o L- .4v 5
3.“: mg“ .
u dumt \
0 .I I . 0 ~ .lo
4‘ ‘ - > 1.0.45:
. an. . ‘ “A
. ‘f‘ ‘3‘ R. 1 . * ..7

~§ . ,
“932$; 9'“.
.«1‘. ‘ ‘ v
’2 o- J‘ ’éuwnbév‘I.‘)'.tw-Il ‘3 ‘ 4
~ . ‘ . \o -

.1“) IX”, - .2 .4 “‘“C'Ti ”\f K “‘."‘;‘.2 -
‘ ‘ .a'.a .‘.>_.......
‘i‘wffi‘ffl: ‘ 'EL sgd‘va" :1” ':~ 12‘:

n‘ V A‘t‘i‘1\ . ‘ {nu

..

A s x

ink-3.5.5? tyg.

.. s. van-yum
‘n‘ wfl‘A

. _. ‘

I V
v ". J 3:13
.4. fig

,4

- tygh

1223;} "a. 3»: :1
533 s._ .
.. 9m."

1.
v x. 5‘4“, .
"“V‘XSJP‘S‘ '(
I s 7.“, {
"=9: ”:21?

.‘. Z

J. 0:.

u‘

“11v.
.

t .r r x .u
«h— z:- :-
‘ “-13%,
l§41 ‘- A
" ‘ 0’! 3‘49
3 ;
U

.1" .
"‘7.,.,

7-)“ .

uu'. ..
..

 

 

 

 

 

WWWMimi
“33:33:“,

 

 

 

This is to certify that the

dissertation entitled

PATH ANALYTIC TEST OF A MODEL OF THE
DETERMINANTS OF TURNOVER INTENTIONS
0F JUNIOR ARMY OFFICERS

presented by

Martha Lappin Teplitzky

has been accepted towards fulfillment
of the requirements for

Ph.D. , Psychology
degree in

 

 

7 ,.
.. 7 /,
7 Iqajor progssor 7

Neal Schmitt

Date &//$(/7[
] /

MSU is an Affirmatiw Action/Equal Opportunity Institution 0— 12771

 

 

PLACE IN RETURN BOX to remove We checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE DATE DUE DATE DUE

      
 

 

 

 

 

L

 

 

 

 

U
E

 

 

 

j
E

 

 

 

 

 

 

JL
C]
E

 

 

 

 

 

 

 

 

 

 

 

 

 

[TC I?

MSU Is An Afflrmetlve ActiorVEquel Opportunity InetItutIon
owns-9.1

 

% fi‘flv

PATH ANALYTIC TEST OF A MODEL OF THE DETERMINANTS
OF TURNOVER INTENTIONS OF JUNIOR ARMY OFFICERS

BY

Martha Lappin Teplitzky

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Psychology

1991

ABSTRACT

PATH ANALYTIC TEST OF A MODEL OF THE DETERMINANTS
OF TURNOVER INTENTIONS OF JUNIOR ARMY OFFICERS

BY

Martha Lappin Teplitzky

Organizational commitment and expected utility models
have dominated recent efforts to identify the direct
determinants of turnover intentions in civilian organiza-
tions. In military organizations there is a stronger
emphasis on family and career concerns in the study of
retention decisions. These approaches were combined in the
present research. A multivariate model of the determinants
of propensity to stay was tested in a large sample of
married, male junior Army officers. Officers were surveyed
late in 1989, and commissioned between 1984 and 1987.

The path model proposed that four variables, perceived
career prospects, organizational identification, anticipated
Army/family conflict and years of service would all have
direct, independent effects on propensity to stay in the
military. The model also predicted that five additional
variables (person/branch match, prior career orientation,
work satisfaction, operational support and inspirational
leadership) would have indirect effects on propensity to
stay through career prospects and identification.

The path analysis supported the importance of the

proposed direct determinants. The four hypothesized direct

determinants all had significant paths to propensity to
stay, and together accounted for half of the variance in the
dependent variable. The effects of organizational
identification and anticipated Army/family were particularly
strong. One of the hypothesized indirect determinants of
propensity to stay, current work satisfaction, displayed
unexpected direct, as well as indirect effects through
career prospects and organizational identification.

Results were interpreted as suggesting that current
turnover models need to be expanded to include career and
family related considerations, particularly when young,
married professionals comprise the population of interest.
The data also point to the need to include both present and
future oriented measures in turnover models; turnover
decisions appear to be influenced by immediate, affective
reactions to the job and the organization, as well as
rational assessments of future outcomes. It was also argued
that organizational identification is a promising,
theoretically useful alternative to traditional

organizational commitment constructs in turnover research.

This dissertation is dedicated to my husband, Victor.

I was so lucky to have found you.

iv

ACKNOWLEDGEMENTS

The day this dissertation was accepted was one of the
most fulfilling of my life. I think that my own sense of
satisfaction is exceeded only by my family's delight.
Throughout all of the ups and downs of the past three years
my family provided unwavering love, support, and
encouragement. With the help of Beverly and Mrs. Mac, they
tolerated my preoccupation, frustrations and absences better
than I did. I owe Victor, Kimberly, Benjamin and my mother
a debt of gratitude I can never repay.

I also owe many thanks to the members of my committee.
Dr. Neal Schmitt, the chair of my doctoral committee and a
valued friend and advisor from my first day in graduate
school, made it possible for me to resume my studies after a
long absence. I always felt that he believed I could do it,
and do it well, and this sustained me. I feel honored to
have been the student of such a distinguished scholar and
exceptional man. I am also grateful to the other members of
my committee, Drs. Kevin Ford, Daniel Ilgen, and Steven
Kozlowski. They willingly agreed to shepard yet another
long—lost graduate student through the final hurdles of the

doctoral program. Most importantly, they took the time to

provide positive feedback, as well as constructive comments
about what needed further thought. I can't leave Michigan
State without also thanking Suzy Pavick. If every school
had such competent, helpful and friendly people in their
graduate offices, I am sure that many more ”long—distance"
students would be able to complete their degrees.

Finally, I owe much to my employer for the last nine
years, the Army Research Institute. In particular, I want
to thank Drs. Nora Stewart and Paul Gade, good friends as
well as mentors. Only now, after having worked for them
both, can I really understand what an important difference a
mentor can make to an individual's personal and professional
development. Thank-you, it means more to me than I will

ever be able to express.

vi

TABLE OF CONTENTS

INTRODUCTION ......... . ................................
Contemporary Models of Organizational Turnover ........
Overview of the Proposed Model ........................
Theoretical Assumptions About Decision Processes ......
Literature Related to the Direct Determinants .........
Career Prospects .................................
Organizational Identification ....................
Anticipated Work/Family Conflict .................
Years of Service .................................
Literature Related to the Indirect Determinants ......
Overview .........................................
Person/Branch Match ..............................
Current Work Satisfaction ........................
Prior Career Orientation .........................
Operational Support ..............................
Inspirational Leadership .........................
Potential Determinants Omitted from the Model .........
Perceived Alternatives ...........................
Interactions Among the Direct Determinants .......
Model Specification and Research Design Issues ........

The Unmeasured Variable Problem ..................

vii

10
14
19
19
25
34
37
42
42
44
47
53
62
68
73
73
78
79

79

Sample Restrictions .............................. 83

Summary and Central Research Questions ........... 87
METHOD ........... . .................................... 91
Sample and Procedure ............................. 91
Measures ......................................... 96
Common Method Variance Issues .................... 107
Data Analysis .................................... 109
RESULTS ............................................... 111
Preliminary Analyses..... ............................. 111
Path Analysis ......................................... 127
Exploratory Regresions with Propensity to Stay ........ 142
DISCUSSION ............................................ 157
LIST OF REFERENCES .................................... 182

APPENDIX: A Priori and Final Scale Items for
Variables in the Propensity to Stay Model ......... 198

viii

LIST OF FIGURES

Figure 1: Proposed Junior Officer Career Intentions....11

Figure 2: Path Coefficients for the Proposed Career
Intentions Model ................................... 128

Figure 3: Path Coefficients for Empirically Derived
Model .............................................. 137

ix

LI ST OF TABLES

Table 1: 1989 LROC Population and Sample Statistics
for ROTC and USMA Male Officers by Year Group ...... 94

Table 2: Frequency Distributions for the Two Propensity
to Stay Items ..................................... 98

Table 3: Scale Statistics and Intercorrelations for
Indirect Determinants ............................. 102

Table 4: Principle Components Analysis Results:
Original Items for Direct Determinants Scales ...... 112

Table 5: Factor Analysis Communality Estimates:
Original Items for Direct Determinants Scales ...... 113

Table 6: Pattern and Structure Matrices for Oblique
Rotation of Common Factors: Original Items for
Direct Determinants Scales ........................ 115

Table 7: Principle Components Analysis: Reduced Set
of Items for Direct Determinants Scales ............ 119

Table 8: Factor Analysis Communality Estimates: Reduced
Set of Items for Direct Determinants Scales ........ 120

Table 9: Pattern and Structure Matrices for Oblique
Rotation of Common Factors: Reduced Set of Items
for Direct Determinants Scales ..................... 121

Table 10: Direct Determinants: Scale Statistics,
Intercorrelations, and Correlations with Propensity
to Stay ........................................... 124

Table 11: Correlations between Indirect and Direct
Determinants..... .................................. 124

Table 12: Hierarchical Multiple Regression Results for
Propensity to Stay Using Predicted and Non-Predicted
Paths .............................................. 131

Table 13: Hierarchical Multiple Regression Results for
Career Prospects Using Predicted and Non—Predicted
Paths .............................................. 133

Table 14: Hierarchical Multiple Regression Results for
Organizational Identification Using Predicted and
Non-Predicted Paths ................................ 134

Table 15: Total, Direct and Indirect Effects on
Propensity to Stay for Empirical Model ............. 138

Table 16: Regression Results Testing Source of Commission
Moderating Effects on Propensity to Stay ........... 145

Table 17: Descriptive Statistics and Intercorrelations
for Exploratory Variables .......................... 151

Table 18: Additional Variance Explained in Propensity
to Stay by Three Sets of Exploratory Variables ..... 152

Table 19: Hierarchical Multiple Regression Results for

Propensity to Stay Using Model and Signigficant
Exploratory Variables .............................. 154

xi

INTRODUCTION

Organizational turnover research has a long history.
The impetus for this work stems from the important
implications of turnover research for both organizations and
individuals. For organizations, an understanding of the
precursors of turnover can facilitate efforts to better
predict and manage the process. At the same time, the
results of turnover research can inform efforts to develop
programs and policies that enhance individual satisfaction,
career development and effectiveness.

Efforts to identify the antecedents of turnover began
with analyses of the bivariate correlations between job
attitudes, particularly satisfaction, and turnover. In the
past decade, however, a number of multivariate models of the
factors and processes involved in turnover decisions have
been proposed and tested. The results of these analyses are
promising, yet the types of variables examined as
antecedents of turnover decisions are still somewhat
limited. Theoretical explanations for observed effects are
also frequently lacking, and few researchers have tried to
refine general models to fit the demographic, occupational
or organizational characteristics of different research
populations.

The present research tests an expanded model of the

determinants of turnover intentions and suggests alternative

2
ways of conceptualizing key variables. Career and family
factors are incorporated into the model along with a
”commitment-like" construct defined as "organizational
identification". The variables in the model also reflect
assumptions about the factors likely to be especially
salient in the target population. The population of
interest consists of married, male Army officers still in
the early stage of their careers.

An applied objective of this research is to provide the
Army with useful information on the career decisions of an
important subgroup of officers. Millions of dollars are
spent annually to attract and train new officers. The Army
spends close to $200,000 just to send one cadet through West
Point, and the costs to the Army of one ROTC cadet range
from $66,000 to $80,000 (Adelsberger, 1988). Premature
separations not only involve high replacement costs, they
also limit the Army's ability to be selective in promoting
and assigning junior officers to key leadership positions.
Impending budget cuts and mandates to reduce the size of the
officer corps make it especially critical that Army manpower
planners today understand the factors that influence the
career decisions of junior officers.

From a theoretical perspective the goal of this
research is to contribute to the development of parsimon-
ious, theoretically defensible models of the factors and
processes involved in turnover decisions. The proposed

model builds on existing theory and research and is designed

3
to test the assumptions of several contemporary models. The
results of the path analyis will provide data on the utility
of some previously unexplored constructs and shed light on
the potential mediating role of two key variables.

To provide a context for the discussion of the proposed
model, the literature review begins with a brief overview of
contemporary models of organizational turnover. Following
this overview, the proposed model is summarized and under-
lying theoretical assumptions are discussed. The pertinent
civilian and military literature is then reviewed in the
context of the hypothesized model linkages. The final
section of the introduction addresses general research
design and model specification issues. For convenience and
because the sample for this and most officer research
consists only of male officers, masculine pronouns are used

throughout this paper to refer to officers.

Contemporary Models of Organizational Turnover
Contemporary turnover models reflect a variety of
sociological, economic and psychological perspectives.
These models differ in terms of the nature of the variables
they examine, their focus on content versus process, and the
emphasis on antecedents or consequences of turnover.
Models reflecting a sociological perspective (Bluedorn,
1982a, 1982b: Price, 1977: Price & Mueller, 1981) tend to be
content oriented. Although psychological variables

(especially job satisfaction and organizational commitment)

4
are often included as intervening variables in these models,
the research focus is on the structural and organizational
antecedents of turnover (e.g., centralization, integration,
routinization). There is little consistency across these
studies with regard to the effects and relative importance
of structural characteristics in individual turnover
decisions. Results do, however, suggest that factors likely
to increase worker autonomy and responsibility (e.g., low
routinization, instrumental communication) are often related
to job satisfaction and commitment, and thus have important
indirect effects on turnover (Bluedorn, 1982b; Martin, 1979:
Price & Meuller, 1981: Thompson & Terpening, 1983).

Models with their roots in the industrial psychology or
organizational behavior literature may also include
structural variables as antecedents. The primary focus of
these models, however, is on the more proximal psychological
processes and attitudinal antecedents involved in turnover
decisions.

Most psychological models are built on an expectancy
theory framework, suggesting that rational evalution
processes and utility maximization principles underlie
turnover decisions (Forrest, Cummings & Johnson, 1977;
Mobley, Griffeth, Hand & Meglino, 1979: Rhodes & Doering,
1983). The Mobley et a1. (1979) model reflects this
orientation, and buttressed by a comprehensive review of the
literature, has had a major impact on turnover research in

the past decade.

5

In the Mobley model, behavioral intentions (to search
for alternatives and to quit one's job) are proposed as the
only direct determinants of turnover. The primary
determinants of turnover intentions are hypothesized to be:
(a) current job satisfaction, (b) the expected utility of
the current job for the attainment of desired outcomes, and
(c) the expected utility of alternatives to the present job.

The specification of satisfaction as a central variable
in the model is based on years of research documenting the
bivariate relationship between job satisfaction and
turnover. The central role of the expected utility
constructs stems from expectancy theory. It is assumed that
individuals assess the likelihood of obtaining desired
(and/or aversive) outcomes in different situations, and then
choose the alternative with the highest expected utility.

The notion that turnover decisions are influenced by
the quality and/or availability of alternatives is not new.
March and Simon (1958) introduced a similar "ease of
movement" construct into the organizational participation
literature 30 years ago. The Mobley model did, however,
spark a renewed interest in the empirical assessment of
alternatives. Unfortunately, the operationalization of and
placement of this construct varies markedly across studies.
A recent review of the literature concluded that the effects
of perceived alternatives are weak and inconsistent (Hulin,

Roznowski & Hachiya, 1985). There is stronger support for

6
the hypothesis that expectations about rewards forthcoming
from the current job influence turnover decisions.

Models based on Fishbein and Ajzen's (1975; Ajzen &
Fishbein, 1980) theory of reasoned action are similar to
expectancy based models in their focus on the consequences
of turnover decisions. Instead of assessing evaluations of
alternative jobs, however, these models examine attitudes
toward the act of leaving the present organization.
Reasoned action models also go beyond the instrumentality
approach by postulating a normative component in the
decision process. Subjective norms, construed as beliefs
about, and the motivation to comply with what important
others think one should do, are also hypothesized to
influence turnover intentions. Although not frequently
examined, subjective norms appear to be a promising
explanatory construct (Hinsz & Nelson, 1990: Hom & Hulin,
1981: Newman, 1974: Prestholdt, Lane & Mathews, 1987).

Another body of research has accumulated around the
role of organizational commitment in turnover decisions.
The dominant stream of research is guided by Porter, Steers
and Mowday's conceptualization of organizational commitment
(Mowday, Porter & Steers, 1982: Porter, Steers, Mowday &
Boulian, 1974). The Porter et a1. (1974) definition of
commitment is multidimensional, including perceptual,
attitudinal and intentional components. Measures of the
construct show consistent and fairly strong correlations

with turnover intentions, and when intentions are not

7
measured, turnover behavior (Cotton 8 Tuttle, 1986: Mathieu
8 Zajac, 1990: Steel 8 Ovalle, 1984).

The theoretical utility of such a broadly defined
construct has been questioned, however. Consequently, a
number of researchers are now examining more specific
components of commitment (McGee 8 Ford, 1987: Meyer 8 Allen,
1984; O'Reilly 8 Chatman, 1986) or different referents of
commitment in organizations (Lachman 8 Aranya, 1986:
Reichers, 1986: Weiner 8 Vardi, 1980). Evidence that
different components and referents of commitment are
differentially related to attitudes and behaviors suggests
that further work on more narrowly defined "commitment-like"
constructs is warranted.

Other pertinent work fits into what might be called the
"too much invested to quit" paradigm (Teger, 1980). Models
based on Becker's (1960) "side bets" or investment theory of
behavioral commitment reflect this approach. Investment
models suggest that the costs of leaving of an organization
play an important role in turnover decisions (Farrell 8
Rusbult, 1981: Rusbult 8 Farrell, 1983; Stevens, Beyer 8
Trice, 1978). As the costs of leaving an organization are
generally assumed to increase over time (e.g., employees
acquire seniority, become more invested in retirement
plans), the oft observed negative effects of age and tenure
on turnover are interpreted as support for this model (Cohen

8 Lowenberg, 1990).

8

In addition to the considerable research on the
antecedents of turnover, concerns about the individual and
organizational consequences of turnover have been addressed
(Muchinsky 8 Morrow, 1980; Steers 8 Mowday, 1981), and the
relationship between job performance and turnover has been
explored (Dreher, 1982; Jackofsky, 1984; Wells 8 Muchinsky,
1985). These studies, and the work on the specific
withdrawal cognitions and search and comparison activities
that precede the turnover decision (Hom, Griffeth 8 Sellaro,
1984: Mobley, 1977; Mobley, Horner 8 Hollingsworth, 1978)
are less relevant to the model proposed here.

In summary, the turnover literature is characterized by
a variety of approaches and theoretical perspectives. A
common theme, however, is that affective reactions to the
job and/or organization are important determinants of
turnover intentions. Reviews of the organizational turnover
literature support this notion, indicating that job
satisfaction and organizational commitment are relatively
strong and consistent correlates of turnover cognitions and
behavior (Cotton 8 Tuttle, 1986: Mobley et al., 1979; Steel
8 Ovalle, 1984: Steers 8 Mowday, 1981). Furthermore,
despite their generally high intercorrelations, satisfaction
and commitment typically account for independent variance in
turnover intentions (Arnold 8 Feldman, 1982; Bluedorn, 1982:
DeCotiis 8 Summers, 1987; Dougherty, Bluedorn, 8 Keon,
1985; Lachman 8 Aranya, 1986: Lee 8 Mowday, 1987; Michaels 8

Spector, 1982; Porter et al., 1974).

9

There is also evidence that multiple psychological
processes underlie turnover decisions: no single explanatory
framework is clearly superior or sufficient. Results
supporting elements of investment and expectancy based
models imply that rational, utility based evaluation and
comparison processes are involved. At the same time,
immediate affective reactions to the job, psychological
bonds to the organization, and normative pressures also
appear to influence turnover decisions.

In the absence of a clearly dominant theoretical
framework, the approach adopted here is somewhat eclectic.
The proposed model builds on research demonstrating the
importance of anticipated outcomes, job satisfaction and
organizational commitment, but attempts to specify more
precisely the critical dimensions of these variables. The
theoretical model also reflects a consideration of the
special characteristics of military organizations and the
career and family life cycle stages of officers in the
target population.

An overview of the model is presented in the next
section. Following the overview, explanations for the
different ways key variables might operate in the decision
process are advanced. The purpose for this is not to
generate or test hypotheses about processes per se, but
rather to more clearly differentiate the nature of the

proposed determinants of turnover intentions.

10
Overview of the Proposed Model

The present research is concerned with the attitudinal
and perceptual determinants of junior officer turnover
intentions. A causal model of the direct and indirect
determinants of "propensity to stay" is tested using path
analytic techniques (see Figure 1).

The dependent variable, propensity to stay, reflects an
officer's intentions or plans with regard to the Army;
specifically, does he intend to make the Army a career or
leave the organization to pursue a career in the civilian
world. Turnover intentions typically exhibit high correla-
tions with actual behavior (Steel 8 Ovalle, 1984), thus the
test of the model can contribute to our understanding of the
causes of actual turnover. At the same time, propensity to
stay is an important dependent variable in its own right.
Whether employees leave or not, organizations need to be
aware of the factors that make individuals want to leave.

Four variables are proposed as direct determinants of
propensity to stay in the model. These variables include:
(a) the officer's assessment of his prospects for a
satisfying career in the Army, (b) the extent to which the
officer identifies with the Army, (c) the level of
work/family conflict the officer anticipates if he stays in
the Army, and (d) the number of years the officer has been
on active duty in the Army.

Five additional variables are hypothesized to affect

prospensity to stay indirectly through effects on perceived

.2602 953:9... .6050 5220 5:5... comoaotn. ”F 93E

                 

11

><hw O...
>._._wzmm0mm

 

._.O_.._n_ZOO
>..___>_<n_\>_>_m<

 

 

ZO_._.<O_n:._.ZMD_
._<ZO_._.<N_Z<GmO

  

 

 

kammwOm—n.

\‘

  

n=Imme<m3
._<ZO_._.<m_me_

 

 

 

— Fran—mam
.._<ZO_._.<mm_n._O

 

 

 

ZO_._.<._.Zm_mO
mmmm<0 IOEn.

 

 

 

ZO_._.O<u_w_._.<w

 

i

 

mmmm<0 >2m<

 

 

w0_>m_mw
Z. wm<m>

 

 

—
_ vac; .

 

 

 

IO._.<S_
IOZ<mm

 

 

12
career prospects and organizational identification. These
antecedent variables include: (a) the match between the
officer's preferred and assigned branch, (b) opportunities
for intrinsic satisfaction in the current work assignment,
(c) prior orientation toward or plans for a military career,
(d) perceptions of the effectiveness of the operational,
work-related support provided by the organization, and (e)
exposure to inspirational leadership.

Two variables, prior career orientation and operational
support, are hypothesized to influence bgth career prospects
and organizational identification. Person/branch match and
work satisfaction are expected to influence only career
prospects, and inspirational leadership is hypothesized to
affect only organizational identification.

The organizational commitment literature indicates that
"commitment-like" constructs often mediate the effects of
organizational perceptions and experiences. This literature
(reviewed later) provides support for the hypothesis that
organizational identification is an important intervening
variable. The theoretical model posits that experiences in
the organization, particularly with respect to leaders and
operational support, will influence the way an individual
feels about and relates to the organization. These
experiences are not, however, expected to influence turnover
intentions net of their effects on identification with the

organization.

13

There is less empirical support for the hypothesized
mediating role of career prospects. Arguments for the
centrality of this variable in the model rest partly on two
unique features of the military. First, the decision to
stay in the Army reflects a career choice as well as an
organizational choice. Second, job assignments in the
military are varied and temporary; the typical assignment
lasts only two to three years. In light of these
considerations and the rational evaluation processes
expected to play a part in turnover decisions, perceptions
of long-term career prospects in the organization are
hypothesized to be stronger, more immediate determinants of
intentions than prior plans or current satisfaction. The
antecedent variables are expected to shape perceptions of
career prospects, but they are not expected to influence
propensity to stay once the effects of the direct
determinants are taken into account.

Possible antecedents of anticipated work/family
conflict are not explored in this research. The military
literature suggests that a variety of personal, family and
job related factors interact to determine perceptions of
work/family conflict. Examination of this complex
constellation of factors is beyond the scope of the present
study. Future work in this area will be clearly warranted,
however, if the results of this research suggest that
work/family conflict is indeed an important factor in

turnover decisions.

14

The primary purpose of the research is to determine the
independence, predictive utility and sufficiency of the four
proposed direct determinants. Specifying additional
variables hypothesized to have only indirect effects on
propensity to stay affords a more convincing, theoretically
useful test of the utility of the proposed direct
determinants. Evidence that the antecedent variables do or
do not have the hypothesized effects on career prospects and
organizational identification will also contribute to future
efforts to identify the determinants of these intervening

variables.

Theoretical Assumptions About Decision Processes

§§I§§I_2IQ§E§2§§

The assumptions underlying the specification of Army
career prospects as a central variable in the model are
that: (a) officers value opportunities for advancement and
interesting, engaging careers, (b) they assess the
likelihood that staying in the Army will provide these
opportunties, and (c) positive evaluations of career
prospects are an inducement to stay in the organization.

These assumptions reflect the evaluation processes and
utility maximization principles implied in the expectancy
theory framework. The premises are also consistent with
data demonstrating the importance of intrinsic job/career

rewards in turnover decisions.

15
mimics

The hypothesized positive effects of years of service
on propensity to stay can be explained within the investment
or "side-bets" framework. Tenure in an organization
represents a time investment likely to have economic
consequences for the employee. This is particularly the
case in the military, where eligibility for retirement
benefits depends on length of serivce. Officers who stay
for 20 years receive annual retirement benefits equivalent
to half of their salary: those who leave before 20 years of
service receive nothing in the way of retirement benefits.
With each passing year then, the perceived opportunity costs
of leaving the organization are likely to increase.

Tenure may also reflect a knowledge or skill investment
in an organization. Organizationally specific expertise
acquired through years of experience is likely to be valued
within the organization, but may not be easily marketed
elsewhere. The expected positive effect of tenure on
propensity to stay is viewed as resulting from an evaluation
of the costs of sacrificing tenure related investments. In
other words, individuals with more time in an organization
may decide to stay simply because they have "too much
invested to quit".

Qrsanizafignal_ldsnfifigatign

The inclusion of a variable tapping feelings about the

organization is clearly warranted in organizational turnover

research. Typically, the relevant organizational variable

16
is conceptualized as "commitment". Yet when commitment
measures include intentions or desire to remain with the
organization (as most do) evidence that commitment is
related to turnover intentions is almost tautological.
Another limitation is that typical conceptualizations of
commitment are not grounded in theories that offer an
understanding of the distinctive nature of the phenomenon
(Scholl, 1981: Weiner, 1982).

A focus on organizational identification may be more
fruitful from a theoretical perspective. The fairly recent
literature on social identity theory (Ashforth 8 Mael, 1989;
Turner, 1987) offers a precise definition of identification
and a theoretical framework that suggests how identification
may influence propensity to stay. Identification is said to
occur when membership in the group becomes a salient element
in the individual's social self-concept, or central to the
way he defines himself. Strong identification with an
organization may make it difficult to leave both because
organizational norms against resigning are internalized, and
because leaving is tantamount to renouncing an important
aspect of one's social identity. In highlighting the
importance of the self-concept and internalized norms,
social identity theory is compatible with theories
emphasizing the importance of internal and external
normative pressures in the turnover decision process

(Fishbein 8 Ajzen, 1975: Fishbein, 1980; Weiner, 1982).

17
W

Despite calls for the inclusion of nonwork or family
factors in turnover models (Mobley et al., 1979; Schein,
1978: Schneider 8 Dachler, 1978: Steers 8 Mowday, 1981),
family factors have been largely neglected in civilian
turnover research. The importance of family considerations
is well documented in the military literature, however
(e.g., Orthner, 1990).

Perceptions of work/family conflict may influence
turnover decisions both through rational utility evaluations
and normative pressures. For some, separations and family
life disruptions may be viewed as costs of staying in the
organization, factors to be weighed against the economic and
career related advantages of staying. Yet internal and
external norms related to work and family may also be a
factor. Family values and role expectations may preclude
staying in an organization where job demands or
organizational needs take precedence over family needs.
§BEEQIY

In summary, the direct determinants proposed in the
model reflect assumptions about the psychological processes,
as well as the critical factors or content areas, that
influence turnover decisions. Consistent with expectancy
theory, decisions are presumed to be guided to some extent
by expectations that desirable career and economic outcomes
will be achieved. If utility is broadly defined to include

family outcomes, utility maximization principles may also

18
account for the influence of work/family conflict on
propensity to stay.

Internalized norms and self-defining values are also
expected to come into play, however. When an officer
strongly identifies with the Army, for example, the need to
be ”true to oneself” or loyal to the organization may lead
to a decision to stay even when career and economic
considerations might dictate leaving. Similarly, if serious
work/family conflicts are anticipated, internal or external
(e.g., spouse) norms related to family roles may induce
individuals to leave even when the utility "balance sheet"
suggests that they should stay.

The distinction between instrumental and normative
influences is consistent with the conceptual frameworks
offered by Scholl (1981) and Weiner (1982). Both argue that
commitment-like constructs are theoretically useful only if
they explain behavior that cannot be predicted within
traditional expectancy and equity based motivational models.
Scholl (1981), in fact, defines commitment as a "stabilizing
force that acts to maintain behavioral direction when
expectancy/equity conditions are not met and do not
function" (p. 593). In light of the different motivational
processes implied by the variables in the model, data on the
magnitude of the effects of the direct determinants may
point to processes, as well content areas, that require

additional attention in the study of turnover decisions.

19

Literature Related to the Direct Determinants

Career Prospegts
g . i E E1 .!.

A career is typically viewed as an occupational,
professional or organizational "path" an individual follows
(Schein, 1978). Implicit in this definition is the notion
that a career involves a developmental sequence of
positions, typically leading to higher levels of expertise
and work role responsibility (Hall, 1976; Rapoport 8
Rapoport, 1978).

The term "career prospects" in this research refers to
perceived opportunities for career-enhancing assignments,
advancement within the individual's chosen career field, and
work that matches the individual's skills and interests.

