continued to show
up in varying degrees with every subsequent model. Regression coefficients for eleven
out of the total of fifteen causal paths between the five job characteristics latent variables
and the three work engagement factors were non-significant (p > 0.05), and seven out of
the eleven coefficients were negative, a violation of the underlying theory of this study
which hypothesizes a positive relationship exists between all predictor and criterion
variables. Estimates of standard errors for the variances of five out of the eight latent
variables in the model were high, and the variance for dedication was non-significant at
p = 0.19. The matrix of standardized residual covariance terms contained numerous high
values exceeding +/-2.58. This pattern of results repeated itself with only slight
variations in each of the additional five structural models that were developed.
Second order terms did not improve the results. In the second order models the
regression path for experienced meaningfulness was statistically significant and positive,
but the regression paths for autonomy and feedback were not significant. Deleting any
one of these variables from a structural model would undermine the core principles of the
job characteristics theory developed by Hackman and Oldham (1975). The goodness-of
fit indexes for the six models are summarized in Table G.11.

222

Figure G.19 Five Factor First Order Job Characteristics

One Factor 9-item Work Engagement Model

223

Figure G.20 Five Factor First Order Job Characteristics

Three Factor First Order 9-item Work Engagement Model

224

Figure G.21 Five Factor First Order Job Characteristics

Three Factor Second Order 9-item Work Engagement Model

225

Figure G.22 Five Factor Second Order Job Characteristics

One Factor 9-item Work Engagement Model

226

Figure G.23 Five Factor Second Order Job Characteristics

Three Factor First Order 9-item Work Engagement Model

227

e10

e11

e12

au2

au3

au1
1
1

resau
er1

er2

er3

sv2

sv3

sv1

autonomy

ressv

var_a

skill variety

distvi

er4

er5

er6

ts1

ts2

ts3

dist
we

task
significance

meaningfulness

vi1

vigor

rests

errvi1

vi2

errvi2

vi3

errvi3

work
engagement

distde

de2

dedication

de3
de4

er7

er8

er9

var_a

distab

ti1

ti2

ti3
absorption

resti

ab3
ab4
ab5

task identity

e13

e14

e15

fbji2

fbji3

fbji1
1

resfb

feedback
job itself

Figure G.24 Five Factor Second Order Job Characteristics

Three Factor Second Order 9-item Work Engagement Model

228

1
errab3

1
errab4

1
errab5

1
errde2

1
errde3

1
errde4

Table G.11 SEM Analysis Results for the Job Characteristics Research Models

Misspecification issues in the job characteristics research structural models were

severe and no model improvements suggested by modification indices or rearrangement
of regression paths were able to produce a valid job characteristics structural model with
work engagement as a criterion variable.

Multi-Group Analysis of Short Version Work Engagement UWES-9 Instrument

Up to this point in the SEM process, multi-group analysis results have been
reported for an unconstrained model. An unconstrained model allows the estimation of
all solution parameters to vary in each individual group solution set without constraints
229

assumed between groups. The assumption is that measurement regression paths,

measurement residual error terms, structural regression paths and structural residual
terms may vary across groups. This solution must be obtained first. In the same analysis
session, the next step is to set up multi-group analysis with a series of nested models. The
nested models progressively test for invariance of subsets of model parameters across
groups. The multi-group analysis settings are shown in Figure G.24. Five models are
developed, starting with the unconstrained model already solved, and each nest model
progressively assuming more constraints between groups. Each nested model tests for a
greater level of invariance between groups. The changes in chi-squared statistics between
each nested model are used to determine a P-value to test the hypothesis of invariance
between groups for each subset of estimated model parameters.

Figure G.25 Work Engagement Multi-Group Analysis Settings

The work engagement model analyzed in this multi-group analysis is shown in
Figure G.25. This model achieved over-identification by setting the structural path
230

coefficient from second order work engagement to first order vigor to a value of 1
(meaning it is not estimated in the analysis), while the residual variances for all three first
order latent variables were all allowed to be estimated parameters.

Figure G.26 Short Version Three Factor Second Order Work Engagement Structural
Equation Model Multi-Group Analysis
The results of this analysis are presented in Tables G.12 and G.13 directly from
the AMOS solution output file. Several goodness-of-fit statistics are presented for each
231

nested model in Table G.12, followed by evaluation of the deltas in chi-squared statistics
in Table G.13. Deltas of the other goodness-of-fit measures are also provided in the
output of the AMOS solution file.

