University of Pennsylvania

ScholarlyCommons
Publicly Accessible Penn Dissertations

2018

Working Off-Track: Adjunct Labor In Higher

Education
Chad Gregory Evans
University of Pennsylvania, [email protected]

Follow this and additional works at: https://repository.upenn.edu/edissertations

Part of the Educational Sociology Commons, Higher Education Administration Commons,
Higher Education and Teaching Commons, and the Organizational Behavior and Theory Commons

Recommended Citation
Evans, Chad Gregory, "Working Off-Track: Adjunct Labor In Higher Education" (2018). Publicly Accessible Penn Dissertations. 2727.
https://repository.upenn.edu/edissertations/2727

This paper is posted at ScholarlyCommons. https://repository.upenn.edu/edissertations/2727

For more information, please contact [email protected].
Working Off-Track: Adjunct Labor In Higher Education
Abstract
The proportion of non-tenure track faculty has grown over the last decade and adjuncts now constitute two-
thirds of the academic workforce. Despite this growth, there remain important limitations to our
understanding of this new faculty majority. For one, typologies for conceptualizing adjunct diversity are often
poorly aligned and make limited use of information valuable for classification. This study addresses these
issues by employing the multivariate typological method of cluster analysis. The analysis implied a “natural
typology” for adjunct faculty and suggested important nuances for fully recognizing adjunct diversity in
higher education. This dissertation also addresses limitations with regard to adjunct job satisfaction and
turnover. With lower earnings and less job security, it has typically been assumed that beginning off the tenure
line carries with it a greater risk of early career departure. However, the empirical evidence of this has been
weak. Using survival analysis and a behavioral measure of career attrition, this study confirmed the risks of
beginning off the tenure track. Furthermore, using a structural equation model, this study examined nuances
in the satisfaction and turnover intentions of different subclasses of contingent faculty members. Satisfaction
with benefits and financial satisfaction are distinct among aspiring academics and career-ending adjuncts and
this has important implications with regard to faculty retention policies.

Degree Type
Dissertation

Degree Name
Doctor of Philosophy (PhD)

Graduate Group
Sociology

First Advisor
Paul D. Allison

Keywords
adjunct, contingent, new faculty majority, non-tenure, tenure, track

Subject Categories
Educational Sociology | Higher Education Administration | Higher Education and Teaching | Organizational
Behavior and Theory

This dissertation is available at ScholarlyCommons: https://repository.upenn.edu/edissertations/2727

 WORKING OFF-TRACK: ADJUNCT LABOR IN HIGHER EDUCATION

Chad Evans

A DISSERTATION

in

Sociology

Presented to the Faculties of the University of Pennsylvania

in

Partial Fulfillment of the Requirements for the

Degree of Doctor of Philosophy

2018

Supervisor of Dissertation
_____________________

Paul Allison

Professor of Sociology

Graduate Group Chairperson

________________________

David Grazian, Associate Professor of Sociology

Dissertation Committee

Frank Furstenberg, Zellerbach Family Professor of Sociology, Emeritus

Jerry Jacobs, Professor of Sociology

 ABSTRACT

WORKING OFF-TRACK: ADJUNCT LABOR IN HIGHER EDUCATION

Chad Evans

Paul Allison

The proportion of non-tenure track faculty has grown over the last decade and

adjuncts now constitute two-thirds of the academic workforce. Despite this growth, there

remain important limitations to our understanding of this new faculty majority. For one,

typologies for conceptualizing adjunct diversity are often poorly aligned and make limited

use of information valuable for classification. This study addresses these issues by

employing the multivariate typological method of cluster analysis. The analysis implied a

“natural typology” for adjunct faculty and suggested important nuances for fully

recognizing adjunct diversity in higher education. This dissertation also addresses

limitations with regard to adjunct job satisfaction and turnover. With lower earnings and

less job security, it has typically been assumed that beginning off the tenure line carries

with it a greater risk of early career departure. However, the empirical evidence of this

has been weak. Using survival analysis and a behavioral measure of career attrition,

this study confirmed the risks of beginning off the tenure track. Furthermore, using a

structural equation model, this study examined nuances in the satisfaction and turnover

intentions of different subclasses of contingent faculty members. Satisfaction with

benefits and financial satisfaction are distinct among aspiring academics and career-

ending adjuncts and this has important implications with regard to faculty retention

policies.

ii
 TABLE OF CONTENTS

ABSTRACT .......................................................................................................... II

LIST OF TABLES ............................................................................................... VI

LIST OF ILLUSTRATIONS ................................................................................ VII

CHAPTER 1: INTRODUCTION............................................................................ 1
Theoretical Framework .................................................................................................................. 2

Problem statement ......................................................................................................................... 2

Research Questions ...................................................................................................................... 3

Research design ............................................................................................................................ 3

Significance .................................................................................................................................... 4

CHAPTER 2: CLASSIFYING NON-TENURE TRACK FACULTY ....................... 6

Abstract ........................................................................................................................................... 6

Introduction .................................................................................................................................... 6

Literature Review ........................................................................................................................... 7

Classifying Adjuncts ..................................................................................................................... 7
Research Problem ....................................................................................................................... 8

Research Questions .................................................................................................................... 10

Method........................................................................................................................................... 10
Measures ................................................................................................................................... 11

Results .......................................................................................................................................... 11
Classification of Full-time Adjunct Faculty ........................................................................... 12
Classification of Part-time Adjunct Faculty .......................................................................... 13

Discussion .................................................................................................................................... 16

Conclusion .................................................................................................................................... 18

References .................................................................................................................................... 18

Appendix ....................................................................................................................................... 21
iii
CHAPTER 3: NON-TENURE TRACK JOB SATISFACTION AND TURNOVER
INTENTIONS ...................................................................................................... 24
Abstract ......................................................................................................................................... 24

Introduction .................................................................................................................................. 24

Literature Review ......................................................................................................................... 25

Turnover .................................................................................................................................... 25
Job Satisfaction ....................................................................................................................... 26
Satisfaction of Adjunct Types ................................................................................................ 28
Limitations of Earlier Work ..................................................................................................... 29
Proposed Model ....................................................................................................................... 30

Research Questions .................................................................................................................... 30

Adjunct Job Satisfaction Hypotheses ......................................................................................... 31
Turnover Hypotheses ................................................................................................................. 32

Method........................................................................................................................................... 33
Measures ................................................................................................................................... 34
Analytic Approach ...................................................................................................................... 35

Results .......................................................................................................................................... 36

Discussion .................................................................................................................................... 41

Conclusion .................................................................................................................................... 43

References .................................................................................................................................... 43

Appendix ....................................................................................................................................... 48

CHAPTER 4: ATTRITION: NON-TENURE TRACK FACULTY LEAVING

ACADEMIA ........................................................................................................ 51
Abstract ......................................................................................................................................... 51

Introduction .................................................................................................................................. 51

Literature Review ......................................................................................................................... 52

Conceptual Framework .............................................................................................................. 52
Theory on Turnover.................................................................................................................... 53
Prior Research on Faculty Career Attrition ................................................................................ 55
Limitations of Earlier Research .................................................................................................. 56
Proposed Model ......................................................................................................................... 57

Research Questions .................................................................................................................... 57

Hypotheses ................................................................................................................................ 58

Method........................................................................................................................................... 60
Sample ....................................................................................................................................... 60

iv
 Measures ................................................................................................................................... 60

Results .......................................................................................................................................... 61

Discussion .................................................................................................................................... 65

Conclusion .................................................................................................................................... 67

References .................................................................................................................................... 67

Appendix ....................................................................................................................................... 74

CHAPTER 5: CONCLUSION ............................................................................. 75

Contribution .................................................................................................................................. 76

Practical implications .................................................................................................................. 76

Future Work .................................................................................................................................. 77

v
 LIST OF TABLES

Table 1: Crosstabulation of Full-time Adjunct Clusters...................................................12

Table 2: Crosstabulation of Part-time Adjunct Clusters ..................................................14

Table 3: Descriptive Statistics of Quantitative Variables in the Analysis.........................22

Table 4: Variables Used in Cluster Analysis ..................................................................23

Table 5: Adjunct Satisfaction .........................................................................................38

Table 6: Bayesian Structural Equation Models Predicting Turnover Intentions ..............40

Table 7: Descriptive Statistics (n=8418) ........................................................................48

Table 8: Standardized Factor Loadings .........................................................................49

Table 9: Latent Factors with Satisfaction Items..............................................................50

Table 10: Cox Proportional Hazards Models..................................................................64

Table 11: Life Table .......................................................................................................74

Table 12: Study Entrants by Year ..................................................................................74

vi
 LIST OF ILLUSTRATIONS

Figure 1: Bayesian Structural Equation Model (Full Model with All Covariates) .............36

Figure 2: Survival Curves ..............................................................................................63

Figure 3: Changes in Tenure Systems (1993-2015) ......................................................74

vii
 CHAPTER 1: INTRODUCTION

Decades ago, the work of higher education was conducted mostly by long-term
faculty with tenured contracts. Visiting, clinical and in-residence faculty had a role, but
that role was limited to short-term course assignments or ancillary support that tenured
faculty were not able to provide. In recent years, however, there has been an eruption of
criticism regarding the use of adjunct faculty, as it has become clear that adjunct faculty
are now permanent fixtures on college campuses. Scholars have documented the
uptick in non-tenure track faculty, tracing back to the late 1960s. At that point, a mere
22% of faculty members were tenure-ineligible (Schuster and Finkelstein 2006).
However, by the Fall of 2009, nearly two-thirds of all faculty were non-tenure track. This
dramatic change has revolutionized the academic workforce: fixed-term faculty are now
the new faculty majority.

This fundamental shift in the academic workforce has attracted substantial

research for decades. Considerable efforts have been made to understand the historical
and economic reasons for this extensive growth (Schuster and Finkelstein 2006).
Others have examined how uncoordinated and decentralized hiring practices relate to
fixed-term faculty (Cross and Goldenberg 2003). Some have focused on the job
conditions and work responsibilities of these faculty members (Baldwin and Chronister
2001, Schell and Stock 2001). Others have examined student persistence and
undergraduate academics under the instruction of adjuncts (Jaeger and Eagan 2009,
Eagan and Jaeger 2008). These are only a few of the many themes undertaken due to
this revolutionary change in the contracts of postsecondary faculty.

The scope of this dissertation is focused on two key aspects of non-tenure track
faculty and their work. First, faculty adjuncts are far too often treated as a homogenous
block with uniform experiences working in academia. They are portrayed as struggling
with poverty wages, disrespected by their tenure-line peers, and universally exploited.
No one would dispute the importance of these topics and possibilities. No doubt, some
members of the adjunct population do work under such regrettable conditions. However,
before implementing policies to ameliorate these conditions, it is of great importance to
consider just how widespread these conditions and experiences are. It is possible that
not all adjuncts experience them—maybe not even a majority. After all, we know that
some adjuncts teach outside of a professional career because they enjoy sharing their
expertise. Others work part-time to stay productive during their early retirement years.
In short, there is surely diversity among postsecondary adjuncts, and an understanding
of this diversity is sorely needed.

The second component of this dissertation relates to the job satisfaction,

turnover, and permanence of non-tenure track faculty in higher education. It is a fact
that non-tenure track faculty earn less and are less likely to receive health and
retirement benefits. By definition, they also do not have tenure—a desirable contractual
guarantee encouraging faculty commitment. Adjuncts also have less autonomy, control
over their work and professional support. It is intuitive that employees working in such
circumstances would be less satisfied and more likely to abandon their jobs and

1
potentially their careers. However, demonstrating this empirically is a formidable
challenge.

Theoretical Framework
Extensive research on employee withdrawal and turnover exists. Researchers in
this line of work typically approach the topic from one of four theoretical angles. Some
focus specifically on the psychological commitment that bonds employee with employer
(Porter, Crampon, and Smith 1976, Porter et al. 1976, Meyer and Allen 1991). Workers
commit due to the implied costs of leaving as well as normative pressures. Another
group of researchers has approached turnover from the theoretical lens of
embeddedness (Mitchell and Lee 2001). This line of research draws on social networks,
employee “fit,” and cost-avoidance to understand worker decisions (Lee, Burch, and
Mitchell 2014). The unfolding model—a third perspective—highlights the importance of
unexpected shocks causing employees to leave (Hom et al. 2017, Holtom et al. 2008,
Weller et al. 2009).

Vroom’s (1964) work on expectancy theory may be the most influential of all.
Expectancy theory views motivation as a cognitive process and focuses on how
structural, attitudinal and environmental factors influence this process. In the context of
work, this perspective implies that employees maintain the work arrangements that fulfill
their job expectations. When unfulfilled, workers pursue other options like exerting less
effort, detaching from their work, or leaving a job altogether (March and Simon 1958).
Job satisfaction is central to expectancy theory because it is viewed as an expression of
“fit.” Satisfied workers feel more aligned and connected to their employer and
colleagues. Dissatisfied workers feel alienated and detached from their work.

Researchers integrate job satisfaction into turnover models in several ways.

Herzberg, Mausner, and Snyderman's (1959) argued that there were two principal
dimensions to job satisfaction. Some work conditions mapped explicitly onto a factor
measuring satisfaction while other work features mapped onto a factor measuring
dissatisfaction. These factors were viewed as entirely independent in their framework.
A second important job satisfaction model is Hackman and Oldham’s (1980) “Job
Characteristics” model. This framework was very important for elaborating on the
intrinsic aspects of job satisfaction like task variety, meaningfulness and task feedback.
In the domain of higher education research, however, the work of Kalleberg (1977) truly
stands out. Like Hackman and Oldham (1980), Kalleberg also recognized the intrinsic
components of job satisfaction. However, his model was valuable for its incorporation of
external factors of job satisfaction as well. Specifically, he found six dimensions to the
job satisfaction construct: intrinsic value, convenience, financial, relationships with co-
workers, career and resource adequacy.