The hypothesized effect of career prospects on propensity to
stay is based on (a) research suggesting that anticipated
outcomes, particularly with respect to opportunities for
advancement and challenge, influence turnover decisions: and
(b) evidence that career goals provide a context for
evaluating the utility of staying in the organization.

mes

The notion that anticipated outcomes influence
employment decisions is a central thesis of models derived
from expectancy theory (Forrest et al., 1977: Mobley et al.,
1979: Rhodes 8 Doering, 1983). As Mobley et al. (1979)
observe, an employee can be unhappy with certain aspects of

a job, yet be attracted to and intend to stay on the job

20
because it is expected to facilitate the attainment of
valued outcomes or goals in the future. Accordingly, Mobley
et a1. (1979) proposed that perceptions of the expected
utility of the present job (along with current job satis—
faction and the expected utility of available alternatives)
would be an important determinant of turnover intentions.

In expectancy models, the "force" toward, or expected
utility of an alternative is presumed to be a multiplicative
function of three factors: (a) the likelihood that different
outcomes will be associated with, or result from an
alternative (”instrumentality" in expectancy theory
terminology): (b) the desirability of the various outcomes
(”valence"): and (c) the expectancy that the individual can
obtain (or keep, in the case of the present job) that
alternative.

Tests of expectancy models of job choice and turnover
have not supported the utility of theoretically derived
multiplicative measures ([Sum IV] x E) of expected utility.
Simple indices obtained by summing instrumentality ratings
(i.e., expectations regarding positive outcomes) generally
predict employment decisions better than the more complex
weighted measures (Hom, 1980; Hom, Griffeth, 8 Sellaro,
1984; Hom 8 Hulin, 1981: Lawler, Kuleck, Rhode, 8 Sorenson,
1975: Mitchell 8 Knudson, 1973: Mobley, Hand, Baker 8
Meglino, 1979; Oldham, 1976: Schneider, 1976). However, the
fact that simple outcome expectation measures are related to

employment decisions supports the hypothesized importance of

21
anticipated outcomes in turnover decisions (Hom, Griffeth, 8
Sellaro, 1984: Hon 8 Hulin, 1981: Mitchell 8 Albright, 1972;
Mobley, Hand, Baker 8 Meglino, 1979: Parker 8 Dyer, 1976:
Schneider, 1976).

A key question in expectancy models concerns the nature
of the outcomes individuals consider when they evaluate
current and alternative jobs. It is assumed here that
opportunities for advancement and intrinsically satisfying
work are the most critical dimensions. Expectations with
regard to other aspects of the job (e.g., pay, co-workers,
supervision, working conditions) are presumed to be much
less important when individuals make turnover decisions.

Support for this assumption comes from meta-analyses
showing that satisfaction with promotion opportunities is a
fairly strong and consistent correlate of turnover (Cotton 8
Tuttle, 1986): and evidence that the mean correlation (r=-
.28) between work satisfaction and turnover is even higher
than the mean correlation (r=-.25) between overall job
satisfaction and turnover (Steel 8 Ovalle, 1984).

Intrinsic satisfaction and advancement opportunities
appear to be particularly strong predictors of retention for
"white collar" workers (Flowers 8 Hughes, 1973: Thompson 8
Terpening, 1983) and college graduates still in the early
exploratory, or ”trial" stages of their careers (DePasquale
8 Lange, 1971; Dunnette, Arvey 8 Banas, 1973: Hall, 1976).

Dunnette et al. (1973), for example, found that

opportunities to use abilities and get ahead were among the

22
most valued attributes of a job in a sample of newly hired
college graduates. Although few in the sample found that
their jobs fulfilled their high expectations, discrepancies
between expected and experienced opportunities for challenge
and advancement were considerably greater among those who
subsequently left the organization.
; : .3 !°

Surprisingly few researchers conceptualize anticipated
work outcomes as career considerations. The shift in
emphasis from expected job or work outcomes to anticipated
career outcomes is subtle, but important.

As noted earlier, a career involves a related sequence
or progression of assignments, typically leading to higher
levels of responsibility and expertise. The notion that
employees base utility assessments on career prospects
suggests a longer term, more goal oriented perspective than
is implied in the anticipated job outcomes framework. In
other words, the present model assumes that turnover
decisions are based on a desire to maximize long-term cereer
satisfaction, not simply day-to-day job satisfaction.

Military researchers have been most likely to frame
questions about future outcomes in a career context,
probably because the decision to leave the military involves
a career, as well as an organizational choice. Perceptions
that a military career offers challenging and meaningful
work, a sense of accomplishment and opportunities for

advancement are strongly and consistently related to

23
military officers' retention decisions (Ashcraft, 1987;
Gibb, Nontasek, Dolgin 8 Helm, 1987: Paris, 1984; Hayden,
1985: Meola 8 Koechel, 1983; Monkus, 1979: Schmidt, 1982:
Shenk 8 Wilbourn, 1971: Steele, 1987: Szoc, 1982). Paris
(1984), for example, found that the belief that a military
career would be more interesting than civilian work was a
very strong predictor of the decision to stay. Perceptions
of differences in pay, on the other hand, had no effect; the
majority of all officers, even those intending to stay,
believed that they would earn less in the military than in a
comparable civilian job.

Jans (1988) developed the career construct further in
his research on the careers of officers in the Australian
Defence Force. His analyses included a measure of "career
prospects" similar to the one proposed here. Jans' measure
addressed promotion prospects, likelihood of getting
satisfying assignments, and satisfaction with the service's
career management practices. In a regression equation that
included eight other theoretically important attitudinal and
perceptual variables, career prospects was the strongest
(albeit, by only a slight amount) predictor of a measure of
career involvement. Career involvement, in turn, was the
strongest predictor of propensity to stay in the military.

Research in civilian organizations also suggests that
career considerations play a role in turnover decisions. In
one of the earliest studies, Farris (1971) asked a group of

scientists and engineers whether it would be more profitable

24
to their career development to remain in their current
technical division or move elsewhere. Across two organiza-
tional samples, the career development variable was more
strongly and consistently related to turnover than any other
job factor examined (e.g., influence, diversity, contacts,
working hours).

Schneider and Dachler (1978) addressed career concerns
in a slightly different way, asking employees the extent to
which the organization facilitated or inhibited progress
toward their career goals. When turnover intentions were
regressed on a variety of work, family and career related
variables, organizational career facilitation entered the
regression first in both halves of a cross-validation
analysis. This variable alone accounted for an impressive
25% of the variance in intentions to leave.

Graen and his colleagues also focused on careers,
asking employees how relevant their present work was to
their career plans. Analyses based on both newly hired
(Graen, Orris 8 Johnson, 1973) and established (Graen 8
Ginsburgh, 1977) nonsupervisory, clerical employees
indicated that when perceived career relevance was high,
employees were more satisfied and less likely to resign.
The results were especially striking in the sample of new
employees. Within 10 months of the beginning of the study,
over twice as many "low career relevance" as "high career
relevance" employees had quit (59% vs. 25%). In the sample

of relatively established employees, an interaction effect

25
was observed: employees with low scores on both career
relevance and leader acceptance measures were particularly
likely to leave.

Scholl (1983) developed an objective measure of
promotion prospects in the organization based on ratios of
(a) the number of vacancies in the position above the focal
position relative to the number of incumbents in the focal
position, and (b) the number of incumbents in the next
higher position relative to the number of incumbents in the
focal position. Both indices of internal career mobility
opportunities were related to intended and actual turnover.
SBEEAIY

In summary, the literature supports the hypothesis that
perceived career prospects will be an important predictor of
propensity to stay. Past research confirms the importance
of career considerations in general, as well as more
specific expectations regarding future opportunties for
advancement and intrinsic satisfaction. The literature also
suggests that perceived prospects for a satisfying career
are especially salient for employees similar to those in the
present study - military officers, white collar employees
and recent college graduates.

0 an zational Iden if n
gemmirment in rhe Ternover Literature

The implications of organizational identification are
best understood in the context of the broader commitment

literature. In their comprehensive review, Mobley et al.

26
(1979) concluded that organizational commitment is
”significantly and negatively related to turnover and more
strongly related to turnover than satisfaction" (p. 508).
Research subsequent to this review has also demonstrated
that organizational commitment is consistently related to
both turnover intentions and behavior (Arnold 8 Feldman,
1982: Bluedorn, 1982: Daugherty, Bluedorn 8 Keon, 1985:
Michaels 8 Spector, 1982: Mowday, Steers 8 Porter, 1979:
Williams 8 Hazer, 1986).

Typically turnover researchers have adopted the Porter
et al. (1974) definition of organizational commitment as
”the strength of an individual's identification with and
involvement in a particular organization" (p. 604). More
specifically, organizational commitment is said to be
characterized by: (a) a strong belief in and acceptance of
the organization's goals and values, (b) a willingness to
exert considerable effort on behalf of the organization: and
(c) a definite desire to maintain membership in the
organization (p. 604). Each of these three components of
commitment is tapped in the Organizational Commitment
Questionnaire (OCQ, Mowday, Steers, 8 Porter, 1979).

The widespread use of the OCQ in turnover research
attests to the popularity of this multi-dimensional
conceptualization of commitment. Over the course of the
past decade, however, a number of theorists have questioned
the analytical utility of such a complex construct (DeCotiis

8 Summers, 1987; Graen 8 Johnson, 1977; Mobley et al., 1979;

27
Morrow, 1983: Reichers, 1985; Scholl, 1981). Typically
critics argue that more specific components or forms of
attachment to an organization need to be examined. Mobley
et al. (1979), for example, suggested that the congruence
between individual and organizational values can vary
independently of the other two components of commitment, and
therefore should be examined separately. Mobley and his
colleagues also noted that the importance of commitment
relative to other antecedents of turnover is difficult to
assess when commitment measures include items tapping
behavioral intentions to remain with the organization.

In a similar vein, Scholl (1981) argued that when
commitment is defined largely in terms of intentions to
exert effort and remain with the organization, it is
superfluous. Antecedents of commitment can be related
directly to the effort or turnover intentions of interest,
obviating the need for the global commitment construct.

Porter and his colleagues made an important
contribution to the literature by encouraging researchers to
look beyond job specific attitudes in their search for the
determinants of turnover. The emerging consensus, however,
is that theory development is best served by research on
more specific, conceptually distinctive dimensions of
attachment to an organization. Organizational identifica-
tion appears to be a promising construct because it is
unidimensional and its operation can be explained within

existing theoretical frameworks.

u: " u an. ov-g- . 0 can' . 'oa __-- .. ion

In a thoughtful review of the literature on social
identity theory, Ashforth and Mael (1989) conceptualize
organizational identification as a specific form of social
identification. Identification in general is defined as
"the perception of oneness with or belongingness to a group,
involving direct or vicarious experience of its successes or
failures" (Ashforth 8 Mael, 1989, p. 34). When individuals
identify with an organization, their membership or role in
the organization becomes an important part of their social
self-concept. Although imbued with an affective component,
identification is viewed primarily as a cognitive
phenomenon, reflecting a tendency to define oneself in terms
of the identification target.

Identification is sometimes viewed as being synonomous
with the internalization of organizational goals and values.
Hall, Schneider and Nygren (1970), for example, suggest that
organizational identification is in fact "the process by
which the goals of the organization and those of the
individual become increasingly integrated or congruent" (p.
176).

Ashforth and Mael (1989), on the other hand, try to
more clearly differentiate the two constructs. They propose
that value congruence is likely to be both an antecedent and
a consequence of organizational identification, but it is
not a necessary component. When norms and values are

perceived to be similar initially, this perceived similarity

29

promotes an enhanced sense of "belonging" to the group,
which is likely to encourage identification. At the same
time, once an individual starts to define himself in terms
of his status as a member of a particular group, the goals
and values of this reference group become more salient,
increasing the likelihood that characteristic norms and
values will be internalized.

O'Reilly and Chatman (1986) also viewed internalization
and identification as related, but conceptually distinct
constructs. Based on their review of the literature on
identification and attitude change, they identified three
possible bases for psychological attachment to an
organization. Identification was conceptualized as
involvement based on a desire for affiliation, and this form
of attachment was differentiated from attachment predicated
on congruence between individual and organizational values,
and attachment based on compliance, or instrumental
involvement for specific, extrinsic rewards (O'Reilly 8
Chatman, 1986). A factor analysis of 12 items designed to
measure these constructs supported the authors' conceptual
distinctions and regression analyses showed internalization
and identification to be independent sources of variance in
turnover intentions.

Despite the slightly different definitions of
identification available in the literature, there is some
consistency in the operationalization of the construct.

Items in O'Reilly and Chatmans' (1986) "identification"

30

factor referred to pride in the organization, feeling a
sense of ownership, and "talking up" the organization as a
great place to work (p. 494). Pride in being part of the
organization and feeling a "sense of belonging" were also
used in the Hall et al. (1970) measure of identification.
Similar items were included in the "affective commitment"
scale Meyer and Allen (1984) developed "to assess commitment
characterized by positive feelings of identification with,
attachment to and involvement in the work organization"
(p.375). Specific items tap emotional attachment to the
organization, feeling a "sense of belonging", and feeling as
though the organization's problems are one's own. Because
this scale has sound psychometric properties and appears to
reflect what Ashforth and Mael (1989) and others (Hall,
Schneider 8 Nygren, 1970; O'Reilly 8 Chatman, 1986) have
called identification it is used to measure identification
in the present research.
Qrgenizerienel Identification in rhe Army

The constellation of attributes characterizing military
organizations suggests that a focus on organizational
identification is especially appropriate in this setting.
The military has a unique mission in society, and a
distinctive, identifiable set of values (e.g., "duty, honor,
country"). Symbols, ceremonies and legends abound in the
culture and are used to highlight the distinctiveness of the
organization. New entrants into the military also undergo

an intensive socialization process (e.g., pre-commissioning

31
military science courses and and basic training) emphasizing
the importance of military values and traditions.

The existence of an identifiable organizational image
and value system increases the likelihood that an individual
can and ultimately will come to identify with the
organization (Ashforth 8 Mael, 1989). The high level of
organizational identification found among U.S. Forest
Service employees in the early 1970's was attributed in
large part to the unique mission and ideology of the
organization (Hall et al., 1970). Similarly, the lower
level of identification observed in the same organization 15
years later was attributed to changes over the intervening
years that reduced the distinctiveness of the organization
(Bullis 8 Tompkins, 1989).

Hall et al. (1970) also observed that when a career is
uniquely associated with a particular organization (e.g.,
forest rangers, priests, military officers) the probability
that the organization will become a focus of identification
is increased.

Igegrifieerien egg Turnover Intentigns

The literature on organizational identification and its
impact on turnover is quite limited. Only one study has
directly examined the effects of organizational identifi-
cation on turnover and turnover intentions. In a small
mixed sample of university employees, O'Reilly and Chatman
(1986) found that identification with the organization was

correlated with both intentions to remain with the

32
organization and actual turnover 16 months later. Moreover,
this relationship was obtained even when compliance and
value internalization measures were simultaneously included
in the regression equation.

Indirect support for the identification/turnover
relationship can also be inferred from data indicating that
organizational identification is related to tenure among
forest rangers and priests (Hall 8 Schneider, 1972). One
would expect a positive relationship if individuals who do
not identify with the organization tend to leave early in
their careers. It must be noted, however, that the overall
turnover rates in these organizations were relatively low,
leading the authors to conclude that factors other than
self-selection were primarily responsible for the observed
identification/tenure relationship. Interestingly, tenure
was not related to identification for organizationally
mobile research and development professionals (Hall 8
Schneider, 1972). This finding was consistent with the
authors' hypothesis that tenure would be more strongly
associated with identification among employees in single-
organization careers.

Studies in which other measures of affective commitment
have been related to turnover intentions lend additional
support to the identification-propensity to stay link
hypothesized in the model. In three different studies,
measures of affective commitment (developed by the authors

in each case to exclude the intentional components of the

33

OCQ) were related to either turnover intentions or turnover
behavior (Cook 8 Wall, 1980: DeCotiis 8 Summers, 1987;
Werbel 8 Gould, 1984). Werbel and Gould's (1984) measure
was especially interesting because it tapped the perceived
moral obligation to be loyal to the organization. Evidence
that this type of measure is a strong predictor of
propensity to stay highlights the potential importance of
internal norms in turnover decisions.
SQEEQIY

In the present research it is hypothesized that
individuals who more strongly identify with the organization
will be more inclined to stay. The model posits that the
effects of identification are independent of the effects of
Army career prospects, work/family conflict and years of
service. Identification is also expected to operate
differently, influencing turnover decisions through
internalized norms against leaving and psychological needs
to maintain a coherent self-concept.

Empirical support for the hypothesized relationship
between organizational identification and retention is
based largely on evidence that affective commitment measures
are consistently correlated with turnover intentions and
behavior. It was argued, however, that theoretical
considerations warrant the focus on a more clearly defined,
theoretically grounded construct, and identification appears

to be particularly relevant in settings like the military.

34
a o m' ct
g i l D E' i!'0

Although research has confirmed the importance of
satisfaction and commitment related variables in voluntary
turnover, much of the variance in turnover decisions remains
to be explained. Meta-analyses indicate that the mean
weighted correlations between turnover and overall job
satisfaction and organizational commitment are only a modest
.25 and .36, respectively (Steel 8 Ovalle, 1984). Under-
lying the present model is the assumption that some of the
unexplained variance in turnover decisions is a function of
work/family conflict.

Work/family conflict has not been a focus of attention
in the civilian turnover research to date despite several
calls for the incorporation of family or non-work factors
into turnover models (Mobley et al., 1979: Schneider 8
Dachler, 1978: Steers 8 Mowday, 1981). The family cannot be
easily ignored in the military, however. The military is a
particularly "greedy instituition", placing tremendous
demands on soldiers and their families (Segal, 1986).
W

Military life is characterized by frequent relocations,
both short and long term family separations, long work
hours, unpredictable schedule and assignment changes, and
”normative pressures on family members regarding their roles
in the military community" (Segal, 1986, p. 16). These

requirements place demands on spouses, as well as service

35
members, and have important implications for the quality of
life family members experience (Derr, 1979: Farkas 8
Durning, 1982: Grace 8 Steiner, 1978: Hunter 8 Shaylor,
1978: Jans, 1988: Orthner 8 Bowen, 1982: Wood, 1982).

Over the years, officers have consistently identified
demands impinging on family life as the career attributes
most likely to make them leave. In a large scale survey
conducted in 1964, separations from family and isolated
tours were identified by officers who stayed, as well as
those who left, as the factors which would most influence
them to leave the Air Force (Shenk and Wilbourn, 1971). In
a sample of Army officers a decade later, family separations
outranked even repetitive tours in Vietnam as an "influence
to leave", and frequency of moves was the third most
negative factor (Lund, 1978). Similarly, junior Naval
officers presented with a list of 45 aspects of military
life identified relocation frequency, family separations,
and time with family as the most important reasons for
leaving the military (Szoc, 1982). Furthermore, in the same
study, officers intending to leave indicated that reducing
the disruptions to family life (fewer/shorter deployments,
regular 40 hour work weeks, fewer moves) would be more
likely than changes in any other job conditions (including
better pay and benefits) to make them stay (Szoc, 1982).

-2 to . -ni- Confli t and i i a i-te t°o

Although the objective demands of a military career are

formidable, the results of several different studies suggest

36
that the extent and frequency of specific demands have less
impact on retention decisions than subjective perceptions of
conflict or problems. Both Szoc (1982) and Jans (1988), for
example, found that only reeeriere to separations (e.g.,
level of satisfaction, family stress and perceived problems)
not the number or duration of separations, were related to
feelings about military life and career intentions.

Similarly, Farkas and Durning (1982) found that a
subjective measure of work interference with family life was
a strong predictor of family pressure to leave the Navy,
which in turn predicted career intentions. Yet the
subjective work/family conflict measure was only weakly
(r=.11 to .15) related to actual time spent away from the
family, the length of the typical work week and number of
hours per week with the spouse.

Results of a survey of over 600 Naval officers
confirmed the importance of enticipareg family outcomes
(Neumann, Abrahams 8 Githens, 1972). Officers who stayed in
the military were much more likely than those who
subsequently left to believe that they could attain a
satisfactory family life in the Navy. In fact, there was
only one item (likelihood of fully using one's abilities) in
the combined set of intrinsic and extrinsic factors where
differences between "stayers" and "leavers" were larger than

they were for "a satisfactory family life".

37

SHEEQIY

The stresses of a military career appear to make
work/family conflict an especially salient factor in
military retention decisions. The data further suggest that
the subjective experience of work/family conflict is not
necessarily related to the extent or frequency of objective
demands. The focus on anticipated conflict in this research
is consistent with the underlying assumption that officers
tend to focus on future outcomes, rather than past or
present conditions when they make career decisions.

geers of Service

Wm

Early reviews of the literature identify tenure as a
consistent correlate of propensity to stay (Mobley et al.,
1979: Porter 8 Steers, 1973). Results in more recent
multivariate studies, however, tend to be mixed. Dalessio,
Silverman and Shuck (1986), for example, dropped tenure from
a revised version of the Mobley et a1. (1978) model after
testing the original model in five different samples. Yet
others continue to find tenure effects even when variables
like commitment and satisfaction are included in a model
(Arnold 8 Feldman, 1982: Jans, 1988: Price 8 Mueller, 1981:
Thompson 8 Terpening, 1983: Werbel 8 Gould, 1985).

Conceptually, it may be most productive to examine
tenure effects in the context of what an employee's time
investment represents in a particular occupational or

organizational sample. Among relatively new employees, for

38
example, tenure accounts for very little variance in
turnover intentions (Rusbult 8 Farell, 1983: Werbel 8 Gould,
1984). A fairly substantial time investment may be required
before tenure begins to affect career decisions.

Career stage and occupational differences may also be
important. In some professions "job hopping" may be the
norm and the primary means of advancement. In such cases,
employees with more tenure may feel more pressure to move on
than those with only one or two years in the organization.

re v stm n s 'n th 1 'ta

In the military, the effects of tenure on retention
decisions are predictable, and in mixed tenure samples, at
least, quite strong. Investment or "side-bet" models of
turnover provide an explanation for these effects (Farrell 8
Rusbult, 1981: Meyer 8 Allen, 1984: Rusbult 8 Farrell,
1983).

Investment models build on Becker's (1960) proposition
that individuals become bound to a particular course of
action as a result of the accumulation of "side bets"
reflecting investments in the organization. Interpreted
broadly, "side bets" refer to "anything of value the
individual has invested (e.g., time, effort, money) that
would be lost or deemed worthless ... if he or she were to
leave the organization" (Meyer 8 Allen, 1984, p. 373).

Investment in an organizational retirement plan may be
a particularly salient investment. In the military, for

example, individuals who complete 20 years of active duty

39
service are eligible to retire at half of their regular
active duty pay. Officers who leave the military prior to
completing 20 years of service receive re retirement
benefits. Obviously, as officers approach the 20 year mark,
it becomes increasingly impractical, from an economic stand
point, to voluntarily leave the Army before retirement.

The military is also an organization that offers very
specialized training and work experiences. Over time,
officers develop skills and specialties that are valued
internally, but not necessarily marketable in the civilian
world. Time invested acquiring organizationally specific
expertise is likely to be lost when the individual leaves
the organization. When the costs of leaving the
organization reach a certain level, employees may stay
primarily to protect their investments.

Measures of "investments" or "continuance commitment"
have predicted propensity to stay in two multivariate
analyses (Meyer 8 Allen, 1984: Rusbult 8 Farrell, 1983).
The measures used in these studies included a variety of
factors, from time invested in retirement plans, to the lack
of alternatives to the present job, to general perceptions
of the hardships, sacrifices and costs that would be
associated with leaving the present job (Meyer 8 Allen,
1983: Rusbult 8 Farrell, 1983). This type of measure does
not allow effects attributable solely to tenure to be
identified, however, it is typically assumed that increased

tenure is associated with the accumulation of investments.

40

Pfeffer and Lawler (1980) offer an alternative
explanation for the relationship between tenure and turnover
intentions. In Pfeffer and Lawler's framework, tenure
represents a behavioral commitment to the organization.
Drawing on the social information processing model of
attitude development (Salancik 8 Pfeffer, 1978) and the
insufficient justification paradigm, a behavioral commitment
to a course of action (i.e., remaining with an organization)
is hypothesized to affect attitudes. Specifically, Pfeffer
and Lawler argue that long-tenured employees need to justify
their behavioral commitment to the organization. They do
this by developing more positive attitudes toward the
organization, and devaluing the importance of extrinsic
rewards when attractive alternatives are available. This
line of thinking led Pfeffer and Lawler to hypothesize that
the effects of pay and the availability of alternatives on
turnover intentions would be non-significant or even
negative among employees with the greatest tenure.

These main effect and interaction hypotheses were
tested in a large sample of university faculty members.
Results indicated that tenure was positively related to
propensity to stay, and the effects of salary and perceived
alternatives on propensity to stay were weaker for those
with more tenure. The main effects of tenure were
attributed to tendencies to become psychologically bound to
a particular course of action (i.e., staying in the

organization) as a result of past behavior. The interaction

41
effects were interpreted as supporting the "weak form" of
the insufficient justification hypothesis: individuals with
the greatest behavioral commitment to the organization
reconstruct the situation so that extrinsic rewards and
alternative opportunities become less important.

Although the insufficient justification paradigm is
intriguing, the investment model offers a simpler
explanation for the observed effects. Alternatives and pay
may be less important to the turnover decisions of more
tenured employees simply because the costs of leaving the
organization are prohibitive. In other words, reluctance to
sacrifice tenure related investments rather efforts to
reduce dissonance can explain both the main and interaction
effects observed by Pfeffer and Lawler (1980). In the
military, for example, officers close to retirement very
seldom voluntarily leave the service because few can afford
to give up their retirement benefits (Jans, 1988).

These arguments imply that there is a point at which
investments may become the dominant consideration, resulting
in weak or non-significant relationships between attitudinal
variables and turnover intentions. Tenure is only likely to
have significant moderating effects, however, in samples
that include a substantial number of individuals who have
significant investments in the organization.

Officers in the present sample are all in the early
stage of their careers and only a small percentage (those

with prior service) have served more than six years.

42
Because their time investment is relatively small, few
officers are expected to base turnover decisions on
investments alone: career prospects, identification and
work/family conflict are expected to have similar effects
across all tenure levels.
£93222!

Years of service is included in the present model to
tap the effects of time investments on turnover decisions.
More time in the organization is associated with a greater
investment in the retirement plan, and in many cases, the
acquisition of organizationally specific skills. Rational
assessments of the costs of sacrificing these investments
are assumed to underlie the positive effects of tenure on
propensity to stay. These effects are expected to be
additive and relatively small because of the restricted
range of tenure in the sample. Exploratory analyses will be
conducted to test the assumption that years of service does
not moderate the effects of the other direct determinants on

propensity to stay.

Literature Related to the Indirect Determinants
Overview
In this section, the conceptual and empirical support
for the antecedent linkages in the model is reviewed. Two
variables in the model (person/branch match and current work
satisfaction) are expected to influence propensity to stay

only through effects on Army career prospects. One variable

43
(inspirational leadership) is expected to affect retention
intentions only through organizational identification. Two
other antecedents (prior career orientation and operational
support) are hypothesized to have effects on both career
prospects and identification. Most importantly, none of the
antecedent variables in the model are hypothesized to have
direct, independent effects on propensity to stay.

There is little direct support in the turnover
literature for the mediating effects proposed in the model.
Tests of intervening variable models are still relatively
rare, and most of the variables in the model have not been
previously examined in the form proposed here. Therefore,
much of the empirical support for the antecedent linkages is
only suggestive, based on research on the antecedents of
organizational commitment or studies examining the direct
effects of related variables on turnover.

In light of the primary focus on the sufficiency of the
direct determinants, however, the lack of strong empirical
support for the proposed indirect effects is not necessarily
a weakness. A great number and variety of variables
demonstrate bivariate correlations with turnover intentions
or withdrawal cognitions. The present model posits that
observed relationships between certain antecedent variables
and turnover intentions are spurious, resulting from the
failure to include key mediating variables. The inclusion
of these antecedent variable allows a stronger test of the

utility of the proposed direct determinants.

44

The discussion of the antecedent variables centers on
(a) establishing that these variables are related to
turnover intentions and/or are particularly relevant in
military settings, (b) presenting any empirical data
suggesting that the effects of these variables might be
mediated by constructs similar to career prospects and
identification, and (c) providing logical or conceptual
arguments to support the hypothesized relationships.

PersonZBrench Merch
W

There are 16 different basic branches in the Army, each
with a different operational specialty (e.g., finance,
infantry, aviation, transportation). Branch assignments are
very important because they largely determine an officer's
occupational specialty and Army career path. Officers are
trained, promoted and given the majority of their
assignments within their branches.

Branch assignments are made when an officer is first
commissioned. Officer preferences are taken into account in
the assignment process; however, Army manpower requirements
take precedence. Therefore a number of officers each year
begin their careers in a branch they did not select.
Mismatches may also occur even when officers get the branch
they originally requested. Inaccurate or incomplete
information about the various branches may lead to poor
initial choices. In other cases, interests and aptitudes

may change, producing person/branch mismatches at a later

45

point in time. Schein (1978) notes that one of the critical
tasks employees face in the early stage of their careers is
to reassess their original decisions to pursue a certain
type of work. Initial choices may be re-evaluated in light
of the individual's developing "occupational self-concept"
and the opportunities and constraints of the organization.

Opportunities for internal transfers must also be
considered in attempts to assess person/job, or in this
case, person/branch fit. Jackofsky and Peters (1983)
suggest that individuals who are unhappy with their current
job situation look for alternatives within the organization
before they look for outside alternatives. If desirable
alternatives are available internally, dissatisfaction is
less likely to translate into a decision to leave the
organization. Jackofsky and Peters found support for this
hypothesis when they included a measure of internal transfer
opportunities in their test of March and Simon's (1958) Ease
of Movement x Desirability of Movement turnover model.
Berienale for Hyperhesized Effecte

The person/branch match measure in this research
reflects both current (as opposed to initial) branch
preferences, and perceived opportunities to transfer into a
more desirable branch. The hypothesis tested in the model
is that a good person/branch match leads to more positive
perceptions of the opportunities for a satisfying career.

The logic for this hypothesis is quite straightforward.