Table G.12 Results Work Engagement Multi-Group Nested Models Analysis

Table G.13 Tests for Invariance Work Engagement Multi-Group Analysis

232

The results show that invariance has been first demonstrated between the pilot and
full study groups for the measurement weights which means the regression path weights
between each measurement variable and its respective latent 1st order variable, Vigor,
Dedication, and Absorption. Next invariance between groups has been demonstrated for
structural weights which are the regression path weights between each 1st order latent
variable and the 2nd order variable, Work Engagement. Thirdly, the structural residuals
for each first order latent variable have then been demonstrated to be invariant between
groups. Finally, invariance between measurement residuals has not been demonstrated.
233

This means that the measurement error terms for each scale item varied across the
two groups.

234

REFERENCES
American Psychological Association, Psychological Healthy Workplace Practices
Award Program, http://www.phwa.org/; 2008.
Andrews, Frank M., Construct Validity and Error Components of Survey Measures:
a Structural Modeling Approach, Public Opinion Quarterly, 1984, Vol. 48,
pp. 409-442.
Arbuckle, James L., AMOS 17.0 Users Guide, AMOS Development Corporation,
1995-2008.
Atkinson, John William, An Introduction to Motivation, Van Nostrand Publishing
Company, Princeton, New Jersey, 1964.
Bakker, Arnold B., Emmerik, Hetty van and Euwema, Martin C., Crossover of Burnout
and Engagement in Work Teams, Work and Occupations, November 2006,
Vol. 33, No. 4, pp. 464-489.
Bandura, Albert, Self-efficacy: Toward a Unifying Theory of Behavioral Change,
Psychological Review,1977, Vol. 84, pg. 191.
Bandura, Albert, Social Foundations of Thought and Action: A Social Cognitive
Theory, National Institute of Mental Health, Prentice-Hall, Inc., Englewood
Cliffs, New Jersey, 1986.
Baranik, Lisa E., Barron, Kenneth E. and Finney, Sara J., Measuring Goal Orientation
in a Work Domain, Educational and Psychological Measurement, 2007,
Vol. 67, No. 4, pp. 697-718.
Baron, R. M. and Kenny, D. A., The Moderator-Mediator Variable Distinction in Social
Psychological Research - Conceptual, Strategic, and Statistical Considerations,
Journal of Personality and Social Psychology, 1986, Vol. 51, No. 6, pp. 1173-1182.
Britt, Thomas W.; Castro, Carl Andrew and Adler, Amy B., Self-Engagement,
Stressors, and Health: A Longitudinal study, Personality and Social Psychology
Bulletin, 2005, Vol. 31, No. 11, pp. 1475-1486.
Buys, M.A., Olckers, C. and Schaap, P., The Construct Validity of the Revised Job
235

Diagnostic Survey, South African Journal of Business Management, 2007,

Vol. 38, No. 2, pp. 33-40.
Byrne, Barbara M., Structural Equation Modeling with AMOS, Psychology Press
Taylor & Francis Group, New York, New York, 2001.
Chen, Gilad, Gully, Stanley M. and Eden, Dov, Validation of a New General
Self-Efficacy Scale, Organizational Research Methods, January 2001, Vol. 4,
No. 1, pp. 62-83.
Coats, Erik J., Janoff-Bulman, Ronnie and Alpert, Nancy, Approach Versus Avoidance
Goals: Differences in Self-evaluation and Well-being, Personality and Social
Psychology Bulletin, October 1996, Vol. 22, No. 10, pp. 1057-1067.
Comrey, Andrew L., Reise, Stephen P. and Waller, Niels G., Factor Analysis and Scale
Revision, Psychological Assessment, 2000, Vol. 12, No. 3, pp. 287-297.
Fang, Eric, Palmatier, Robert W. and Evans, Kenneth R., Goal-setting Paradoxes?
Trade-offs between Working Hard and Working Smart: The United States versus
China, Journal of the Academy of Marketing Science, 2004, Vol. 32, No. 2,
pp. 188-202.
Eccles, Jacquelynne S. and Wigfield, Allan, Motivational Beliefs, Values, and Goals,
Annual Review of Psychology, 2002, Vol. 53, pp 109-132.
Grant, Adam M., Christianson, Marlys and Price, Richard H., Happiness, Health, or
Relationships? Managerial Practices and Employee Well-being Tradeoffs,
Academy of Management Perspectives, August 2007, pp. 51-63.
Grawitch, M.J., Gottschalk, M. and Munz, D.C., The Path to a Healthy Workplace: A
Critical Review Linking Healthy Workplace Practices, Well-being, and
Organizational Improvements, Consulting Psychology Journal: Practice and
Research, 2006, Vol. 58, No. 3, pp. 129-147.
Gonzalez-Roma, Vicente, Schaufeli, Wilmar B., Bakker, Arnold B. and Lloret, Susana,
Burnout and Work Engagement: Independent Factors or Opposite Poles?
Journal of Vocational Behavior, 2006, Vol. 68, pp. 165-174.
Hackman, J. Richard, Work Redesign and Motivation, Professional Psychology, June
1980, Vol. 11, No. 3, pp. 445-455.
Hackman, J. Richard and Oldham, Greg R., Development of the Job Diagnostic
Survey, Journal of Applied Psychology, 1975, Vol. 60, No. 2, pp. 159-170.
Hallberg, Ulrika E. and Schaufeli, Wilmar B., Same Same but Different? Can Work
Engagement be Discriminated from Job Involvement and Organizational
236