Problem statement
While research often treats non-tenure track faculty as a homogenous block,
there is an important line of typological scholarship recognizing adjunct heterogeneity.
Notably, Gappa and Leslie (1993) identified career-enders, adjunct experts, freelancers
and aspiring academics. Baldwin and Chronister (2001) classified full-time adjuncts
according to principal work responsibilties. While these studies (and others) have done
a great service, they have important limitations that this dissertation aims to address.
For one, this line of research has led to a proliferation of labels for adjuncts without
2
clearly demonstrating that actual differences exist in new classes. As a result, there has
been redundancy in some cases and conceptual vagueness in others. To truly carve out
distinct classes of adjunct faculty members, this line of research would benefit from a
method inferring how many clusters are justified in the adjunct population.

The second limitation this study aims to address relates to the failure to utilize
information that is potentially valuable for creating an adjunct typology. Commonly,
researchers build typologies around two or sometimes three characteristics they believe
to be self-evident. The problem is that this approach potentially excludes variables
important for classification. Ethnographers get around this to some extent by pouring
over fieldnotes. However, even there, it is not certain that the right information is
employed for the purposes of creating a robust typology. Ideally, a method would be
employed that systematically extracts and draws on all information useful for creating
distinct faculty types.

This dissertation also aims to address limitations in research on adjunct

institutional and career turnover. For one, the timing of faculty attrition is rarely
examined and when it has, the focus has been mainly on traditional, tenure-line faculty.
Investigating the departure of adjunct subclasses is even more rare. Research on the
turnover of adjunct experts or freelancers, for example, is unheard of. Secondly, this line
of work commonly oversimplifies job satisfaction by measuring it on a unidimensional
scale (e.g., “overall satisfaction”). It is well-known that this overlooks important aspects
of job satisfaction and produces upwardly biased estimates (Oshagbemi 1999). Better
research designs can correct this.

Research Questions
Regarding adjunct classification, what can a large dataset with extensive
information on non-tenure track faculty tell us about the types of adjuncts working in
higher education? Do natural clusters emerge from the data among full-time and part-
time adjunct faculty? How does the typology suggested by a cluster analysis compare to
earlier findings based on traditional typological methods?

This study also aims to address research questions related to adjunct turnover
and career attrition. It seems likely that non-tenure track faculty leave their careers at
higher rates than tenure-line faculty. Can this be demonstrated empirically? How does
job satisfaction fit into the institutional turnover of postsecondary adjuncts? Are there
differences among different kinds of non-tenure track faculty (i.e., adjunct experts,
career-enders, aspiring academics, and freelancers)? Which work activities and
institutional characteristics are related to turnover?

Research design
This dissertation addresses the problem of adjunct classification by employing
cluster analysis—a multivariate method of classification. Specifically, a k-means
algorithm is utilized, which extracts natural structure in a data source by clustering
according to similarity. Doing this successfully requires a rich and relevant data source.
To this end, this project acquired data from the HERI Faculty Survey (2010)—a unique,
3
cross-sectional survey that contains extensive information on the job experiences,
demographic characteristics, and institutional features of 8,418 non-tenure track faculty
members.

This rich dataset was also extremely valuable for the examination of turnover
intentions among non-tenure track faculty members. To do so, these data were
analyzed in a structural equation model. The choice of this methodology was driven by
its capacity to incorporate a more comprehensive and appropriate measure of job
satisfaction. This allowed for a test of whether adjunct faculty subclasses may
experience job satisfaction in different ways and how this all relates to their decisions to
leave their institutions.

Finally, recognizing the importance of career attrition, this study also acquired
longitudinal data from the Survey of Doctorate Recipients (1993-2013). By utilizing
information from doctorate recipients pursuing academic appointments, this study was
able to calculate nonparametric estimates of career permanence among faculty
beginning on and off the tenure line. Furthermore, these data were examined using Cox
proportional hazards regression, allowing for deeper exploration of how work and
institutional features relate to attrition behavior.

Significance
Higher education has not been isolated from the range of important social and
economic changes over the last decades (Schuster and Finkelstein 2006). The structure
of the economy has shifted from a material foundation to one based on information and
knowledge. Education has changed in the public’s eye from being a public good to
resembling more of a private good. Coupled with a demographic transformation, new
technologies, and marketplace pressure, administrators in higher education have
consistently been forced to reexamine costs. As it is the largest institutional cost,
instruction has often been the target. The usage of lower-cost, adjunct faculty has
slowly grown over decades and, today, non-tenure track faculty members constitute over
two-thirds of the American professoriate.

Of the many research questions engendered by the new faculty majority, this
study prioritized questions related to adjunct classification and turnover. By classifying
adjuncts better, this study helps do better science. This is because distinguishing the
objects of inquiry is fundamental to the scientific method. A clearer typology also
improves public policy by distinguishing subgroups that can be targeted. If there are
inadequate supports and conditions among the adjunct population (as most believe),
identifying who those adjuncts are and tailoring interventions with their interests in mind
would be prudent.

This study is also justified by the fact that contingent faculty—not tenured
faculty—are now the ones most responsible for executing the primary mission of higher
education. The working conditions and careers these teachers experience, many
believe, must have some kind of impact on their ability to teach and mentor young
people. Are non-tenure track faculty satisfied with their work? Do they have enough job
security to feel connected to their institutions? Are they capable of participating in and
contributing to the positive learning environments we aim to establish on campuses?
With as many as 12 million undergraduates taking coursework under the direction of a
non-tenure track faculty member, it is high time to develop a more thorough
4
understanding of these individuals and their work environments (National Center for
Education Statistics 2015).

There is also concern that an increased use of non-tenure track labor may open
up new modes of stratification of educational quality and opportunity in the United
States. If tenured faculty perform better at promoting student learning, then institutions
more reliant on adjunct labor are more likely to produce inferior educational results.
Students at elite colleges and universities (where there are more tenured faculty), by
contrast, may have their advantages compounded thanks to stable and more
experienced faculty members. A greater understanding of non-tenured faculty will help
us assess the role of contingency in the stratification of educational opportunity.

Finally, institutions also have a very large stake in the careers and job
satisfaction of their contingent faculty. Today, contingent faculty are a stopgap, allowing
administrators to cope with fluctuations in enrollment and the withdrawal of government
support. But there are subtle costs associated with this system that may add up in a
substantial way. It is not clear how colleges and universities can continue to attract “the
best and the brightest” if positions are increasingly off the tenure line. There are
potentially large losses in the process of hiring, orientating and training short-term
faculty. By elucidating the kinds of adjuncts in higher education and how the turnover
process unfolds for them, this dissertation makes a substantial contribution to
scholarship.

5
 CHAPTER 2: CLASSIFYING NON-TENURE TRACK FACULTY

Abstract
With large increases in the number of faculty members working off the tenure
track, scholars have developed numerous typologies to conceptualize this
heterogeneous group of academic laborers. However, the proliferation of typologies has
sometimes obscured patterns more than elucidated them. These studies are also
limited in the information they use for classification. The following study employs cluster
analysis—a multivariate typological approach—to identify natural groupings of adjunct
faculty in higher education. By applying this procedure to a dataset with extensive
information (2010 HERI Faculty Survey, n=8,418), this study brings greater information
to bear on the task of adjunct classification. The findings of this study suggest an eight-
fold typology with important connections to past typological research.

Introduction
Schuster and colleagues recently considered updating their influential book The
American Faculty (2006); however, they concluded that an update was impossible.
Postsecondary education “had transformed to such an extent [that an update] could not
adequately capture the magnitude and significance of what [had] transpired (Finkelstein,
Conley, and Schuster 2016).” This view is broadly shared. Colleges and universities
have changed in many ways, including how they are financed, how they use technology,
the students who attend them and the programs of study they offer. They have also
changed with regard to their workforce. Beginning in the 1960s, postsecondary
institutions began hiring increasing numbers of faculty off the tenure line. As “adjuncts”
earn lower salaries and receive fewer benefits, the broader use of them has resulted in
important cost-savings. However, few would have predicted the formation of a “new
faculty majority.” Indeed, over two-thirds of academic faculty today work off the tenure
line (Finkelstein et al. 2016).

An important observation is the diversity among non-tenure track faculty (Kezar

and Sam 2010). Adjuncts have different job responsibilities and many reasons for
working off the tenure line. They experience diverse work environments and have a
multitude of job titles and academic backgrounds. Some work full-time and some work
part-time. For this reason, researchers have developed adjunct typologies in order to
better conceptualize and understand this heterogeneous group of academic laborers.
As the number of contingent faculty has grown, so too have the number of typologies for
describing them. However, in many instances, additional typologies have not resulted in
greater conceptual clarity or deeper insights. Often, typologies are redundant or re-
describe earlier adjunct classes in trivial ways. This line of research is also limited by
not making full use of the entire span of information available for classification.

This study addresses these issues by employing a multivariate approach to

classification called cluster analysis. Utilizing a k-means algorithm, the procedure
derives structure in an appropriate data source by grouping units according to degree of
similarity. The adjunct clusters that emerge constitute a natural typology. What can a
large dataset tell us about the types of non-tenure track faculty working in higher
education? Is there natural structure in the data that would be useful for a typology
6
scheme? How would such a typology compare to earlier work that has sought to classify
non-tenure track faculty? An important strength of this design is its use of data with
unparalleled information on non-tenure track faculty (HERI Faculty Survey 2010). This
dataset includes a broad range of variables related to non-tenure track job experiences,
demographic characteristics, institutional features, and more. As cluster analysis is most
useful when applied to broad information on the objects of classification, the
expansiveness of this dataset is integral. Drawing on contemporary data also has the
advantage of updating a field that has changed rapidly over the last decade.

It is crucial to develop a better typology of non-tenure track faculty.

Distinguishing the objects of inquiry is part of science, and better categories means
better science. By improving how we conceptualize types of adjunct faculty, we advance
our knowledge and understanding rooted in and around those categories. Typologies
also suggest different occupational needs and different potential responses to
interventions. As higher education is currently undergoing serious reforms, having better
typologies will help policy-makers and administrators implement targeted policies that
are efficient and effective. It is well known that many non-tenured track faculty work in
challenging environments, and it is important for policy changes to be tailored to the
varying needs of different subgroups.

Literature Review
Classifying Adjuncts
Classification1 is the method of systematically arranging objects into groups
(McQuitty 1987). This process is fundamental to learning and understanding, and it is
only through classification that scholars, scientists, and people in general come to
identify, differentiate and understand the objects of knowledge. The set of categories
that emerge from this process is called a typology. Typologies often pertain to some
domain like appearance, role, function, behavior, or activity. They are also typically
developed with some purpose or goal in mind. However, there is also great interest in
developing typologies that can be applied more generally.

The classification of postsecondary adjuncts has a long history. Tuckman's

(1978) seminal work surveyed nearly four thousand part-time faculty members,
classifying adjuncts according to their reasons for working part-time. This variable-
centered analysis suggested seven principal types of part-time faculty: semiretireds,
graduate students, hopeful full-timers, full-mooners, part-mooners, homeworkers, and
part unknowners. The largest category was full-mooners (27.6%) who, in addition to
their adjunct position, held full-time careers outside of academia. Graduate students
were another large class (21.2%), completing their graduate studies while working part-
time at a nearby institution.

The adjunct typology created by Tuckman was subsumed into a more general
framework in Gappa and Leslie’s Invisible Faculty (1993). Interviewing part-time faculty
at eighteen colleges and universities, Gappa and Leslie identified career enders,
experts2, freelancers, and aspiring academics. Career enders referred to faculty who

1More specifically, taxonomic classification.

2Gappa and Leslie specifically called these faculty members “specialists, experts, and
professionals.” This study refers to them as “experts” to simplify the language.
7
were in the process of retiring from the workforce. Many of these individuals were not
career academics, but instead had worked in the private sector. Part-time faculty in this
class worked for supplemental income or simply because they enjoyed teaching.
Experts were hired for their specialized knowledge or experience. Freelancers were
mostly faculty who wanted to supplement the income earned from a career outside of
academia. These faculty were also commonly homemakers, taking care of children and
domestic chores on the side. The final category consisted of aspiring academics3.
These were Ph.D. recipients who were seeking full-time, tenure-track appointments.

In Teaching without Tenure (2001), Baldwin and Chronister conducted interviews

with faculty members at twelve colleges and universities. This study was important for
its incorporation of full-time, non-tenure track faculty—a group that had been growing
rapidly since the early 1990s. Their typology was centered on principal work
responsibilities. Teachers spent most of their time teaching and researchers spend most
of their time in research. They also identified significant numbers of adjuncts working in
an administrative capacity (“administrators”). A residual category was retained for
faculty spending the majority of their time on something else. This included lab
technicians, programmers, and faculty members in community service.

Recognizing the rising number of adjuncts providing online instruction, Schnitzer

and Crosby (2003) identified additional classes of adjunct faculty. Philosophers were
faculty members employed in a field outside the area of study they took a degree in.
While some of these faculty members did hold philosophy degrees, this category also
included individuals with backgrounds in the humanities, religion and other fields with
weak academic job markets. Full-time part-timers were adjuncts who created full-time
work schedules by piecing together multiple part-time positions at different institutions.
Bedford and Miller (2013) found that full-time part-timers constituted the largest category
of adjuncts in their study. They also found substantial numbers of full-time instructors
who, in addition to their full-time job, provided part-time instruction online. Other faculty
groups they identified included full-time administrators, recent graduates, and
employment seekers.