Individuals try to choose work that will allow them to

46
”fulfill their interests, meet their needs and express
themselves" (Hall, 1976, p. 11). Branch preferences are
assumed to reflect the perceived fit between opportunities
available in a branch and the individual's occupational
interests and aptitudes.

Occupational and organizational choice theories
emphasize the importance of a good person/job or
person/organization fit for overall career satisfaction
(Hall, 1976: Schein, 1978). Assignment to a branch that is
not compatible with occupational interests or goals should
thus have a negative impact on perceived prospects for a
satisfying career in the organization.

Stumpf and Hartman's (1984) results confirm the
importance of person-job conguence for both satisfaction
and, indirectly, turnover intentions. A path model
specifying perceived person-job congruence ("this is not the
right type of job for me") as an indirect determinant of
turnover was tested in a small sample of recent college
graduates. The results indicated that the effects of
person-job congruence were mediated primarily by work
satisfaction. Despite the substantial correlation between
congruence and turnover intentions (r=-.48), person-job fit
had no direct effect on intentions net of the effects of
work satisfaction and organizational commitment.

A similar model is proposed here: specifically,
person/branch match is expected to influence career

prospects, but exhibit no independent effects on propensity

47

to stay once the effects of the direct determinants in the
model are controlled.
§BEEQIY

In summary, an officer's branch largely determines his
military specialty and the kinds of training and assignments
he is likely to receive. The type of branch an officer
prefers is presumed to reflect a desire for a particular
kind of specialty or career. Consistent with models
suggesting the importance of a good person/job match, it is
hypothesized that officers not in (and not expecting to get
into) their preferred branches will have less positive
perceptions of their career prospects in the Army.

gerrent Work Satisfacrion

milieu

Motowidlo (1983) observed that the relative predictive
power of the different facets of job satisfaction seems to
vary from one study to another. He further suggested that
determining which particular facet will best predict
turnover in any given situation is the central issue in
turnover research. The focus in this research is on
satisfaction with the work itself. Work satisfaction
appears to drive overall job satisfaction, demonstrating
much higher correlations with global satisfaction than other
facets of the job (Ironson, et al., 1989).

Work satisfaction is also a consistent and fairly large
correlate of intentions to remain with an organization

(Cotton 8 Tuttle, 1986: Mobley et al., 1979). Meta-analyses

48
further indicate that work satisfaction is an even stronger
correlate of turnover than overall job satisfaction (Steel 8
Ovalle, 1984).

Most models of the turnover process specify job or work
satisfaction as a direct determinant of turnover intentions
(e.g., Mobley et al., 1979: Price 8 Mueller, 1981: Steers 8
Mowday, 1981). In contrast, the model tested here proposes
that the effects of current work satisfaction are indirect,
mediated through perceptions of career prospects.

o e H o hes ed Ef e

Opportunities for intrinsic satisfaction in the current
assignment are expected to influence expectations regarding
future satisfaction (i.e., career prospects) in two ways.
First, the nature of the work in the current assignment is
likely to be viewed as an indicator of the type of work the
individual can expect in the future. Thus a satisfying,
developmental assignment may lead to expectations that a
career in the organization will continue to be rewarding.
Second, an assignment to a responsible, challenging job may
be construed as evidence that the organization values the
individual and thus will ensure his advancement in the
organization. In either case, a satisfying current
assignment is expected to lead to more positive perceptions
of the prospects for a military career.

This hypothesis is supported by evidence that job
satisfaction is related to attitudes toward and perceptions

of one's career. Strong correlations have been observed

49
between measures of career relevance and job satisfaction
(Graen et al. 1973: Graen 8 Ginsburgh, 1977), career
satisfaction and satisfaction with the work itself
(Schneider 8 Dachler, 1978) and career prospects and job
satisfaction (Jans, 1988).

Reactions to the immediate job are not, however,
expected to influence turnover intentions net of their
effects on career prospects. This hypothesis is based on
the assumption that employees are most concerned with
maximizing their long-term career satisfaction, not their
immediate job satisfaction. In most occupations, and
professional level occupations in particular, individuals
typically hold a number of different jobs or assignments
over the course of their careers. An assignment in an
organization may last months or years, but it can generally
be construed as temporary stage in the career life-cycle.
This is especially true in the military where job rotation
is the norm and officers can count on being reassigned every
two to four years. Thus, dissatisfaction with the current
job is not expected to be an important direct of a turnover
decision with such long-term career implications.

There is some empirical support for this hypothesis in
data suggesting that work satisfaction fails to exhibit
significant independent effects when examined in conjunction
with more future oriented or global attitudinal variables.

The results of Motowidlo and Lawton's (1984) study are

instructive in this respect. In a large sample of first-

50
term enlisted soldiers, Motowidlo and Lawton examined three
different models of the way job satisfaction, expectations
about the present job, and expectations about civilian
alternatives could influence turnover intentions. To assess
expectations, they had soldiers rate a list of 13 outcomes
in terms of the likelihood that the event would occur (a) if
they re-enlisted in the Army, and (b) if they left the Army.
The outcomes included only factors that were clearly either
favorable or unfavorable and had been identified in
interviews as being important to re-enlistment decisions.
Two types of satisfaction measures were also obtained. One
was an 11 item summed index of satisfaction with various
aspects of the job, and another was a measure of the extent
to which soldiers perceived Army life to have desirable
attributes.

Hierarchical multiple regression and path analyses were
used to determine the best fitting model. Results failed to
confirm the Mobley et al. (1979) hypothesis that job
satisfaction and expectations have independent effects on
turnover intentions. Current satisfaction influenced
reenlistment intentions only through effects on expectations
about future outcomes. Expectations regarding the current
job were also stronger predictors of reenlistment intentions
than perceptions of alternatives.

Motowidlo and Lawton's (1984) interpretation of the
results was that present affective states may bias beliefs

or expectancies about the future. It is not necessary to

51
assume that affective bias is responsible for these effects,
however. It is often noted that "the best predictor of
future behavior is past behavior". Individuals who are
treated poorly or underutilized in their current assignments
may simply see no reason to believe that things will be
different in future.

Hom, Katerberg and Hulin (1979) also examined
alternative models of the factors responsible for reenlist-
ment decisions. In their research on Army National
Guardsmen they compared the predictive power of commitment,
satisfaction, and reasoned action (i.e., Fishbein-Ajzen)
models of the determinants of the decision to reenlist. The
commitment model was tested using the Occupational
Commitment Questionnaire (Porter et al., 1974). The job
satisfaction model included five job facet satisfaction
measures and one organizational satisfaction item. For the
test of the Fishbein-Ajzen model, subjective norms were
assessed by having members indicate whether or not "people
whose opinions are important to me and whose opinions I
value think I should reenlist in the Guard at the next
opportunity". Attitudes toward the act were measured by
having members rate "reenlisting in the Guard at the next
opportunity" on three semantic differential scales.

A comparison of the results of three separate
regression equations indicated that the variables derived
from the Fishbein-Ajzen model accounted for more variance in

reenlistment intentions than the organizational commitment

52

and job satisfaction measures. More relevant to the
mediating hypothesis proposed here, however are the
additional analyses conducted to examine the effects of work
satisfaction in the context of more global attitudes toward
the act of leaving and subjective norms. The correlations
between the satisfaction measures and reenlistment
intentions dropped sharply (e.g., from .55 to .19 for work
satisfaction) when the effects of the Fishbein model
predictors were partialled out.
EQEEQIY

The results of several studies indicate that current
work satisfaction is related to satisfaction with, or
perceptions of career opportunities. In addition, two
multivariate studies suggest that work satisfaction may not
be an important predictor of turnover intentions when more
global, future oriented measures are also included in a
model. Thus, despite the general assumption that current
work satisfaction directly influences turnover decisions,
there is some empirical support for the mediational model
proposed here. Utility maximization principles also suggest
that current satisfaction should be less important when, as
is the case in the military, the organizational career is
characterized by a variety of short-term, temporary jobs.

The relative importance of present versus future
oriented concerns is often difficult to ascertain because
the global measures typically employed tend to blur the

distinction between present and future outcomes. As Ironson

53
et al. (1989) observe, the more global or general the
measure of job satisfaction, the more likely it is to tap
long-term, as well as current, evaluations of the costs and
benefits of the employment situation. The use of context
and time specific measures of work satisfaction and career
prospects in this research should allow the independent
effects of current and future oriented variables to be more
clearly identified.
ior ree Or'enta '
'o 'ew

"Prior career orientation" reflects respondents' early,
pre-entry career plans. The measure used here is
retrospective, asking respondents to indicate to the best of
their recollection if they were planning on or leaning
toward a military or civilian career when they were (a)
still in pre-commissioning training (i.e., USMA, ROTC): and
(b) first commissioned as officers.

Results of several longitudinal studies show that
measures of intentions and commitment obtained prior to, or
immediately after entry are related to later turnover
intentions and behavior. It is proposed here that these
effects are not direct, but rather are mediated through
intervening variables. Specifically, it is hypothesized
that pre-entry career intentions influence subsequent
turnover intentions only through effects on career prospects

and organizational identification.

54

The hypothesis is based on research and theory on the
reasons individuals initially select certain organizations,
and arguments that the goals and values that produce pre-
entry intentions also affect organizational identification
and subsequent perceptions of career prospects.

omm I e ' ove

Several longitudinal studies have demonstrated that
early, or pre-entry attitudes and intentions are remarkably
stable over time. Research in civilian settings has focused
on organizational commitment, and studies in military
populations have examined the relationship between early and
later intentions to stay in the military.

In two civilian studies of recent college graduates,
organizational commitment measures were obtained on the
first day with the organization (Porter, Crampon 8 Smith,
1976) and after only one month on the job (Meyer 8 Allen,
1987). These early commitment measures were strongly
related to subsequent measures of both commitment and
turnover. Bateman and Strasser (1984) also found evidence
for the stability of commitment over time: earlier
commitment was the best predictor of commitment measured
five and twelve months later. Furthermore, although
commitment and job attitudes (e.g., satisfaction) were
correlated within and across time periods, job attitudes had
no apparent influence on subsequent commitment once the

effects of prior commitment were controlled. In all three

55
studies, then, early commitment to the organization was
strongly related to later commitment.

The military literature in this area focuses on early
intentions or expectations with regard to turnover rather
than commitment. In their research on Marine Corps
recruits, Mobley, Hand, Baker and Meglino (1979) asked
recruits about to begin basic training whether or not they
expected to complete the program. This variable was
included along with a variety of satisfaction, expected
utility and desirability of alternatives measures in a
regression equation to predict training attrition. Results
indicated that the recruit's expectation that he would
complete training was the best predictor of actual program
completion. The results of the follow-up study were even
more impressive. Youngblood, Mobley and Meglino (1983)
found that pre-entry intentions to reenlist for a second
tour were a powerful predictor of actual re-enlistment
decisions four years later.

Navy researchers (Bachman 8 Blair, 1976: LaRocco 8
Jones, 1980: LaRocco, 1983) have obtained similar results
regarding the predictive power of early intentions and
commitment. LaRocco and Jones (1980), for example, found
that the reenlistment intentions of inexperienced sailors
(sailors with less than six months of service who had never
been to sea) were correlated almost as strongly with re-
enlistment two to three years later as intentions measured

after an eight to ten month deployment.

56
e o ized ffects

ee ' tion 1 a ' . Evidence
that early attitudes and intentions predict subsequent
retention have prompted military researchers to re-examine
assumptions about the importance of situational determinants
of retention decisions. Youngblood et al. (1983) concluded
that "individual and anticipatory socialization differences
exist prior to entry, ... and such differences predict
turnover" (p. 514). In a similar vein, Bachman and Blair
(1976) speculated that actual experiences in the military
may be less important to retention decisions than "deep-
rooted perceptions and ideology related to the military
life-style and mission" that exist prior to entry. Meyer
and Allen (1987) also interpreted their results as
confirming Mowday's (1982) hypothesis that individuals enter
organizations with different propensities to become
committed.

These interpretations are consistent with person-
environment fit theories of organizational choice (Hall,
1976: Schein, 1978). Schneider and Hall, for example, have
long maintained that individuals tend to select careers and
organizations that are consistent with important facets of
their identity (Hall, 1976: Hall, Schneider 8 Nygren, 1970:
Hall 8 Schneider, 1972: Schneider, 1987: Schneider, Hall 8
Nygren, 1971). These authors, among others (Ashforth 8
Mael, 1989) further assert that individuals who select

organizations on the basis of value congruence enter with a

57
propensity to identify with the organization. This
proposition is consistent with the results of their own
research as well as the post hoc explanations others have
offered for the stability of commitment and intentions over
time (Bachman 8 Blair, 1976: Meyer 8 Allen, 1987: Youngblood
et al., 1983).

In organizations like the military where there is a
distinctive mission and ideology, self-selection on the
basis of value congruence is particularly likely to occur
(Ashforth 8 Mael, 1989: Hall et al., 1970). College
students enrolled in ROTC, for example, are more likely than
non-ROTC students to perceive the Army as a prestigious
institution, view military service as a duty, and value the
opportunity to perform as a leader (Gilbert 8 Wilson, 1983).
Similarly, high school students who foresee a career in the
military not only rate the military higher in terms of job
opportunities and work conditions, they are also much more
likely to have conservative, pro-military values (Bachman,
Siegelman, 8 Diamond, 1987).

McAllister and Smith's (1989) results are also
consistent with the value congruence model. Their survey of
young officer cadets in the Australian Defence Force
indicated that "institutional" or reasons for joining the
service (e.g., to do something worthwhile for my country,
Comradeship in the Defence Force) were strongly related to
military values (e.g., duty must come before personal

interests, military service is a 24 hour a day

58
responsibility). Both institutional reasons and military
values were strongly correlated with expected years of
service.

In the present research, the hypothesized influence of
prior career orientation on identification is expected to
stem from the military or institutional values that lead
some individuals to choose a military career in the first
place. Both theory and research suggest that these values
are relatively enduring and likely to foster subsequent
identification with the military.

0 entat'on r ros . The
distinction between occupational and value congruence
motives parallels the distinction military sociologists make
between ”occupational" and "institutional" orientations
among service members (Moskos, 1977: Wood, 1982). Officers
with an institutional orientation tend to see the military
almost as a "calling" and endorse a host of values
associated with importance of military service and military
traditions (Wood, 1982). In contrast, members with an
occupational orientation tend to view military service as
simply a job, and personal career goals and financial
considerations tend to be more central to their career
decisions than the "service ethic".

Although institutional values have been invoked most
Often as the factor responsible for pre- and post-entry
military career intentions, instrumental and occupationally

oriented motives should not be overlooked. As noted

59
earlier, individuals also base entry decisions on the
perceived fit between occupational interests and abilities
and the career opportunities offered by the organization
(Hall, 1976: Schein, 1978).

The military offers some unique career opportunities:
certainly some individuals join because they want to drive
tanks, fire guns, jump out of airplanes or procure weapons
systems for a living. Assuming some stability in
occupational goals, prior career intentions based on a
desire for a uniquely military career should be positively
related to subsequent perceptions of Army career prospects.

For others, military service is a means to an end
rather than an end in itself. In exchange for a limited
period of service, soldiers can receive civilian educational
benefits, as well as military training and leadership
experience. Individuals with no desire to remain in the
military often join because it is instrumental to their
civilian career goals (Blair, 1975: McAllister 8 Smith,
1989). Blair (1975), for example, found that many high
school students not interested in a military career would be
willing to serve three to four years in exchange for
financial assistance for college.

When prior career intentions reflect primarily
instrumental motives for joining, the same motives and
civilian career goals should influence later perceptions of
one's career prospects in the military. Once the effects of

prior intentions on career prospects and organizational

60
identification are taken into account, prior intentions are
not expected to have any additional direct effects on
subsequent turnover intentions.
Summary

It is proposed here that the influence of prior career
plans on subsequent career intentions can be captured by the
two intervening variables in the model. Pre-entry career
intentions are assumed to reflect personal values and
occupational interests that in most cases, at least, are
stable over the first few years in the organization. These
underlying goals and values are expected to be responsible
for the relationhips expected between prior orientation and
later career prospects and organizational identification.
When the values of the organization are the basis for the
initial attraction, there is a greater propensity to
identify with the organization. Thus early commitments
influence subsequent propensity to stay, in part at least,
through identification. When occupational or instrumental
motives shape early plans, these same considerations should
influence subsequent perceptions of the prospects for a
satisfying career in the Army.

These hypotheses offer a mediational explanation for
the observed effects of early commitment and intentions on
later intentions. If prior career intentions are not
related to career prospects and identification, assumptions
about the stability of pre-entry values and career goals

will be challenged. The lack of a relationship could also

61
be explained by a marked change in pre- and post-entry
perceptions of the organization (e.g., values and career
opportunities). Should prior intentions exert direct, as
well as indirect effects on propensity to stay, two
interpretations suggest themselves. Perhaps, as Pfeffer and
Lawler (1980) argue, and other studies suggest, commitment
is self-maintaining (Batement 8 Strasser, 1984: Meyer 8
Allen, 1987: Porter et al., 1976) and at least partly
independent of organizational experiences and expectations.
Another possibility is that pre-entry intentions are shaped
by factors other than specific career goals and
institutional values. Subjective norms, in the form of
parents' wishes or friends' evaluations may influence early
decisions, then lose their power or be replaced other
reference group (e.g., family, peers) norms.

A problem with the test of this portion of the model is
the retrospective nature of the prior orientation measure.
There is no way to assess the extent to which faulty
recollections or current attitudes and intentions have
biased the measure. Bias stemming from current attitudes or
turnover intentions is likely to inflate the correlations
between these variables and the prior orientation measure.
Therefore, even strong correlations between prior intentions
and the intervening identification and career prospects
variables can only provide weak support for the hypothesized
antecedent effects. However, correlations between the

retrospective and current career intentions measures should

62
also be inflated if retrospective bias is a strong factor.
Since the model predicts that prior intentions will re; have
independent effects on current intentions, the analyses can
still provide a fairly strong test of the "no direct effect"

hypothesis.

QEQEQELQDQI_§BEBQ£§
Waist:

Operational support in this model taps perceptions that
the organization is effective in providing the information,
resources and feedback individuals require in order to
perform their jobs effectively. It is proposed that
perceptions that the organization is supportive will have a
positive impact on both identification and career prospects.
These linkages have not been tested in previous research,
however, similar variables (organizational dependability,
service effectiveness, appropriateness of policies and
practices) have displayed positive relationships with
commitment and satisfaction measures.

Operational support was selected for inclusion in the
model because this dimension of organizational functioning
is expected to be particularly salient in military settings.
There is a constant emphasis on "mission accomplishment" in
the military and recent operations are likely to have
heightened the importance of effective performance (surveys
were administered one month after Operation "Just Cause" was

launched in Panama, but prior to Desert Storm).

63

Individual performance is also constantly under
scrutiny. Yearly performance evaluations are the primary
basis for promotion decisions, and just one poor evaluation
can seriously limit training, assignment and advancement
opportunities. Under the current "up or out" policies,
junior officers not selected for promotion to captain and
major are forced to leave the Army. The dual emphasis on
mission accomplishment and individual performance is
expected to "sensitize" officers to the support they receive
from the organization. When the expected level of
operational support is not forthcoming, it is expected to
have a negative impact on both perceived career prospects
and organizational identification.

DeCotiis and Summers (1987) invoked a similar
"sensitization" concept to explain an unexpected result in
their research on restaurant managers. Satisfaction with
the time demands of the job had a stronger impact on
organizational commitment than any other type of
satisfaction, including overall job satisfaction. In
explaining these results, the authors noted that this firm
"had a published goal/value that stresses the importance of
its employees having a life that balances work, social and
family demands" (DeCotiis 8 Summers, 1987, p. 465). In
reality, however, the organization demanded extremely long
hours (averaging 65 hours per week) from its managers.

The authors suggested that the public emphasis on the

balance between work and personal life sensitized employees

64
to the time demands of the organization. The strong time
demands/commitment relationship was attributed primarily to
discrepancies between stated organizational values and
actual practice.

-9 .,; . 0 .1, a 01: , ionin- a . cuu'tmei

Measures of organizational functioning somewhat related
to the one proposed here were used in studies by Jans (1988)
and Buchanan (1974). For his research on the careers of
officers in the Australian Defence force, Jans (1988)
developed an eight item measure (alpha=.72) of "service
effectiveness”. Items tapped fairly global perceptions of
organizational effectiveness, focusing on the level of
morale in the organization, perceptions that units are well
run, the sense of purpose in the organization, and the
quality of the personnel management system and top level
management. Multiple regression analyses indicated that
service effectiveness was one of eight significant
predictors of career involvement.

Buchanan (1974) developed an organizational
dependability measure he described as reflecting the
individual's perception that the organization had come
through with promised rewards, and had done the things it
said it would do for the individual. In Buchanan's sample
and three subsequent samples where this measure was used,
organizational dependability was a significant, independent
predictor of organizational commitment (Meyer 8 Allen, 1987:

Spencer 8 Steers, 1980: Steers, 1977).

65

Other studies have examined employees' perceptions of
the appropriateness of organizational policies and
practices. Parasuraman and Futrell (1983) assessed
satisfaction with company policies and support along with
satisfaction with the job, co-workers, supervision, pay,
promotion and development. They found that satisfaction
with company policies and support (.41) was second only to
job satisfaction in terms of the strength of the bivariate
correlation with turnover intentions (.47).

In two other studies, attitudes toward company policies
and practices were examined as antecedents of organizational
commitment. In an unusual study of the referents of
commitments in organizations, Reichers (1986) asked
community service providers to report (a) what they thought
should be done in a series of hypothetical situations, and
(b) what they thought top management would say was the
appropriate course of action. Results indicated that
congruence between top management and employee judgements
was related to organizational commitment.

Lee and Mowday (1987) tapped a similar construct. They
asked employees from a number of different banks how
consistent quality standards and operational procedures were
with their own values and professional judgements. When
organizational commitment was regressed on a very large set
of potential predictors, the measure tapping consistency
between organizational and personal standards was the second

strongest determinant.

66

In summary, variables reflecting perceptions of how the
organization functions have demonstrated independent effects
on commitment and turnover intentions in multivariate
studies. Taken together, these studies suggest the
plausibility of the operational support/organizational
identification linkage in the model.
WW

One way the lack of effective operational support might
influence identification is through frustration. In testing
their work facilitation hypothesis, Peters and O'Connors
(1980: Peters, O'Connors 8 Rudolph, 1980) found that
inadequate resources led to greater frustration, stress and
dissatisfaction. Jans (1988) extensive interviews with
military officers confirmed that many officers become
frustrated by what they perceive to be deficiencies in the
quality of their service's management. Interviews led Jans
to conclude that when management deficiencies thwart
personal effectiveness, career involvement tends to decline.
When frustration mounts, individuals may try to psychologi-
cally distance themselves from the organization, thereby
reducing identification. Discrepancies between stated
values and the level of support for performance may also
lead to lower levels of commitment (DeCotiis 8 Summers,
1987) or identification.

The hypothesized effect of operational support on
career prospects is based largely on logic suggesting a

possible relationship between operational support,

67
individual performance, and promotability. An individual's
ability to perform his job effectively is likely to be
contingent on the level of operational support provided by
the organization. If support is lacking, and performance is
hindered by lack of information or resources, an officer's
promotion prospects may be affected. In addition,
perceptions that the organization is ineffective in
supporting its members may lead to doubts about the
organization's ability to utilize members effectively and
provide appropriate assignments.
SBEEQIY

In the present research, the perceived level of
operational support provided by the organization is proposed
as an antecedent variable for two reasons. First,
perceptions of organizational effectiveness are likely to be
especially important in the military because of the constant
emphasis on "mission accomplishment" and performance.
Second, this variable is similar to other constructs (e.g.,
dependability, organizational support, organizational
practices and policies) shown to be determinants of
organizational commitment or intentions to stay.

Perceived operational support is expected to influence
propensity to stay indirectly through effects on perceived
career prospects and organizational identification.

Evidence that variables like service effectiveness and
organizational dependability are related to commitment

suggests that organizational support is likely to influence

68
identification. The proposed operational support/career
prospects relationship is based on the argument that the
lack of effective support may adversely affect performance,
leading to negative assessments of promotability and career

prospects in general.

Inepirationel Leadership
Miss
The present model proposes that leadership plays an

important role in the development of organizational
identification. Specifically, it is hypothesized that
exposure to an "inspirational" leader in the current
assignment will be associated with a greater tendency to
identify with the organization. This hypothesis is based on
both theoretical propositions about the processes underlying
identification, and the limited empirical literature
addressing the role of leadership in relation to commitment
and turnover.

In an important book on leadership, Bass (1985)
attempts to expand current thinking about leadership beyond
the traditional exchange and expectancy models of
motivation. He contends that by thinking about leader-
subordinate relationships primarily in terms of path-goal or
cost-benefit models of influence we are missing some of the
more important effects leaders can have on subordinates.
Bass (1985) describes a type of "transformational"
leadership observed in the "charismatic movers and shakers

of our time" (p. 10). Charismatic or inspirational leaders

69
"relate the work and mission of their group to strongly held
values, ideals, and aspirations shared in common by their
organization's culture...they paint for their subordinates
an attractive picture of what the outcomes of their efforts
could be” (p. 40). Bass goes on to say that through their
influence, transformational leaders imbue work with a higher
meaning, and arouse "enthusiasm, excitement, emotional
involvement and commitment to group objectives" (p. 40).
Inspirational leaders are thus expected to enhance the
organizational identification of subordinates through the
organizational values, norms, and vision they transmit.
Ashforth and Mael (1989) suggest that the growing interest
in symbolic management and transformational leadership in
civilian organizations reflects in part, at least, the
recognition that identification with the organization can
occur through identification with a charismatic leader. The
strong emphasis on motivational leadership in the Army also
appears to reflect this assumption.
e ' ov

As Graen (Graen, Liden 8 Hoel, 1982) observed, very few
studies have explored the role of leadership in the turnover
process. Typically, if leadership is addressed at all in
turnover studies, measures of satisfaction with one's
supervisor are simply correlated with turnover. This
approach sheds little light on the types of leader behaviors
that affect withdrawal decisions, and fails to address the

possible mediating processes underlying leadership/turnover

7O

relationships. The perspective adopted by Graen and
Ginsburgh (1977) suggests that employees are linked to their
organizational roles through "leader-member exchange"
relationships as well as the nature of their work.
Specifically, employees in relationships characterized by
mutual trust, attention, sensitivity and support are
expected to be more attached to their organizational roles,
and hence less likely to consider leaving the organization.

Two studies provide empirical support for Graen's
model. In a sample of predominantly clerical employees
(Graen 8 Ginsburgh, 1977) and a small sample of systems
analysts and computer programmers (Graen, Liden 8 Hoel,
1982), the quality of the leader-member exchange
relationship was an important predictor of turnover.

Krackhardt, McKenna, Porter and Steers (1981) also
found that supervisory interactions with employees were
related to turnover. First, measures of supervisory support
exhibited significant correlations with the length of time
bank tellers intended to stay at the bank. Second, compared
to branches in a control group, turnover rates were
significantly lower in branches where supervisors
implemented an experimental program of individual and group
meetings with tellers.

Although these studies indicate that leader behaviors
are directly related to turnover, there is also evidence
that the effects of leader behavior on turnover may be

mediated by organizational commitment. Michaels and Spector

71
(1982) tested a causal model of the determinants of
turnover. Their results indicated that (a) both job
satisfaction and commitment had direct effects on intentions
to leave: and (b) leader consideration had stronger
independent effects on organizational commitment and job
satisfaction than "met expectations" and the "motivating
potential” of the job.

Bateman and Strasser (1984) found that measures of
leader reward behavior were strongly related to concurrent
measures of organizational commitment. In addition, cross—
lagged regression analyses suggested that leader punishment
behavior had small significant negative effects on
commitment measured five months later. DeCotiis and
Summers (1987), on the other hand, found that none of their
three measures of leader behavior (including consideration)
were independently related to either commitment or
satisfaction when more global perceptions of organizational
climate were included in a regression equation.

Eel; Imperrenee and Qommitment

Bass (1985) contends that one of the most important
ways inspirational leaders "transform" subordinates is by
increasing their self-confidence and imparting in them a
sense of the importance of their jobs. Buchanan's (1974)
research suggests that employees who feel that they are
important to the organization are more likely to feel

committed to the organization. Specifically, Buchanan asked

72
managers from eight different organizations to indicate how
important they felt they were to the organization.

Multiple regression was used to examine the effects of
"felt importance" and 12 other work experience variables on
organizational commitment (defined as identification,
involvement and loyalty). Results indicated that perceived
importance to the organization was the best predictor of
organizational commitment. In two subsequent multivariate
studies, felt importance to the organization was also a
significant, independent predictor of organizational
commitment (Spencer 8 Steers, 1980: Steers, 1977).

These studies do not address Bass's contention that
leaders play an important role in making employees feel more
important. The results do, however, suggest a plausible
explanation for how an inspirational leader might enhance
organizational identification.

SQEEQIY

In summary, the present model proposes that
inspirational leaders influence propensity to stay
indirectly through effects on organizational identification.
Results from several studies indicate that leader behaviors
are often related to organizational turnover. In addition,
there is also empirical evidence that leadership is related
to organizational commitment. Taken together, these studies

provide some support for the proposed model linkages.

73
Potential Determinants Omitted from the Model
Wigs

A number of models include employee perceptions of
alternatives as a factor likely to influence turnover
decisions (e.g., Mobley et al., 1979). In models derived
from expectancy theory, the focus tends to be on the geeliry
of employment alternatives, or expectations regarding the
positive attributes associated with alternatives. Typically
the effects of evaluations of alternatives on turnover
decisions are hypothesized to be additive and direct (e.g.,
Forrest et al. 1977: Mobley et al., 1979).

Other models focus on the perceived eyeilebiliry of
alternatives and predict that outside opportunities irrereer
with satisfaction (Jackofsky 8 Peters, 1983: Martin, 1979:
Muchinsky 8 Morrow, 1980: Price, 1977: Rusbult 8 Farrell,
1983: Rusbult 8 Lowery, 1985) or intentions (Steers 8
Mowday, 1981) to determine turnover. In other words,
dissatisfaction is expected to lead to turnover only if
satisfactory alternatives are perceived to be available.

The explanations for these effects are intuitively
appealing, yet measures of the availability or quality of
alternatives have consistently failed to demonstrate the
hypothesized independent or interactive effects (Arnold 8
Feldman, 1982: Bluedorn, 1982: Hom, Griffeth 8 Sellaro,
1984: Hulin et al., 1985: Michaels 8 Spector, 1982).