Commitment? European Psychologist, 2006, Vol. 11, No. 2, pp. 119-127.

Harter, James K., Schmidt, Frank L. and Hayes, Theodore L., Business-unit-level
Relationships between Employee Satisfaction, Employee Engagement, and
Business Outcomes: a Meta-analysis, Journal of Applied Psychology, 2002,
Vol. 87, No. 2, pp. 268-279.
Halbesleben, Jonathon R. B., Burnout and Engagement: Correlates and Measurement,
University of Oklahoma Graduate College, Norman, Oklahoma, 2003.
Hakanen, Jari J., Schaufeli, Wilmar B. and Ahola, Kirsi, The Job Demands-Resources
Model: A Three Year Cross-lagged Study of Burnout, Depression, Commitment
and Work Engagement, Work & Stress, 2008, Vol. 22, No. 3, pp. 224-241.
Haudan, James A. and MacLean, Donald, E is for Engagement: Transforming your
Business by Transforming your People, Journal of Change Management, 2002,
Vol. 2, No. 3, pp. 255-265.
Hu, Li-tze and Bentler, Peter M., Cutoff Criteria for Fit Indexes in Covariance Structure
Analysis: Conventional Criteria Versus New Alternatives, Structural Equation
Modeling, 1999, Vol. 6, No. 1, pp. 1-55.
Idaszak, Jacqueline R. and Drasgow, Fritz, A Revision of the Job Diagnostic Survey:
Elimination of a Measurement Artifact, Journal of Applied Psychology, 1987,
Vol. 72, No. 1, pp. 69-74.
Iffaldano, Michelle T. and Muchinsky, Paul M., Job Satisfaction and Job Performance:
A Meta-analysis, Psychological Bulletin, 1985, Vol. 97, No. 2, pp. 251-273.
Kahn, William A., Psychological Conditions of Personal Engagement and
Disengagement at Work, Academy of Management Journal, 1990, Vol. 33,
No. 4, pp. 692-724.
Kahn, William A., To be Fully There: Psychological Presence at Work, Human
Relations, April 1992, Vol. 45, No. 4, pg. 321.
Kaplowitz , Michael D., Hadlock, Timothy D. and Levine, Ralph, A Comparison of
Web and Mail Survey Response Rates, Public Opinion Quarterly, 2004,
Vol. 68, No. 1, pp. 94-101.

Koyuncu, Mustafa, Burke, Ronald J. and Fiksenbaum, Lisa, Work Engagement Among
Women Managers and Professionals in a Turkish Bank: Potential Antecedents
and Consequences, Equal Opportunities International, 2006, Vol. 25, No. 4,
237