The studies presented here are only a few of the adjunct typologies proposed
over the last decades. Researchers have also distinguished non-tenure track faculty
according to their preference for tenure-line or full-time work, institutional titles4 (Shavers
2000), skill-level (Wagoner 2007) and professional ambitions (Bedford 2009, Carnevale
2004).

Research Problem
This study aims to address two main limitations of the typological research on
postsecondary adjuncts. First, with the growth of adjunct faculty, there has been a
proliferation of adjunct typologies. Some of these were reported above. However, more
typologies have not always resulted in more insights. Maynard and Joseph (2008), for
example, examined “underemployed adjuncts”—a category described earlier as “hopeful

3 In this formulation, aspiring academics also include graduate students, itinerants (a.k.a freeway
fliers), and those aspiring only to a full-time position (tenure-track or not).
4 Berry (2005) found over 50 different titles for adjunct faculty.

8
full-timers” (Tuckman 1978) and “aspiring academics5” (Gappa and Leslie 1993).
Schnitzer and Crosby (2003) and Bedford and Miller (2013) described the experiences
of full-time part-timers working in online education. However, this is just a nuanced spin
on the concept of “itinerants” or “freeway fliers” defined years earlier (Cross and
Goldenberg 2009, Schell and Stock 2001). In short, it is not really adjunct classes that
have proliferated as much as the labels used to describe them. This line of research
would benefit from a study that systematically and objectively determines how many
distinct classes of adjuncts are justifiable. Proposed adjunct types would then avoid
redundancy and provide greater conceptual leverage.

The other important limitation of this line of research is its constrained use of
information. In simple bivariate designs, two variables of theoretical importance are
cross-tabulated, resulting in a four-quadrant classification schema. In other studies,
researchers pour over interview data or field notes and ultimately settle on a number of
classes that seems appropriate and consistent with the data. In both cases,
classification fails to make full use of available information. In the first case, the
procedure utilizes two principal variables but ignores any additional components that
may be of importance for classification. The second case is limited by human
perceptivity. At some point, the human mind is incapable of recognizing patterns in
dense information. Such approaches are often incapable of capturing the naturally
occurring interactions between components that constitute an entity.

For these reasons, Robins et al. (1996) and others have advocated for broader
use of multivariate methods of classification like cluster analysis. Cluster analysis
utilizes information from a high-dimensional dataset and separates units into groups
according to similarity. There are many algorithms for accomplishing this task. One
popular class that performs this task particularly well is the method of k-means. Jain
(2010) attributes the popularity of k-means to its “ease of implementation, simplicity,
efficiency, and empirical success.” This algorithm has been used in the natural and
social sciences for decades (Le Roch et al. 2003, Clatworthy et al. 2005, Klemmack et
al. 2007), although it has never been applied to the problem of classifying non-tenure
track faculty.

There are limitations of course. This work does not and cannot replace
foundational research conducted by domain experts in the field of higher education.
However, by approaching the problem of adjunct classification with a new methodology
and a new dataset, this study helps to validate what has been found in the past. The
exploratory nature of this study may also suggest novel ideas and fruitful avenues for
future research on adjunct faculty. This study is also limited in its classification of a
subset of faculty—non-tenure track faculty—instead of all professors in higher education.
While it was necessary to limit the scope for this chapter, a larger study that integrates
classifications of tenure-line faculty is certainly merited. This is particularly important on
account of the role of tenure in the highly differentiated environment of academia.

5 Gappa and Leslie (1993) argued that their concept stressed the desire to be “fully participating,

recognized, and rewarded members of the faculty” of similar status.

9
Research Questions
What can a large dataset tell us about the types of non-tenure track faculty
working in higher education? Is there natural structure in the data useful for classifying
full-time and part-time non-tenure track faculty? How does the emergent typology
scheme compare to earlier work? This study employs a k-means clustering algorithm to
discover a natural typology of non-tenure track faculty. A strength of this study design is
its use of a high-dimensional dataset and cluster analysis—a method that has not been
utilized before in this line of research. By drawing on new tools and new data, this study
complements earlier typological research, working towards an adjunct typology of
general use. As this method is exploratory, there are no formal hypotheses to be tested.

Method
The goal of cluster analysis is to discover a natural grouping within a collection of
data objects, patterns, or points. Operationally, for a set of data points and a pre-
specified number of clusters, the basic k-means algorithm iteratively6 minimizes the
within-cluster sum of squared errors. This study specifically employed the Hartigan and
Wong (1979) algorithm, which makes improvements to the computational efficiency of
this process. To validate resultant clusters, the groups were cross-tabulated across
various domains of non-tenure track faculty characteristics, including: demography,
institutional, departmental, and employment features7. This study conducted two
separate cluster analyses, one for full-time faculty and one for part-time faculty. This
was because the survey instrument contained an important battery of questions that only
pertained to part-time faculty8. For more details on the method, see Aldenderfer and
Blashfield (1984).

The data in this study came from the Faculty Survey (2010), a cross-sectional
survey conducted by the Higher Education Research Institute (HERI). HERI has been
collecting faculty survey data since 1978, however, the focus of earlier iterations has
been on traditional tenured and tenure-track faculty teaching undergraduate students in
the United States. Beginning in 2007-2008, HERI began developing a separate series of
questions for part-time faculty and expanded their part-time sample, recognizing the
important role of part-timers in higher education. These changes were of extreme
importance to this study because the performance of the clustering algorithm and the
representativeness9 of the resultant typology depend on it10. When data representation
(choice of variables) is good, resultant clusters are more likely to be compact and

6 Another popular class of methods is agglomerative.

7 Variation across work characteristics was most apparent and therefore the focus of this paper is
on work variables.
8 Exploratory analysis was also conducted on the pooled data.
9 The variables that are chosen should reflect the content of the domain that the researcher is

interested in uncovering a natural typology Mandara (2003).

10 The survey created for 2007-2008 was the first to introduce a module with approximately thirty
questions specifically for part-time faculty. This included questions about preference for full-time
work, whether the faculty member has an outside career, and how many other institutions the
faculty member teaches at (among others). At the time this dissertation research began, the
2010 instrument was the most recent instrument available that contained this important
information about part-time faculty.
10
distinguishable from other clusters. This study utilizes data from the instrument
distributed in 2010.

The dataset included information on 4,527 full-time non-tenure track faculty

members and 3,891 part-time non-tenure track faculty members. In general, missing
information was very low11. However, because the analysis was conducted on
numerous faculty variables, a missing data method was necessary to retain the near-
complete data. To this end, data were singly imputed. See the methodological
appendix for more information on this procedure.

Measures
A strength of this study is its inclusion of extensive information on non-tenure
track faculty. The cluster analysis utilized 55 variables for full-time faculty and 82
variables for part-time faculty. All of these variables are tabulated in the appendix. In
short, variables included a range of work features, including: number of courses,
principal activities, stress, and productivity (to name a few). All categorical variables
were coded into binaries for the cluster analysis. Quantitative variables maintained their
numeric values. See the appendix for descriptive statistics on the quantitative variables.

Results
In order to discuss the emergent clusters in this study, it was necessary to label
them. However, this is somewhat paradoxical because the goal of this study and its
methodology was to avoid unnecessarily reducing phenomena to a single dimension.
Cluster analysis is novel specifically because of how it utilizes a multitude of interactions
and correlations to identify “natural groups” in a data source. The labels proposed in this
study, thus, are made with a light hand. What is most important in the section that
follows are the clusters of interrelated variables. Unlike univariate and bivariate
typological approaches, cluster analysis allows for great complexity and nonlinearity in
the formation of natural groups. Three clusters resulted from the analysis of full-time
faculty and five clusters emerged among the part-time faculty.

11Only four full-time faculty variables had more than 10% missingness. Only ten part-time faculty
variables had more than 10% missingness.
11
Classification of Full-time Adjunct Faculty

Table 1: Crosstabulation of Full-time Adjunct Clusters

Cluster 1 (n=2686) Cluster 2 (n=716) Cluster 3 (n=1125) Sign.

Teaching 92% 87% 45% **
Research 4% 6% 2% **
Administration/Other 4% 7% 53% **
Instructor 27% 32% 23% **
Lecturer 26% 31% 20% **
Professor 48% 38% 56% **
Union member 14% 16% 11% **
Health benefits 95% 94% 96%
Retirement 94% 90% 98% **
Avg. Salary $55,277 $49,720 $75,026 **
Avg. Courses 3.1 3 1.8 **
Prof. Dev. Rating 0 -0.3 0.2 **
PhD 37% 45% 33% **
Faculty very respectful 53% 49% 47% **
Administators very considerate 21% 14% 27% **
Research valued 70% 57% 69% **
Teaching valued 91% 85% 89% **
Sciences 22% 22% 10% **
Soft/Applied 33% 26% 50% **
Humanities/Arts 22% 31% 17% **
Health Sciences 13% 6% 14% **
Social Sciences 10% 14% 9% **
Public 40% 43% 30% **
Highly Selective 10% 11% 9%
Age 48.28 45.38 51.37 **
Male 45% 39% 43% *
Married 77% 74% 80% *
Number of Children 1.47 1.29 1.72 **
Core Teaching Peripheral Administrative
Suggested Label
Faculty Teaching Faculty Adjunct
* p-value<0.05 ** p-value<0.01

Administrative Adjunct: A distinct faculty type identified in this study was the
“administrative adjunct.” These faculty members work full-time and, unlike other full-time
adjunct types, tend to report administration12 as their principal work responsibility (Table
1). Administrative adjuncts also typically have higher academic rank (although
untenured), higher salaries (over $75,000 annually), more benefits and greater
opportunities for professional development. These adjuncts also tended to be older.
This all supports the assertion that this is the managerial class of adjuncts. Interestingly,
administrative adjuncts tended to have professional credentials (although some have

12 Administration or activity other than teaching or research.

12
Ph.D.’s as well). They were also overrepresented in applied fields like public
administration, law, journalism, and education.

Two Classes of Full-time Teaching Faculty

Core Teaching Faculty: From Table 1, like administrative adjuncts, core teaching
faculty are full-time employees and central to their institution’s mission. This centrality is
reflected in their academic rank (which tends to be higher), their salaries (they earn ten
percent more than peripheral teaching faculty), their benefits, and opportunities for
professional development. Core teaching faculty constitute the largest group of adjunct
faculty members in this study.

Peripheral Teaching Faculty: Peripheral teaching faculty (Table 1) teach full-time for a
college or university but exhibit less integration in their institutions. These adjuncts
generally have lower rank, receive less compensation, and undergo less professional
development. However, compared to administrative adjuncts and core teaching faculty,
peripheral instructors are actually more likely to possess Ph.D.’s. They are also more
likely to be women working in the arts and humanities. Probably because they are less
integrated into their institutions, the workplace attitudes of peripheral teaching faculty
tend to be more critical of colleagues, administrators and their institutions more
generally.

Classification of Part-time Adjunct Faculty

13
 Table 2: Crosstabulation of Part-time Adjunct Clusters
Cluster 1 (n=352) Cluster 2 (n=1405) Cluster 3 (n=1029) Cluster 4 (n=916) Cluster 5 (n=189) Sign.
Teaching 99% 98% 97% 98% 73% **
Research 0% 0% 2% 0% 2% **
Administration/Other 1% 2% 1% 1% 25% **
Instructor 64% 60% 34% 58% 41% **
Lecturer 23% 21% 46% 27% 31% **
Professor 13% 19% 21% 15% 28% **
Union member 15% 14% 35% 18% 24% **
Health benefits 45% 32% 79% 37% 70% **
Retirement 49% 36% 77% 40% 69% **
Total Institutional Salary $12,216 $11,049 $23,666 $11,605 $24,117 **
Payment Per Course $3,105 $3,241 $4,626 $3,265 $3,487 **
Avg. Courses 2.5 1.8 3.1 2.1 2.1 **
Prof. Dev. Rating -0.1 -0.1 0.2 -0.2 0.4 **
PhD 21% 14% 35% 18% 29% **
Involuntary PT 64% 49% 72% 59% 60% **
Sought FT 50% 32% 72% 44% 53% **
Career Outside Academia 49% 54% 11% 47% 24% **
# Other Institutions 0.8 0.5 0.7 0.6 0.6 *
Faculty very respectful 50% 64% 47% 55% 53% **
Administators very considerate 15% 27% 14% 19% 22% **
Research valued 48% 69% 61% 55% 64% **
Teaching valued 83% 95% 87% 88% 89% **
Soft/Applied 40% 48% 25% 41% 48% **
Humanities/Arts 29% 18% 37% 24% 22% **
Other field 31% 35% 37% 34% 30%
Public 28% 29% 52% 36% 35% **
Highly Selective 3% 3% 10% 4% 7% **
Age 50 52 50 50 53 **
Male 45% 53% 41% 46% 49% **
Married 74% 79% 77% 79% 78%
Number of Children 1.6 1.9 1.5 1.7 1.7 **
Aspiring Aspiring PT Admin
Suggested Label freelancer freelancer2
Academic II Academic Adjunct
* p-value<0.05 ** p-value<0.01

Adjunct Administrator: Unlike other classes of part-time faculty, the principal work
responsibilities of adjuncts in this class are often administrative (Table 2). To this end,
they are offered more opportunities for professional development than other part-time
faculty. They also receive greater institutional compensation than other part-timers
(salary) and most report receiving health and retirement benefits. With this kind of
security, only a quarter of adjunct administrators hold positions outside of academia.
Adjunct administrators work in academically selective institutions at higher rates.

Academic Aspirant: Although part-time employees, Academic Aspirants tend to be

more integrated into their departments than any other class of part-time adjunct. They
teach the most courses on average and are more likely than other teaching-focused
adjuncts to receive professional development (Table 2). They tend to receive better
salaries and earn much more on a per-course basis than any other class of part-time
adjunct. Most of them (about three-quarters) receive health and retirement benefits.
They also are more likely to hold the title of lecturer—rather than instructor—which
generally connotes higher status.