Multivariate studies in military settings also indicate

that perceptions of alternatives explain little independent

74
variance in turnover intentions. Steele (1987), for
example, found that the probability of finding a good
civilian job did not influence retention intentions in any
of the four occupational subgroups of Army and Marine Corps
officers he examined. Schmidt (1982) obtained similar
results in a sample of Naval officers: a measure of civilian
job prospects was the least important and least consistent
(across subsamples) predictor of career intentions of any
variable in the model. Ashcraft (1987) found that
perceptions of civilian job prospects had significant, but
very small effects on retention intentions.

One reason for the unexpectedly small effects of
perceived alternatives may be the lack of variance in
perceptions. Wood (1982) noted that on the whole, officers
with both military and civilian career orientations tend to
have positive perceptions of their alternatives in the
civilian employment market. This optimism is also apparent
in ARI's 1988 survey of junior officers (conducted in a time
period with a fairly string economy). Only 9% of the male
officers indicated that it would be "difficult" or "very
difficult" to ”find a good civilian job right now
considering both your own qualifications and current labor
market conditions". None of the correlations between this
variable and items reflecting propensity to leave, Army
career prospects and feelings about the organization

exceeded .11, and the mean correlation was less than .06.

75

Similar results were obtained in the 1985 D00 Survey of
Officers and Enlisted Personnel (Griffeth et al., 1986).
Responses to an item about civilian job prospects were
nearly all positive, and not systematically related to
career intentions. In fact, when the junior officer
subsample was trichotomized on this variable, officers who
reported the beer prospects for a good civilian job were the
most likely to intend to make the Army a career.

These latter results are consistent with Jans' (1988)
observation that outside job opportunities may be a measure
of the service member's value to the military as well as his
value to potential civilian employers. In other words, the
officers with the best civilian career prospects may also be
the officers with the most promising careers in the
military, confounding the effects of internal and external
career oppportunities. Wood (1982) offered another
explanation for inconsistent effects. On the basis of
interviews with junior Air Force officers, Wood concluded
that external employment opportunities affect retention
decisions only for the small proportion of officers with
clearly technical/ specialist, civilian career orientations.
Limited civilian opportunities in particular specialty areas
may constrain these officers to stay in the military. Other
post hoc explanations for the lack of predicted
relationships have also been offered. Some, for example,
suggest that perceived alternatives operate by influencing

perceptions of, or reactions to the current job (Hulin et

76
al., 1985: Rhodes 8 Doering, 1983). An alternative
explanation is that negative job experiences affect
evaluations of alternatives. Employees who dislike their
present jobs may simply assume that other jobs are much
better (i.e., ”hope springs eternal..."): or employees
determined to leave an organization may lower their
standards so that a greater number of alternatives are
deemed acceptable.

A study of Marine Corps training attrition (Mobley,
Baker, Hand 8 Meglino, 1979) suggests the plausibility of
this alternative explanation. Recruits who disliked the
Marine Corps and subsequently left were more positive about
their civilian job prospects than the recruits who completed
training. The traditional explanation for this effect is
that those who left were "pulled" away from their jobs by
exceptionally good alternatives. In this case, however,
recruits who left the organization had lower general
aptitude scores than those who stayed, and were much more
likely to be high school dropouts. With these dubious labor
market qualifications, it is highly unlikely that the
quitters left because of the attractive job opportunities
available to them, but not the high school graduates who
stayed. Negative experiences in the organization may have
led those who wanted to leave to perceive anything outside
the military as a relatively desirable alternative.

Other have suggested that we focus attention on

concrete rather than perceived alternatives. Real

77

alternatives, in the form of specific job offers, may affect
turnover (but not desire or intention to leave) by allowing
those who are inclined to leave to act on their desires
(Bowen, 1982: Hon et al., 1984). Typical measures of the
quality or availability of alternatives may simply be too
subjective or ambiguous to capture the effects of real labor
market constraints and opportunities. If an employees has
been actively searching for a job, perceptions of
alternatives may reflect concrete opportunities. In other
cases, however, positive perceptions may reflect wishful
thinking, or the fact that the current job is unbearable and
ery alternative looks good. Further work on the
operationalization and determinants of perceptions of
alternatives is required before the role of this variable in
the turnover decision process can be clarified. In the
meantime, an examination of the moderating effects of job
offers on the relationship between turnover intentions and
behavior might be a fruitful avenue for future research.
SBEEQEY

In spite of the recent attention devoted to the role of
perceived alternatives in turnover decisions, theoretically
derived hypotheses about the role of perceived alternatives
in turnover research have not been well supported. Revised,
post hoc models, particularly those concerned with the
effects of alternatives on intenrions to leave are
speculative at this point, and there is no clear empirical

support for the independent, causal effects of perceived

78
alternatives on turnover intentions. The ambiguity
surrounding the role of perceived alternatives in turnover
research led to the decision to omit this variable from the
model. For exploratory purposes, however, the additional
variance explained by a single item measure of civilian
opportunities will be examined.

0 t ' ec De 'n

In the proposed model, the effects of the direct
determinants are presumed to be additive. No moderating
effects are proposed, and the interaction terms implicitly
hypothesized to be nonesignificant could be viewed as
potentially relevant omitted variables.

Tenure is the most plausible source of interaction
effects. As noted earlier, the effects of attitudinal
variables on propensity to stay may be much weaker when
tenure related investments (i.e., "golden handcuffs") bind
employees to the organization. Given the restricted range
of experience in the present sample, however, tenure is not
expected to have significant moderating effects.

Conceptual arguments for the plausibility of other
interaction effects could also be advanced, and in fact are
implied in explanations of some of the proposed effects.
For example, career prospects may have stronger effects on
retention decisions when identification is low. However,
most efforts to find empirical support for logically deduced
interactions are unsuccessful, and at this point, there is

no empirical basis for positing interaction effects.

79
Nevertheless, exploratory tests for interactions are

appropriate and included in the preliminary analyses.

Model Specification and Research Design Issues
W

A critical assumption in path analysis is that the
residuals of the variables in a causal path are not
correlated. James (1980) observes that

”this assumption implies that the causal variables
in a theoretic path equation should be unrelated to
unmeasured causes of the dependent endogenous variable
inasmuch as the unmeasured causes are included in the
error term" (p. 416).

If an important causal variable has been left out of a
causal chain, the estimates of the path coefficients will be
biased. This is what is referred to as the unmeasured, or
omitted variables problem.

In his discussion of the unmeasured variable problem in
path analysis, James (1980) notes that this problem cannot
be completely solved because we can never be sure certain
that ell the relevant causes of an endogenous variable have
been identified. He observes that:

"Consequently, the operative question is not whether
one has an unmeasured variables problem but rather
the degree to which the unavoidable unmeasured
variables problem biases estimates of path
coefficients and provides a basis for alternative
explanations of results" (James, 1980, p. 415).

James goes on to identify the conditions under which

the unmeasured variable problem is per likely to be a

significant source of bias. He points out first, that from

80
a practical standpoint, if an unmeasured variable is only a
minor cause of the dependent variable, bias is not likely to
be serious. Second, an unmeasured cause must be correlated
with a measured cause for bias to result. Yet if the
correlation is either very small or very high, the biasing
effect is not likely to be serious. A very high correlation
between an unmeasured cause and a measured cause suggests
that the unmeasured cause is either redundant with, or
linearly dependent on the measured cause. When this is the
case, the unmeasured variable is unlikely to make a enigge
contribution to the prediction of the dependent variable,
therefore will not seriously bias path coefficients of the
measured causes. Similarly, a trivial cause will not
account for enough variance in the endogenous variable to be
a significant source of bias.

There is no simple way to avoid excluding relevant
causes in a model. Practical constraints and goals of
parsimony and elegance in scientific theory preclude simply
incorporating every conceivable causal factor in causal
models. Efforts to identify and eliminate possible causal
factors through logic, however, are clearly warranted
(James, Mulaik 8 Brett, 1982). The rationale for excluding
perceptions of alternatives rests on the lack of consistent
empirical support for the relationship. Further arguments
for the sufficiency of the variables in the present model

are presented below.

81
°Ln‘ ‘ o, I‘ u ,i fan 0 - ' o--seg ode

In the development of the officer career decision
model, an effort was made to either control for or include
variables likely to make an independent, nontrivial
contribution to the criterion variable, propensity to stay.
The direct determinants in the model were proposed because
they appear to tap the critical components of variables that
often predict turnover intentions or behavior.

The career prospects variable, for example, captures
the expected utility (for career satisfaction, at least) of
staying in the Army. The operationalization of the
construct includes both advancement prospects and intrinsic
satisfaction, outcomes shown in previous research to be
important when individuals make career decisions.

The measure of organizational identification taps what
is hypothesized to be the most critical of the emotional or
psychological bases of attachment to an organization.
Related constructs like value congruence or involvement are
likely to be so highly correlated with identification that
their ommission will not be a serious source of bias. More
global, multidimensional constructs like organizational
commitment are likely to be a linear function of several of
the more specific measures in the model, hence redundant.

The third direct determinant of propensity to leave,
anticipated work/family conflict, is typically not included
in turnover models. Therefore, the inclusion of this

variable in the present model may correct an "unmeasured

82
variable” problem in the current turnover literature. If
family conflicts influence turnover decisions, and at the
same time are correlated with job attitudes, previous
estimates of the effects of job and organizational variables
may have been inflated.

Years of service in the military reflects tenure
related investments. Years of service is also strongly
correlated with age in the military because, with rare
exceptions, new entrants to the organization are all young.
Age is therefore not expected to be an independent source of
variance in propensity to stay. Other potentially important
demographic factors (e.g., gender, marital status and
educational level) are controlled in this research.

The antecedent variables in the model will allow the
potential importance of five additional classes of factors
to be examined. If variations in affective reactions to the
immediate job, perceptions of the organization, experiences
with supervisors, person-job fit, and pre-entry plans or
values directly affect turnover intentions, their effects
should be captured by the antecedent variables in the model.
The analysis strategy, which includes estimating the direct
effects of all variables in the model on propensity to stay,
will reveal whether or not the antecedent variables are more
appropriately conceptualized as direct determinants.

Summary
In summary, the present model includes variables from

each of the major categories of determinants identified in

83
the turnover literature. Different operationalizations of
variables reflecting perceptions and attitudes in the areas
of work, organization, career and family are likely to be
strongly correlated with, and hence largely redundant with
the measures specified in this model. Variables tapping
perceived alternatives are omitted from the model because
they typically add little explanatory power to turnover
intention models and the source of these perceptions is
unclear. The lack of empirical support for moderating
effects also led to the decision to omit interaction terms
from the model. The effects of these omitted variables will
be examined in exploratory analyses, however.
Sample Restrietipps
Dalessio, Silverman and Schuck (1986) recently
suggested that ”more attention should be given to possible
differences in the turnover process among members of
different occupations or groups within the organization" (p.
257). In designing the present research, possible group
differences were an important consideration. Although Army
officers are a fairly homogeneous group in some respects,
there are several personal and military characteristics that
might influence the retention decision process. For
example, men and women, married and single officers, and
officers in the early and later stages of their careers are
likely to consider different factors, or weight factors
differently when they make decisions about staying in the

Army. Accordingly, the proposed model was developed for and

84

is tested within a specific subgroup of Army officers -
married, male officers in the early stage of their careers.
Once the utility of the model is assessed in this particular
group, future research can focus on assessing the
generalizability of the model across different groups. The
rationale for the sample restrictions is provided below.
mm

The branch restriction limits the sample to officers
currently assigned to one of the Army competitive branches
(i.e., combat arms, combat support, combat service support
branches). The small percentage of officers in special
branches (e.g., doctors, lawyers, nurses, and chaplains in
the Medical Corps, Judge Advocate General Corps, and
Chaplain Corps) were excluded from the analysis sample
because of their markedly different backgrounds and career
paths.
Gender

Different factors may also be important for female
officers because of the unique family pressures and career
constraints they are likely to face. Women in the Army, for
example, are excluded from the combat arms branches, and
hence have fewer opportunities for important leadership
positions. Most importantly, however, the majority of
married female officers have spouses who are also in the
military (Teplitzky et al., 1988). The present model is not
designed to address the unique set of career and family

complications (e.g., incompatible specialties, joint

85

domicile, child care in the event of deployment) that might
influence the career decisions of officers in dual military
career marriages.
W

The exclusion of single officers was designed to allow
the importance of family factors to be assessed. Although
single officers may anticipate work/family conflict if they
remain in the Army, family considerations are likely to be
most salient for officers who are currently married. The
focus on married officers is also consistent with the Army's
efforts to understand the implications of the changing
demographic composition of the Army. Young soldiers are
more likely than ever before to be married, and the impact
of family considerations on retention decisions is a major
concern of Army policy makers.
Q§I§§I_§§§Q§

The sample is also limited to officers who are still in
the early stage of their careers, specifically officers who
were commissioned two to six years ago (1984 through 1987).
This restriction is important for both theoretical and
practical reasons. First, inexperienced employees have
little basis for their assessments of the organization.
Their attitudes may reflect immediate reactions to temporary
situations, or global impressions formed prior to entry.
The attitudes and perceptions of the respondents in this
sample, however, are based on at least two years of

experience, and thus are likely to be more accurate, stable,

86
and differentiated. Once attitudes have crystallized, one
can have more confidence that observed effects are stable
and attributable to actual experiences in the organization,
rather than pre-entry expectations.

The career stage of the analysis subsample also makes
this group an especially appropriate sample for a test of a
turnover or career decision model. Most officers in these
year groups are approaching or have just reached the point
at which they are first eligible to leave the Army (i.e.,
the end of their three, four, or five year active duty
obligations). Officers near the end of their obligations
are likely to have started thinking about what they will do
once they are free to leave. At the same time, officers who
have completed their obligations are not likely to have
dismissed the possibility of leaving solely because of the
time they have invested toward retirement. This should
allow the determinants of voluntary, relatively
unconstrained retention decisions to be more clearly
identified. In mixed tenure samples, the effects of career,
organizational and family factors on junior employees may be
obscured by the overwhelming importance of economic factors
for the more senior employees (Jans, 1988).

There may also be differences in the way employees who
are heavily invested in an organization evaluate and respond
to work experiences. Some theorists (O'Reilly 8 Caldwell,
1981: Pfeffer 8 Lawler, 1980: Salancik 8 Pfeffer, 1978)

argue that commitment to a course of action may influence

87

subsequent attitudes. If this is the case, even statistical
controls for the effects of tenure or investments on
turnover decisions will not allow unambiguous interpretation
of the causal effects of noneconomic factors.
W

A final research design consideration is a practical
one. Officers commissioned between 1984 and 1987 are still
in the critical "retention window". These officers are the
logical target for Army force management strategies, whether
the goal is to encourage or discourage the retention of
certain types of officers. With the current and projected
force reductions, the career decisions of these officers
will largely determine if the Army falls short of, meets, or
excedes its reduced requirements for captains and majors in
the years ahead. Testing the proposed model on a
subpopulation that is easily identifiable and meaningful
from a force management perspective should increase the
utility of the results for Army leaders.

Summery end Qentral Researeh Quesriops

Ten years ago, Schneider and Dachler (1978, p.3)
observed that:

"The limited and frequently confused understanding

of organizational participation and withdrawal may

well be the result of having looked for determinants

of these behaviors only within the narrow context of

isolated work or individual difference variables".

The situation has improved since 1978. Contemporary

turnover researchers include a broader spectrum of variables

in their analyses and are increasingly likely to test

88
multivariate models of the turnover process. At the same
time, however, the more holistic approach to turnover
research advocated by Schneider and Dachler (1978) has not
been widely adopted. Schneider and Dachler conceptualized
turnover as a response to a mplriegipepeippel affective
state, and suggested that one's long term work career and
extra-work (primarily family) environment might be as
important as the immediate work environment in determining
an individual's propensity to leave an organization. This
broader perspective on the antecedents of turnover has
guided the development of the model proposed in the present
research.

Another important consideration in turnover research is
the nature of the psychological processes or motivational
forces implied in the specification of variables (Weiner,
1982). Accordingly, an effort was made to identify
different types of underlying processes, and ensure that
effects due to affective reactions, instrumental motivations
and normative pressures can all be captured by variables in
the model.

The theoretical model proposes that an officer's
propensity to leave the Army is a direct, additive function
of perceptions of career prospects, identification with the
organization, anticipated work/family conflict, and years of
service. The rational decision processes implied in the
expectancy theory framework (e.g., Mobley et al., 1979) are

assumed to underlie the effects of career prospects and

89
years of service on propensity to stay. Organizational
identification, on the other hand, is assumed to influence
propensity to stay through more subtle, psychological needs
and processes, including needs to preserve one's self-esteem
and maintain a coherent self-concept. The effects of
work/family conflict on retention decision may stem from
both the implied costs of conflict for families, and the
desire to comply with norms surrounding an individual's
family roles. The differential effects of the direct
determinants may point to the relative importance of
different psychological processes in turnover decisions.

The model also includes variables expected to have only
indirect effects on propensity to stay. The effects of
person/branch match, current work satisfaction, prior career
orientation, operational support and inspirational
leadership are hypothesized to be mediated through
perceptions of Army career prospects and/or organizational
identification.

The primary questions addressed in this research can be
summarized as follows:

1. Do the variables in the model account for a
significant, and practically important proportion of the
variance in propensity to stay?

2. Do the four variables proposed as direct
determinants of propensity to stay have significant

independepr effects on the dependent variable?

90

3. Do the two intervening variables in the model
mediate the effects of the five antecedent variables, or do
these variables have direct, independent effects as well?

The first and second research questions basically
address the utility of the constructs proposed as direct
determinants of propensity to leave. The third question
addresses the validity of the proposed causal linkages from
the antecedent to the endogenous variables in the model.
Path analysis is used to examine the predictive utility of
the direct determinants in the model and the validity of the
hypothesized indirect causal linkages in the model.

The test of the present model is expected to contribute
to literature on both military retention and organizational
turnover in general, by suggesting the relative importance
of career, organizational and nonwork factors in retention
decisions. At a broader level, the results of this research
may suggest the appropriateness of developing integrated
turnover models that simultaneously consider the effects of:
(a) utility expectations with respect to both economic
factors and long-term career goals, (b) anticipated family
costs of staying in a particular job or organization, and
(c) one's social self-concept and both internal and external

norms related to organizational membership.

METHOD
W

The sample for this research consists of a subset of
the officers surveyed in December, 1989 for the Army
Research Institute's Longitudinal Research on Officer
Careers (LROC). The sampling strategy was based on
procedures used for the first LROC survey in 1988.
MW

For the 1988 sample, individuals in the target
population were identified from records in the Officer
Master File - the personnel database maintained by the
Army. The target population consisted of active duty Army
officers commissioned between 1980 and 1987 from either the
Reserve Officers' Training Corps (ROTC) or the United States
Military Academy (USMA) at West Point. A stratified random
sampling approach was adopted to ensure adequate sample
sizes for subgroup analyses of USMA and female officers.
The target numbers of USMA and ROTC male and female officers
were randomly selected from the population of officers
within each commissioning year group. Approximately 1000
officers were selected from each year group.

The 1989 LROC sample included all of the officers from
the 1988 sample who were still in the Army, plus additional,
randomly selected officers within each stratum (e.g., ROTC
males commissioned in 1980) to replace those who had left
the Army. The final target sample consisted of just over

1000 officers per year group (1980-1988).

91

92
'n's ro ed 3

Mailing labels with current home addresses were
obtained from the Military Personnel Center in October,
1989. Survey packets were mailed to officers' homes during
December 1989 and January 1990.

Each survey packet included a machine scannable survey
booklet, a return envelope, instructions for returning the
survey, and cover letters from the Commander of the Army
Research Institute (ARI) and the Deputy Chief of Staff for
Personnel. The letters summarized the overall objectives of
the research project, encouraged voluntary participation and
guaranteed that individual responses would be held in strict
confidence. Respondents were asked to use the envelope
provided to return the survey directly to the civilian
contractor responsible for survey distribution and data base
management.

In April, letters with updated addresses went out to
all officers who had not yet responded. The letter stressed
the importance of representative data and asked officers to
please complete the survey or notify the contractor (using
the postcard provided) if they needed another survey. Over
500 second surveys were mailed out, mostly to officers who
indicated that they had never received the first mailing.

By the end of May 1990, the survey cut-off date for
this set of analyses, 4900 surveys had been entered into the

database for an overall response rate of 47%.

93
W

Population statistics for officers in the 1984-1987
year groups (those still on active duty as of the end of
fiscal year 1989) were examined to determine how the
composition of the sample differed from the population.
Table 1 shows the breakdown of ROTC and USMA male officers
by year group in the population, the LROC target sample
(i.e., officers who were mailed surveys), and the final 1989
respondent sample.

ROTC respondents comprised 6% (n=755) of the ROTC male
officer population in the 1984-1987 year groups. The
overall response rate for ROTC officers was 42%, with only
minor variations across year groups.

For USMA males, response rates ranged from a low of 36%
in 1986 to a high of 43% in 1984, with an overall response
rate of 39%. The 500 USMA respondents represent 15% of all
USMA males in the 1984-1987 year groups.

The most notable difference between the population and
the final LROC sample is the overrepresentation of officers
commissioned from USMA. USMA officers comprise 40% of the
1989 LROC sample, but only 21% of the population of officers
in the 1984-1987 year groups. In light of this discrepancy
and the demographic and military differences associated with
USMA and ROTC commissions, exploratory analyses were
conducted to see if the variables in the model operated

differently in the two groups.

94

 

 

 

 

 

Table 1
1989 IKJC
POpulation and Sanple Statistics for
mammmleOffioersbyYeaerup
W
_ PUpulatim Target Respondent Response
Year FY89 Sanple Sanple Rate
Group N % N % N % %
1984 2738 23% 505 28% 200 26% 40%
1985 2853 24% 421 23% 176 23% 42%
1986 3056 26% 417 23% 180 24% 43%
1987 am 273 4.5.5 Zéi .122 23 .453
Totals 11893 100% 1798 99% 755 99% 42%
USMA Male Officers
POpulatim Target Respondent Response

Year FY89 Sanple Sanple Rate
Group N % N % N % %
1984 617 19% 276 21% 120 24% 43%
1985 844 26% 330 26% 128 26% 39%
1986 890 27% 352 27% 128 26% 36%
1987 §2§ 2&3 §2§ 2.53 124 223 §§3
'Dotals 3249 100% 1286 100% 500 101% 39%

 

Note. USMA Officers = 21% of population, 40% of sanple:
1984-85 Year Groups = 47% of population, 50% of sanple.

95
W

There were a total of 757 officers in the LROC 1989
sample who met the criteria for inclusion in the sample for
this research. All married, male, 1984-1987 USMA or ROTC
graduates in one of the basic branches were included in the
analysis sample. The USMA (41%) versus ROTC (59%) split in
the analysis sample is nearly identical to that in the total
LROC 89 sample. The ROTC group includes 25% of the total
sample who were ROTC scholarship recipients and 34% who were
ROTC non-scholarship officers.

Just over half of the officers in the sample were
captains (52%) and the remainder were first lieutenants.
Almost two thirds (64%) of the officers are in combat arms
branches, with 25% in combat support and 11% combat service
support branches.

In terms of year groups, the analysis sample has a
higher percentage of officers from the 1984 year group (29%)
than either the population or the total LROC sample. This
is largely due to the fact that a higher percentage of
officers in the earlier year groups are married and thus
eligible for inclusion in the analysis sample.

The large majority of officers (79%) had been in the
Army for three to six years: only 9% had more than six years
of active duty service. Approximately one fourth had
completed their active duty service obligations (almost all
within the past year), and close to one third had less than

one year remaining in their obligations.

96

With respect to personal characteristics, the analysis
sample consists largely of 25 to 29 year olds (82%). The
mean age is 27. All of the officers are college graduates,
and almost 20% have had at least some graduate level work.
Just under half of the officers in the sample (47%) have
children, and 16% are expecting a child.

The wives of these officers are well educated, and
typically plan on pursuing their own careers. The majority
of the wives (56%) had college degrees, and all but 14% of
the officers said that their wives planned to get additional
education. Although only 38% of the wives were currently
working full-time, fully three fourths of the officers said
that their wives had long-term plans for a career (as
opposed to plans for just occasional jobs or temporary
work). Fourteen percent (14%) of the officers had wives who
either were (5%) or still are in the military (9%).

Of the 757 respondents, complete data were available on
all model variables for 714 (93%) of the officers. This
sample of 714 officers was used in the test of the model.

Measures

With the exception of "years of service", multi-item
a priori scales were proposed for each variable in the
model. In some cases, items or entire scales were used in
the 1988 LROC survey. Other items and scales are new, and
were included in the 1989 survey specifically to measure
constructs in the career decision model. When reliabilities

(Cronbach's alpha) are available for a comparable subgroup

97
from the 1988 survey they are reported in parentheses. The
exact wording of each item can be found in Appendix A.

All scales used in the analyses are composites created
by simply averaging responses to the component items.
Negatively worded items were recoded before item analyses
were conducted and scales were computed. Items were also
recoded where necessary so that higher scores on all scales
reflect "more" or a higher level of the construct. For
example, higher scores on Propensity to Stay reflect a
stronger probability of staying, whereas higher Conflict
scores indicate greater anticipated conflict.
W

The propensity to stay scale consists of two items:
intentions with regard to leaving the Army, and plans for a
military versus civilian career. The items, response
alternatives, and response frequencies are shown in Table 2.
Responses to the two items were averaged to obtain scale
scores. For the six officers with missing data on one of
the two items (no one had missing data for both items)
propensity to stay was measured using the single item for
which there was a valid score.

Scores on propensity to stay ranged from 1 to 5.5
(higher scores reflecting a stronger propensity to stay),
with a mean of 3.72 and a reliability of .93 (LROC 1988
reliablity was .92). The distribution of scale scores is
somewhat skewed (.53) because of the high proportion of

officers intending to make the Army a career.

98
Table 2

Frequency Distributicns for the Two Propensity to Stay Items

 

1. Which of the following best describes your current

career intentims?
(6) PlantostayintheArmybeyond
20 years

(5) Plan to stay in the Army until
retiranent at 20 years

(4) PlantostayintheArmybeycnd
my obligation but am undecided
about staying until retirement

(3) I am undecided whether
will stay in the Army upm
oanpletion of my obligaticn

(2) I will probably leave the Army
upon oarpleticn of my obligaticn

(1) I will definitely leave the Army
upon ccnpleticn of my obligation

2. RightnowIam

(5) Planning on an Army career

(4) leaning toward an Army career
(3) Undecided

(2) leaning toward a civilian career

(1) Planning an a civilian career

Missing

'Ibtal

'2

165

175

192

67

69

84

757

IZ

241
154
132

90
136

Ice

21.8%

23.5%

25.4%

8.9%

11.1%
.3%

100%

3
31.8%

20.3%
17.4%
11.9%
18.0%

4 .5%

757

100%

 

99
a s o e s' S a

Eerepnzprenen_neren. A dichotomous person/branch match
index was constructed from items asking what branch the
officer is currently in, what branch he would like to be in,
and whether or not he intends to, and expects to be able to
transfer into a more desirable branch. A score of "1",
indicating a good person/branch match was assigned if the
officer was either in the branch he wanted or expected to
get into that branch. The 34 officers with missing data on
any of the three items used to construct the index (31
officers failed to indicate what branch they would like to
be in) were also assigned the modal value "1". All other
officers were assigned a score of "0". Sixty-five percent
(65%) of the officers were either in, or expecting to get
into (8% of the total) their preferred branch.

Prior gereer Orientation. The prior career orientation
measure consists of two items asking about career plans at
two earlier points in time. Respondents were asked to use
the 5-point ”planning on an Army career" to "planning on a
civilian career" response scale to indicate, to the best of
their recollection, what their plans were before they began
pre-commissioning training (i.e., USMA or ROTC) and when
they were first commissioned. The mean for these two items
was 3.77, indicating that most officers had been planning on
an Army career. The reliability for these two items was .83

(1988 LROC alpha=.85).

100

Qnrrenr_flerk_§er1ereerpn. Four items were used to
measure satisfaction with the nature of the work in the
officer's current assignment. One item is a global work
satisfaction measure ("How satisfied are you with the kinds
of work you do in your current assignment?") based on a 5-
point satisfaction scale. The other three items ask
respondents to use a 5-point "very good" to "very poor"
scale to rate opportunities for intrinsic work satisfaction
(learn/develop skills, do interesting work, exercise
initiative) in their current assignment. The mean of 3.83
indicates that most officers find their current work
assignments satisfying, interesting and developmental. The
reliability of the scale was .84 (1988 LROC alpha with only
the last three items was. 83).

Inspirational_Leader§hin- The measure of inspirational
leadership was designed to reflect the "inspirational"
qualities of the peer leader in the officer's current
assignment. Five items on the Multifactor Leadership
Questionnaire (Bass, 1985) with high loadings on the
"charismatic” or "inspirational" leadership factor were
adapted to describe inspirational leadership in this
research.

After identifying the "best" leader in their current
assignment (e.g., rater, senior rater, NCO, officer not in
their chain of command), respondents were asked to indicate

how often (5-point "always" to "never" scale) this leader

101
inspires enthusiasm, loyalty, and trust, and helps others
understand the unit mission and the Army's goals.

Not surprisingly, the mean for the inspirational
leadership scale was quite high (4.30) and the distribution
was markedly skewed. Only the "best" leader in the current
assignment was evaluated and this leader was typically
characterized as "always" or "usually" able to inspire and
help people. The reliability of this scale was .84.

Opererienel_§nppprr. The operational support items ask
respondents to indicate how effective (S-point "very
effective" to "very ineffective" scale) the Army is in
providing various types of management and operational
support. The seven areas evaluated include organizational
support, communication channels, information resources,
materiel resources, personnel resources, feedback on duty
performance, and recognition for a job well done.

Prior evidence that employees tend to form global
impressions of organizations with respect to support
(Eisenberger et al., 1986) and effectiveness (Jans, 1988)
suggested that these items would form a single factor. The
feedback and recognition items displayed lower overall item—
total correlations than others in the set, however, their
inclusion increased the internal consistency of the scale
(alpha=.80). The mean (3.27) suggests that officers are
neutral to somewhat positive in their evaluations of the

operational support the Army provides.