pp. 299-310.
Latham, Gary P. and Pinder, Craig C., Work Motivation Theory and Research at the
Dawn of the Twenty-first Century, Annual Review of Psychology, 2005,
Vol. 56, pp. 485-516.
Latham, Gary P., Work Motivation: History, Theory, Research, and Practice, Sage
Publications, Thousand Oaks, California, 2007.
Lance, Charles E., Butts, Marcus M. and Michels, Lawrence C., The Sources of Four
Commonly Reported Cutoff Criteria, What did They Really Say? Organizational
Research Methods, April 2006, Vol. 9, No. 2, pp. 202-230.
Lawler, Edward E. and Porter, Lyman W., The Effect of Performance on Job
Satisfaction, Industrial Relations, A Journal of Economy and Society, 1967,
Vol. 7, No. 1, pp. 20-28.
Lee, Cynthia and Bobko, Philip, Exploring the Meaning and Usefulness of Measures of
Subjective Goal Difficulty, Journal of Applied Social Psychology, 1992, Vol. 22,
No. 18, pp.1417- 1428.
Lee, Thomas W., Locke, Edwin A. and Latham, Gary P., Goal Concepts in Personality
and Social Psychology, Lawrence Erlbaum Associates, Hillsdale, New Jersey,
1989, pp. 291-326.
Locke, Edwin A. and Latham, Gary P., Building a Practically Useful Theory of Goal
Setting and Task Motivation: A 35-Year Odyssey, American Psychologist,
September 2002, Vol. 57, No. 9, pp. 705-717.
Louison, Claura P., Convergent and Discriminant Validity of Employee Engagement,
Alliant International University, 2007.
MacCallum, Robert C., Widaman, Keith F., Zhang, Shaobo and Hong, Sehee, Sample
Size in Factor Analysis, Psychological Methods, 1999, Vol. 4, No. 1, pp.84-99.
MacLennan, Richard N., Inter-rater Reliability with SPSS for Windows 5.0, The
American Statistician, November 1993, Vol. 47, No. 4, pp.292-296.
Marcum, James W., Out with Motivation, in with Engagement, National Productivity
Review, Autumn 1999, pg. 43.
Marsh, Herbert W., Hau, Kit-Tai and Wen, Zhonglin, In Search of Golden Rules:
Comment on Hypothesis-Testing Approaches to Setting Cutoff Values for Fit
Indexes and Dangers in Over Generalizing Hu and Bentlers (1999) Findings,
Structural Equation Modeling, 2004, Vol. 11, No. 3, pp. 320-341.
238

Maslach, Christina, Schaufeli, Wilmar B. and Leiter, Michael P., Job Burnout, Annual
Review Psychology, 2001, Vol. 52, pp. 397-422.
May, Douglas R., Gilson, Richard L. and Harter, Lynn M., The Psychological
Conditions of Meaningfulness, Safety and Availability and the Engagement of the
Human Spirit at Work, Journal of Occupational and Organizational
Psychology, 2004, Vol. 77, pp. 11-37.
Mayer, John D., Faber, Michael A. and Xu, Xiaoyan, Seventy-five Years of Motivation
Measures (1935-2005): A Descriptive Analysis, Motiv Emot, 2007, Vol. 31,
pp. 83-103.
MINITAB Release 14 for Windows, 2003-2004, Mintitab Inc.
Muchinsky, Paul M., Psychology Applied to Work, Eighth Edition, Thomson
Wadsworth Corporation, Stamford, Connecticut, 2006, pg. 98.
Mundfrom, Daniel J., Shaw, Dale G. and Ke, Tian Lu, Minimum Sample Size for
Conducting Factor Analyses, International Journal of Testing, 2005, Vol. 5,
No. 2, pp. 159-168.
OBrien, Kelly, Factor Analysis: An Overview in the Field of Measurement,
Physiotherapy Canada, 2007, Vol. 59, No. 2, pp. 1432-155.
Oldham, Greg R., Job Characteristics and Internal Motivation: The Moderating Effect of
Interpersonal and Individual Variables, Human Relations, 1976, Vol. 29, No. 6,
pp. 55-569.
PASW 17.0 Statistics Gradpack, 2009, SPSS, Inc.
Rigotti, Thomas, Schyns, Birgit and Mohr, Gisela, A Short Version of the Occupational
Self-efficacy Scale: Structural and Construct Validity across Five Countries,
Journal of Career Assessment, 2008, Vol. 16, No. 2, pp. 238-255.
Rothbard, Nancy P., Enriching or Depleting? The Dynamics of Engagement in Work
and Family Roles, Administrative Science Quarterly, 2001, Vol. 42, pp. 655-684.
Ryan, Richard M. and Deci, Edward L., Self-Determination Theory and the Facilitation
of Intrinsic motivation, Social Development, and Well-Being, American
Psychologist, January 2000, Vol. 55, No. 1, pp. 68-78.
Ryan, Richard M. and Deci, Edward L., On Happiness and Human Potentials: A
Review of Research on Hedonic and Eudaimonic Well-being, Annual Review of
Psychology, 2001, Vol. 52, pp. 141-166.
Saks, Alan M., Antecedents and Consequences of Employee Engagement. Journal of
239

Managerial Psychology, 2006, Vol. 21, no. 7, pp. 600-619.