It is not clear why Academic Aspirants are afforded this relatively privileged
position among part-time faculty. Perhaps it is because more of them hold traditional
faculty credentials (Ph.D.’s) or because they maintain a more concentrated presence
within academia (nine out of ten do not pursue careers outside of academia). Or
perhaps it relates to their higher participation in faculty unions. In any case, these
faculty certainly are not satisfied with their part-time positions. Of all part-time faculty,
14
they are the most likely to have pursued full-time faculty positions in the recent past and
the least likely to be working part-time by their own volition. These setbacks may be why
they express more negative views about colleagues and administrators than other
adjunct types. Academic Aspirants work disproportionately in the humanities and at
selective, public colleges and universities.

Journeyman Aspirant: Journeyman aspirants tend to be postsecondary instructors—

not lecturers. As a class, they bear resemblance to Academic Aspirants in some ways.
They teach a great deal—nearly two and a half courses every term—and their classes
are often in the humanities (Table 2). Among part-timers, only Academic Aspirants
teach more courses. Journeyman Aspirants have often pursued full-time academic work
in the recent past and been unsuccessful in that endeavor. Like Academic Aspirants,
they often have negative perspectives with regard to their faculty co-workers and
administrators. They also tend to say that research may not be a priority for their
institution.

They differ from Academic Aspirants principally in their engagement with outside
employment. While only a small percentage of Academic Aspirants have careers
outside of academia, nearly half of Journeyman Aspirants do. They are also more likely
to teach at other academic institutions—sometimes referred to as itinerancy. Coupled
with the workload they report at their surveyed institution, it is clear that these adjuncts
work a lot. With fewer children and a slightly lower marital rate, these adjuncts seem to
be placing a strong emphasis on their careers—even if these are not traditional career
pathways.

Freelancing Aspirant: Freelancing Aspirants tend to be professionals teaching a couple

of classes for their institution (Table 2). Among other part-time adjuncts, Freelancing
Aspirants resemble Journeyman Aspirants in important ways. Like Journeymen, they
also work as teachers in academia and tend to hold the title of instructor. It is also not
uncommon for them to be pursuing full-time careers outside of academia. However,
compared to Journeyman Aspirants, they hold outside careers at lower rates. They also
teach fewer classes and have fewer institutional employers. Thus, while many aspire to
full-time work in academia, it may not be their principal focus. Perhaps the flexibility of
freelance work helps them manage family responsibilities or gives them a better work-life
balance.

Adjunct Expert: Unlike other part-time adjuncts, most Adjunct Experts (54%) have full-
time careers outside of academia (Table 2). Probably for this reason, they tend to teach
fewer academic courses than other adjunct types. Many Experts seem satisfied with this
kind of part-time work arrangement. More so than other adjunct classes, they report
working part-time by choice and fewer of them have pursued full-time academic work.
The agreeableness of this work arrangement is probably related to their positive
attitudes towards colleagues and administrators, relative to other adjunct types. Adjunct
Experts tend to receive lower compensation, benefits and professional development,
suggesting they may have different motives for academic work. Perhaps these adjuncts
enjoy sharing their expertise, mentoring young people, or simply enjoy the intellectual
environment of higher education.

15
Discussion
Now the “new faculty majority,” non-tenure track faculty constitute over two-thirds
of academic faculty members today. On account of the considerable diversity among
these postsecondary adjuncts, researchers have created numerous typologies for
conceptualizing them. However, this line of research has historically been unable to
make full use of all information available for classification and faces limitations inherent
to traditional methodologies of classification. These studies also on many occasions
have proliferated adjunct labels in ways that have not always enhanced our
understanding. By applying cluster analysis to the HERI Faculty Survey, this study
approached the problem of classification from a new perspective. Model diagnostics
suggested three natural groupings of full-time adjunct faculty and five natural clusters of
part-time faculty.

Among the full-time adjuncts, administrative adjuncts were more established in

their institutions and had extensive administrative responsibilities. Core teaching faculty
and peripheral teaching faculty were the other two types of full-time faculty, distinguished
by their level of institutional integration. Among part-time faculty, the analysis also
identified an adjunct class with a greater emphasis on administrative responsibilities:
part-time Adjunct Administrators. This study also identified Adjunct Experts. These
were part-time faculty members with careers outside of academia. Experts held some of
the most positive attitudes about their work, even though they were poorly compensated
(in traditional ways) for their efforts. This study also found three classes of aspirants.
Academic Aspirants often resembled traditional faculty with Ph.D. credentials and very
few of them pursued careers outside academic walls. Journeyman Aspirants tended to
have professional backgrounds and often worked extensively both within and outside of
academia. Their career pathways were diversified but many still seemed to hold out
hope for a full-time academic position. Finally, Freelancing Aspirants resembled
Journeymen, except that they worked less in the formal economy. Perhaps they
dedicate more of their time to volunteering, caring for family members or pursuing
additional education.

The typology discovered in this study is similar to what had been established in
seminal research on this topic. Gappa and Leslie (1993) recognized four classes of
part-time adjuncts: aspiring adjuncts, experts, freelancers, and career enders. This
study found similar classes, albeit with important nuances. Gappa and Leslie’s
conceptualization of aspiring academics was that of a doctorate recipient seeking full-
time, tenure-track appointments. This study identified three particular types of aspiring
academics: Academic Aspirants, Journeyman Aspirants and Freelancing Aspirants. The
main way these aspirants differed from one another was their involvement in extra-
academic work. Academic Aspirants seemed focused on acquiring a full-time position
by dedicating themselves entirely to academia. Journeyman Aspirants, while also
interested in full-time academic work, were professionals working extensively both inside
and outside of academia. Freelancing Aspirants were between these two extremes.
Identifying subclasses of aspiring academics was an important contribution of this study.
Rather than lump all aspirants together, it was important in many instances to factor in
their level of engagement in careers outside of academia. The cluster analysis also
identified Academic Experts—an adjunct class that seems similar to the original type
identified in Gappa and Leslie’s work. Experts were professionals with outside careers

16
working in academia presumably for the intellectual climate of the classroom or for an
opportunity to share their expertise with young people.

In contrast to the findings of Gappa and Leslie, the cluster analysis uncovered a
distinct part-time adjunct type focusing on administration—the so-called Adjunct
Administrator. It is unclear why this adjunct type would emerge from the cluster analysis
and not in the Gappa and Leslie study. One explanation is that part-time faculty perform
more administrative duties today than in the past. Potentially institutions today are using
part-time labor to lower the costs of administration—similar to the way they do for
instruction. Alternatively, this category may have been subsumed into the career-ender
type in Gappa and Leslie’s study. The career-ender type did not emerge as a type in the
cluster analysis. However, this might have been due to the exclusion of part-timers
holding tenure. Indeed, exploratory research suggested this may be the case.

The cluster analysis of full-time faculty also revealed types that were similar to
foundational work. Like Baldwin and Chronister (2001), this study found that full-time
faculty were principally distinguished according to their main work responsibilities. Some
of these adjuncts were administrators and the other two types of full-time adjuncts were
teaching faculty. Unlike Baldwin and Chronister (2001), however, this study did not
identify a strictly “research” adjunct. This is in part because the HERI sample excluded
faculty without a classroom presence. An important nuance overlooked by Baldwin and
Chronister (and other researchers) was the core versus periphery distinction among full-
time adjunct teaching faculty. Core teaching faculty have reasonable salaries and more
opportunities for professional development. Peripheral teaching faculty, while employed
full-time, undergo less professionalization in their departments and receive less
compensation for their work. This probably leads to a lower level of workplace
integration. Of course, this study did not find strong separation between core and
peripheral teaching adjuncts. However, this point is still worth noting.

With the dramatic growth and diversification of adjunct faculty, it is imperative for
researchers to consistently update how they conceptualize and think about
postsecondary adjuncts. This is because better conceptual tools provide greater
leverage over unwieldy and ever-changing social phenomena. While the clusters that
emerged in this study were similar to earlier classifications, this study identified important
nuances. These subtleties can help researchers articulate their ideas more clearly and
test their hypotheses more rigorously. Improved typologies also help frame the “adjunct
problem” among policy-makers, administrators and in the popular media. This work
suggests that the role of administration should factor into classification, as well as the
engagement of aspirants in outside work. Policies and programs should be created that
target peripheral teaching faculty, integrating them better into their workplaces. Policy-
makers should also be aware of the value that many adjuncts place on flexible, part-time
work. It would be a mistake to dismantle the part-time system altogether.

Cluster analysis is an exploratory technique, and there were no formal

hypotheses tested in this study: the method simply extracted patterns that were naturally
present among variables in the data. This is not to say that this study is without
limitations. The sample used, for example, was limited. Without information on adjuncts
working exclusively in research or administration, this study was unable to incorporate
information from the entire population of adjuncts. Regrettably, this makes the

17
representativeness of the proposed typology sub-optimal, even if only a fraction of
adjuncts were missing from the study.

This method also raises concerns with regard to cluster partitioning. Ideally, the
natural groupings that emerge from a data source are compact and highly distinctive.
When this occurs, the implied typology has greater integrity. In this study, the clusters
that emerged had some degree of fuzziness or vagueness: there was substantial
overlap between some of the clusters. Core teaching faculty and peripheral teaching
faculty, for example, did exhibit similarity across many dimensions. Nevertheless,
overlap in the partitioning may not be a serious problem. It may simply suggest that two
or more clusters are very similar or that there is a hierarchical nature to the ordering of
groups. Future work should seek out what may be responsible for partitioning issues.
Useful strategies might include using more variables, better measurements or k-fold
cross-validation to improve model performance and help with validation.

There are many other ways for researchers to build on this line of research. This
study focused on how faculty clustered with regard to job function and workplace
characteristics. However, cluster analysis could provide insights into other aspects of
faculty work-life as well. Past research, for example, utilized work titles for classification
of adjuncts (Shavers 2000, Hollenshead et al. 2007). This information could also be
used in a cluster analysis to see whether there may be natural status-based groupings
among postsecondary adjuncts. Researchers may also want to consider a data source
with information on the entire population of adjuncts (with research-oriented and career-
ending adjuncts). NSOPF data13, for example, may be useful for its generality, although
it is important to understand that it lacks the extensive set of important adjunct variables
contained in the HERI sample.

Conclusion
Adjunct typologies have stagnated in recent years, due to the application of
traditional methods with inherent limitations. Using a new method and a new dataset,
this study approached adjunct classification from a fresh direction. Importantly, this
study helped validate typological findings from foundational work in this line of research.
However, it also shed new light onto nuances associated with these traditional adjunct
groupings. By honing our understanding of adjunct classes, this study strengthens our
conceptual framework for thinking about and understanding postsecondary adjuncts.
With better conceptual tools, researchers, administrators, and policy-makers have better
instruments for designing efficient and effective policies for the postsecondary workforce.

References

Aldenderfer, M. S. and R. K. Blashfield. 1984. Cluster Analysis. Beverly Hills, CA: SAGE
Publications, Inc.

13 The NSOPF instrument was discontinued in 2004, but still may hold relevant information for the

purposes of adjunct classification.

18
Baldwin, Roger G. and Jay L. Chronister. 2001. Teaching without Tenure: Policies and
Practices for a New Era. Baltimore, MD: Johns Hopkins University Press.

Bedford, Laurie A. 2009. “The Professional Adjunct: An Emerging Trend in Online

Instruction.” Online Journal of Distance Learning Administration 12(3).

Bedford, Laurie and Heather Miller. 2013. “All Adjuncts Are Not Created Equal: An
Exploratory Study of Teaching and Professional Needs of Online Adjuncts.”
Online Journal of Distance Learning Administration 16(1).

Berry, Joe. 2005. Reclaiming the Ivory Tower: Organizing Adjuncts to Change Higher
Education. New York: Monthly Review Press.

Carnevale, D. 2004. “For Online Adjuncts, a Seller’s Market.” Chronicle of Higher

Education, (50)34.

Clatworthy, J., D. Buick, M. Hankins, J. Weinman, and R. Horne. 2005. “The Use and
Reporting of Cluster Analysis in Health Psychology: A Review.” British Journal of
Health Psychology 10:329–58.

Cross, John and Edie Goldenberg. 2009. Off-Track Profs: Nontenured Teachers in
Higher Education.

Finkelstein, Martin J., Valerie Martin Conley, and Jack H. Schuster. 2016. The Faculty
Factor: Reassessing the American Academy in a Turbulent Era. Baltimore,
Maryland: Johns Hopkins University Press.

Gappa, Judith M. and David W. Leslie. 1993. The Invisible Faculty. Improving the Status
of Part-Timers in Higher Education. San Francisco, CA: Jossey-Bass Inc.

Hartigan, J. A. and M. A. Wong. 1979. “Algorithm AS 136: A K-Means Clustering

Algorithm.” Journal of the Royal Statistics Society 28(1):100–108.

Hollenshead, Carol et al. 2007. Making the Best of Both Worlds: Findings from a
National Institution-Level Survey on Non-Tenure Track Faculty. Center for the
Education of Women: University of Michigan.

Jain, Anil K. 2010. “Data Clustering: 50 Years beyond K-Means.” Pattern Recognition
Letters 31(8):651–66.

Kezar, Adrianna and Cecile Sam. 2010. Understanding the New Majority of Non-Tenure-
Track Faculty in Higher Education: Demographics, Experiences, and Plans of
Action: ASHE Higher Education Report. Jossey-Bass, Inc.

Klemmack, D. L. et al. 2007. “A Cluster Analysis Typology of Religiousness/Spirituality

Among Older Adults.” Research on Aging 29(2):163–83.