102

Snnnery. Scale statistics and intercorrelations for
the five proposed indirect determinants of propensity to
stay are presented in Table 3. All a priori items for these
scales increased the internal consistency of the measures,
thus all were retained. Scale reliabilities were high,
ranging from .80 to .84, and means were all above the
midpoints of the scales, reflecting generally positive Army
attitudes and experiences.

The indirect determinants are not strongly correlated.
Only two correlations exceed .20: work satisfaction exhibits
correlations of .36 and .24 with operational support and
inspirational leadership. Prior career orientation is not

related to the other antecedent variables.

Table 3

Ekzdetflzmisth:anilmmenznzedathxrsfer
Inrfirectlxmenmummms

 

N DEED SD 1 2 3 4 5

 

1. BranchMatd'l 757 .65 .48 --

2. Work Sat. 749 3.83 .87 .17 (.84)

3. Prior Orient. 755 3.77 1.05 .04 .02 (.83)

4. q). Support 752 3.27 .61 .18 .36 -.01 (.80)

5. Insp.Idrship 743 4.30 .58 .03 .24 .02 .13 (.84)

 

Ikme: Scauarelfinmlitksz«armbmjvs:iume)auecmithecfiegmruq
correlations over .10 are significant at p<.01.

103
n e s' o a

A priori scales were proposed for three of the four
variables hypothesized to be direct determinants of
propensity to stay (years of service is a single item
measure). However, because of the centrality of these
variables in the model, the common response format for the
items, and uncertainty about the empirical distinctiveness
of the constructs, factor analyses were conducted prior to
final scale construction.

The a priori item sets and the final scales constructed
on the basis of the factor analysis are described below.
For convenience, final scale reliabilities are reported
here. The factor analysis results and reasons for dropping
items are reported in the first part of the Results section.
Item names are included in parentheses below to facilitate
reference in the next section to the factor analysis tables.

gereer_£reepeere. The proposed career prospects scale
was designed to tap expectations regarding promotion
opportunities and the likelihood of getting career enhancing
and satisfying assignments should the officer decide to
pursue an Army career. The one global satisfaction item (5-
point ”very satisfied" to "very dissatisfied" scale) reads:
”All in all, how satisfied are you with your career
prospects in the Army?" (CARSAT). The other six items are
more specific and perceptual in nature. Opportunities for
advancement in the officer's branch (OPPADV) are assessed

using a 5-point "excellent" to "very limited" scale. A

104
scale ranging from ”I'd have a strong advantage" to "I'd be
at a strong disadvantage" measures how competitive officers
believe they are with regard to promotions and schooling
opportunities (HOWCOM).

The four remaining items employ a 5-point "strongly
agree" to "strongly disagree" response scale. These items
address the officer's confidence or belief that he will (a)
be promoted as high as his ability and interest warrant if
he stays in the Army (AGHIGH): (b) get the kinds of
assignments he needs in order to be competitive for
promotions (AGASGN): (c) get assignments that match his
skills and interests (AGSKIL): and (d) get ahead in the Army
doing the kinds of work he likes best (LIKWRK).

All items except the last were retained for the final
measure, resulting in a six item scale with a mean of 3.48
and a reliability of .82 (1988 LROC alpha with all seven
items was also .82).

Id n ' ic io . Items proposed for the
identification measure include seven of the eight items in
Meyer and Allen's (1984) Affective Commitment scale, with
the word "Army" substituted for "organization". One item (I
would be very happy to spend the rest of my career with this
organization) was excluded from the scale because of its
conceptual overlap with the dependent variable.

The remaining items are consistent with Ashforth and
Mael's (1989) definition of identification as "the

perception of oneness with or belongingness to a group,

105
involving direct or vicarious experience of its successes or
failures” (p. 34). Specific items refer to the personal
meaningfulness of the Army (ARMEAN), sense of belonging to
the organization (ARBLNG), feeling like the Army's problems
are one's own (ARPROB), feeling like "part of the family" in
the Army (ARPART), taking pleasure in talking about the Army
(ARTALK), feeling an emotional attachment to the Army
(AREMOT), and the belief that one could easily become as
attached to another organization as the Army (ARATCH). All
items used a 5-point "strongly agree" to "strongly disagree"
response format.

One item (ARPART) was dropped on the basis of the
factor analysis, resulting in a six item scale with a mean
of 3.60 and an alpha of .79. In previous research using
four different samples, reliabilities of .88, .84, .88 and
.74 were obtained for the full eight item scale (McGee 8
Ford, 1987: Meyer 8 Allen, 1984).

Antieipeted WQIBZEQEILY Qonfliet. This measure was
designed to tap the extent to which the officer anticipates
conflicts between family demands or the kind of family life
he would like, and the demands of an Army career. Three of
the proposed items were included in the 1988 LROC Survey
(alpha=.84): "If I were to make the Army a career, I could
maintain the kind of balance I want between my work and
personal life" (WRKBAL): "The demands of an Army career
would make it difficult to have the kind of family life I

would like" (CARCMD): and "If I were to stay in the Army, I

106
could provide my family with the opportunities and
experiences I think are most important" (CARFAM). An
additional item was added to the 1989 survey to make
conflict more salient in the scale. This item states: "I
foresee a lot of conflict between the demands of my job and
my family life if I make the Army a career" (CARCON). A 5-
point agree/disagree response format was used for all items.

The factor analysis led to the decision to drop the
item referring to opportunities for important experiences
(CARFAM). The final three item scale had a mean of 3.17 and
a reliability of .86.

Xeere_pj_§eryiee. The fourth direct determinant, years
of service, is simply the response to the question, "How
many years of active duty service have you completed
(including any enlisted or warrant officer time)?".
Responses ranged from 2 to 15 years with a mean of 4.52.
Because the sample was restricted to officers with no more
than six years of epmmissioned service, active duty service
exceeding six years reflects prior enlisted or warrant
officer service. Over half (57%) of the sample had three
and five years of service: 9% had more than six and 3% had
more than ten years of service.

finnnery. The scales used to tap the proposed direct
determinants of propensity to stay have fairly high
reliabilities (.79 to .86) and means above the mid-points of
the scales. Officers tend to feel a sense of identification

with the Army and are generally positive about their career

107

prospects in the Army. There is also a slight tendency to
agree that an Army career is likely to produce conflicts
with family life. Scale intercorrelations and correlations
with the antecedent and dependent variables in the model
(Tables 10 and 11) are presented after the discussion of the
factor analysis in the Results section.

e V ' sues

Organizational researchers are increasingly concerned
with the effects of common method variance in survey
research (e.g., Williams, Cote 8 Buckley, 1989). When both
independent and dependent variables are measured using a
self-report, survey methodology, it is difficult to rule out
common method variance as a plausible explanation for
observed relationships. In other words, measures may be
correlated simply because people have a tendency to respond
in a consistent way to questions asked in a particular
manner, regardless of the content.

Common method variance is clearly a potential source of
bias in any survey, however, several factors are expected to
mitigate the problem in this research. First of all,
feedback from interviews and pre-tests indicates that the
career and family issues covered in the survey are important
to the officers in the target population. In addition,
approximately 40% of the 1988 LROC respondents took the time
to write in comments at the end of the survey, and comments

pertaining to the survey itself were nearly all positive.

108
When a survey is interesting and personally relevant,
respondents are likely to respond more thoughtfully.

Second, the questionnaire is divided into a number of
different sections, and section titles and introductions are
designed to focus attention on the content of the items.
Item and response formats are also varied throughout the
questionnaire. The cognitive shifts required by numerous
format changes should inhibit unconscious tendencies to fall
into a set pattern of responding.

The most important consideration with regard to method
variance, however, is the type of bias it is likely to
produce in the test of a theoretical model. For example,
when identical response formats are used to measure
variables hypothesized to be correlated, method variance
will inflate correlations in the hypothesized direction.
Consequently, results construed as support for a conceptual
model may simply reflect method effects.

In the present research, this is less likely to be a
problem because variables hypothesized to be causally
related are not measured using the same response formats.

If method variance is a serious problem, it is the
correlations among the direct determinants that are most
likely to be inflated because of the common agree/disagree
response format. Spuriously high intercorrelations among
these variables should decrease, rather than increase the
likelihood of observing the predicted independent effects on

propensity to stay. Common method variance is thus a less

109
plausible explanation for results that support the proposed
model.
mm

Data analysis proceeded in three steps: (a) preliminary
analyses, (b) the path analytic test of the model, and (c)
exploratory regression analyses. All analyses were
conducted using SPSS-X, release 3.0.

Preliminary analyses included a factor analysis of the
18 items proposed as measures of the direct determinants.
The common factor model was used with prinicpal axis
factoring, and factors were rotated to an oblique solution.
Results of the factor analysis were used to assess the
severity of the common method variance problem, test the
validity of a priori notions about the underlying factors,
and identify complex or ambiguous items. Items with high
loadings on more than one factor were dropped to facilitate
interpretation and preserve the conceptual distinctiveness
of the scales.

Preliminary analyses also included an examination of
scatterplots to assess the linearity of the relationships
among the central variables in the model. In addition,
hierarchical moderated multiple regression analysis (Cohen 8
Cohen, 1983) was used to test for interactions among the
direct determinants of propensity to stay.

The path analytic test of the model included several
steps. First, Ordinary Least Squares (OLS) regression was

used to estimate path coefficients for the causal linkages

110
proposed in the model. Second, the "omitted parameters
test" recommended by James, Mulaik and Brett (1982) was
performed to see if the paths hypothesized to be nnimportenr
(i.e., the paths omitted from the model) are, in fact, non-
significant or trivial. When omitted paths are significant
(and large enough to be of some substantive importance), the
restricted theoretical model is not supported even if all
hypothesized paths are significant. When omitted paths also
markedly alter the coefficients for predicted paths, serious
specification errors are indicated. The overall "goodness-
of-fit" of the model relative to the unrestricted model was
tested by computing the Q statistic (Pedhazur, 1982).

The final step in the assessment of the model involved
an examination of the residuals from the regression. Both
the distribution of the residuals and the demographic
characteristics associated with large prediction errors were
examined.

The exploratory analyses consisted of hierarchical and
moderated multiple regression analyses with propensity to
stay as the dependent variable. These analyses were guided
by the results of the residuals analysis as well as
unresolved issues in the literature. They were designed to
explore both the possible moderating effects of source of
commission, and the contribution to propensity to stay of
plausible determinants not included in the theoretical
model. The exploratory analyses provide a stricter test of

the model and suggest hypotheses for future research.

RESULTS
Preliminary Analyses
a sis ' e eterm' a ts
Anelyeee en A Priori Item Se;

The 18 items proposed as measures of organizational
identification, career prospects and work/family conflict
were factor analyzed to refine the initial conceptualization
of the constructs and identify inappropriate scale items.
Although the common factor model was used for the primary
analysis, principal components analysis was used first to
determine the appropriate number of factors to retain.
Eigenvalues and the percent of variance accounted for by
each component are displayed in Table 4. Both the Kaiser
criterion (retaining factors with eigenvalues greater than
one) and the scree plot indicated that the three factor
solution was best. Together, the first three factors
explain 56% of the total variance in the set of 18 items.

In the second step, the three factors were extracted
using the common factor model with principle axis factoring
(PAF). The common factor model excludes the unique variance
associated with each item from the analysis and is most
appropriate when the primary purpose is to examine the
underlying structure of the items (Ford, MacCallum 8 Tait,
1986). SPSS-X estimates communalities through an iterative
procedure beginning with the squared multiple correlation of
each variable with all other variables. Table 5 shows the

initial and final communality estimates.

111

112

 

 

Table 4
Principal Outpatents Analysis Results:
Original Items for Direct Determinants Scales
Factor Eigenvalue Pct of var CUm.Pct
1 6.25 34.7 34.7
2 2.13 11.9 46.6
3 1.68 9.3 55.9
4 .91 5.1 61.0
5 .80 4.5 65.4
6 .75 4.2 69.6
7 .70 3.9 73.5
8 .61 3.4 76.9
9 .58 3.2 80.1
10 .51 2.9 82.9
11 .49 2.7 85.7
12 .46 2.6 88.2
13 .46 2.5 90.8
14 .37 2.1 92.9
15 .35 2.1 94.9
16 .31 1.9 96.8
17 .31 1.7 98.5
18 .26 1.5 100.0

 

113

 

 

Table 5
Factor Analysis Ocnmmality Estimates:
Original Itats for Direct Determinants Scales
Initial Final
Variable Canmmality Camunality
Career m Itens
CARSAT .55 .60
OPPADV .33 .33
AGIIG-I .42 .46
AGASCN .51 .54
AGSKIL .57 .59
W .29 .26
LD<WRK .40 .40

gm. Identification Items

AEMEAN . 47 . 55
ARBLI‘I; . 34 . 35
ARI'ALK . 26 . 26
AMP . 52 . 63
ARPROB . 32 . 33
ARA‘IU-i . 33 . 33
ARPARI‘ . 47 . 50
WorkZFamily Conflict Itens

WRKBAL . 54 . 62
CARD!) . 60 . 71
CARLIN . 55 . 62

CARFAM . 49 . 53

 

114

Initial communality estimates ranged from .26 to .60
and final communalities (the squared multiple correlation of
each item with each factor) ranged from .26 to .71. Items
in the work/family conflict scale tended to have the highest
communalities, however, variables from all the a priori
scales shared a substantial amount of variance with others
in the set.

Oblique (OBLIMIN) rather than orthogonal rotation was
used in interpreting factors because it was assumed that the
underlying factors would be correlated. The resulting
pattern and structure matrix coefficients are presented in
Table 6. (See pages 2-4 of Appendix A or the description of
measures in the Method section for the actual questionnaire
items corresponding to the variable names used in the factor
analysis tables.)

The pattern matrix values represent the nnigne
contribution of each factor to the variance of each item.
These values support the a priori conceptualization of the
structure of the items. Loadings for the first factor are
highest for Organizational Identification items, factor 2
loadings are highest for the Career Prospects items, and
factor 3 loadings are highest for the Work/Family Conflict
items. Factor loadings for two items, however, (LIKWRK from
the Career Prospects scale and CARFAM from the Work/Family
Conflict scale) suggest that these variables are influenced

by more than one factor.

u .m o N muoooom 1%.-

... .m o F muoooom am.

y .N w P mwouomm "gaumgoonoufi Hannah .ouoz

 

115

 

mo. Nmi mmi 5% mm. moi mm. 2.4%
on. mmi moi .2533 ms. moi oo. am:
on. vmi mmi g oo. oo. mo. 56
oo. omi mmi 9950 mm. 5i mo. 9940
moi om. om. r85: mmi mm. mm. Ema
moi mm. mo. Eng 3. mm. mo. 2825:
5... mm. on. Somme mp. mm. or. 5&0
omi ho. E. 55 mmi oo. oo. 9%
5.: mp. R. g boi mo. Si 5%
5i 2.. No. 9% 5i oo. oo. AU?
5... mp. mm. 5 moi 3.. poi 5mg
omi or. om. a mo... mo... 9‘. E
mm... on. mm. 65% moi 3. E. 94%
mmi hm. mm. g oo.... oo. mm. 62%
mmi mm. mm. gag Poi mo. mm. E
3i oo. mo. 53% moi 3. oo. Emma
omi on. 3.. Zen—Ed no. moi or. g
mNi mm. on. 9% oo. moi mm. 9992
”63.4480 uncommoum .983 033.85 ”63.880 30360.8 .883 o§8>
.m\3 H8980 .95 m} .8980 .90
m N F m N P

fifllllo: Elm 6:51.. ..la: Empolao

 

888 855538 8883 you 908 8:885
"68806.8 5:8 no 83808 0838 88 08388 08595.6. our good

m 0.3mm.

116

Structure matrix coefficients are typically higher than
pattern matrix loadings because they reflect correlations
among the factors in addition to the unique variance shared
by factors and variables. In other words, structure
coefficients show the total correlations between the
variables and the factors. Table 6 indicates that two
variables in addition to LIKWRK and CARFAM are complex:
ARPART and CARSAT both exhibit total correlations of .40 or
more with factors they were not intended to measure.

The complex variables are both a cause and a
consequence of the correlations among the factors. The
Organizational Identification and Work/Family Conflict
factors are most highly correlated (-.47). The Career
Prospects factor has slightly lower correlations with the
other two factors (.40 with Organzational Identification and
-.34 with Work/Family Conflict).

In an effort to simplify the factor structure, the
factor analysis was re-run excluding from each factor the
item with the strongest correlation with a second factor.
The excluded variables were ARPART (Organizational
Identification), LIKWRK (Career Prospects), and CARFAM
(Work/Family Conflict). Examination of the content of these
items suggests that there are conceptual as well as
empirical grounds for deleting these particular items.

c n e for Dro in ems
The ARPART item reads: "I do not feel like I am 'part

of the family' in the Army". The word "family" in this item

117
was intended to refer to the symbolic "Army family". This
may not have been clear to respondents, however. The
organizational identification items followed a long series
of items on spouse and family issues. This context may have
led some to interpret the item as asking if they felt less a
part of their own families because they were in the Army
(e.g., separations or long hours made them feel left out of
family life). This could explain the especially high
correlation (.41) between ARPART and the Work/Family
Conflict factor.

The LIKWRK item from the Career Prospects scale is also
somewhat ambiguous, asking if the officer can get ahead in
the Army "dong the kinds of work I like best". Other items
in the Career Prospects scale are more specific, referring
to things like advancement opportunities in one's branch and
likelihood of getting assignments compatible with skills and
interests. The kind of work an officer "likes best" could
easily depend on its impact on his family life (e.g., "I
don't like work that interferes with family life") or the
way he feels about the Army (e.g., "as long as it's Army
work I like it"). This could explain the considerable
overlap between LIKWRK and the conflict and organizational
identification factors.

Re-examination of the CARFAM item suggests that despite
its high correlations with the other Work/Family Conflict
items, it is qualitatively different. The other three items

in the scale focus specifically on conflicting demands or

118
the balance between work and family life. CARFAM, on the
other hand, reflects an officer's belief that if he stays in
the Army he can provide his family with the opportunities
and experiences he thinks are most important. The item was
intended to tap the opposite of conflict, or the potential
for an Army career to enhance, rather than detract from
family life. However it appears that the item also (or
instead) taps general feelings about the organization.

In each case, dropping the most ambiguous or complex
item from the a priori set appears to enhance the conceptual
as well as empirical distinctiveness of the scales.

Analyses with the Reduced Set of Items

Results of the analyses with the three ambiguous items
excluded are presented in Tables 7 through 9. Eigenvalues
and the percentage of variance associated with the principle
components (Table 7), and the communality estimates for the
common factors (Table 8) are very similar to those obtained
using the original set of 18 items.

The pattern and structure matrices, along with the
factor intercorrelations are show in Table 9. Examination
of the factor intercorrelations indicates that eliminating
the complex items produced the desired effect. The
correlation between the Organizational Identification and
Work/Family Conflict factors was reduced from -.47 to -.42,
and correlations between the other factors also dropped

slightly.

119
Table 7

Principal components Analysis:
Reduced Set of Items far Direct Determinants Scales

 

 

Factor Efignnwalue Pct of var Chm Pct
1 5.01 33.4 33.4
2 2.05 13.6 47.0
3 1.57 10.5 57.5
4 .91 6.1 63.6
5 .80 5.3 68.9
6 .72 4.8 73.7
7 .67 4.5 78.2
8 .55 3.7 81.9
9 .51 3.4 85.2

10 .46 3.1 88.3
11 .43 2.9 91.2
12 .38 2.5 93.7
13 .35 2.4 96.0
14 .31 2.1 98.1
15 .28 1.9 100.0

 

120

 

 

Table 8
Factor Analysis Communality Estimates:
Reduced Set of Items for Direct Determinants Scales
Initial Final
Variable Oamunality Oamnmality
Career mm Items
CARSAT .54 .60
OPPADV .32 .32
AGHIGH .41 .46
AGASQI .50 .57
AGSICIL .54 .57
W .28 .26

(2g . Identification Items

ARMEAN . 47 . 60
ARBLIB . 32 . 34
ARI‘AIK . 25 . 26
AREVUI‘ . 47 . 61
m . 30 . 33
ARA‘ICfl . 32 . 33

Work/Family Conflict Items

WRKBAL .52 .60
CARDD .58 .72
CAREIN .55 .65

 

121

Nm.- - u .m a N $808 “Nvf - H .m a P museums 6m. - u .N a F EBB-N ”6339.85.53 “Bums .302

 

 

 

E. «m..- S.- g 2.. 8. mo.- gas
5. ".Nr wmf 895 2.. 8. 8... 686
mm. mN.- SK 996 E. No.- No.- 9.86
Si om. N. g :. B. 8. g
S.- cm. mN. 86% NF. mm. 8. 8&8
me S. 2. smug FN... E. mo. 9%
a? 2.. 3. gamma 2.- 8. 8. Hog
mm.- 3. mm. g B.- 8. Sr 569
RE ms. .vN. g 8.- 2. Sf 53
8f 5. cm. a Sr 3: 8. a
RF mm. mm. 995 Sf NP. 9.. Emma
mN.- mN. S. 5.2% 8.- 8. mm. 52mm
oNr 5N. S. g Sf oo. 3. g
8.. R. 2.. 24924 8. Si 8. g
R.- 8. 9.. .554 oo. S.- 8. 85%
839.8 388% 383 230.3 829.8 3% . 083 «Eons,
“<3 $8 .95 “<3 .8095 .90
m N F m N P
33...: mfiuumfim fihm: Emfimm

 

 

830m magnum yoga 8m 23H «0 8m 888m
"30mm SE60 mo 5330a 033.30 you 808m: 032m can gown

m manna.

122

The structure matrix coefficients indicate that two
items (CARSAT and WRKBAL) still display substantial total
correlations with other factors: however, there are no
compelling reasons to drop these items from their respective
scales. CARSAT reflects overall satisfaction with Army
career propects and WRKBAL asks officers if an Army career
would allow them to maintain the kind of balance they want
between work and family life. Although perhaps more global
than other items in the career prospects and conflict
scales, conceptually, both items appear to reflect the
constructs they were intended to measure. Both items also
contribute to the reliabilities of their scales. Dropping
CARSAT, for example, would slightly reduce the correlation
between the Career Prospects and Identification factors
(from .38 to .36), but only at the expense of a marked
decrease in the reliability of the career prospects scale
(from .82 to .77).

In summary, the factor analysis results suggest that
common method variance is not a serious problem in this
research. Three clearly distinct factors emerged, and their
loadings were consistent with the a priori conceptualization
of the constructs. Empirical and conceptual considerations
led to the decision to drop one complex item from each
scale, resulting in a simpler factor structure and slightly
lower factor intercorrelations.

The factor analyses also highlighted the negative

relationship between organizational identification and

123

work/family conflict. In light of the overlap between these
two consructs, the name of the work/family conflict scale
was changed to Army/Family Conflict. This label appears to
more accurately reflect the measured construct. Specifying
"Army" instead of "work" as the referent clarifies the
source of the conflict tapped by these items (i.e., Army
career demands rather than specific work assignments) and
suggests the relationship of conflict to feelings about, or
identification with the Army.

c t 's ics and Interco elations

Table 10 displays the means for the final measures of
the direct determinants, scale reliabilities and
intercorrelations, and correlations with the dependent
variable. Correlations between the direct and indirect
determinants are shown in Table 11.

Intercorrelations for the direct determinants range
from .05 to -.41. Years of service is only weakly related
to the other direct determinants. Career prospects,
Army/family conflict and organizational identification are
all moderately intercorrelated (.35 to -.4l). The two most
strongly correlated variables, organizational identification
and Army/family conflict, also display the largest bivariate
correlations (.58 and -.55) with propensity to stay. The
correlation between career prospects and propensity to stay
is slightly lower (.43) and years of service exhibits the

weakest relationship with the dependent variable (.28).

Table 10

124

Direct Determinants: Scale Statistics, Intercorrelations and
Correlations with Propensity to Stay

 

 

N than SD 1 2 3 4 5
Career Pros. 750 3.48 .73 (.82)
Org. Ident. 754 3.60 .69 .35 (.79)
A/F Calflict 750 3.17 .99 -.36 —.41 (.86)
Years Service 757 4.52 1.99 .05 .17 —.14 -——-
Prop. to Stay 757 3.72 1.49 .43 .58 -.55 .28 (.93)

 

Note. Scale reliabilities (Crcznbach's alpha) are on the diagonal,
oonelaticns of .10 or greater are significant at p<.01.

 

 

Table 11

Correlations between Indirect and Direct Determinants

Indirect Determinants

Direct Branch Work Prior Op. Inspir.
Determinants Match Sat. Orient. Suppo Leadership
A/F Ocnflict —.11 -.27 —.10 —.41 —.11
Career Pros. .22 .49 .01 .47 .19
Org. Ident. .10 .34 .18 .34 .22
Years Service .01 .04 .10 .05 —.06
Prop. to Stay .16 .39 .17 .39 .13

 

Note. Correlations .10 and above are significant at p<.01

125

Table 11 shows that within the set of indirect
determinants, work satisfaction and operational support
exhibited the largest correlations with the direct
determinants (ranging from -.27 to .49 when correlations
with years of service are excluded). Branch match, prior
career orientation and inspirational leadership are only
weakly related to the direct determinants, with none of the
correlations exceeding .22. Correlations between the
indirect determinants and years of service are negligible.

The antecedent variables with the strongest
relationship to the hypothesized direct determinants (work
satisfaction and operational support) also exhibit the
strongest correlations with propensity to stay (r=.39 for
both). The path analysis results will indicate whether or
not the intervening variables capture the effects of work
satisfaction and operational support on propensity to stay.

0 cs edast' ' Va ' c

Prior to conducting the path analyses, assumptions
about the linearity of the proposed relationships were
examined. Scatterplots showing the relationship between
each exogenous and endogenous variable in the model were
visually examined. None of the patterns in these plots
suggested non-linear relationships.

Heteroscedasticity also appeared not to be a problem.
The variance in propensity to stay was fairly consistent
across levels of all the independent variables except years

of service. As one might expect, there was less variance in

126
career intentions at the highest tenure level because the
large majority of officers with more than six years of
service intend to stay in the Army. This is not likely to
seriously reduce the predictive power of the model, however,
since the linear trend will still be captured.
Intezngtinns gnong tne Ditegt Qetetninants

The model assumes that the effects of the direct
determinants (organizational identification, career
prospects, Army/family conflict and years of service) are
additive. None of the determinants is expected to have
effects that vary across levels of the other three direct
determinants. The assumption that interaction effects are
not significant was tested through hierarchical moderated
multiple regression analysis (Cohen & Cohen, 1983).

The test for interaction effects involves computing
interaction terms (e.g., ORGID x YEARS) for all combinations
of the four direct determinants. A regression analysis is
then performed in which interaction terms are entered into
the regression after variables representing main effects
have been entered.

Following Cohen and Cohen's (1983) recommendation, a
significance testing strategy analogous to Fisher's
"protected t" test was employed to protect the significance
level and reduce the likelihood of Type I errors. This
involves entering the exploratory variables (i.e., the
interaction terms) as a set in the final step of the

analysis, and assessing the significance of the effects of

127
individual variables only if the overall increase in the
explained variance is significant.

In the present case, the R2 for the set of four direct
determinants was .500 (F=181: df=4,725) and the set of the
six interaction terms resulted in an R2 change of .011.
This increment was not significant at p<.01 (F change=2.66;
df=5,724), suggesting that interactions among the proposed
direct determinants are not likely to be significant,

independent sources of variance in propensity to stay.

Path Analysis
e ’ 'c. e H o 'zed at

The first step in the path analysis involved computing
standardized regression coefficients (i.e., path
coefficients) for each hypothesized causal linkage.
Ordinary least squares (OLS) regression was used to regress
the three endogenous variables on their hypothesized
antecedents. The resulting path coefficients are displayed
in Figure 2, along with the R? for each endogenous variable.
A significance level of p <.Ol is used for all coefficients,
and the sample (n=714) for all analyses includes only those
officers with complete data for all variables.

The results of the propensity to stay regression show
that all of the paths from the hypothesized direct
determinants in the model are significant. The effects are
also substantial, with path coefficients ranging from .17

(for years of service) to .35 (for identification).

128'

Etc—2 mcozcot: tootmo omwoooi of .2 3:22:80 fan. ”m 059“.

    

      

 

._.O_.._n_ZOO
>.._:>.<n_\>2m<

n:I.wmm_O<m-_
-_<ZO_._.<m_me_

  

 

 

    

...mOnEDw
-_<ZO_._.<mm_aO

    
   
   
           

 

84...”. zo:.<o_n_:zwo_ .

4<ZO_._.<N_Z<GmO

 

           
 

 

ZOrbE-ZMEO
mmwm<0 mOEn.

><._.w O...
>._._wzma0md

 

man... u a

. Z
wFOMQwOm—n—

mmmm<0 >_>_m<

ZO_._.O<.-._w_._.<.w

 

vEO>>

 

 

 

NV.

IO._.<_>_
Ioz<mm

  

w0_>mww
Z_ wm<w>

 

 

 

129

Together, the four direct determinants account for
about half of the variance in propensity to stay (R2=.495,
Adjusted R%=.493). Organizational identification and
Army/family conflict have the strongest independent effects
in this sample.

With regard to the intervening variables, six out of
seven of the hypothesized paths from antecedent variables
are significant. Operational support, prior career
orientation and inspirational leadership all demonstrate the
expected effects on organizational identification.

Together, these variables explain 18% of the variance in
identification (R2=.183, Adjusted R2=.l80).

Work satisfaction, operational support and branch match
are significant predictors of career prospects, together
accounting for about 34% of the variance (R;=.338, Adjusted
RA=.334). The predicted path from prior career orientation
to career prospects was not significant.

te a 5 es

A two step hierarchical multiple regression analysis
was used for the omitted parameters test. First, each
endogenous variable was regressed on its hypothesized
antecedents (Step 1). In the second step, the antecedent
variables that were ngt expected to have significant, direct
effects on the endogenous variables were added to the
regressions. The purpose of this analysis was to see if
paths originally excluded from the model (i.e., implicitly

hypothesized to be zero) had significant effects.

130

Results for the regressions with propensity to stay as
the dependent variable are summarized in Table 12. The
first column in the table shows the regression coefficients
and R2 after Step 1, when only the predicted paths are
included. These coefficients are identical to those
displayed in the path diagram in Figure 2. The second
column shows the regression coefficients after Step 2, when
all remaining antecedent variables in the model have been
entered.