Salanova, Marisa, Agut, Sonia and Peiro, Jose Maria, Linking Organizational
Resources and Work Engagement to Employee Performance and Customer
Loyalty: The Mediation of Service Climate, Journal of Applied Psychology,
2005, Vol. 90, No. 6, pp. 1217-1227.
Schaufeli, Wilmar B., Salanova, Marisa, Gonzalez-Roma, Vicente and Bakker, Arnold B.,
The Measurement of Engagement and Burnout: A Two Sample Confirmatory
Factor Analytic Approach, Journal of Happiness Studies, 2002, Vol. 3, pp. 71-92.
Schaufeli, Wilmar B. and Bakker, Arnold B., Job Demands, Job Resources, and their
Relationship with Burnout and Engagement: A Multi-sample Study, Journal of
Organizational Behavior, 2004, Vol. 25, pp. 293-315.
Schaufeli, Wilmar B., Bakker, Arnold B. and Salanova, Marisa, The Measurement of
Work Engagement with a Short Questionnaire, Educational and Psychological
Measurement, August 2006, Vol. 66, No. 4, pp. 701-716.
Schyns, Birgit and von Collani, Gernot, A New Occupational Self-efficacy Scale and
Its Relation to Personality Constructs and Organizational Variables, European
Journal of Work and Organizational Psychology, 2002, Vol. 11, No. 2, pp. 219-241.
Seijts, Gerard H., Latham, Gary P., Tasa, Kevin and Latham, Brandon W., Goal Setting
and Goal Orientation: An Integration of Two Different yet Related Literatures,
Academy of Management Journal, 2004, Vol. 47, No. 2, pp. 227 239.
Senko, Corwin and Harackiewicz, Judith M., Achievement Goals, Task Performance,
and Interest: Why Perceived Goal Difficulty Matters, Personality and Social
Psychology Bulletin, December 2005, Vol. 31, No. 12, pp.1739-1753.
Shimazu, A., Schaufeli, W.B., Nashiwa, H., Kato, A. and Sakamato, M., Work
Engagement in Japan: Validation of the Japanese Version of the Utrecht Work
Engagement Scale, Applied Psychology: An International Review, 2008,
Vol. 57, No. 3, pp. 510-523.
Sims Jr., Henry P. and Szilagyi, Andrew D., Job Characteristic Relationships:
Individual and Structural Moderators, Organizational Behavior and Human
Performance, 1976, Vol. 17, pp. 211-230.
Sims Jr., Henry P., Szilagyi, Andrew D. and Keller, Robert T., The Measurement of Job
Characteristics, Academy of Management Journal, 1976, Vol. 19, No. 2, pp.195-212.
Sonnentag, Sabine, Recovery, Work Engagement, and Proactive Behavior: A New Look
at the Interface Between Non-work and Work, Journal of Applied Psychology,
240

2003, Vol. 88, No. 3, pp 518-528.

Steers, Richard M., Factors Affecting Job Attitudes in a Goal-Setting Environment,
Academy of Management Journal, 1974, Vol. 19, No. 1, pp. 6-16.
Trochim, William M.K., Research Methods Knowledge Base, Web Center for Social
Research Methods, http://www.socialresearchmethods.net/, 2006.
Tucker, L. R., Koopman, R. F. and Linn, R. L, Evaluation of Factor Analytic Research
Procedures by Means of Simulated Correlation Matrices, Psychometrika, 1969,
Vol. 34, pp. 421-459.
VandeWalle, Don, Development and Validation of a Work Domain Goal Orientation
Instrument, Educational and Psychological Measurement, 1997, Vol. 57, No. 6,
pp.995-1015.
Waterman, Alan S., Two Conceptions of Happiness: Contrasts of Personal
Expressiveness (Eudaimonia) and Hedonic Enjoyment, Journal of Personality
and Social Psychology, 1993, Vol. 64, No. 4, pp. 678-691.
Wright, Bradley, E., The Role of Work Context in Work Motivation: A Public Sector
Application of Goal and Social Cognitive Theories, Journal of Public
Administration Research and Theory, 2004, Vol. 14, No. 1, pp.59-78.
Wright, Patrick M., A Theoretical Examination of the Construct Validity of
Operationalizations of Goal Difficulty, Human Resources Management Review,
1992, Vol. 2, No. 4, pp. 275-298.
Ximnez, Carmen, A Monte Carlo Study of Recovery of Weak Factor Loadings in
Confirmatory Factor Analysis, Structural Equation Modeling, 2006, Vol. 13,
No. 4, pp. 587-614.
Your Dictionary.com, http://www.yourdictionary.com/engaged, 2009.

241