Le Roch, K. et al. 2003. “Discovery of Gene Function by Expression Profiling of the

Malaria Parasite Life Cycle.” Science 301(5639):1503–8.

19
Mandara, Jelani. 2003. “The Typological Approach in Child and Family Psychology: A
Review of Theory, Methods, and Research.” Clinical Child and Family
Psychology Review 6(2):129–46.

Maynard, Douglas C. and Todd Allen Joseph. 2008. “Are All Part-Time Faculty
Underemployed? The Influence of Faculty Status Preference on Satisfaction and
Commitment.” Higher Education: The International Journal of Higher Education
and Educational Planning 55(2):139–54.

McQuitty, L. L. 1987. Pattern-Analytic Clustering: Theory, Method, Research and

Configural Findings. Lanham, MD: University Press of America.

Milligan, G. W. 1980. “An Examination of the Effect of Six Types of Error Perturbation on
Fifteen Clustering Algorithms.” Psychometrika 45:325–42.

Milligan, G. W. 1996. “Cluster Validation: Results and Implications for Applied Analyses.”
in Clustering and classification, edited by P. Arabie, L. J. Hubert, and G. De Soto.
River Edge, NJ: World Scientific.

Milligan, G. W. and L. M. Sokol. 1980. “A Two-Stage Clustering Algorithm with Robust

Recovery Characteristics.” Educational and Psychological Measurement 40:755–
59.

Robins, R. W., O. P. John, A. Caspi, T. E. Moffitt, and M. Stouthamer-Loeber. 1996.

“Resilient, Overcontrolled, and Undercontrolled Boys: Three Replicable
Personality Types.” Journal of Personality and Social Psychology 70(1):157–71.

Schell, Eileen E. and Patricia Lambert Stock. 2001. Moving a Mountain: Transforming
the Role of Contingent Faculty in Composition Studies and Higher Education.
Urbana, IL: National Council of Teachers of English.

Schnitzer, Maria and Lynne Crosby. 2003. “Recruitment and Development of Online
Adjunct Instructors.” Online Journal of Distance Learning Administration 6(2).

Schuster, Jack H. and Martin J. Finkelstein. 2006. The American Faculty: The
Restructuring of Academic Work and Careers. Baltimore, MD: The Johns
Hopkins University Press.

Shavers, Frances. 2000. “Academic Ranks and Titles of Full-Time Non-Tenure-Track

Faculty.” Pp. 110–40 in Policies on faculty appointment: Standard practices and
unusual arrangements.

Tuckman, Howard P. 1978. “Who Is Part-Time in Academe?” AAUP Bulletin 64(4):305–

15.

Van Buuren, Stef and Karin Groothuis-Oudshoorn. 2011. “Mice: Multivariate Imputation
by Chained Equations in R.” Journal of Statistical Software 45(3):1–67.

20
Wagoner, Richard L. 2007. “Globalization, the New Economy, and Part-Time Faculty.”
New Directions for Community Colleges 2007(140):21–27.

Appendix

Clustering Algorithm

K-means clustering requires the specification of three parameters: the number of

clusters, how clusters are initialized and how distance is measured. The number of
clusters was guided by hierarchical agglomerative clustering (Milligan 1980) and an
elbow plot associated with the within-cluster sum of squared errors14. Three clusters
were optimal for full-time faculty and five clusters were optimal for part-time faculty.
Before conducting k-means clustering, all variables were scaled and normalized15 so
that variables with the greatest ranges did not have undue influence on the formation of
clusters. To conduct the k-means analysis, it is important to choose random starting
points for the means (Milligan and Sokol 1980). This helps prevent convergence on
suboptimal means. This study used 10 different sets of starting points to identify the
means that best summarized the information in the data. The Hartigan-Wong algorithm
employed in this study uses the Euclidean measure of distance.

Imputation

Listwise deletion across a large number of variables led to an unacceptably small

sample. To rectangularize the matrices, this study utilized the mice package in R to
produce a single imputation (Van Buuren and Groothuis-Oudshoorn 2011). As
calculating standard errors16 was not important to this study, single imputation was
simple and sufficient. This method was implemented specifically using a CART
algorithm. Each target column was imputed drawing on information from all of the other
columns in the data. As most other columns had some degree of missingness as well,
the algorithm used the most recent round of imputations for each. The iteration
maximum was set to five, giving the chained equations multiple attempts to converge on
a good imputed value for each cell.

14 It is still not established which way is best for determining the number of clusters.
15 The importance of this step is unresolved in the literature (Milligan 1996).
16 When standard errors are important, there are better missing data methods available like full-

information maximum likelihood and multiple imputation.

21
 Table 3: Descriptive Statistics of Quantitative Variables in the Analysis

Quantitative Variable Mean S.D.

Besides this institution, how many other institutions 0.6 0.9
# of courses teaching this term (all institutions) 2.6 1.7
Students enrolled in course 31.6 40.2
Year of birth 1960 12
Year of highest degree now held 1995 12
Year of appointment at present institution 2003 8
Career related Stress 45.0 9.1
Institutional Selectivity 1107.5 123.0
Salary: % Base salary from this institution 64.3 37.4
Salary: % Other income from this institution 3.7 11.3
Salary: % Income from another academic institution 8.4 21.6
Salary: % Non-academic income 24.6 34.8
Base institutional salary** $59,271 $36,977
Total salary from teaching at institution* $15,270 $13,034
Pay per course at institution* $3,596 $3,067
* Variable only pertained to part-time non-tenure track faculty
** Variable only pertained to full-time non-tenure track faculty

22
 Table 4: Variables Used in Cluster Analysis
Code Description Code Description
PRINACT Principal activity PASTACT01 Considered early retirement
ACADRANK Academic rank PASTACT02 Considered leaving academe for another job
TENURE Tenure status PASTACT03 Considered leaving this institution for anothe
ADMCHAIR Serves as department Chair PASTACT04 Changed academic institutions
ADMDEAN Serves as Dean STRESS11 Stress: Colleagues
ADMPRES Serves as President STRESS13 Stress: Research or publishing demands
ADMVP Serves as Vice-President STRESS15 Stress: Teaching load
ADMPROVOST Serves as Provost STRESS20 Stress: Job security
ADMOTHER Serves as other STRESS25 Stress: Institutional budget cuts
ADMNA Serves as not applicable APPTYR Year of appointment
COURSENUM Courses taught current term STATE State
CRSTYPE01 Type of course OBEREG Region
CRSENROLL01 Number of students SALARY FT Base salary
PROFDEV01 PD: outside workshops PTSALARY PT Part-time salary
PT
PROFDEV02 PD: paid sabbatical PTPAY Pay per course
PROFDEV03 PD: travel funds PTCHOICE PT (in)voluntary part-time status
PROFDEV04 PD: internal grants PTWORKFTPT Sought full-time position in past
PT
PROFDEV05 PD: administrative training PTCAREER Full-time career outside of academia
PROFDEV06 PD: course development PTREASON01 PT Part-time for income
PT
PROFDEV07 PD: new technology PTREASON02 Part-time for compensation
PT
PUBLISH01 Number of articles PTREASON03 Part-time for career
PUBLISH02 Number of chapters PTREASON04 PT Part-time for benefits
PT
PUBLISH03 Number of books PTREASON05 Part-time for lifestyle
PT
PUBLISH04 Number of patents PTREASON06 No full-time work available
PUBLISH05 Number of exhibitions PTREASON07 PT Part-time to offer expertise
PT
PUBLISH06 Number accepted for publication PTRESOURCES01 Private office
PT
HPW01 Hours per Week: Scheduled teaching PTRESOURCES02 Shared office
HPW02 Hours per Week: Preparing for teaching PTRESOURCES03 PT Computer
PT
HPW03 Hours per Week: Advising and counseling of students PTRESOURCES04 Email account
PT
HPW04 Hours per Week: Committee work and meetings PTRESOURCES05 Phone/voicemail
HPW05 Hours per Week: Other administration PTOPN01 PT Part-timers receive training
PT
HPW06 Hours per Week: Research and scholarly writing PTOPN02 Part-timers rarely hired full-time
HPW07 Hours per Week: Other creative products/performances PTOPN03 PT Part-timers respected by students
HPW08 Hours per Week: Consultation with clients/patients PTOPN04 PT Part-timers mainly in introductory classes
PT
HPW09 Hours per Week: Community or public service PTOPN05 Part-timers have no employment security
HPW10 Hours per Week: Outside consulting/freelance work PTOPN06 PT Part-timers have access to support survices
HPW11 Hours per Week: Household/childcare duties PTOPN07 PT Part-timers compensated for advising
PT
HPW12 Hours per Week: Commuting to campus PTOPN08 Part-timers must attend meetings
HPW13 Hours per Week: Other employment, outside of academia PTOPN09 PT Part-timers have good workshop relationship
GENACT01 Member of a faculty union PTOPN10 PT Part-timers respected by full-time colleagues
PT
GENACT03 Plans to retire within the next three years PTTEACH Addition institutions of employment
GENACT04 Scholarship addresses local community needs
FT PT
- Variable only available/relevant for full-time faculty members. - Variable only available/relevant for part-time faculty members
Note: Variables with the roots of "PUBLISH" and "HPW" were discretized by HERI prior to provisioning.

23
 CHAPTER 3: NON-TENURE TRACK JOB SATISFACTION AND TURNOVER INTENTIONS

Abstract
Although most postsecondary faculty work off the tenure line, there is little
empirical research to guide administrators on retention strategies tailored for the “new
faculty majority.” Utilizing the 2010 HERI Faculty Survey (N=8418), this study examines
the job satisfaction and organizational turnover intentions of non-tenure track college
teachers. Importantly, this study considers multiple facets of job satisfaction and
examines adjunct types ignored in earlier research. As adjuncts have different
motivations for working off the tenure track, the findings of this study have important
implications for the design of supportive policies and effective retention strategies.

Introduction
Research on the retention of postsecondary faculty members has been an
important line of study for over half a century. As its faculty may be higher education’s
greatest resource, the importance of creating attractive work environments is clear.
However, higher education began an important transition in the 1970s when institutions
began hiring larger numbers of faculty off the tenure line. Today, faculty adjuncts17
number over a million and constitute over two-thirds of the academic work force
(National Center for Education Statistics 2015). Despite this dramatic transformation,
retention research has maintained its focus on how to retain traditional, tenure-line
faculty and generally ignored the growing number of adjuncts. Surprisingly little is
established about how to support and retain the new faculty majority—non-tenure track
professors.

This study examines the organizational turnover intentions of non-tenure track

faculty. As job satisfaction is the key mediator of the turnover process (Mobley 1977,
Price and Mueller 1981), this variable is given first order consideration. However, unlike
earlier work that relies on a single measure of job satisfaction, this study examines
multiple facets of job satisfaction in a structural equation model (SEM) framework. It
identifies four key components (work and supports, co-worker relationships, financial
satisfaction, and benefit satisfaction), mapping fairly well onto early theoretical work
(Kalleberg 1977). This study is also novel in its examination of adjunct “experts”,
“career-enders” and “freelancers.” Described by Gappa and Leslie (1993), these adjunct
subgroups work off the tenure line for different reasons and this probably impacts their
job satisfaction (Morse 1953) and perhaps turnover intentions as well.

Which adjuncts are at greatest risk of leaving their institution? Which aspects of
their work are most enjoyable and how important are those experiences and perceptions
with regard to leaving their institution? These questions are examined using the 2010
Higher Education Research Institute (HERI) Faculty Survey. The Faculty Survey is a
national, cross-sectional survey with extensive information on the characteristics and
experiences of postsecondary faculty. It is somewhat special in that it made a concerted
effort to collect extensive information on both part-time and full-time non-tenure track
faculty (n=8,418). It also, importantly, contains a rich set of job satisfaction items which,

17 This study refers to all non-tenure track faculty as “adjuncts.”

24
in a structural equation model, reveal important nuances about the organizational
attrition of non-tenure track faculty.

There are many reasons to care about the job satisfaction and turnover process
of non-tenure track faculty. For one, providing agreeable work conditions should be an
institutional goal in its own right. However, institutions have selfish motives as well.
Faculty turnover—even of those off the tenure track—is expensive because
replacements must be identified, hired, oriented and trained. Turnover also
accompanies inevitable disruptions as departments reallocate teaching assignments and
some faculty teach new course material for the first time (Smart 1990). Sensible
personnel strategies, then, aim to minimize these costs. By clarifying which factors drive
the turnover process of adjuncts, this study suggests effective ways for administrators to
support and retain a stable workforce. Better working environments have also been
linked to a variety of other important outcomes like institutional effectiveness,
functionality, and productivity (Hom et al. 2017).

Literature Review
Turnover
Four interrelated lines of theory define our knowledge of employee turnover
(Hom et al. 2017). Commitment theory generally focuses on the psychological link
bonding worker to employer (Porter, Crampon, and Smith 1976, Porter et al. 1976).
Meyer and Allen (1991) argued that workers also experience cost-related18 forms of
commitment and normative commitment. The theory of embeddedness concentrates on
the contextual forces motivating workers to stay (Mitchell and Lee 2001). This includes
network connections, perceptions of fit, and forms of cost-avoidance (Lee, Burch, and
Mitchell 2014). A third body of theory, termed “the unfolding model”, focuses on the role
of shocks or triggers that are external to the job itself. Certain life-events, like
parenthood or unsolicited job offers, force a reconsideration of circumstances and
sometimes lead to decisions to leave an employer (Hom et al. 2017, Holtom et al. 2008,
Weller et al. 2009).