In the propensity to stay regressions, the change in
the R2 produced by the addition of non-predicted paths is
small (less than 2%), but significant. The increase in the
explained variance is largely attributable to the effects of
work satisfaction. Work satisfaction, with a path
coefficient of .11, is the only hypothesized indirect
determinant to exhibit significant direct effects on the
dependent variable. Work satisfaction alone accounts for an
additional 1% of the variance in propensity to stay.

The addition of the path from work satisfaction does
not, however, substantially alter the path coefficients or
the relative importance of the other variables. The effects
of organizational identification and Army/family conflict
(.32 and -.28, respectively) are slightly smaller with work
satisfaction in the equation. Yet, these two variables
still have substantially larger effects than the other three

significant predictors.

131
Table 12

Hierarchical Miltiple Regression Results for Propensity to Stay
Using Predicted and Nam-Predicted Paths

 

 

 

Step 1: Step 2:
Predicted Paths Onitted Paths

Independent Variables B Beta B Beta
Org. Identification .76 .35 * .68 .32 *
A/F Ocnflict -.47 -.31 * -.43 -.28 *
Career Prospects .38 .19 * .25 .12 *
Years Service .12 .17 * .12 .16 *
Work Satisfacticn .19 .11 *
Prior Qrientatim . 09 . 07
Op. Support .14 .06
Branch Match .13 .04
Insp. Leadership —.06 -.02
End of Step: R=.704 122:.495 * R=.717 R2=.514 *

R2 change=.018 *

 

Note. Regressicn coefficients in the Step 2 oolunn are final weights with
all variables included.

* Significant at p (.01

132

Tables 13 and 14 summarize results of the omitted
parameters test for the two intervening variables, career
prospects and organizational identification. When all
possible antecedents are included in the regression there
are three additional significant paths to organizational
identification and one new path to career prospects.

The one unexpected path to career prospects comes from
Army/family conflict. Although work satisfaction and
operational support are still the most influential
variables, the weight assigned to Army/family conflict is
higher than weights for two of the predicted paths (branch
match and prior orientation). In combination, the set of
additional variables results in an R2 increase of .021.

With regard to organizational identification, three of
the four additional variables (Army/family conflict, work
satisfaction and years of service) have significant,
independent effects on organizational identification. The
inclusion of these omitted parameters substantially
increases the explained variance (from 18% to 30%). It also
results in a markedly lower path coefficient for operational
support (the coefficient drops from .32 to .15).

Of the three additional variables, Army/family conflict
has the strongest independent effect on identification.
Paths from both conflict and work satisfaction (another
omitted path) are stronger than the paths from the predicted
antecedents. This is further evidence of a poorly specified

set of original predictors.

133
Table 13

Hierarchical mltiple Regressicn Results for Career Prospects
Using Predicted and Non-Predicted Paths

 

 

 

Step 1: Step 2:
Predicted Paths Omitted Paths

Independent Variables B Beta B Beta
Work Satisfacticn .30 .35 * .27 .32 *
Op. Support .37 .31 * .30 .25 *
Branch watch .18 .12 * .17 .11 *
Prior Orientation -.01 -.01 -.02 —.03
A/F Conflict —.11 -.15 *
Insp. Leadership .07 .06
Years Service .00 .01
End of Step: R=.581 R2=.338 * R=.599 122:.359 *

R2 change=.021 *

 

Note. Regressim coefficients in the Step 2 coluun are final weights with
all variables included.

* p (.01

134
Table 14

Hierarchical Miltiple Regressicn Results for Organizational
Identificatim Using Predicted and Nth-Predicted Paths

 

 

 

Step 1: Step 2:

Predicted Paths Onitted Paths
Independent Variables B Beta B Beta
Op. Support .36 .32 * .17 .15 *
Prior Orientation .13 .19 * .10 .15 *
Inspir. Leadership .21 .17 * .16 .14 *
A/F Oaiflict —.18 —.25 *
Work Satisfaction .14 .18 *
Years Service .05 .13 *
Branch Match .00 .00
End of step: R=.428 R2=.183 * R=.544 R2=.296

R2 change=.113 *

 

Note. Regressim coefficients in the Step 2 column are final weights
with all variables included.

* Significant at p <.01

135
o dn s

In the omitted parameters test the focus is on path
coefficients. The "goodness-of-fit" test, on the other
hand, is concerned with the proportion of variance explained
by the two alternative models. The magnitude of "Q" (the
goodness-of-fit index) reflects the predictive power of the
restricted, theoretical model relative to the power of the
full set of potential predictors. This test provides
another assessment of the utility of the model.

The goodness-of-fit index (also called the "generalized
multiple regression coefficient" (Mathieu & Hamel, 1989))
was computed according to procedures recommended by Pedhazur
(1982, pp. 617—628). The first step is to estimate squared
residual path coefficients by subtracting the R? for each
endogenous variable from one. The product of these squared
residual path coefficients is then subtracted from one.
Completing these calculations for both the full and
restricted models produces two values, the ratio of which
gives Q. This index can take on values ranging from zero to
one, with larger values indicating a better fit. Values in
the neighborhood of .90 indicate a reasonable fit between
the restricted model and the data (Mathieu & Hamel, 1989).

The value of Q in the present case was .89, suggesting
a "reasonable" fit. The fact that Q was less than 1.00 was
due primarily to the missing paths from work satisfaction

and Army/family conflict. The R2 for organizational

136
identification was considerably larger when these variables
were included in the set of predictors.

The Q statistic can also be tested for significance by
deriving a value based on sample size and the natural log of
Q. However, as Pedhazur (1982) points out, with large
samples, even very small discrepancies between the full and
restricted models are likely to be significant. Pedhazur
thus suggests that researchers focus on the magnitude of Q
(which is not a function of sample size) rather than tests
of significance.

ot._ 08,: .19 direct __ e . '_o--ns5t . . .

In order to compare the relative importance of the
empirical determinants of propensity to stay, regressions
for each endogenous variable were re-computed with only
significant paths retained. The path coefficients for this
empirically derived model are displayed in Figure 3. These
coefficients (carried to three decimal places) were used to
compute the total and indirect effects reported in Table 15.

Three variables, Army/family conflict, work
satisfaction and operational support have substantial
indirect effects on propensity to stay. When direct and
indirect effects are summed, Army/family conflict emerges as
the most important determinant of propensity to stay.
Caveats in interpreting these effects because of the lack of
a priori causal hypotheses are presented in the discussion

section.

137'

 

Boo—2 om>tmo >=moEaEm .8 3:22:80 Eon. "m 2:9“.

._.0_-_--_ZOO n:I.wmmo<w-_
>-:_>_<n_\>_>_m< -_<ZO_._.<m_n_wz_

 

 

 

 

     

m7-
... VP.
._.mOn_n=._w
-_<ZO_._.<EMn_O

 

89.x. zo_e<o_.._:zmo_

 

       

-_<ZO_._.<N_Z<0mO

 

 

   

ZO_._.<._.2w_mO
mmwm<0 mOEm

><._.m O...
>k_w2mn_0mn_

 

 

   

wH0mawOmn.
mmmm<0 >S_m_<

ZO_._.0<u_w_._.<w
vEO>>

 

3..

 

:.

   

10.2.2
IOZ<mm

 

m0_>m_wm

 

Z_ mm<m>

 

 

138

 

mvo.
oro.

mmo.
woo.

vow.l
mop.l

mvo.

moo.
moo.

vwo.
mmo.
mmo.l

WC WC
see. m:
“C WC
one. me
me e,
wee. NP.
ONe.- em.-
«e.
mm.

mo.
no.
mo.
mo.

FN.
NN.
ov.

me.
mm.

mr.
or.
or.
mm.

mm.
mm.
mm.l

mm.
mm.

QflSMHQUGQQ .QMGH
nuumz SUGmHm
.ucnmfinmchog
£835 do

8g a8“
caeuomumeemmixepz
63.880 a?

muonmmonm gnu
5383353 .95

 

Hmnon. DH DHSB .90.:HSB muocmum
muocmmm HOOHflUGH

muooumm
buoyed Nance “mono oumN

..HH8

 

boom 3 be; so 300mg pooh? Em #083 .395

Hmaoz_amoeueesmnuau

mF manna

139

Work satisfaction also has substantial total effects,
with a small direct effect as well as indirect effects
through both intervening variables. When both direct and
indirect effects are taken into account, work satisfaction
appears to be a more powerful determinant of propensity to
stay than career prospects. Of the remaining antecedent
variables, only operational support had sizable total
effects, also operating through both intervening variables.

' u a s'

The final step in the assessment of the model was the
examination of residuals. First, standardized residuals
were plotted against predicted values to test assumptions
about normality, linearity and homoscedasticity of error
terms. The only residuals analyzed in detail are those from
the regression of propensity to stay on the four direct
determinants proposed in the model (career prospects,
organizational identification, Army/family conflict, and
years in service). The addition of work satisfaction did
not result in any noticeable change in the distribution of
residuals.

Qisttinntign_gfi_tg§ignnl§. The regression of
propensity to stay on the direct determinants resulted in a
distribution of residuals that was not quite normal. First,
there was a larger than expected concentration of residuals
between 0 and +1. This slight deviation from normality
appears to be a function of the distribution of observed

scores. More officers than one would expect (given a

140
normally distributed variable) have extremely high
propensity to stay scores, and fewer than one would expect
have scores in the less extreme "leaning toward the Army"
range. The model tends to predict the less extreme (and
less probable, in this sample) score for many officers
definitely intending to stay, resulting in a cluster of
small positive residuals.

More interesting was the fact that the tails of the
distribution of residuals were not balanced. Most of the
extreme residuals (more than plus or minus two standard
deviations from the mean) were negative. In fact, of the 34
cases with residuals exceeding two standard deviations, all
but six (or 82%) were negative. The extreme negative
residuals typically occurred when officers were definitely
intending to lgnyg the Army, but had predicted scores in the
"neutral" or ”inclined to stay" range. In other words, the
model was less successful predicting decisions to leave the
Army than it was in predicting decisions to stay.

Ch ' s associate w'th ar r siduals. In a
second set of analyses, officers with residuals more than
one standard deviation from the mean were isolated and the
demographic characteristics of these officers were compared
to those of the sample as a whole. The purpose of this
analysis was to see if there were any personal or family
characteristics systematically associated with positive or

negative prediction errors.

141

Officers who were more likely to Iggy; than predicted
were different from the total sample in several ways. First
of all, they were considerably more likely to be USMA
graduates (50% vs. 41% in the total sample) and less likely
to be ROTC non-scholarship graduates (21% vs. 29% in the
total sample). These officers were also more likely to have
time remaining in their active duty service obligations and
to have majored in engineering or a physical science in
college. These latter characteristics, however, are much
more typical of USMA than ROTC non-scholarship officers, and
thus their association with negative residuals may be simply
a result of the relatively high proportion of USMA officers
in this group.

In terms of family characteristics, officers more
inclined to leave than predicted were less likely to have
children (65% childless vs. 53% overall) and slightly more
likely to have highly educated, career oriented spouses
currently working in professional level jobs.

Officers who were more likely to gtny than predicted
were largely similar to the sample as a whole. The primary
difference was that a greater proportion had spouses who
were not currently employed (53% not working vs. 46% in the
total sample) and whose long-term work plans did not include
a career (33% not career oriented vs. 24% in the total
sample). There were also 5% fewer USMA graduates in this

group than there were in the total sample.

142
In summary, the examination of residuals suggested that
family and military characteristics might account for some
of the variance not explained by the attitudinal and
perceptual variables specified in the model. The possible
contribution of these variables to the prediction of

propensity to stay is examined below.

Exploratory Regresions with Propensity to Stay
r e ss'o ffec s

For the first of the exploratory analyses, effects
associated with source of commission were examined. This
variable warrants separate attention for two reasons.

First, source of commission is a meaningful way to group
officers in the Army, capturing differences across USMA and
ROTC officers in academic and military training, as well as
length of obligation and branch assignments. Second, as
noted earlier, USMA officers are overrepresented in the
sample, thus it is important to establish that the model
does not operate differently in this subgroup.

Following procedures recommended by Cohen and Cohen
(1983) hierarchical moderated multiple regression analysis
was employed to test both the direct and possible moderating
effects of source of commission on propensity to stay. This
type of analysis essentially affords a test of the
differential validity of the predictors in the model within

the USMA and ROTC groups (Cohen & Cohen, 1983).

143

Two separate regressions were conducted, one using only
the hypothesized direct determinants of propensity to stay,
and one using all antecedent variables in the model. The
steps for each regression (source of commission, then model
variables, then interaction terms) were identical except for
the number of model variables included in the analyses.

A source of commission dummy variable (coded "1" for
USMA and ”0" for ROTC) was created first to assess the
direct effects of USMA versus ROTC commissions. This
variable had a correlation of -.25 with propensity to stay
(USMA officers are less likely to intend to stay) and
significant correlations with three of the nine antecedent
variables in the model. Being a USMA officer was associated
with more anticipated Army/family conflict (r=.27), less
organizational identification (r=-.19) and slightly fewer
years of service r=-.12). As indicated in Table 16, the
USMA/ROTC dichotomy alone accounted for 6% of the variance
in propensity to stay.

For Step 2, the model variables were entered into the
equation (hypothesized determinants only for Regression 1,
and the full set of model variables for Regression 2).
Entering source of commission into the equation first
partials out the effects of differences in the mean scores
of USMA and ROTC officers on the model variables. Therefore
the regression coefficients for the model variables at Step
2 reflect pooled within group effects, or a kind of

”average" of the effects of these variables within the USMA

144
and ROTC subgroups (Cohen 8 Cohen, 1983). These "averaged"
coefficients (not reported here) are not substantially
different from those obtained when source of commission is
not included in the regression. This does not, however,
rule out the possibility that the model variables have
markedly different effects within the USMA and ROTC
subgroups.

The possibility that the optimal regression
coefficients for two groups are significantly different is
tested by adding interaction terms to the regression (Cohen
8 Cohen, 1983). Because of the 0/1 coding for commission
source, interaction terms for all ROTC officers are "0" and
interaction terms for all USMA officers are simply their
scores on the independent variables. A significant
interaction term indicates that the optimal weight for that
variable is significantly different across groups.

Table 16 indicates that in both regressions, the R2
change with the addition of the entire set of interaction
terms is not significant (Regression 1, F(9,704)=1.62 for
the increment in R2; in Regression 2, F(18,695)=1.67). This
obviates the need to test the individual interaction terms
for significance.

Two conclusions can be drawn from these results.

First, the effects of source of commission on propensity to
stay are basically captured through the attitudinal
variables (particularly Army/family conflict, based on the

bivariate correlations). The cumulative R2 at the end of

145

 

Table 16
Regression Results Testing Source of Commission
Moderating Effects on Propensity to Stay
Variable Entered on Each Step Change R2 Cum R2

 

Regression 1

Step 1: Source of Commission .062 * .062
Step 2: Predicted Determinants .438 * .500
Step 3: 4 Interaction Terms .005 .505

Regression 2

Step 1: Source of Commission .062 * .062
Step 2: All Model Variables .458 * .520
Step 3: 9 Interaction Terms .012 .532

 

3*15 change significant at p <.01; N=714.

146
Step 2 for each regression is less than 1% larger than the
Ig for the same regression when source of commission is
excluded (see Table 12).

Second, because path coefficients for the model
variables are not significantly different across groups, the
same model can be applied to both USMA and ROTC officers.
Thus, the overrepresentation of USMA officers in the sample
does not bias the test of the theoretical model. There are
however, important differences in the means of the model
variables across groups, so descriptive statistics need to
be weighted or reported separately for the two groups to
accurately reflect attitudes in the target population.
W

The residuals analysis suggested several demographic
variables in addition to source of commission that might
enhance the predictive power of the model. Variables
selected for examination in the exploratory analyses include
two spouse career variables (employment status and career
aspirations) and four variables reflecting personal (college
major and presence of children) and military (obligation
status and Army scholarship) characteristics of the officer.

The exploratory analyses also examine the possibility
that three perceptual variables excluded from the model have
independent effects on propensity to stay. Perceptions of
alternatives, for example, were omitted from the model
because there is little evidence, in military samples in

particular, that such perceptions are independent sources of

147
variance in turnover intentions. Two single item,
perceptual measures of alternatives are available, however,
and will be used to test the assumption that these effects
are not important.

The model also failed to include a perceptual measure
of the constraints or difficulties associated with leaving
the organization. Tenure was included as a surrogate for
tenure related investments, but this type of measure is
unlikely to capture constraints stemming from other
investments (e.g., having just bought a home in the area) or
situational factors. Situational constraints might include
such things as the stress or financial demands associated
with the recent or expected birth of a child, a spouse's
involvement in the military community (either socially or
through employment), or a family member's dependence on
military health care. There is an item in the survey that
might tap these factors, and this variable is also included
in the exploratory analyses.

Measures. In the set of officer characteristics, the
the officers who received Army scholarships (ROTC
scholarship recipients and all USMA officers, 66% of the
sample) were given scores of "1" and non-scholarship
officers were given scores of "0" (34%). Scholarship status
largely reflects the Army's evaluation of the academic and
leadership potential of officer candidates. Only candidates

with exceptional high school records (e.g., good grades,

148
high SAT's, participation in sports or student government)
are selected for USMA and ROTC scholarships.

Obligation status was determined by asking "How many
months do you have remaining in your obligated period of
active duty service (including additional obligations
incurred from PCS, military training, civilian schooling)?
Asking officers to include additional obligations incurred
from training or relocation (a permanent change of station,
or PCS move) meant that a number of officers who had
completed their initial obligations would still be included
in the "under obligation" groups. Codes for obligation
status ranged from -2 to +2, positive scores indicating that
the officer still has obligated time to serve. The five
categories include: (-2) obligation completed more than one
year ago (4.5%); (-1) obligation completed within the past
year (19%): (1) less than one year remaining in the
obligation (32%); (2) one or more years to go (38%). A
score of "0" was assigned to the officers (6.5%) whose
obligation status could not be determined because of missing
or inconsistent data.

The variable "children" was coded "1" for officers with
children (47%) and "0" for officers without children. The
"hard science major" variable was coded "1“ for officers
whose college majors fell into physical science, computer
science, or engineering categories (36%), and "0" for all

others.

149

For the spouse variables, employment status was coded
"2" if the officer said that his wife was employed in a
"professional level" job, "1" if the wife was working but
no; in professional job, and "0" if the wife was not
working. Career orientation was measured by asking the
officer to report his spouse's lgng;tg;m work/career
aspirations. Four categories were created from the five
options: (1) not interested in working (10%): (2) interested
in working, but not career oriented (14%); (3) career
oriented, but willing to accept career interruptions (67%);
and (4) career oriented and not willing to accept
interruptions (9%).

The three perceptual measures include two items
measured using a 5-point "very easy" to "very difficult"
scale. One (”easy to find civilian job") asked "How
difficult do you think it would be for you to find a good
civilian job right now, considering both your own
qualifications and current labor market conditions?". High
scores reflect perceptions that it would be easy to find a
job. The second item ("personal/family constraints on
leaving") asked "How difficult would it be for you to leave
the Army in the next year or so given your current personal
or family situation?". Higher scores on this item indicate
greater constraints or difficulties associated with leaving.

The third perceptual item uses a 5-point "much better
in the Army" to "much better in civilian life" response

scale to tap perceptions of pay in "a civilian job you could

150
realistically expect to get". A higher score indicates that
pay is perceived to be better in civilian life.

Means, standard deviations, intercorrelations, and
correlations with propensity to stay for the exploratory
variables are displayed in Table 17. Bivariate correlations
with propensity to stay suggest that scholarship officers,
officers still under obligation and those with a background
in the hard sciences are more likely to intend to leave.
Similarly, officers who believe that pay is much better in
civilian life and who expect that they can easily find a
good job are more likely to plan on leaving the Army.

Officers with children, and those whose wives are less
career oriented are slightly more likely to intend to stay
in the Army. The strongest correlate of propensity to stay
is the perception that personal or family factors would make
it difficult to leave.

The first set of exploratory regression analyses
focused on the additional variance explained by the three
sets of factors. Three hierarchical regressions were
conducted, each with a different ordering of the sets of
exploratory variables. The significant predictors from the
test of the theoretical model (the four hypothesized
determinants plus work satisfaction) are entered first in
each regression.

The results presented in Table 18 indicate that the
officer characteristics and spouse career variables have the

weakest independent effects on propensity to stay.

151

5:12 30.06 um ucmoEuQm 9.8 S. 5.5 “88.5 983mg a .302

 

0v.

1%.- mm.- 5r 2f 2.- 2. mm.- emf 2.; ohm sfimoufiagaooumé

 

283ml, 086868

§>§ :0 3:835

8f mmf mo.- 2.- EV 2. «fi- mm.- 6: mm: 32E\§ .m
- 8. mo. 8. cm. 8f 8. 5. mm. 8.4 .838 Em 5.34.98 .m
- mo. 3. a. mo.- 8. mm. ms. 2.5 now 58 65m 3 swam .1.

355mg 3,330.82

- mm. mo. 85 3. 3. 3.. mtm 833830 38 6
- mof mmf mo. 8. mm. 8: mawa “35.5396 ....
33333398... a

- mo. 3. am. 9.. mm. 8.32 8:38. 3mm .4

- 8.- mn- cm. 3. 58:8 .m

- mm. R; S. 83838 .N

- 5. 8. 638293 E .F

molilflmanon-udallmug-13:-ill. m.-

 

m N. o m w. m N P am amm¥a

 

moafiflg Engage you 98335 at moflmBBm «>388me
t 638..

Table 18

152

Additional Variance Explained in Propensity to Stay

by Three Sets of Exploratory Variables

 

 

Variables Entered on Each Step Change R2 Cum R2
Regression 1
Step 1: Model Predictors .502 * .502
Step 2: Spouse Career Variables .009 * .511
Step 3: Alternatives/Constraints .050 * .561
Step 4: Officer Characteristics .006 .567
Regression 2
Step 1: Model Predictors .502 * .502
Step 2: Alternatives/Constraints .053 * .555
Step 3: Spouse Career Variables .006 * .561
Step 4: Officer Characteristics .006 .567
Regression 3
Step 1: Model Predictors .502 * .502
Step 2: Officer Characteristics .018 * .520
Step 3: Spouse Career Variables .006 .526
Step 4: Alternatives/Constraints .042 * .567

 

1*18 change significant at p <.01; N=710.

153

The alternatives/constraints variables, on the other
hand, account for a substantial increment in the explained
variance even when they are entered after the demographic
and spouse variables. Together the three sets of variables
account for an additional 6.5% of the variance; however, the
alternatives/constraints variables alone add 5.3% to the
explained variance. The variable tapping personal/family
constraints on leaving was by far the most important
variable in the alternatives/constraints set of items. The
constraints item had a path coefficient of .21 (when this
set was entered before the other two), compared to non-
significant coefficients of -.03 and -.05 for the civilian
job and civilian pay items.

When the spouse items were entered into the equation
first, the spouse career orientation item just achieved
significance with a slightly higher coefficient (-.07) than
the employment status item (-.06). Two items were
significant when the officer characteristics were entered
first: scholarship status (-.10) and presence of children
(.07). Only scholarship status remained significant once
the spouse variables were entered.

For the final analysis, an empirically derived model
was tested using only the three consistently significant
exploratory variables: constraints on leaving, Army
scholarship and spouse career orientation. The results of
adding these three variables to the significant model

predictors are presented in Table 19.

154
Table 19

Hierarchical Multiple Regressicn Results for Propensity to Stay
Using Nbdel and Significant Exploratory Variables

 

 

 

Step 1: Step 2:

Predicted Paths Onitted Paths

Independent Variables B Beta B Beta
Org. Identificatim .71 .33 .57 .26
A/F Conflict -.47 -.31 -.38 -.24
Career Prospects .26 .13 .27 .13
Years of Service .12 .17 .10 .14
Work Satisfaction .20 .12 .21 .12
Constraints (11 leaving .29 .22
Army Scholarship -.26 —.08
Spouse Career Orientation —.14 -.07

End of Step: R=.708 R2=.502 R=.750 R2=.562
R2 change=.061

 

Note. Regressicn coefficients in the Step 2 column are final weights
with all variables included; all coefficients are significant p (.01.

155

The model variables accounted for 50% of the variance
in propensity to stay, and the three new variables added 6%
to the explained variance. Most noteworthy is the strong
effect of personal and family constraints on leaving. One
would expect at least a small relationship simply because
the item refers to leaving the Army. The strength of the
effect is surprising, however. The path coefficient for
personal/family constraints is comparable in magnitude to
coefficients obtained for organizational identification and
Army/family conflict. These results suggest that efforts to
identify non-work constraints on employment decisions could
be useful in turnover research.
§BEE§I¥

In summary, the theoretical model accounted for
approximately 50% of the variance in propensity to stay.

All of the hypothesized direct determinants had significant
independent effects on propensity to stay, and only one of
the five antecedent variables (work satisfaction) had
unexpected direct, as well as indirect effects. Including
work satisfaction as a direct determinant of propensity to
stay increased the explained variance in the dependent
variable from 49.5% to 50.5%.

With respect to the intervening variables, the
hypothesized antecedents explained 34% of the variance in
career prospects. The addition of the unspecified path from
Army/family conflict to career prospects resulted in a small

increase (just under 2%) in the explained variance.

156

In contrast, the omitted parameters test for
organizational identification indicated the possibility of
three additional influences (Army/family conflict, work
satisfaction, years of service) on identification. Together
these variables increased the R2 for organizational
identification from .18 to .30. With the addition of the
three significant omitted parameters, all of the
hypothesized antecedents still had significant coefficients,
but their effects were not large. Clearly, several
important determinants of organizational identification were
not specified in the model.

Despite the problems predicting organizational
identification, the goodness-of-fit test suggested that the
overall model fit the data quite well. Residuals analyses
indicated that the largest errors occurred when officers who
were actually intending to leave the Army were predicted to
be neutral or inclined to stay. Officers more inclined to
leave than predicted had several distinctive personal and
family characteristics, the effects of which were examined
in the exploratory analyses.

The supplementary analyses indicated that the direct
determinants specified in the model do not have interactive
effects, nor does source of commission interact with other
variables in the model to determine propensity to stay. The
exploratory analyses also suggested that academic
background, spouse career orientation and personal/family

constraints warrant examination in future research.

DISCUSSION

Weiner (1982) observed that it is unusual to find a
study that simultaneously relates instrumental motivation,
job satisfaction and commitment to an outcome variable so
that their relative contributions can be ascertained.
Scholl (1981) also emphasized the need to conceptually and
empirically distinguish different types of motivational
processes or forces. Yet it was noted in the introduction
that the turnover literature is still largely characterized
by separate streams of research focusing either on
commitment (affective or behavioral) or job satisfaction and
expected utility constructs. Schneider and Dachler (1978)
offered a more content oriented critique of the turnover
literature, noting that the search for antecedents of
turnover is typically limited to specific job or work
context variables.

The model tested here reflects the broader perspective
advocated by these theorists. Variables reflecting clearly
different cognitive and affective processes were
simultaneously examined, and career, organizational, family
and economically oriented factors were proposed as
antecedents of turnover intentions. Results supporting the
importance of these factors have several implications for
the way we conceptualize the determinants of turnover

intentions. These implications are discussed below.

157

158

Qrganizatignal_ldentifi2ation

Organizational identification is defined as a process
whereby the individual's status as a member of the
organization becomes a salient aspect of his or her social
identity (Ashforth & Mael, 1989; Turner, 1987). Strongly
identified individuals tend to experience the organization's
successes and failures as their own and to attribute to
themselves the prototypical characteristics of the
organization. An intention to remain with or exert effort
on behalf of the organization is likely to follow
identification, but it is not a necessary component.

Although organizational commitment is a more commonly
examined antecedent of turnover intentions, the focus on
organizational identification was deemed promising for
several reasons. First, identification is typically viewed
as a critical component of affective commitment (Buchanan,
1974: Meyer & Allen, 1984: Porter et al., 1979), and thus
may capture much of the variance in turnover intentions
typically explained by commitment. The strong bivariate
correlation between identification and propensity to stay in
this research (r=.58) suggests that this is the case. In a
recent meta-analysis the mean weighted correlation between
attitudinal commitment and turnover intentions was slightly
lower (-.52), even when corrected for attentuation (Mathieu
& Zajac, 1990). An advantage of the identification
construct is that there is no intentional component to

confound efforts to compare effects across variables.

159

Second, identification is both conceptually and
empirically distinct from job satisfaction. The correlation
between identification and work satisfaction in this sample
(r=.34) is considerably smaller than the typical correlation
between attitudinal commitment measures and work
satisfaction (the mean weighted, corrected correlation was
.63 in the Mathieu and Zajac (1990) meta-analysis).

Evidence that organizational identification is a
distinctive phenomenon and strongly related to turnover
intentions, even when competing with other important
determinants, is encouraging from an empirical perspective.
There are also theoretical advantages to a focus on
identification. Social identity theory offers a useful
framework for defining the construct and analyzing the
antecdents, consequences and processes involved in
identification (Ashforth & Mael, 1989). Related social
psychological literature on values, group attraction and
group effects may provide further insights into the nature
and causes of the phenomenon.

The focus on identification also highlights the need to
consider organizational characteristics in the study of
turnover. Reichers (1985) attributes the lack of progress
in the commitment literature, in part, to the failure to
consider the "organization" in organizational commitment
research. Identification, on the other hand, is grounded in
theory that emphasizes the importance of the objective and

perceived attributes of the identification target.

160

Some organizations, for example, cultivate a
distinctive image, embody an identifiable value system, or
articulate their mission in terms of some higher social
good. The potential for identification in this type of
organization is especially high, particularly if
organizational training and socialization experiences are
geared toward inculcating and reinforcing organizational
values (Hall & Schneider, 1970: Ashforth & Mael, 1989).

The military epitomizes this type of organization, and
this may account for the strong identification/turnover
intentions relationship observed in this sample. Yet many
public and service organizations (e.g., schools, law
enforcement agencies, advocacy groups) may share at least
some of these attributes. Private firms are also
increasingly concerned with establishing strong, value-
oriented corporate cultures (Rousseau, 1990). Differences
across organizations in distinctiveness, image and prestige,
and socialization practices may account for sample
variations in both the level of identification and the
importance of identification as a determinant of turnover
(Ashforth & Mael, 1989).