While each of these theories offers important insights into the turnover process,
the most developed line of work is rooted in Vroom's (1964) expectancy theory. In this
perspective, motivation is viewed as a cognitive process resulting from a combination of
structural, attitudinal and environmental forces. Individuals identify desirable goals,
rationally evaluate behavioral options to accomplish those goals, and then select the
option with the greatest expectancy of bringing about desired ends. March and Simon
(1958) incorporated expectancy into their work on organizational management. In their
view, workers possess expectations about what a job is and what it should be. When a
job meets expectations, employees seek to maintain the work arrangement and perform
requisite behaviors. When it falls short of expectations, workers consider alternative
behaviors with a greater likelihood of bringing about work expectations. One behavioral
option is to quit and pursue a new employer.

Price (1977), Price and Mueller (1981), Mobley (1977) and Mobley et al. (1979)
advanced this line of work by creating measures and empirically validating many of the
ideas. Importantly, they operationalized job desirability using measures of job

18 Also referred to as continuance commitment in the literature.

25
satisfaction. Job satisfaction, in other words, measured just how aligned the work
environment was with the individual’s expectations. More sociological in nature, Price’s
work clarified the influence of contextual features (e.g., workplace integration, pay, social
relationships, professionalism, perceived labor market opportunities) on those
expectations. Mobley’s work was more psychological in his treatment, examining in
great detail the cognitive processes preceding a decision to quit19. For both Price and
Mobley, job satisfaction was viewed as the key mediator of the turnover process;
differences in job expectations drive the turnover process and decisions to quit.

Job Satisfaction
Locke (1969) called job satisfaction the “pleasurable emotional state resulting
from the appraisal of one’s job as achieving or facilitating the achievement of one’s job
values.” It encompasses a diverse set of feelings and emotions, including stress,
anxiety, boredom, inspiration and excitement. While satisfaction is highly subjective, it is
critical for understanding turnover. This is because structural features and external work
conditions impact turnover indirectly: their effects are channeled through job satisfaction
and other socio-psychological variables (e.g., commitment, equity, justice).

Research on job satisfaction is extensive. There exist thousands of publications

on job satisfaction and it is the most frequently studied variable in all of organizational
research (Spector 1997). Scholars typically draw on one of three frameworks in this line
of research. The most influential is probably Herzberg, Mausner, and Snyderman's
(1959) motivator-hygiene theory. According to this theory, external features—termed
“hygienes”—map onto a job dissatisfaction component. Intrinsic processes (e.g.,
achievement, job recognition, and performing interesting and challenging tasks), on the
other hand, map onto an independent factor associated with satisfaction. These were
called “motivators.” Achievement, recognition, work itself, responsibility, advancement
and (to a lesser degree) salary were found to be the most important characteristics in
this process (Herzberg et al. 1959).

Scholars in this line of research also commonly cite Hackman and Oldham
(1980) and their “Job Characteristics” model. This framework is important for its
nuanced theorization of the intrinsic dimension of job satisfaction. In this view, job
satisfaction is the consequence of features intrinsic to the work itself, including task
variety, task identity, task significance, autonomy, and task feedback. Jobs high on
these features result in motivation, satisfaction, performance, lower absenteeism, and
lower turnover. This line of work has found substantial empirical support.

While both of these theories have been very influential, research has suggested
that a third model (Kalleberg 1977) may be most appropriate for research on faculty
satisfaction (Seifert and Umbach 2008, Eagan, Jaeger, and Grantham 2015).
Kalleberg’s theory of job satisfaction recognizes the intrinsic dimension of satisfaction;
however, it is more expansive with regard to the external factors impacting job
satisfaction and turnover. As this paper draws strongly on Kalleberg’s sociological work,
I present a summary below. His theory was validated using a principal component

19 Mobley's (1977) attrition process: thinking about quitting, evaluating the utility of a job search

and the costs of quitting, intention to search for alternatives, search for alternatives, evaluation of
alternatives, comparing options, and intentions to quit.
26
analysis of thirty-four job values and characteristics (Kalleberg 1977). Kalleberg found
six latent components to the job satisfaction construct (one intrinsic and five extrinsic).

Intrinsic

Work intrinsic to the job itself is an important component of job satisfaction.

Research has shown that employees want tasks that are interesting, varied and require
them to develop and employ new skills. They experience greater satisfaction when
charged with completing whole tasks, rather than partial tasks characteristic of assembly
lines. Workers also want the challenges, responsibility and autonomy of determining the
best strategy to complete a project or task. Feedback is also important as workers want
to know when they have completed tasks correctly and to receive recognition for it.
Work on adjunct satisfaction has found strong support for this dimension (Antony and
Hayden 2011, Rosser 2004, Eagan et al. 2015, Antony and Valadez 2002, Seifert and
Umbach 2008, Maynard and Joseph 2008).

Convenience

Convenience is another important dimension of job satisfaction. Having a

workplace close to where they live, available parking, and “good hours” make it easier to
accomplish work. Workers also value jobs that give them the flexibility to deal with
competing time-demands (Rice, Frone, and McFarlin 1992). Sick children, doctors’
appointments, and other “emergencies” are common and workers appreciate jobs that
allow them to quickly resolve these problems when they arise. Research has found that
adjunct faculty value convenience (Seifert and Umbach 2008, Rosser 2004).

Financial

Financial compensation (rewards) is another important facet of job satisfaction.

Compensation includes one’s base pay, bonus, and fringe benefits like health insurance,
child care subsidies and retirement funds. It also includes job security, as this implies
future income and benefits. Research on adjunct faculty has consistently found a
financial dimension to job satisfaction. Smart (1990), Antony and Hayden (2011),
Toutkoushian and Bellas (2003), Antony and Valadez (2002), Maynard and Joseph
(2008), and Johnsrud and Rosser (2002)20 examined pay satisfaction. Toutkoushian
and Bellas (2003), Rosser (2004), Johnsrud and Rosser (2002), and Maynard and
Joseph (2008) have found support for the role of benefits and security.

Relationships with Co-workers

20 Johnsrud and Rosser’s (2002) work specifically examined the “quality of work-lives.”
27
 Work relationships constitute another important component of job satisfaction.
Workers spend extensive time interacting with their colleagues and they seek a sense of
community in their workplaces (Murphy, Athanasou, and King 2002, Carnevale and Rios
1995, Klassen, Usher, and Bong 2010). Research on non-tenure track faculty has found
a social dimension to satisfaction with administration, colleagues, and student
subordinates (Ryan, Healy, and Sullivan 2012, Seifert and Umbach 2008, Johnsrud and
Rosser 2002, Antony and Valadez 2002, Waltman et al. 2012).

Career

Perceived career opportunities constitute another important facet of job

satisfaction. Workers want to know that their efforts advance their status and
responsibility in an organization. So, promotional opportunities and the fairness of that
process is an important form of job satisfaction. Career satisfaction is particularly
important for adjunct faculty, who often work without formal promises of future
employment (Seifert and Umbach 2008, Maynard and Joseph 2008, Smart 1990).

Resource adequacy

Resource adequacy is the final dimension of satisfaction in Kalleberg's

framework (1977). Workers need the right materials, support, tools and equipment to do
their jobs completely and efficiently. This dimension of job satisfaction also involves
possessing, when necessary, adequate authority to effectively manage subordinates.
Guidance from competent leaders and supervisors can also be viewed as an important
resource. When tools, materials or staff are inadequate, workers feel ineffectual and
frustrated with obstacles. This leads to a drop in morale and lower levels of job
satisfaction (Peters, O’Connor, and Rudolf 1980). Some research (Eagan et al. 2015)
has identified the role of resources in the satisfaction of adjunct faculty.

Satisfaction of Adjunct Types

Adjunct faculty are “heterogeneous” (Kezar and Sam 2010) and have many
different motives for working off the tenure line (Gappa and Leslie 1993, Conley and
Leslie 2002). As motives have important consequences for satisfaction (Morse 1953,
Kalleberg 1977), many insights are only possible upon disaggregating this diverse group
of professors. Toutkoushian and Bellas (2003) distinguished between full-time and part-
time faculty, examining faculty job satisfaction using the National Study of
Postsecondary Faculty (NSOPF). They found that part-time faculty were much less
satisfied with their job benefits but were otherwise comparable to full-time faculty.
However, using a more nuanced model21 (and the same data), Antony and Valadez
(2002) found that faculty members were comparable with regard to “role demands and
rewards”—a factor measured by satisfaction with benefits (workload, job security,
opportunities for advancement, and pay). This research did reveal, however, that part-

21 This analysis was a structural equation model.

28
time faculty were significantly less satisfied with their autonomy and student
relationships than full-time faculty.

Maynard and Joseph (2008) were the first in this line of research to recognize the
importance of distinguishing between faculty working part-time by choice and those who
would prefer full-time employment. Examining faculty at a single institution, they
compared the job satisfaction of voluntary part-time, involuntary part-time, and full-time
faculty. The strength of this study was its use of the Minnesota satisfaction
questionnaire, an instrument measuring twenty satisfaction factors with one-hundred
items. They reasoned that these faculty classes would differ in their satisfaction with
regard to advancement, compensation, recognition, and job security. These hypotheses
were mostly supported by their analysis. Involuntary part-time faculty were less satisfied
with advancement than other faculty types. They also had lower satisfaction with
compensation and job security than full-time faculty. Full-time faculty were more
satisfied with their job security than all other faculty types.

Antony and Hayden (2011) and Eagan et al. (2015) made the same faculty
distinctions and examined similar research questions using national datasets. Using
National Study of Postsecondary Faculty (NSOPF), Antony and Hayden (2011) found
that involuntary part-time faculty were less satisfied with pay, benefits and authority to
make decisions (autonomy) than full-time peers. The only way that voluntary part-time
faculty were less satisfied than full-time faculty was with their benefits. Using a more
sophisticated model, Eagan et al. (2015) examined workplace satisfaction—a factor
measured by satisfaction with autonomy, professional relationships, competency of
faculty, departmental leadership, and course assignments. They found that involuntary
part-time faculty experienced significantly lower workplace satisfaction than their
voluntary part-time peers. However, this relationship disappeared once controlling for
institutional resources and supports, suggesting that resource adequacy is probably
important to these relationships.

Limitations of Earlier Work

In their seminal study, Invisible Faculty, Gappa and Leslie (1993) interviewed
faculty members at 18 American and Canadian colleges and universities. They found
different motives for working off the tenure track and created an influential adjunct
typology consisting of experts, career-enders, freelancers and aspiring academics.
Experts22 held full-time positions outside of academia and taught in order to contribute
their knowledge and expertise to students at the institution. Career-enders were
retirement age and seeking fewer work hours and responsibilities. Freelancers worked
part-time by choice and often held part-time work outside of higher education. Aspiring
academics23 wanted full-time academic employment but, for one reason or another,
were unable to arrange this. This group is comparable to the “involuntary part-time
adjunct” explored in later work.

While research now makes finer-grained adjunct distinctions than in the past,
Gappa and Leslie's (1993) work suggests that the job satisfaction of other adjunct
subclasses may be important to consider as well. This study draws on these empirically-

22 Also referred to as specialists or professionals in Gappa and Leslie’s work.

23 Gappa and Leslie included graduate students in this class.
29
derived adjunct classes to more fully recognize the heterogeneity among faculty working
off the tenure line. In addition to full-time and aspiring academics, this study examines
the satisfaction and turnover intentions of adjunct experts, career-enders and
freelancers. These are subclasses of adjuncts working part-time by their own volition.

Another limitation of earlier work is related to the measurement of job

satisfaction. Despite the fact that a single measure greatly oversimplifies a complicated
concept (Oshagbemi 1997) and produces upwardly biased estimates (Oshagbemi
1999), much of the earlier work relies on a single measure of job satisfaction. Clearly,
faculty can be more satisfied in some ways and less satisfied in others. This study joins
the efforts of Antony and Valadez (2002), Maynard and Joseph (2008), and Eagan et al.
(2015) in recognizing and modeling latent components of job satisfaction. By using a
multidimensional measure, this study establishes a nuanced understanding of the
cognitive and evaluative processes associated with job satisfaction and turnover
intentions.

While many researchers have used sophisticated models to examine faculty

turnover, nearly all of them focus on traditional, tenure-line faculty (Ryan et al. 2012,
Barnes, Agago, and Coombs 1998, Daly and Dee 2006). When non-tenure track faculty
were included in studies, part-time adjuncts have typically been excluded (Smart 1990,
Xu 200824, Zhou and Volkwein 2004, Rosser 2004). Of all the literature reviewed, only
Johnsrud and Rosser (2002) used a structural equation model to examine the
satisfaction and turnover intentions of non-tenure line faculty. However, these authors
never actually indicate whether any of their sample were part-time faculty members.
Based on their cross-tabulations, they probably were not. In any case, only a fraction of
their sample was non-tenure track and there was little consideration for the nuanced
groupings of non-tenure track faculty that this study examines.

Proposed Model
This study examines the turnover intentions of adjunct faculty. Turnover
intentions—a cognitive measure—measures whether or not a faculty member has
considered leaving his or her institution. A behavioral measure (like actual turnover)
might have produced important insights with regard to these individuals; however, data
on faculty turnover behavior is rare due to the complications of following up with mobile
survey participants in transition. In any case, turnover intention is a good proxy and the
single best predictor of actual turnover (Steers and Mowday 1981, Bluedorn 1982).

Research on faculty turnover has produced causal models with some

differences, however Smart (1990) attributes this mostly to the authors’ diverse
disciplinary orientations—not fundamental differences in how theorists believe attrition
occurs in academia. In general, turnover is the function of individual characteristics,
contextual variables and external conditions. Multiple dimensions of job satisfaction
mediate this process, expressing an employee’s alignment or adjustment to her or his
work environment.

Research Questions
While non-tenure track faculty earn less, have fewer benefits and fewer
amenities (Kezar and Sam 2010), research has shown them to have comparable levels

24 Technically, Xu used hierarchical linear models.

30
of job satisfaction in the aggregate. This study follows the lead of earlier researchers
who produced important insights by disaggregating non-tenure track faculty and
considering different facets of job satisfaction. This study specifically employs the
Gappa and Leslie (1993) adjunct typology in its assessment of adjunct satisfaction and
turnover.

1) How do adjunct types compare across latent measures of job satisfaction?

2) How do adjunct types compare with regard to organizational turnover intentions?
3) Do contextual features (e.g., institutional characteristics) matter with regard to
turnover intentions?

These questions were investigated in an exploratory factor analysis and through

the use of a Bayesian probit structural equation model. These methods will be
discussed in more detail in the methods section. However, very briefly, the exploratory
factor analysis revealed four dimensions of job satisfaction that map well onto the
components identified in Kalleberg (1977): work and supports, co-workers, financial, and
job benefits.

Adjunct Job Satisfaction Hypotheses

Earlier work suggests that aspiring adjuncts experience lower satisfaction with
work and supports (Antony and Hayden 2011, Eagan et al. 2015). This is tied to the
involuntary nature of their part-time employment. While freelancers work part-time by
choice, there is still reason to believe that they may be less satisfied with work and
supports than other types of voluntary part-timers. Potentially, having less autonomy
and influence over curriculum and course assignments results in lower satisfaction for
them on this dimension.

Research has not found differences among adjunct subgroups with regard to
satisfaction with co-workers. However, now disaggregated, we know that the role of
work relationships is different for career-enders than for other faculty types. Workplace
relationships are an important part of why career-enders decide to “transition” into
retirement. So, there is an expectation that career-enders will express more satisfaction
on this dimension than other types of voluntary part-time adjuncts.

Earlier research has shown aspiring academics to be much less satisfied with
financial aspects of their work25 (Antony and Hayden 2011, Maynard and Joseph 2008).
However, there is also reason to believe that freelancers may be lower on this dimension
as well. Unlike adjunct experts with full-time careers in the private sector, freelancers
have a strong dependency on their institutional salary—which is often quite low. Thus,
their satisfaction on this dimension is expected to be lower as well. Career-enders, on
the other hand, generally do not teach part-time for financial reasons. Thus, the financial
satisfaction they experience is expected to be higher than that of other voluntary part-
time faculty members.

25 The job satisfaction of full-time adjuncts and voluntary part-timers is for the most part

comparable. Only with regard to job security have differences been clearly identified (Maynard
and Joseph 2008).
31
 Research has found that both aspiring academics and voluntary part-time faculty
tend to be less satisfied with job benefits (Antony and Hayden 2011). It is possible,
however, that the heterogeneity of volunteer part-time adjuncts obscures important
relationships—especially for career-enders. As extending benefits into later years is
often a priority for those at the end of their careers, one might expect the job satisfaction
of career-enders on this dimension to be distinctly higher.

Turnover Hypotheses
Turnover research has shown that faculty who are less satisfied are more likely
to leave their organizations. This is likely true for all dimensions of job satisfaction: work
and supports, co-workers, financial satisfaction, or job benefits.

Hypothesis A: Lower levels of satisfaction with work and supports, co-workers, financial
satisfaction, and job benefits all cause turnover.

Research has shown that satisfaction is the key mediator between structural
conditions and turnover intentions. Upon controlling for satisfaction, relationships
between structural and environmental variables should dissipate (partial mediation).
This includes a partial mediation of relationships between turnover intentions and
adjunct faculty types. However, the motivations for working off the tenure-line are
fundamentally different for career-enders. Independent of their satisfaction, I still expect
that career-ender status remains important to organizational turnover intentions.

Hypothesis B: After controlling for dimensions of satisfaction, career-enders will be more

likely to leave their organizations.

Research has shown that turnover is mostly an individual-level phenomenon.

That is, most of the variation in turnover intentions happens between individuals, not
between organizations (Seifert and Umbach 2008). However, two structural
characteristics are theorized to matter: discipline and institutional control.

With regard to discipline, earlier work has shown that faculty from applied and
professional fields are subject to a greater “pull” from employment opportunities in the
private sector, where demand and salaries are higher (Ehrenberg, Kasper, and Rees
1991, Conley and Leslie 2002). This is particularly the case for faculty in “hard-applied”
fields like biomedical/health sciences, engineering and math/statistics (Ryan et al.
2012)26. So, while faculty in different departments may have similar job satisfaction
(Seifert and Umbach 2008), external factors make it more likely that faculty in applied
fields leave their institutions.

Following earlier work (Eagan et al. 2015, Smart 1990, Xu 2008), this study also
examines the impact of institutional control. Eagan et al. (2015) found that part-time
faculty were less satisfied in public institutions, perhaps because of institutional

26 “Soft-pure” disciplines linked to organizational turnover (Ryan, Healy, and Sullivan 2012).
32
problems with funding. Lower satisfaction implies that adjuncts at public institutions may
be more likely to leave for other organizations. No other institution-level characteristics
have been linked to differences in organizational turnover (Johnsrud and Rosser 2002).

Hypothesis C: Faculty in applied disciplines and public institutions will have higher
organizational turnover.

Method
This study utilized information from non-tenure track faculty in the 2010 HERI
Faculty Survey27. The full HERI sample (with tenured and tenure-line faculty) is
nationally representative of undergraduate teaching faculty in the United States. For
more information on this instrument, see Hurtado et al. (2012). The sample contained
information on 8,418 non-tenure track faculty members, approximately half of whom
worked part-time. Consistent with Dobbin and Simon (2011), models were trained using
a random 60% split of these data (n=5,050). Models were tested on the remaining 40%
(n=3,368)28. Missingness was handled using full information maximum likelihood in the
analyses.

There are several important caveats regarding this sample. First, faculty
sometimes work for multiple institutions (Martin Conley & Leslie 2002). We must
assume then, that study participants reported specifically on their main academic
employer (and not their general experiences in academia). The HERI instrument used
clear language asking the respondent to respond only about the institutional employer
participating in the study. Another important caveat is the timing of data collection. This
study uses the 2010 wave and this period is not far removed from the Great Recession.
March and Simon (1958) explained that "when jobs are plentiful, voluntary movement is
high; when jobs are scarce, voluntary movement is small." Thus, I expect that turnover
intentions may be somewhat suppressed during this particular time period.

Fortunately, the HERI Faculty Survey has some important advantages that make
it the best dataset for investigating these research questions. For one, it contains
contemporary information relevant to today’s academic workforce. This is important
because academia has changed rapidly over the last decades and it is important to
know what is happening in recent times. HERI also made it a priority to recruit non-
traditional faculty and to ask questions particularly relevant to contingent workers. This
gives us detailed information on the “invisible faculty” commonly overlooked in other data
sources.

27 Most researchers in this line of work have utilized the NSOPF panel study (discontinued in

2004). This includes Antony and Hayden (2011), Seifert and Umbach (2008), and Toutkoushian
and Bellas (2003).
28 Training a model on more data generally results in a better fit model. Testing a model on more
data generally results in a better estimate of model accuracy. There is no “ideal” split—only a
trade-off between fit and accuracy. This study opted for a 60-40 split, similar to the
recommendation of Dobbin and Simon (2011).
33
Measures
Dependent Variables.

Dimensions of Faculty Job Satisfaction29: This study used nineteen job satisfaction items
to identify four satisfaction subscales. For each job satisfaction item, respondents
indicated whether they were “not satisfied,” “marginally satisfied,” “satisfied,” or “very
satisfied.” These were coded numerically (1-4) for the analysis. There were four
dimensions of satisfaction identified in the exploratory factor analysis: work and
supports, co-workers, financial, and benefits.

Turnover intentions: Three questions were used to identify the turnover intentions of
faculty: During the past two years, have you considered early retirement? Considered
leaving academe for another job? Considered leaving this institution for another?
Faculty who responded yes to any of these questions were coded 1 as having turnover
intentions. The rest were coded as 0.30

Independent Variables.

Faculty Type: Part-time faculty who would prefer to be working full-time for their
institution were coded as “aspiring academics.” Part-time faculty who were not
interested in full-time work at their institution were coded as experts (when full-time
professional career was outside academia), freelancers (when full-time professional
career was not outside academia), and career-enders (when the faculty member
planned to retire within the next three years). This study also included full-time adjuncts.

Contextual Variables: Consistent with earlier research, this study coded disciplines
according to the Biglan system of departmental classification (Biglan 1973). This study
also coded institutional control as public or private.

Covariates.

This study controlled for a range of individual characteristics and contextual

variables. Final specification was guided by Cotton and Tuttle’s meta-analysis (Cotton
and Tuttle 1986) and was also influenced by an automated variable selection process31.

29 As a mediating variable, job satisfaction serves both as a dependent and independent variable

in this study.
30 To clarify, a faculty member planning on a traditional retirement does not experience turnover

intentions, because a traditional retirement is not ”early,” for another job, or for another institution.
31 This study used training data and a backward selection process to identify which variables to

include in the final model. Covariates that were unable to show a statistically significant
relationship with turnover intentions (alpha>.05) were, one by one, backward deleted from the
logistic training model (with robust standard errors). The final specification was tested using
independent test data withheld from training. This approach allowed for an empirically-driven
simplification of the model that substantially reduced the computational demands of the final
analysis.
34
Eight variables were removed from the model by this process32. Joining the independent
variables listed above, the final model included the following individual and work
controls: age, gender, number of children, native English speaker, highest degree
(Ph.D., professional degree, Master’s degree, B.A. or less), productivity, time employed,
Carnegie research intensity and region. Age and time-employed also included a
second-order polynomial.

Analytic Approach
There were two main analytical parts to this study: 1) a factor analysis and 2) a
series of structural equation models. These parts were related. The exploratory factor
analysis determined which latent factors were suggested by the data and how the
nineteen job satisfaction items loaded onto the factors. This factor structure informed
the development of the structural equation models. While the first two structural
equation models were simpler models, the final structural equation incorporated the
measurement model identified in the exploratory factor analysis stage. This full model—
a Bayesian probit structural equation model—was used to formally tested hypotheses in
this study.

For the factor analysis, early diagnostics33 suggested a four-factor structure to

job satisfaction. In the factor analysis itself, axes were rotated obliquely as correlation
was expected among latent factors. Principal axis factoring was employed to help
identify underlying constructs in the data. Item loadings suggested the following latent
dimensions of job satisfaction: work and supports (.76), co-workers (.79), financial (.79),
and benefits (.71) (Cronbach’s alpha reported in parentheses). According to standards
set down by Spector (1992), the internal consistency of these factors was excellent.
Each factor and its respective items are tabulated in the appendix.

The second part of the analysis consisted of a series of three nested structural
equation models (SEM), each predicting the outcome of turnover intentions. The first
SEM simply regressed turnover intentions on adjunct typology. In the second SEM,
covariates were introduced to the model to control for background characteristics. The
final model included adjunct typology, covariates, interaction terms and the
measurement model of job satisfaction factors. An independent holdout sample of data
(n=3,368) was used during testing to help prevent overfitting. The specific estimator
used in these models was Bayesian, instead of the traditional maximum likelihood
estimator. This was because of the complexity of the model and the resulting
computational demands that inhibited convergence. Bayesian methods often perform
better in these circumstances (Muthen and Asparouhov 2012), as was the case in this
study. The Bayesian structural equation models (BSEMs) used diffuse (noninformative)
priors. Future work should incorporate informative priors that can take advantage of
earlier findings in this line of research.

32 Union status, salary, professional development, selectivity, institutional type, citizenship status,
marital status, and race. Stress was not included in the model for reasons described by
Hagedorn (2000). In short, job satisfaction is viewed as the consequence of stressors, so it was
important to not “condition out” the variance in the model.
33 Scree plots, eigenvalues, parallel analysis and optimal coordinates.

35
 Figure 1: Bayesian Structural Equation Model (Full Model with All Covariates)

Note: In addition to the causal pathways indicated by arrows, the model also allowed for
unrestrained correlations between all independent variables (variables categorized under Adjunct
Type, Demographic Features and Work Background/Characteristics). The model also allowed for
inter-correlation among the latent satisfaction variables.

Results
Descriptive statistics of the HERI sample of non-tenure track faculty are available
in the appendix. Generally, these statistics were consistent with expectations. The only
point worth noting is that “aspiring academics” were older in this sample than expected.
In Gappa and Leslie's framework (1993), aspiring academics tended to be younger and
pursuing traditional faculty positions. The descriptive statistics suggested that this group
may be older and more “settled into” their part-time institutional roles than in Gappa and
Leslie’s original conceptualization.

Four latent factors were identified in the exploratory factor analysis. The “work
and supports” factor subsumes Kalleberg's intrinsic and resource adequacy dimensions.
Satisfaction with co-workers is analogous to the factor identified in the Kalleberg model.
Financial satisfaction subsumes Kalleberg’s financial and career dimensions. However,
in my analysis, the benefits dimension splits off into its own factor. Kalleberg’s model
viewed benefits as part of the financial dimension. This difference may be related to
unique benefit patterns among adjuncts. When the measurement model suggested by
the EFA was specified in a confirmatory factor analysis, the model fit was adequate (CFI
.889, TLI .882, RMSEA 0.061). Researchers typically wish to see CFI and TLI values
above .95 and an RMSEA less than 0.05 (Allison, Williams and Moral-Benito 2017).