The nature of the psychological contracts organizations
establish with employees (e.g., relational vs. transactional
per Rousseau's (1990) distinction) may also be an important
explanatory variable. Rousseau (1990) found that employees
who perceived the contract in relational terms, where

loyalty and to development within the organization are

161

stressed, typically intended to stay longer. Relational
versus transactional contracts are likely to correspond to
"institutional" versus "occupational“ organizational
structures and orientations (Moskos, 1986: Wood, 1982). In
a related vein, the "commitment norm" in an organization may
capture a host of related constructs (Meyer & Allen, 1988).
In a longitudinal, multiorganizational study, the only one
of ten work context variables (assessed at six months in the
organization) to predict later commitment (at eleven months)
was the perceived commitment norm (Meyer & Allen, 1988).

Reichers' (1985) multiple constituencies approach to
organizational commitment could also be usefully extended to
the study of identification and turnover. In some cases,
client groups or subgroups within the organization may be
the focus of identification (Reichers, 1986). In the Army,
for example, branch or unit identification may supercede
organizational identification. In other cases, employees
may identify with professional or occupational groups
outside the organization (Lachman & Aranya, 1986). The
implications of these potentially conflicting social
identities on organizational identification and turnover
remain to be explored. The extent or nature of an
employee's identification with family roles also warrants
examination, particularly in light of the unexpectedly
strong relationship between Army/family conflict and

organizational identification in this research.

162
W193;

The global measure of anticipated Army/family conflict
had strong direct and indirect effects on propensity to stay
in this research. These results clearly suggest the
potential importance of family factors in the turnover
decisions of young, married professionals.

One could argue that family factors influence turnover
only in "greedy institutions" (Segal, 1986) like the
military, where time, separation and relocation demands are
extreme. Yet the military literature suggests that it is
the subjective experience of conflict, not the extent or
frequency of time and separation demands that influences
turnover intentions (Jans, 1988; Szoc, 1982). Therefore,
any job may be experienced as precipitating conflict if
requirements are incompatible with family obligations or
values. Variations in family obligations associated with
different family life cycle stages (e.g., single, just
starting a family, "empty nest") may be a more important
determinant of subjective conflict than objective job
demands.

The movement in corporate America toward more family
oriented programs (e.g., childcare) and policies (e.g.,
flexible work schedules) also suggests that work/family
conflict occurs across a broad spectrum of organizations.
Moreover, there is evidence that conflicts, particularly
with respect to time, influence a variety of employment

decisions. In a large, national survey of Fortune 500

163
employees, 20% of the married male respondents indicated
that they had sought a less demanding job for family
reasons, and 30% had refused a job, promotion or transfer
because it would have meant less time for their families
(Eertgne, February, 1987). Time for the family was an even
more important factor in the employment decisions of married
women with children.

The type of global Army/family conflict measure used in
this research may be adequate to capture conflicts stemming
from the day-to-day time and schedule demands of the job. A
different type of measure may be required, however, to
capture conflicts stemming from conflicting demands in two
career households. The present model tended to underpredict
propensity to leave the Army for officers who fit the "dual
career family" profile (i.e., fewer children and more highly
educated, career oriented wives). The addition of spouse
career aspirations in the exploratory analyses added a
significant, but very small increment to the explained
variance in propensity to stay. Measures focusing
specifically on relocation and travel conflicts and the
overall implications of staying in the Army for the spouse's
career might have explained more variance.

The importance of work/family conflict also suggests
that a more detailed assessment of subjective norms may be
useful. When family conflicts arise or are anticipated,
pressures from the spouse to leave the organization may

become an important factor in turnover decisions. A

164
spouse's willingness to accomodate the demands of the
employee's career may also determine the level of conflict
experienced or moderate the effects of conflict on turnover
decisions. Additional research on the types or sources of
work/family conflict, as well as the consequences of
conflict is clearly warranted.
W

The implications of career prospects for propensity to
stay are best examined in the context of the effects of work
satisfaction. Mobley et al. (1979) called for research on
the joint contribution of job satisfaction (present affect)
and job attraction (expected future affect) to turnover
intentions or behavior. The call was heeded, however,
methodological problems have hindered efforts to compare the
relative effects of present and future oriented outcome
measures.

One problem is that traditional job satisfaction
measures typically reflect more than simply reactions to the
immediate work situation. Global satisfaction measures
capture evaluations of the future as well as the present
situation (Ironson et al., 1988): and most summed facet
satisfaction measures include an evaluation of promotion
opportunities.

In contrast, the present model incorporates a narrower,
more time specific measure of current satisfaction. Only

the most critical dimension of overall satisfaction was

165
assessed (satisfaction with the work itself) and respondents
were instructed to evaluate only their current assignment.

Outcome expectation measures in turnover research also
tend to blur distinctions between present experiences and
long-term expectations. Typical methods involve having
employees rate a researcher generated list of outcomes in
terms of desirability and attainability (e.g., Hom & Hulin,
1981: Hom et al., 1984: Mobley, Hand, Baker & Meglino, 1979:
Motowidlo & Lawton, 1984), or simply expected satisfaction
(Hinzs & Nelson, 1990: Motowidlo, 1983). This method
implies that parallel dimensions determine both current
satisfaction and the expected utility of remaining in the
organization. The common method also produces large
correlations between current and expected satisfaction
measures, reducing the likelihood that current and future
oriented measures will be independent sources of variance in
turnover intentions (e.g., Hinzs & Nelson, 1990; Motowidlo &
Lawton, 1984).

The focus on career prospects in the present research
reflects a different perspective - namely, that long-term
career goals, not desires for specific job attributes,
provide the context for evaluating future opportunities in
the organization. The relevance of career related
constructs in turnover research has been demonstrated in a
variety of samples (Farris, 1971; Graen et al., 1973; Graen
& Ginsburgh, 1977: Jans, 1988; Schneider & Dachler, 1978).

However, a career prospects measure focusing specifically on

166
career development and advancement opportunities has not
been tested in a multivariate model of turnover intentions.

The results of this research lend strong support to the
utility of the construct. Perceptions of career prospects
influenced intentions independently of the effects of other
powerful antecedents. A more traditional summary measure of
expected job outcomes might have had similar effects on
propensity to stay. Career prospects appears to be a more
useful construct, however, because it specifies both the
context for evaluating future opportunities and the
dimensions likely to be important.

Current work satisfaction also demonstrated direct
effects on propensity to stay, net of the influence of
career prospects. This contradicts the present model but
supports the hypothesized joint contribution of satisfaction
and expected utility in the Mobley et al. (1979) model. An
independent relationship was not expected in this sample
primarily because of the temporary nature of military job
assignments and the long-term career implications of the
decision to leave. In this case, in particular, utility
maximization principles would predict that immediate
satisfaction would be far less important than prospects for
longer-term career satisfaction in determining intentions.

The contrary results suggest that rational, utility
based decision models do not adequately capture the effects
of job perceptions on turnover intentions. Multiple

processes are involved either within or across individuals.

167
Mobley et al. (1979), suggested that satisfaction (present
oriented) may be more important than "attraction" (future
oriented) for individuals with stronger needs for immediate
gratification. A short-term perspective may also reflect
the absence of a strong career orientation or the lack of
clearly formulated career goals.

In future research, satisfaction and expectation
measures that are time specific (i.e., clearly current or
future oriented) and focus on a limited, theoretically
derived set of dimensions may be most useful. Specific
measures reduce the conceptual and empirical overlap across
constructs and point more clearly to the areas organizations
need to address if they are to influence turnover.
W

Years of service had a small direct effect on
propensity to stay. Even a small effect is noteworthy,
however, in a sample with such a restricted range of
experience.

In line with research derived from the investments, or
side-bet framework, tenure was conceptualized as a surrogate
for the individual's level of investment in the
organization. Accordingly, the observed effects are
interpreted as support for the hypothesis that the costs of
leaving the organization are an independent source of
variance in turnover intentions. This interpretation seems
justified in military settings because eligibility for

retirement benefits is solely a function of years of

168
service, and much of the knowledge acquired through
experience is specific to the organization. Where the
relationship between time and investments is more tenuous,
however, perceptual measures of the costs of leaving the
organization are required to test the investments model
(Meyer & Allen, 1984).

Regardless of how costs are measured, the investments
model fits within the expected utility framework. Evidence
that investment or "side bet" losses are related to turnover
suggests that economic utility maximization principles
influence turnover decisions. The critical question then
centers on the relative importance of economic costs in the
larger domain of factors (utility based or not) that
influence turnover decisions. In this sample, tenure
related effects are quite small relative to other factors.
This is likely to be a function of the restricted variance
in tenure, however.

A second, related question concerns the point at which
tenure related investments in an organization become the
dominant factor in turnover decisions. Tenure may moderate
the relationship between other variables and turnover
intentions, but not necessarily in a continuous linear
function. It would be useful to try to identify
organizational tenure stages in which (a) investments are
negligible and have little or no influence on turnover
decisions, (b) investments are small to substantial, and

variations in investment levels contribute additively to

169
turnover intentions, and (c) investments are so large that
voluntarily leaving the organization before retirement is no
longer percieved to be a viable option. Very different
models might be required to explain turnover decisions in
these organizational subgroups.

It might also be fruitful to look at tenure as an
outcome variable. Although typically treated as an
antecedent, demographic type variable, from a longitudinal
perspective tenure can be viewed as the outcome of an
earlier decision to remain in the organization (i.e., a
decision made before researchers arrived to measure
intentions). In other words, accumulated time and tenure
related investments in an organization may be by products
rather than causes of decisions to stay.

From this perspective, relationships between tenure and
identification, here, and tenure and attitudinal commitment
elsewhere (Mathieu & Zajac, 1990: Meyer & Allen, 1984) could
result from the impact of early identification on propensity
to stay. If early attitudes cause people to remain in
organizations, total effects of attitudinal variables may be
underestimated in typical models. Longitudinal analyses are
required to identify the role played by tenure related
investments in turnover decisions. However, when only cross
sectional samples are available, it might be instructive to
conduct separate analyses within different tenure groups and
to interpret findings in light of ghee employees say they

made decisions to stay or leave.

170
s n n r

Analyses of the effects of the antecedent variables in
the model also have implications for future research.

Wezk_§etiefieetien. As noted above, work satisfaction
had indirect effects on propensity to stay through both
intervening variables. The path from satisfaction to career
prospects was hypothesized to be significant, and is
consistent with Motowidlo and Lawton's (1984) conclusion
that the effects of job satisfaction on turnover intentions
are mediated by outcome expectations. Yet, contrary to the
prediction of the model and Motowidlo and Lawton's (1984)
results, work satisfaction also had direct effects on
propensity to stay.

Whether or not satisfaction displays effects
independent of expectations is likely to be largely a
function of the overlap across measures. The relatively
specific work satisfaction and career prospects measures
used in this research were only moderately correlated
(r=.49). In samples where independent effects were not
found, correlations between satisfaction and expectation
measures exceeded .70 (Motowidlo & Lawton, 1984) and .80
(Hinzs & Nelson, 1990). Thus, the predicted independent
effects of present and future satisfaction in the Mobley et
al. (1979) model are most likely to be supported only when
fairly specific, independent measures are used.

The significant path from work satisfaction to

organizational identification was not predicted by the

171
model. Operational support, inspirational leadership and
prior career orientation were hypothesized to be the
important determinants of identification. The significant
path from work satisfaction to identification is, however,
consistent with most conceptualizations of the relationship
between satisfaction and commitment (Meyer 8 Allen, 1987;
Williams & Hazer, 1986). Identification appears to be
determined, in part, at least, by intrinsically satisfying
work experiences in the organization. Work satisfaction
itself may color feelings about the organization, or the
employee's sense that he is important, and able to make a
meaningful contribution to the organization may foster
identification and commitment (Buchanan, 1974).

The possibility of causal effects in the opposite
direction should also be entertained, however. Individuals
who identify with the organization may be more inclined to
emphasize the positive aspects of their work assignments.
Longitudinal analyses designed to test directional as well
as reciprocal effects are required to sort out these
effects.

s or . Operational support is the only
other antecedent variable with significant paths to both
intervening variables. The effects of support on
identification parallel the effects of organizational
dependability (Buchanan, 1974: Spencer & Steers, 1980:
Steers, 1977), service effectiveness (Jans, 1988) and

organizational support (Eisenberger et al., 1986) on

172
commitment. It is clear that perceptions of organizational
functioning influence feelings toward the organization.
When organizational practices are inconsistent with stated
organizational values (e.g., mission accomplishment) the
effects on identification or commitment may be especially
strong (DeCotiis 8 Summers, 1987).

The hypothesized link between operational support and
career prospects was predicated on the assumption that level
of support affects performance, and performance evaluations
largely determine promotion prospects. Although this
specific mediational hypothesis was not tested, the
relatively strong path from operational support to career
prospects suggests that efforts to confirm this link would
be useful.

Prie;_1n§entiege. In light of previous research on the
stability of prior career intentions in the military
(Bachman & Blair, 1975: LaRocco & Jones, 1980; Youngblood et
al., 1983), the fact that retrospective pre-entry intentions
were only weakly related to propensity to stay two to six
years later is surprising. The weak paths from prior
intentions to the variables hypothesized to mediate this
relationship were also unexpectedly small. Only the path to
identification was significant. These results suggest that
to some extent, pre-entry intentions reflect values that
promote subsequent identification. However, the assumption
that early plans influence career prospects (through stable

occupational goals) was not supported.

173

The lack of strong direct or indirect effects for prior
career orientation suggests that situational factors and
organizational experiences are more important determinants
of attitudes and perceptions than pre-entry values or career
expectations. Hall (1976) notes that as employees acquire
work experience, their career goals and occupational self-
concept are likely to change. The career goals of recent
college graduates may be especially unstable, resulting in a
re-evaluation of the opportunties for a satisfying career in
the organization.

At the same time, changes in the organization itself
may alter initial perceptions of internal career
opportunities. A particularly significant organizational
change experienced by officers in the present sample is the
"downsizing” of the Army. Most of the officers studied here
were commissioned before it was apparent that the military
would suffer massive budget cuts and manpower reductions.
Perceptions of career prospects in the Army may have changed
markedly in light of reduced prospects for advancement and
job security.

Breneh_meeeh. The branch match variable was
conceptualized as a type of internal person-environment fit
measure. In this research the measure was dichotomous and
skewed (the majority of officers reported a good match), but
even with these limitations the measure displayed the
predicted relationship to career prospects. Comparable

measures in other organizations might focus on preferences

174

for a different department or career track in the
organization. Internal fit measures, particularly in
conjucntion with perceived opportunities for internal
transfers, could help to explain movement within, as well as
across organizations (Jackofsky & Peters, 1984).

Inepizetienel_leege;ehip. Inspirational leadership was
related to organizational identification, however the
effects were not as large as expected. The lack of variance
in the measure might be responsible for the small amount of
variance explained. Nearly everyone in the sample could
identify at least one inspirational leader in their current
assignment. A more discriminating inspirational leadership
measure, or one focusing on immediate supervisors might have
stronger effects. The strong identification/propensity to
stay relationship suggests that experiences with
inspirational or "transformational" leaders (Bass, 1985)
warrant further study as indirect determinants of turnover
decisions.
W

When direct and indirect effects are summed,
Army/family conflict appears to have larger total effects
than any other variable in the model. These results may be
misleading, however. In the omitted parameters test,
Army/family conflict is treated as a potential omitted
"cause" of the intervening variables. The path coefficient
from conflict to identification is substantial, yet, because

the relationship was not predicted there is no a priori

175
basis for assuming that the implied causal direction is
correct.

It is certainly conceivable that a high level of
conflict could inhibit tendencies to identify with the
organization. When work demands interfere with family life,
employees may have to psychologically distance themselves
from the organization in order to cope with and resolve
conflicts. Conflicts that force employees to prioritize
their obligations may enhance the salience of family
obligations at the expense of organizational roles.

Yet the direction of causal influence may depend on
which develops first, identification with family roles or
identification with the organization. If identification
with the organization is already well-established,
subsequent family role obligations may be defined in ways
that reduce the experience of conflict. Organizational
demands may be accepted as legitimate and necessary, and
families may be expected to accomodate these demands.
Furthermore, strongly identified officers will most likely
marry women whose values are compatible with their own.
Thus, a strong early identification with either family or
organizational roles may partly determine how one's role in
the other domain is defined. To the extent that Army/family
conflict is determined by, rather than a cause of
organizational identification, the total effects of conflict
will be overestimated and the effects of identification will

be underestimated in the implicit causal model tested here.

176

Another, related limitation of the analyses and the
model itself, is that organizational identification and
career prospects are treated as exogenous to each other.
If, in fact, one variable is a cause of the other the total
effects of the causal variable will be underestimated.
Gaertner and Nollen (1989) for example, found that employee
perceptions that the firm supported individual training,
development and promotion from within were strongly related
to measures of affective commitment. They suggested that
employees given this type of support derive a positive self-
concept from the employment relationship, and tend to feel
an obligation to reciprocate by supporting the organization.
Both the sense of competence associated with the
organizational role and reciprocity norms were hypothesized
to account for the relationship between supportive personnel
practices and identification. Positive perceptions of one's
career prospects in the organization may encourage
identification in a similar manner. This would suggest that
the total effects of career prospects are underestimated in
the present model because the possible indirect effects
through identification are not analyzed.
Limitations

An important theoretical limitation of the present
research was noted above: causal linkages in the model are
incompletely specified. The model also fails to specify
paths that in the complicated web of real world relation-

ships are likely to be reciprocal. More sophisticated and

177
complete causal hypotheses may emerge as theory development
progresses, and hopefully can be tested through longitudinal
analyses. Path analysis in cross-sectional samples can
confirm the plausibility of theoretically derived causal
linkages, but longitudinal analyses are required to
convincingly demonstrate causal effects.

The restricted nature of the sample in this research is
also a limitation, yet one that also has certain advantages.
The focus on young, married, males in a unique type of
organization limits the generalizability of the research
results. On the other hand, testing models in clearly
identifiable subpopulations may point to the organizational
and demographic characteristics that need to be addressed in
broader turnover research.

W!

The strong effect of organizational identification on
turnover intentions appears to suggest that organizations
should make a concerted effort to foster organizational
identification among their employees. Schneider (1985),
however, warns that selection and socialization practices
that promote homogeneity and conformity in the work force
risk diminishing the potential for creativity and
innovation. Intensive, value-oriented socialization efforts
may also polarize attitudes in both directions, alienating
those who don't "buy into" the process.

A wiser approach might be for organizations to work on

establishing a climate that encourages identification. A

178
combined emphasis on quality and operational support for
workers is likely to facilitate performance and engender
pride in the organization. At the same time, policies and
practices that demonstrate respect for family needs may
mitigate both objective work/family conflicts and psychic
conflicts produced by incompatible organizational and family
values. Flexibility may be the key to both enhancing work
satisfaction and reducing work/family conflicts (Teplitzky,
1989). When demands that interfere with family life are
unavoidable, allowing employees the flexibility to sometimes
put their families first may go a long way toward mitigating
the negative effects of demanding careers.

The strong total effects of work satisfaction on
turnover intentions lend further support to the well
documented importance of satisfying work assignments.
Attempts to ensure that particularly satisfying assignments
coincide with critical career decision points (e.g., end of
obligation) could pay off in terms of lower turnover.

Career development programs that include efforts to help
employees formulate, as well as achieve, career goals should
also reduce turnover. Career oriented counseling could both
enhance employee perceptions of their career prospects in
the organization, and make it easier to tolerate temporary,
less satisfying assignments.

The more general implications of the test of this model

for future research are summarized below.

179
W

In summary, the results of this research suggest that
the direct determinants proposed in the model may help to
explain and predict employee turnover decisions.

Work/family conflict appears to be a pervasive phenomenon in
the working population today, thus its impact on turnover is
not likely to be limited to military settings. Career
conflicts in dual career families may warrant separate
treatment in professional samples in particular. Organiza-
tional identification also appears to be a promising
construct, although it may be less relevant in organizations
that are not characterized by a distinctive mission or
ideology. The utility of a career oriented perspective in
turnover research was also supported, although the
importance of career versus immediate job considerations may
vary across occupational groups. A continued focus on
tenure, particularly in the context of the investments
framework, is also recommended. As others have noted,
however, the effects of investments on employment decisions
may captured best by broader, perceptual measures.

In addition to supporting the importance of the
hypothesized direct determinants, results also suggested
ways the model might be improved. First of all, work
satisfaction is clearly an influential factor in turnover
decisions, and direct as well as indirect effects should be
posited. Measures that are clearly present oriented and

focus on intrinsic satisfaction are most likey to display

180
effects independent of those associated with expected
outcome and commitment measures. Second, the exploratory
analyses indicated that many officers felt constrained to
stay in the military because of personal or situational
factors. In this and other research, employee evaluations
of the availability of acceptable alternatives have failed
to explain much variance in turnover intentions. The single
item measure of personal/family constraints, however, was
among the most powerful determinants of intentions in this
research. Measures of alternatives may not tap time or
financial constraints that may limit an employee's ability
to conduct a thorough job search or relocate the family.
The recent birth of a child, for example, or a spouse's ties
to a job, school or friends in the community, or a large
investment in a new home may all operate to constrain an
individual's options. Additional effort to determine more
precisely the types of personal and family related
constraints employees experience appears warranted.

It would also be interesting in future longitudinal
research to assess the impact of constraints on the
relationship between turnover intentions and behavior. In
the military, many service members who say they intend to
leave wind up staying (Szoc, 1982). It would be useful from
both an applied and a theoretical perspective to try to
learn why and when individuals fail to follow through on
intentions to leave. This type of analysis would be

particularly useful to organizations facing requirements to

181
reduce the size of the work force. Instead of laying off
employees, management might be able encourage voluntary
turnover by removing obstacles that prevent some individuals
from acting on their inclinations to leave.

Important issues that remain unresolved in this
research concern the determinants of organizational
identification and work/family conflict, and the nature of
the relationship between these two variables. Longitudinal
analyses will be required to determine the dominant
direction of influence. More discriminating measures of
leadership and measures that assess career goals and
work/family priorities may prove to be useful in efforts to
understand the complex turnover decision process. Mobley et
al. (1979), for example suggested that the centrality of
work in an individual's life could moderate the effects of
satisfaction and expected outcomes on turnover intentions.

A final consideration in the design of future turnover
research is the nature of the population or organization of
interest. Numerous studies focus on specific occupational
groups, yet discussions linking group characteristics to
components of the proposed models are rare. The model
tested in this research reflected an analysis of the factors
likely to be particularly salient to young, married Army
officers. Tests of similar models in different
subpopulations should clarify the extent to which the
results obtained here are attributable to the special

characteristics of the research population.

LIST OF REFERENCES

LI ST OF REFERENCES

Adelsberger, B. J. (1988, May 9). West Point: Worth the
cost? The Army Times, p. 12.

Ajzen, I., & Fishbein, M. (1980). gnderstanding attitudes
and predicting eocial behavior. Englewood Cliffs, NJ:

Prentice-Hall.

Alwin, D. F., & Hauser, R. M. (1975). The decomposition
of effects in path analysis. American Sociological
Review, 40, 37—47.

Arnold, H. J., & Feldman, D. C. (1982). A multivariate
analysis of the determinants of job turnover. Journal
of Applied Ps cholo , 61, 350—360.

Ashcraft, R. (1987). An Analysis of the Factore Affecting
the Career Orientation ef Junior URI Naval Officers
(Research Report, Master Thesis). Monterey, CA: Naval
Postgraduate School.

 

Ashforth, B. E., & Mael, F. (1989). Social identity
theory and the organization. Academy of Management
Review, 15, 20- 39.

Bachman, J. G., & Blair, J. (1975). "Citizen force" or
"career force"? Implications for ideology in the all-
volunteer Army. Armed Force and Societ , 2, 81-96.

Bachman, J. G., Seigelman, L., & Diamond, G. (1987).
Strategic limitations of "A Nation In Arms". Armed
Forces and Society, 1;, 169—187.

Bannister, B. D., & Griffeth, R. W. (1986). Applying a
causal analytic framework to the Mobley, Horner, and
Hollingsworth (1978) turnover model: A useful
reexamination. Journal of Management, 1;, 433-443.

 

Bass, B. M. (1985). Leaderehip and Performance Beyond
Expectations. New York, NY: The Free Press.

Bateman, T. S., & Strasser, S. A. (1984). A longitudinal

analysis of the antecedents of organizational commit-
ment. Academy of Management Journal, 21, 95-112.

182

183

Baysinger, B. D., & Mobley, W. H. (April, 1982).
Employee turnover; Individual and organizational
analysis. Arlington, VA: Organizational Effectiveness
Research. (TR—ONR-6)

Becker, H. S. (1960). Notes on the concept of commitment.
American Journal of Sociology, 66, 32-40.

Bluedorn, A. C. (1979). Structure, environment, and
satisfaction: Toward a causal model of turnover from
military organizations. Journal 6f Political and

Military Sociology, 1, 181-207.

Bluedorn, A. C. (1982a). The theories of turnover:
Causes, effects, and meaning. Research in the
Sociology of Organizatione, 1, 75-128.

 

Bluedorn, A. C. (1982b). A unified model of turnover from
organizations. Human Relations, 66, 135-153.

Bowen, D. E. (1982). Some unintended consequences of
intention to quit. Academy of Management Review, 1,
205-211.

Buchanan, E., II. (1974). Building organizational

commitment: The socialization of managers in work

organizations. Adminietrative Science Quarterly, 16,
533- 546.

Bullis, C. A. & Tompkins, P. K. (1989). The forest
ranger revisited: A study of control practices

and identification. gommunications Monographs,
66, 287-305.

Carsten, J. M., & Spector, P. E. (1987) . Unemployment,
job satisfaction, and employee turnover: A meta-
analytic test of the Muchinsky model. Journal of
Applied Ps cholo , 1;, 374—381.

Cohen, J., & Cohen, P. (1983). Applied Multiple
RegressionZCorrelation Analysis for the Behavioral

Sciences. Hillsdale, NJ: Lawrence Erlbaum Associates.

 

Cohen, A., & Lowenberg, G. (1990). A re—examination of the
side—bet theory as applied to organizational
commitment: A meta-analysis. Heman Relatione, 1;,
1015-1050.

Cook, J., & Wall, T. D. (1980). New work attitude
measures of trust, organizational commitment and
personal need non- fulfillment. Journal of
Occu ati nal Ps cholo 66, 39-52.

184

Cotton, J. L., & Tuttle, J. M. (1986). Employee turnover:
A meta-analysis and review with implications for
research. Academy of Management Review, 11, 55-70.

Dalessio, A., Silverman, W. H., & Schuck, J. R. (1986).
Paths to turnover: A Re-analysis and review of
existing data on the Mobley, Horner, and Hollingsworth
turnover model. Human Relations, 62, 245—263.

DeCotiis, T. A., & Summers, T. P. (1987). A path analysis
of a model of the antecedents and consequences of
organizational commitment. Human Relatiope, 1, 445-
470.

DePasquale, J. A., & Lange, R. A. (1971). Job hopping and
the MBA. Harvard Business Review, 62, 4-13.

Derr, C. B. (1979). MarriageZFamily Issues and Life
§tyles Across Naval Officer Career Stages: Their

Implications for Cereer Success (Final Rep. No. NP554-
79-003 TR 2). Monterey, CA: Naval Postgraduate School.

 

Dougherty, T. W., Bluedorn, A. C., & Keon, T. L. (1985).
Precursors of employee turnover: A multiple—sample
causal analysis. Journal of Occu ational Behavior, 6,

 

259-271.

Dreher, G. F. (1982). The role of performance in the
turnover process. Academy of Management Journal, 26,
137-147.

Dreher, G. F., & Dougherty, T. W. (1980). Turnover and

competition and expected job openings: An exploratory
analysis. Academy of Management Journal, 26, 766-772.

Dubin, R., Champoux, J. E., & Porter, L. W. (1975).
Central Life Interests and Organizational Commitment of
Blue-Collar and Clerical Workers. Adminietrative

Sgience Quarterly 26, 411—421.

Dunnette, 3. D., Arvey, R. D., & Banas, P. A. (1973). What
do they leave? P r onnel, pp. 29-39.

Eisenberger, R., Huntington, R., Hutchison, S., & Sowa, D.
(1986). Perceived organizational support. Journal of
Applied Ps h l , 11, 500-507.

Faris, J. H. (1984). Economic and noneconomic factors of
personal recruitment and retention in the AVF. Armed
Forces and Society, 16, 251-275.

185

Farkas, A. J., & Durning, K. P. (1982). Characteristics
end Needs of Navy Families: Policy Implications (Tech.
Rep. No. NPRDC SR 83-1). San Diego, CA: Navy Personnel
Research and Development Center.

Farrell, D., & Petersen, J. C. (1984). Commitment,
absenteeism, and turnover of new employees: A
longitudinal study. H m n Relations, 61, 681-692.

Farrell, D., & Rusbult, C. E. (1981). Exchange variables
as predictors of job satisfaction, job commitment, and
turnover: The impact of rewards, costs, alternatives,
and investments. Organipetional Behavior and Humep
Performance, 11, 78-95.

Farris, G. F. (1971). A predictive study of turnover.
Pereonnel Ps cholo , 16, 311-328.

Feather, N. T. (Ed.). (1982). Expectations and Actions:
Expectancy valuepmodele in psychology. Hillsdale, NJ:
Lawrence Erlbaum Associates.

Fishbein, M., & Ajzen, I. (1975). Belief. attitude1

intention and behevior; An Introduction to theory and
research. Reading, MA: Addison-Wesley.

Flowers, V. S., & Hughes, C. L. (1973) . Why employees
stay. Harvard Bueiness Review, 61, 48-60.

Ford, J. K., MacCallum, R. C., & Tait, M. (1986). The
application of exploratory factor analysis in applied
psychology: A critical review and analysis. Personnel
Ps cholo 66, 291-314.

Forrest, C. R., Cummings, L. L., & Johnson, A. C. (1977).
Organizational Participation: A critique and model.

Academy 6f Manegement Review, pp. 586-601.

Gaertner, K. N. & Nollen, S. D. (1989). Career
experiences, perceptions of employment practices, and
psychological commitment to the organization. Human
Relations, 61, 975-991.

Gibb, G. D., Nontasak, T., Dolgin, D. L., & Helm, W. R.
(1987). Factors ha influen career motivation in
the fighter cgmmunity (Tech. Rep. No. NAMRL-1326).
Pensacola, FL: NAS.

Goldman, N. L., & Segal, D. R. (Eds.). (1976). The

Social Psychology of Military Service. Volume VI.
Beverly Hills, CA: Sage Publications.

186

Grace, G. L., & Steiner, M. B. (1978). Wives' attitudes
and the retention of Navy enlisted personnel. In E.
Hunter & D. Nice (Eds.), Military Families: Adaptation
to ehange. New York: Praeger.

Graen, G., & Ginsburgh, S. (1977). Job resignation as a
function of role orientation and leader acceptance: A
longitudinal investigation of organizational
assimilation. Organizational Behavior and Humem
Performance, 16, 1-17.

Graen, G. B., Liden, R. C., & Hoel, W. (1982). Role of
leadership in the employee withdrawal process. Journal
of Applied Peychology, 61, 868—872.

Graen, G. B., Orris, J. B., & Johnson, T. W. (1973). Role
of assimilation processes in a complex organization.
Journal of Vocational Behavior, 6, 395-420.

Griffith, J. D., Doering, Z., & Mahoney, B. S. (1986).
Deecription of epeusee of officereyend enlisted
personnel in the U.S. armed forcee, 1985: A report
based on the 1985 DoD surveys of officer and enlieted
personnel andpmilitary spousee (3 vols.). Research
Triangle Park, NC: Research Triangle Institute.

Hall, D. T. (1976). Qereere in Qrgenizations. Santa
Monica, CA: Goodyear Publishing Company, Inc.

Hall, D. T., & Schneider, B. (1972). Correlates of
organizational identification as function of career
pattern and organizational type. Administrative
Science Quarterly, 11, 340-350.

Hall, D. T., Schneider, B., & Nygren, H. T. (1970).
Personal Factors in Organizational Identification.
Agministrative Quarterly Review, 16, 176-190.

Hayden, L. P. (1985). Fagtors Influencing the Career
Orientation of Junior Officers in the United States
Army (Research Report/Master Thesis). Monterey, CA:
Naval Postgraduate School.

Hinsz, V. B., & Nelson, L. C. (1990). Testing models of
turnover intentions with university facility. Journal

of Applied Social Psyehology 16, 68—84.

Hom, P. W. (1980). Expectancy predictions of reenlistment
in the National Guard. Journal of Vocational Behavior,
16, 235-248.

187

Hom, P. W., Griffeth, R. W., & Sellaro, C. L. (1984). The
validity of Mobley's (1977) model of employee turnover.

Organizational Behavior and Human Performance, 66, 141-
174.

Hom, P. W., & Hulin, C. L. (1981). A competitive test of
the prediction of reenlistment by several models.
Journal of Applied Ps cholo , 66, 23-39.

Hom, P. W., Katerberg, R., Jr., & Hulin, C. L. (1979).
Comparative examination of three approaches to the

prediction of turnover. Journal of Applied Psychology,
66, 280-290.

Hulin, C. L., Roznowski, M., & Hachiya, D. (1985).
Alternative opportunities and withdrawal decisions:
Empirical and theoretical discrepancies and an
integration. Ps cholo i al Bulletin, 61, 233—250.

Ironson, G. H., Smith, P. C., Brannick, M. T., Gibson, W.
M., & Paul, K. B. (1989). Construction of a job in
general scale: A comparison of global, composite, and
specific measures. Journal of Applied PsychologY, 16,
193-200.

Jackofsky, E. (1984). Turnover and job performance: An
integrated process model. Aeedemy of Management
Review, 6, 74—83.

Jackofsky, E. F., & Peters, L. H. (1983). Job turnover
versus company turnover: Reassessment of the March and
Simon participation hypothesis. Journal of Applied

Psychology, 66, 490-495.

James, L. R. (1980). The unmeasured variables problem in
path analysis. Journal of Applied Psychology, 66, 415-
421.

James, L. R., Mulaik, S. A., & Brett, J. M. (1982). Causal

Analysis: Assumptions, models, and data. Beverly
Hills, CA: Sage.

Jans, N. A. (1988). Careers in conflict. Belconnen,
Australia: Canberra College of Advanced Education.

Jans, N. A. (1989). Organizational commitment, career
factors and career/life stage. Journal of
Or anizational Behavior, 16, 247-266.

Koch, J. L., & Steers, R. M. (1978). Job attachment,
satisfaction, and turnover among public sector
employees. Journal of Vocational Behavior, 11, 119-128.

188

Krackhardt, D., McKenna, J., Porter, L. W., & Steers, R. M.
(1981). Supervisory behavior and employee turnover: A
field experiment. Academy of Management Journal, 16,
249-259.

Kraut, A. I. (1975). Predicting turnover of employees
from measured job attitudes. Organizational Behavior
and Human Performance, 16, 233-243.

Lachman, R., & Aranya, N. (1986). Evaluation of
alternative models of commitments and job attitudes of

professionals. gournal of Qccupational Behavior, 1(1),
227- 243.

LaRocco, J. M. (1983). Job attitudes, intentions, and
turnover: An analysis of effects using latent
variables. Human Reletione, 66, 813-826.

LaRocco, J. M., & Jones, A. P. (1980). Organizational
conditions affecting withdrawal intentions and
decisions as moderated by work experience.

Peychglogical Reports, 66, 1223-1231.

LaRocco, J. M., Pugh, W. M., & Gunderson, E. K. E. (1977).
Identifying determinants of retention decisions.
Personnel Ps ch lo , 66, 199-214.

Lawler, E. E., III, Kuleck, W. J., Rhode, J. G., & Sorenson,
J. E. (1975). Job choice and post decision

dissonance. Qrganizetienal Behevier and Human
Performan , 1 , 133-145.

Lee, T. W., & Mowday, R. T. (1987). Voluntarily leaving
an organization: An empirical investigation of Steers
and Mowday's model of turnover. Academy of Management
gournal, 66, 721-743.

Lund, D. A. (1978). Junior Officer Retention in the
Modern Volunteer Army: Who Leaves and Who Stays? In E.
J. Hunter & D. S. Nice (Eds.), In Military Familiee;
Adaptation to Qhange (pp. 32-41). New York, NY:

Praeger Publishers.

March, J. G., & Simon, H. A. (1958). Organizations. New
York, NY: Wiley.

Marsh, R., & Mannari, H. (1977). Organizational
commitment and turnover: A predictive study.
Adminietrative Science Quarterly, 11, 57-75.

Martin, T. N. (1979). A contextual model of employee
turnover intentions. Academ of Mana ement Journal,
16, 313- 324.

189

Martin, T. N., & Hunt. (1980). Social influence and
intent to leave: A path analytic process model.

Personnel Psyghology 66, 505—527.

Mathieu, J. E. (1988). A causal model of organizational
commitment in a military training environment. Journal
of Vocational Beh vior, 66, 299—317.

Mathieu, J. E., & Hamel, K. (1989). A causal model of the
antecedents of organizational commitment among
professionals and nonprofessionals. Journal of
Vocational Behavior, 66, 299-317.

Mathieu, J. E., & Zajac, D. M. (1990). A review and meta-
analysis of the antecedents, correlates, and
consequences of organizational commitment.
Psychological Bulletin, 108, 171—194.

McAllister, I., & Smith, H. (1989). Selecting the
guardians: Recruitment and military values in the
australian officer corps. Journal of Political and

Military Sociolegy, 11, 27-42.

McGee, G. W., & Ford, R. C. (1987). Two (or more?)
dimensions of organizational commitment: Reexamination
of the affective and continuance commitment scales.

gournal of Applied Psychology, 11, 638-642.

Meola, J. (1983). An investigation of the process by which

Air Force offieers view and eveluete their perceived

availability of jgb alternetives. Wright-Patterson
AFB, OH: Air University.

 

Meyer, J. P., & Allen, N. J. (1984). Testing the "side-
bet theory" of organizational commitment: Some
methodological considerations. Jomrnal of Applied
£§ychology, 66, 372-278.

Meyer, J. P., & Allen, N. J. (1987). A longitudinal
analysis of the early development and consequences of
organizational commitment. Qenedien Journal of
Behavioral Sci n , 16, 199-215.

Meyer, J. P., & Allen, N. J. (1988). Links between work
experiences and organizational commitment during the
first year of employment. Journal of Occupational
Ps cholo , 61, 195-209.

Meyer, J. P., Paunonen, S. V., Gellatly, I. R., Goffin, R.
D., & Jackson, D. N. (1989). Organizational
commitment and job performance: It's the nature of the

commitment that counts. gournal of Applied Psychology,
16, 152—156.

190

Michaels, C. E., & Spector, P. E. (1982). Causes of
employee turnover: A test of the Mobley, Griffeth,
Hand, and Meglino Model. Journal of Applied
Egychology, 61, 53-59.

Miller, H. E., Katerberg, R., & Hulin, C. L. (1979).
Evaluation of the Mobley, Horner, and Hollingsworth
model of employee turnover. Journal of Applied
P cholo , 4, 509- 517.

Mitchell, T. R., & Albright, D. W. (1972). Expectancy
theory predictions of the satisfaction, effort,
performance and retention of naval aviation officers.

Organizatignal Behavior and Human Perfgrmance, 6, 1—20.

Mitchell, T. R., & Beach, L. R. (1976). A review of
occupational preference and choice research using
expectancy theory decision theory. Journal of
Occu ational Ps holo , 66, 231-248.

Mitchell, T. R., & Knudson, B. W. (1973). Instrumentality
theory predictions of students' attitudes towards
business and their choice of business as an occupation.
Academy of Management Journal, 16, 41-52.

Mobley, W. H. (1977). Intermediate linkages in the
relationship between job satisfaction and employee
turnover. Journal of Applied Ps cholo , 66, 237-240.

Mobley, W. H. (1982). Some unanswered questions in
turnover and withdrawal research. Academy of
Management Review, 1, 111-116.

Mobley, W. H., Griffeth, R. W., Hand, H. H., & Meglino, B.
M. (1979). Review and conceptual analysis of the

employee turnover process. geychological Bulletin, 66,
493-522.

Mobley, W. H., Hand, H. H., Baker, R. L., & Meglino, B. M.
(1979). Conceptual and empirical analysis of military
recruit training attrition. Journal of Applied

Psychology, 66, 10-18.

Mobley, W. H., Horner, S. 0., & Hollingsworth, A. T.
(1978). An evaluation of precursors of hospital

employee turnover. gournal of Applied Psychology, 66,
408-414.

Mohr, D. A., Holzbach, R. L., & Morrison, R. F. (1981).

§urface Warfare Junior Officer Retention: Spouses'

Influence en Career Decisions (Tech. Rep. No. NPRDC TR
81-17). San Diego, CA: Navy Personnel Research and
Development Center (AD A120807).

191

Monkus, S. W. (1979). Attitudes and Motivatione of Newly
Cgmmissioned Officers (Research Report No. 1625-79).
Maxwell AFB, AL: Air University.

Morrow, P. C. (1983). Concept Redundancy in
organizational research: The case of work commitment.
Aeademy of Management Review, 6, 486-500.

Moskos, C. (1986). Institutional/Occupational trends in
armed forces: An update. Armed Forces and Society,
11, 377-382.

Motowidlo, S. J. (1983). Predicting sales turnover from
pay satisfaction and expectation. Journal of Applied
Psychology, 66, 484-489.

Motowidlo, S. J., & Lawton, G. W. (1984). Affective and
cognitive factors in soldiers' reenlistment decisions.
Journal of Applied Ps cholo 66, 157-166.

Motowidlo, S. J., Packard, J. S., & Manning, M. R. (1986).
Occupational stress: Its causes and consequences for

job performance. Journal of Applied Psycholo , 11,
618-629.

Mowday, R. T., Koberg, C. S., & McArthur, A. W. (1984).
The psychology of the withdrawal process: A cross-
valiation test of Mobley's intermediate linkages model
of turnover in two samples. Aeedemy of Management
Journal, 11, 79-94.

Mowday, R. T., Porter, L. M., & Steers, R. M. (1982).

Employee—grganizational linkages. New York: Academic
Press.

Mowday, R. T., Steers, R. M., & Porter, L. W. (1979). The
measurement of organizational commitment. Journal of
Vocational Behavior, 16, 224—227.

Muchinsky, P., & Tuttle, M. L. (1979). Employee Turnover:
An Empirical and methodological assessment. Journal of
Vocational Behavior, 16, 43-77.

Muchinsky, P. M., & Morrow, P. C. (1980). A
multidisciplinary model of voluntary employee turnover.
Journal of Vocatipnal Behevior, 11, 263-290.

Muchinsky, P. M., & Tuttle, M. L. (1979). Employee
turnover: An empirical and methodological assessment.

gournal of Vocational Behavior, 16, 43-77.

192

Neumann, I., Abrahams, N., & Githens, W. (1972). IQQ
Values of Junior Officers. Part II. The Relationship
Between Career Values and Retention (Research Report
No. SRR72-35). San Diego, CA: Naval Personnel and
Training Research Laboratory.

Newman, J. E. (1974). Predicting absenteeism and turnover:
A field comparison Fishbein's model and traditional job
attitude measures. Journal of A lied Ps cholo , 66,
610-615.

"The number 1 cause of executive guilt: Who's taking care of
the children and how will they turn out?" Fortune,
February, 1987.

O'Reilly, C., & Chatman, J. (1986). Organizational
commitment and psychological attachment: The effects
of compliance, identification, and internalization on
prosocial behavior. gournal of Applied PsychologY, 11,
492-499.

O'Reilly, C. A., & Caldwell, D. F. (1981). The commitment
and job tenure of new employees: Some evidence of
postdecisional justification. Adminietrative Science

Quarterly 16, 597-616.

Oldham, G. R. (1976). Organizational choice, and some
correlates of individuals' expectancies. Decision
5212222. 1. 873-884.

Orthner, D. K. (1990). Femily Impacts on the Retention of
Military Personnel (Research Report 1556). Alexandria,
VA: Army Research Institute.

Orthner, D. K., & Bowen, G. L. (1982). Familieg—in-Blue:
Insights fromeAir Force families in the Pacific.
Greensboro, NC: Family Development Press.

Parasuraman, A., & Futrell, C. M. (1983). Demographics,
job satisfaction, and propensity to leave of industrial
salesmen. Jgurnal of Business Reeearch, 11, 33-48.

Parker, D. F., & Dyer, L. (1976). Expectancy theory as a
within-person behavioral choice model: An empirical
test of some conceptual and methodological refinements.

Organizational Behavior and Human Performance, 11, 97-
117.

 

Pedhazur, E. J. (1982). Multiple Regression in Behavior
Reeearch. New York: Holt, Rinehart and Winston.

193

Peters, L. H., & O'Connor, E. J. (1980). Situational
constraints and work outcomes: The influences of a
frequently overlooked construct. Academy of Management
Journal, 6, 391-397.

Peters, L. H., O'Connor, E. J., & Rudolph, C. J. (1980).
The behavioral and affective consequences of
performance-relevant situational variables.

Organizational Behavier and Human Performance, 66,
79-96.

 

Pfeffer, J., & Lawler, J. (1980). Effects of job
alternatives, extrinsic rewards, and behavioral
commitment on attitudes toward the organization: A
field test of the insufficient justification paradigm.

Administrative Seience Querterly, 16, 38—56.

Porter, L. W., Crampon, W. J., & Smith, F. J. (1976).
Organizational commitment and managerial turnover: A
longitudinal study. Organizational Behavior and Humem
Performance, 16, 87-98.

Porter, L. W., & Steers, R. M. (1973). Organizational,
work, and personal factors in employee turnover and
absenteeism. Ps cholo ical Bulletin, 66, 151-176.

Porter, L. W., Steers, R. M., Mowday, R. T., & Boulian, P.
V. (1974). Organizational commitment, job
satisfaction, and turnover among psychiatric

technicians. gournal of Applied Peychology, 66, 603—
609.

Prestholdt, P. H., Lane, I. M., & Mathews, R. C. (1987).
Nurse turnover as reasoned action: Development of a

process model. ggurnal of Applied Psychology, 11, 221—
227.

Price, J. L. (1977). The §tudy of turnover. Ames, IA:

The Iowa State University Press.

Price, J. L., & Mueller, C. W. (1981). A causal model of
turnover for nurses. Academy of Management Journal,
66, 543- S65.

Rapoport, R., & Rapoport, R. N. (1976). Dual—career
femilies re-examined. New York, NY: Harper.

Reichers, A. E. (1985). A review and reconceptualization
of organizational commitment. Academy of Management

194

 

Reichers, A. E. (1986). Conflict and organizational
commitments. Journal of Applied Peychologv, 11,
508-514.

Rhodes, S. R., & Doering, M. (1983). An integrated model

of career change. Academy of Management Review, 6,
631—639.

Rousseau, D. M. (1990). New hire perceptions of their own
and their employer's obligations: A study of
psychological contracts. Journal of Organizational
Behavior, 11, 389—400.

Rusbult, C. E., & Farrell, D. (1983). A longitudinal test
of the investment model: The impact on job
satisfaction, job commitment, and turnover of
variations in rewards, costs, alternatives, and
investments. Journal of A lied Ps cholo , 66, 429—
438.

Rusbult, C., & Lowry, D. (1985). When bureaucrats get the
blues: Responses to dissatisfaction among federal
employees. Journal of Applied Social Psychology, 16,
80-103.

Salancik, G. R., & Pfeffer, J. (1978). A social
information processing approach to job attitudes and

task design. Administrative Science Quarterly, 66,
224-253.

Schein, E. H. (1978). Qereer Dynemieez Matching
Individual and Organizational Needs. Reading, MA:
Addison- Wesley Publishing Company.

Schmidt, W. H. (1982). Factore Influencing the Career
Qrientation of gunior Offieers in the United States
Navy (Research Report/Master Thesis). Monterey, CA:
Naval Postgraduate School.

Schneider, B. (1987). The people make the place.
Pereonnel Ps h lo , 66, 437-453.

Schneider, B., & Dachler, H. P. (1978). Work, family and
career consideratione in understanging employee
turnover intentions (Research Report No. 19).
Arlington, VA: Personnel and Training Research
Programs, Office of Naval Research.

Schneider, B., Hall, D. T., & Nygren, H. T. (1971). Self-
image and job characteristics as correlates of changing
organizational identification. Human Relations, 61,
397-416.

195

Schneider, J. (1976). The "greener grass" phenomenon:
Differential effects of a work context alternative on
organizational participation and withdrawal intentions.
Organizational Behavior and Human Performance, 16, 308-
333.

Scholl, R. W. (1981). Differentiating organizational
commitment from expectancy as a motivating force.
Academy of Management Review, 6, 589-599.

Scholl, R. W. (1983). Career lines and employment
stability. Aeagemy of Management Journal, 16, 86-103.

Segal, M. W. (1986). The military and the family as
greedy institutions. Armed Force§1§nd Society, 16,
9-38.

 

Shenk, F., & Wilbourn, J. (1971). Officer Attitudes
Related to Career Decieione (Research Report No. AFHRL—
TR-71- 45). Lackland AFB, TX: Air Force Human
Resources Laboratory.

Spencer, D. G., & Steers, R. M. (1980). The influence of
personal factors and perceived work experiences on
employee turnover and absenteeism. Academy of

Manegement gemrnal, 66, 5567-572.

Spencer, D. G., Steers, R. M., & Mowday, R. T. (1983). An
empirical test of the inclusion of job search linkages
into Mobley's model of the turnover decision process.

rm 1 of ccu a ional P holo 66, 137-144.

Steel, R. P., & Ovalle, N. K. (1984). A review and meta-
analysis of research on the relationship between
behavioral intentions and employee turnover. Journal
of Applied Ps cholo , 66, 673-686.

Steele, J. D. (1987). 60cioeconemic Factors and Pereonal
Characterietice Affecting the Retention of Officers in
he United States Arm nd Uni ed tes Marine Corps
(Research Report/Master Thesis). Monterey, CA: Naval
Postgraduate School.

Steers, R. M. (1977). Antecedents and outcomes of
organizational commitment. Adminietrative Science

Quarterly, 61, 46-56.

Steers, R. M., & Mowday, R. T. (1981). Employee turnover
and post-decision accommodation processes. Research in
Organizational Behavior, 6, 235-281.

196

Stevens, J. M., Beyer, J. M., & Trice, H. M. (1978).
Assessing personal, role and organizational predictors
of managerial commitment. Academy of Management
Journal, 61, 380-396.

Stumpf, S. A., & Hartman, K. (1984). Individual
exploration to organizational commitment or withdrawal.

Academy ef Menegement Jgurnal, 61, 308-329.

Szoc, R. (1982). Family Factors Critical to the Retention
of Naval Personnel (Final Report). Columbia, MD:
Westinghouse Public Applied Systems Division.

Tajfel, H. (1981). Social identity and intergroup
relations. Cambridge: Cambridge University Press.

Teger, A. I. (1980). Too Much Invested to Quit. New York:
Pergamon Press.

Teplitzky, M. L., Thomas, S. A., & Nogami, G. (1988). Dual
Army Career Officers: Job Attitudes and Career
Intentions of Male and Female Officers. Alexandria,
VA: Army Research Institute (TR-805).

Thompson, K. R., & Terpening, W. D. (1983). Job-
type variations and antecedents to intention to leave:
A content approach to turnover. Human Relations, 66,
655-682.

Tom, V. R. (1971). The role of personality and
organizational images in the recruiting process.
Organizational Behavior and Human Performance, 6,
573-592.

Turner, J. (1982). Toward a cognitive redefinition of the
social group. In H. Tajfel (Ed.), Social Identity and
Intergroup Behavior (PP. 15-40). Cambridge, England:
Cambridge University Press.

Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., &
Wetherell, M. S. (1987). Regiscovering the Social
Group. Oxford, England: Basil Blackwell.

Wells, D. L., & Muchinsky, P. M. (1985). Performance
antecedents of voluntary and involuntary managerial
turnover. Journal of Applied Psychology, 16, 329—336.

Werbel, J. D., & Gould, S. (1984). A comparison of the
relationship of commitment to turnover in recent hires
and tenured employees. Journal of Applied Ps cholo ,

66, 687- 690.

197

Wiener, Y. (1982). Commitment in Organizations: A
normative view. Academy of Management Review, 1,
418-428.

Wiener, Y., & Vardi, Y. (1980). Relationships between
job, organization, and career commitments and work
outcomes - An integrative approach. Organizational
Behavior and Human Performence, 66, 81-96.

Williams, L. J., Cote, J. A., & Buckley, M. R. (1989).
Lack of method variance in self—reported affect and
perceptions at work: Reality or artifact? Journal of
Applied Psychology, 16, 462-468.

Williams, L. J., & Hazer, J. T. (1986). Antecedents and
consequences of satisfaction and commitment in turnover
models: A reanalysis using latent variable structural
equation methods. Journal of Applied PsychologY, 11,
219- 231.

Wood, F. R. (1982). U.S. Air Force junior officers:
Changing professional identity and cgmmitment
(Unpublished dissertation). Evanston, IL: Northwestern
University.

Youngblood, S. A., Mobley, W. H., & Meglino, B. M. (1983).
A longitudinal analysis of the turnover process.
Journal of A lied Ps cholo , 68, 507—516.

APPENDIX

APPENDIX

A PRIORI AND FINAL SCALE ITEMS
FOR VARIABLES IN OFFICER CAREER DECISION MODEL

 

PROPBNBITY TO STAY

Var Name Survey Item

INTEND Which of the following best describes your current
career intentions?

a) I plan to stay in the Army beyond 20 years

b) I plan to stay in the Army until retirement at
20 years

c) I plan to stay in the Army beyond my
obligation, but am undecided about staying
until retirement

d) I am undecided whether or not I will stay in
the Army upon completion of my obligation

e) I will probably leave the Army upon completion
of my obligation

f) I will definitely leave the Army upon
completion of my obligation

PLNNOW Right now I am ...

a) planning on an Army career

b) leaning toward an Army career

c) undecided

d) leaning toward a civilian career
e) planning on a civilian

The instructions for the set of items including PLNNOW read
as follows:

For some officers, career plans change over time, while for
others, career plans remain constant. Here we are
interested in finding out whether or not your own plans have
changed. Please use the following scale to indicate (to the
best of your recollection) how you felt at tme time of each
event/experience described below.

198

CARSAT

OPPADV

HOWCOM

199
CAREER PROSPECTS

All in all how satisfied are you with your career
prospects in the Army?

a) very satisfied

b) satisfied

c) neutral

d) dissatisfied

e) very dissatisfied

How good are the opportunities for advancement in
your branch for someone who has had the types of
assignments you have had?

a) excellent

b) very good

c) good

d) limited

e) very limited

How competitive for schools and promotions would
you be if you were to be evaluated right now

taking tpe matute of your assignmentsl as wel1 es
yet; petformance, into account?

a) I'd have a strong advantage

b) I'd have an advantage

c) No advantage or disadvantage

d) I'd be at a disadvantage

e) I'd be at a strong disadvantage

The remining 4 items use the following scale:

AGHIGH

AGASGN

AGSKIL

LIKWRK

a) strongly agree

b) agree

c) neither agree nor disagree
d) disagree

e) strongly disagree

I am confident I will be promoted as high as my
ability and interest warrant if I stay in the
Army.

I am confident I will get the kinds of assignments
I need to be competitive for promotions.

I am very likely to get assignments that match my
skills and interests if I stay in the Army.

I can get ahead in the Army doing the kinds of
work I like best. (DELETED)

200
ORGANIZATIONAL IDENTIFICATION

(Meyer & Allen, 1984)

All items use the following scale:

ARBLNG

ARTALK

AREMOT

ARPROB

ARPART

ARATCH

a) strongly agree

b) agree

c) neither agree nor disagree
d) disagree

e) strongly disagree

The Army has a great deal of personal meaning for
me.

I do not feel a strong sense of belonging to the
Army. (Reverse)

I enjoy discussing the Army with people outside

I do not feel "emotionally attached" to the Army.
(Reverse)

I really feel as if the Army's problems are my
own.

I do not feel like "part of the family" in the
Army. (Reverse: DELETED)

I think I could easily become as attached
to another organization as I am to the Army.
(Reverse)

201

ANTICIPATED WORN/FAMILY CONFLICT

All items use the following scale:

WRKBAL

CARCMD

CARCON

CARFAM

a) strongly agree

b) agree

c) neither agree nor disagree
d) disagree

e) strongly disagree

If I were to make the Army a career, I could
maintain the kind of balance I want between my
work and personal life. (Reverse)

The demands of an Army career would make it
difficult to have the kind of family life I would
like.

I forsee a lot of conflict between my work and
family life if I make a career of the Army.

If I were to stay in the Army, I could provide my
family with the opportunities and experiences I
think are most important. (Reverse; DELETED)

202

PERSON/BRANCH HATCH

BRANIN What branch are you in (your permanent assignment,
not the branch you may have been detailed to)?

(list of branches)

BRANWA If you could be in any branch you wanted, which
branch would you select?

(list of branches)

TRANBR Do you intend to try to transfer into a different
branch?

a) me - not interested in changing branches
b) me - I cannot get into the branch I want
c) yes - but I go not expect to get branch I want
d) yes - and I 66 expect to get the branch I want

BRANCH MATCH scores:
1 = Good Match

Branch in equals branch want, or branch in does
not equal branch want, but officer intends to
transfer and expects to get branch he wants
(response "d" for TRANBR); also officers with
missing data for any of the three items were
assigned the modal value of "1".

0 = Poor Match

Branch in does not equal branch want, and TRANBR
does not equal "d".

203
PRIOR CAREER ORIENTATION
Instructions for Item Set

For some officers, career plans change over time, while for
others, career plans remain constant. Here we are
interested in finding out whether or not your own plans have
changed. Please use the following scale to indicate (to the
best of your recollection) how you felt at the time of each
event/experience described below.

a) planning on an Army career

b) leaning toward an Army career

c) undecided

d) leaning toward a civilian career
e) planning on a civilian

PLNPRE When I began precommissioning training (e.g.,
USMA, ROTC) I was...

PLNCOM At the time I received my commission I was ....

204

CURRENT IORR SATISFACTION

SATWRK How satisfied are you with the kinds of work you
do in your current assignment?

a) very satisfied

b) satisfied

c) neutral

d) dissatisfied

e) very dissatisfied

Respondents were also asked to evaluate the nature of the
work in their emttemt_ess1gmmemt using the following scale:

a) very good
b) good
c) fair
d) poor
e) very poor

OPPLRN Opportunity to learn/develop skills relevant to
your career.
OPPWRK Opportunity to do work that interests you.

OPPACT Opportunity to exercise initiative/put your ideas
into action.

205

OPERATIONAL SUPPORT

Respondents were asked to provide their assessments of the
Army's effectiveness in 7 areas using the following scale:

a) very effective
b) effective

c) borderline

d) ineffective

e) very ineffective
f) don't know

PRVSUP Providing organizational support

PRVCOM Providing communication channels
PRVINF Providing information resources
PRVMAT Providing materiel resources
PRVRES Providing personnel resources

PRVFDB Providing feedback on duty performance

PRVRCG Providing recognition for a job well done

Note: This set of items was originally construed as an
organizational effectiveness scale, however the items did
not appear to do justice to such a complex construct.
Accordingly, the eighth item "Accomplishing the Army
mission" was dropped, and the items were interpreted more
narrowly as tapping the level of operational, or work-
related support provided by the Army.

Depending on the item, 1 to 14 respondents used the "Don't
Know” alternative. These responses were recoded to
"Borderline" prior to computing scale scores to avoid losing
additional cases for the analysis. The "borderline"
response (3) was the closest to the overall mean and similar
conceptually (i.e., no strong opinion either way) to "don't
know".

206

INSPIRATIONAL LEADERSHIP

Respondents were first asked the following question:

Who do you consider to be the pest leader in your current
assignment?

a) your rater

b) your senior rater

c) a peer

d) an officer not in your chain of command
e) an NCO

f) an instructor

9) other

Ratings of this leader on five dimensions were used to
compute the scale score; all ratings were based on the
following scale:
How often does this leader ...
a) always
b) frequently
c) usually
d) seldom
e) never
LDRENT Inspire enthusiasm.
LDRLOY Inspire loyalty.
LDRTRU Inspire trust.
LDRHLP Help others to understand the unit mission.

LDRGOL Help others to understand the Army's goals.