This study specified three Bayesian structural equation models—increasingly

complex models terminating with the full specification of the measurement model
36
suggested by the factor analysis. To assess model fit, this study relied on posterior
predictive checking (PPC). PPC utilizes the likelihood ratio chi-square test as the
discrepancy function between the actual HERI data and data simulated by the model. A
chi-square value closer to zero indicates better fit. On this metric, the structural equation
models fit the data with different levels of goodness. The first two models (the simpler
models without the measurement model of satisfaction factors) was excellent. Their
replicated chi-square values overlapped substantially with observed chi-square values.
However, the fit of the full model (with the measurement model) was considerably
weaker. A 95% confidence interval for the difference between the observed and the
replicated chi-square values in this model was [5178.5, 5471.7].

Mplus calculates posterior predictive p-values for each of the models’ difference
in chi-squares. While similar to traditional p-values, Bayesian p-values account for the
variability in model parameters and do not require asymptotic theory (Asparouhov and
Muthen 2010). Poorly fit models reject the hypothesis that the actual and replicated chi-
squared statistics are the same. This was the case of the full model (p-value <0.001).
The first two models in this sequence, on the other hand, fit the data well according to
posterior predictive p-values (.52 and .41).

Satisfaction Hypotheses
While the factor analysis stage was important for the development of a
measurement model, all hypotheses in this study were formally tested using the final
structural equation model. This is the Bayesian probit SEM that includes the
measurement model of the latent satisfaction factors. However, as this is a complex
model simultaneously estimating hundreds of parameters, some organization of results
was called for. The first table (Table 5) presents only the results specifically pertaining
to the satisfaction-related hypotheses. Importantly, this is the part of the model
estimating the conditional medians associated with each of the adjunct classes,
controlling for individual and work characteristics.

37
 Table 5: Adjunct Satisfaction

Estimate Post. S.D.

Work and Supports a
Freelancer -0.03 0.05
Career-ender 0.03 0.06
Aspiring Academic -0.11 0.04 **
Full-time -0.04 0.04

Co-workers a
Freelancer 0.08 0.07
Career-ender 0.09 0.09
Aspiring Academic 0.03 0.06
Full-time 0.07 0.06

Financial a
Freelancer -0.03 0.05
Career-ender 0.17 0.07 *
Aspiring Academic -0.33 0.04 **
Full-time -0.07 0.04 *

Benefits a
Freelancer -0.01 0.10
Career-ender 0.31 0.14 *
Aspiring Academic -0.32 0.08 **
Full-time 0.47 0.08 **
a
Reference category = Adjunct Experts
* one-tail
The p-value
full model <0.05
also controlled **
forone-tail
a range p-value<0.01
of
individual and work characteristics not tabulated

With regard to work and supports, this study hypothesized that freelancers would
be less satisfied than other types of voluntary part-time faculty. This was because
freelancers possess less leverage over key aspects of their work like autonomy and
course assignments. However, from Table 5, the test statistic did not support this
hypothesis. The coefficient34 (-0.03) was not found to be significantly different than zero
(the satisfaction of freelancers on this dimension is comparable to the reference group of
adjunct experts). Aspiring academics, on the other hand, reported significantly less
satisfaction with regard to work and supports (-0.11**). This was fully expected based
on earlier research.

With regard to satisfaction with co-workers, this study expected career-enders to

exhibit greater satisfaction. This was because a key reason career-enders “transition”
into retirement relates to the importance of relationships in the workplace. However,
after controlling for confounders, the co-worker satisfaction of career-enders was
comparable to that of adjunct experts. In fact, this study did not find any differences in
co-worker satisfaction between any of the adjunct classes (voluntary or involuntary).

Freelancers were reasoned to be lower on financial satisfaction on account of

their greater dependency on a meager part-time teaching salary. However, this study

34 This contrast is specifically in relation to adjunct experts. The results are the same when

contrasting with career-enders.

38
found no statistical evidence that they differ from the reference category with regard to
financial satisfaction. The study did find differences, however, between adjunct experts
and career-enders. Career-enders reported much greater financial satisfaction.
Specifically, their median satisfaction was estimated to be .17 higher on this metric. The
one tailed p-value (indicated with * when <0.05 and ** <0.01) has a special interpretation
in Bayesian analysis. When the parameter estimate is positive, the p-value represents
the proportion of the estimated posterior distribution that is less than zero (Muthen
2010)35. Here, with a p-value 0.011 (not tabulated but represented with a single *), only
1.1 percent of the estimated posterior distribution falls below zero. That is strong
evidence of a positive effect. More than likely, financial compensation is not a main
reason career-enders maintain their part-time work—so they are more satisfied with the
compensation they do receive.

As predicted, aspiring adjuncts experienced the lowest financial satisfaction

(-0.33**). This almost certainly relates to their involuntary part-time status. However,
full-time adjuncts also tended to be less satisfied than adjunct experts with regard to
financial satisfaction (-0.07*). Perhaps, with a full-time position, full-time adjuncts expect
or would hope for greater financial compensation for such a strong commitment to their
employer.

Finally, with regard to work benefits, this study expected a higher degree of
satisfaction among career-enders. Job benefits are part of the reason faculty members
transition into retirement. Indeed, the analysis produced evidence that career-enders
were more satisfied on this dimension than the reference category of adjunct experts
(0.31*). The analysis also showed benefits to be a strong source of satisfaction for full-
time adjuncts (0.47**). This is probably because, as full-time employees, they probably
receive more comprehensive coverage from their institutions. Aspiring academics were
less satisfied with their benefits (-0.32**), likely tied to their underemployment.

Turnover Hypotheses
Table 6 presents the results of the Bayesian structural equation models with
different levels of complexity. Again, all hypotheses in this study were formally tested in
the final model (Model 3). The first two models are displayed for their value as reliability
checks and for interpretation of the overall turnover process. On account of the full
model’s complexity, the only coefficients presented are those with direct effects on
turnover intentions.

35 Or alternatively, when the parameter estimate is negative, the p-value represents the

proportion of the posterior distribution that is greater than zero.

39
 Table 6: Bayesian Structural Equation Models Predicting Turnover Intentions

Model 1 Model 2 Model 3

Estimate Post. S.D. Estimate Post. S.D. Estimate Post. S.D.
Freelancer 0.106 0.055 * 0.051 0.110 0.029 0.125
Career-Ender 0.045 0.055 0.414 0.155 ** 0.675 0.172 **
Aspiring Academic 0.181 0.082 * 0.32 0.088 ** -0.026 0.103
Full-time -0.288 0.082 ** 0.286 0.091 ** 0.248 0.098 **
Age -0.011 0.003 ** -0.009 0.003 **
Age2 -0.001 0.000 ** 0.000 0.000
Female 0.03 0.046 -0.043 0.051
Number of Children -0.059 0.018 ** -0.028 0.02
Native English 0.423 0.076 ** 0.61 0.084 **
Professional -0.031 0.088 -0.038 0.092
Master's -0.131 0.057 ** -0.064 0.066
B.A. or Less -0.361 0.099 ** -0.191 0.104 *
Research -0.027 0.160 0.186 0.171
Administration/Other 0.082 0.075 0.215 0.089 **
Productivity 0.02 0.004 ** 0.019 0.004 **
Time Employed 0.01 0.005 * 0.006 0.005
Time Employed2 -0.001 0.000 ** 0.000 0.000
Hard-Pure -0.066 0.131 -0.129 0.143
Soft-Applied -0.046 0.061 0.000 0.068
Soft-Pure 0.093 0.054 * -0.072 0.060
Publicª 0.127 0.051 ** 0.003 0.055
Research II 0.096 0.131 -0.052 0.137
Research III/PhD 0.135 0.148 -0.053 0.149
Bachelors/Masters 0.162 0.126 0.015 0.132
Associates/Other 0.119 0.263 -0.014 0.280
West/Other 0.069 0.064 0.115 0.068 *
Midwest 0.035 0.064 0.071 0.068
South 0.085 0.087 0.238 0.101 *
Satis. w/ Work and Supports -0.411 0.081 **
Satis. w/ Co-workers -0.057 0.040
Financial Satisfaction -1.005 0.087 **
Satis. w/ Job Benefits -0.008 0.049

Number of Free Parameters 19 463 636

Fit Statistic: Chi-square C.I. [-19.039, 21.269] [-66.471, 109.169] [5178.478, 5471.727]*
Posterior Predictive P-Value 0.47 0.6 0.000
* one-tail p-value <0.05 ** one-tail p-value<0.01

Hypothesis A predicted that each dimension of job satisfaction would be

inversely related to turnover intentions. Tabulated under Model 3, this study found
support for “work and supports” and financial satisfaction. As satisfaction with work and
supports and financial compensation increases, the faculty member is less likely to think
about leaving her or his principal academic employer (-0.411** and -1.005**,
respectively). This study did not find support for satisfaction with co-workers or job
benefits.

For Hypothesis B, it was reasoned that career-enders would be more likely to

leave their organizations, even after controlling for job satisfaction (and other
40
covariates). This is because career-enders have different motives than other adjunct
classes. They are transitioning to retirement and part of this may be unrelated to their
job satisfaction. This hypothesis was supported in the analysis. Career-enders had the
highest probability of turnover intentions (p-value <0.01). The study also revealed that
full-time adjuncts were also more likely to leave their institutions, independent of the job
satisfaction they experience. Potentially, their full-time commitment to academia is
rewarded by other institutions with tenure-line appointments. Further research would be
required to confirm this possibility.

For Hypothesis C, institutional control and discipline (Biglan classification) were

expected to be related to turnover intentions. Model 2 presents evidence that faculty in
public institutions think more regularly about leaving their institutions (0.127, p-value
<0.01). However, after controlling for components of job satisfaction in Model 3, this
relationship disappears36. This suggests that faculty think about leaving public
institutions specifically because they experience lower satisfaction on one or more facets
of job satisfaction.

Contrary to expectations, this study produced no evidence that faculty in applied

fields were any more likely to leave their institutions after controlling for background
characteristics (Model 3). However, the nested models revealed important features of
faculty in the “soft-pure” (liberal arts and social sciences) fields. Model 2—which did not
control for job satisfaction—showed that faculty in “soft-pure” fields were more likely to
consider turnover (than faculty in “hard-applied” fields). However, upon controlling for
job satisfaction in Model 3, soft-pure disciplines were no longer associated with turnover
intentions. This suggests that faculty may be leaving soft-pure disciplines specifically
because they are less satisfied with one or more dimensions of job satisfaction.

Discussion
Most researchers examining the work experiences and turnover of academic
faculty have focused on traditional tenure-line faculty. However, with non-tenure track
faculty now constituting the majority of academic professors, it is time to seriously
consider how to retain non-tenure track faculty. This study drew on Gappa and Leslie's
(1993) empirically-derived subclasses of adjuncts to examine the job satisfaction and
turnover intentions of adjuncts. It also examined the role of work and organizational
features (institutional control, Biglan classification) in this process. Like earlier work,
modeling the different dimensions of job satisfaction clarified how the wants and
expectations of adjuncts influence their attitudes, thoughts and turnover behaviors.

While research has shown that involuntary part-time faculty (aspiring academics)
are generally less satisfied with their work, there has been some debate regarding which
aspects of their work are more disagreeable. Maynard and Joseph (2008) found
aspiring academics to be less satisfied on Kalleberg’s intrinsic, career and financial
dimensions. Eagan et al. (2015) and Antony and Hayden (2011) had conflicting results
with regard to benefits satisfaction. With the exception of co-worker satisfaction, this
study found aspiring academics to be less satisfied across all other facets of job

36To be clear, this is not to say that institutional characteristics (like public/private) are
unimportant with regard to turnover intentions. It is that there is evidence that their effects are
channeled through mediators introduced in this study.

41
satisfaction. This finding seems more consistent with unemployment theory than earlier
work, linking surplus education, experience, and skill to various aspects of job
satisfaction.

The job satisfaction of adjunct experts and freelancers was comparable across
all latent factors. While this degree of similarity was not expected a priori, it is sensible.
These two classes of adjuncts do maintain similar structural relationships, even if one
holds a full-time career outside academia (academic experts) and one does not
(freelancers). Career-enders, on the other hand, differed in important ways. In fact, they
experienced higher satisfaction than adjunct experts on both financial and job benefits
scales. This was true independent of age—an important confounder controlled for in the
model. Perhaps career-enders are “grandfathered” into better salary and benefits and
this makes them more satisfied with those conditions. Through internal networks, they
may be able to acquire more generous salaries and benefits than could ever be
negotiated outside their home institutions. Alternatively, maybe their salary and benefits
expectations are different, given that their motives for working part-time are not
fundamentally material.

The findings with regard to full-time adjuncts are also important. Earlier work has
found that volunteer part-time faculty and full-time faculty are generally comparable in
their job satisfaction. The two exceptions are with regard to satisfaction with benefits
(Toutkoushian and Bellas 2003, Antony and Hayden 2011) and autonomy (Maynard and
Joseph 2008, Antony and Valadez 2002). This study corroborated the findings with
regard to benefits, but also found evidence that full-time adjunct faculty have less
financial satisfaction as well (compared to adjunct experts or career-enders). It is
interesting that, while adjunct faculty differed significantly in their satisfaction with
benefits, benefits was found to be unrelated to turnover intentions.

Finally, this study tested a key finding of Ehrenberg et al. (1991), Conley and
Leslie (2002) and Ryan et al. (2012), which found that faculty in applied fields were more
likely to leave their institutions. This was on account of the connection between applied
work and the private sector. However, this study was unable to replicate these findings.
After factoring in job satisfaction, applied faculty generally consider leaving their
positions at the same rate as any other adjunct faculty type. The analysis did show
however that faculty in the “soft-pure” sciences were more likely to leave their
institutions. The nested models suggested that this was because of the lower job
satisfaction in those fields—not a private sector connection per se.

The findings in this study have important implications for policy-makers and
administrators. For one, the most important way that postsecondary institutions can limit
their turnover is by developing policies targeting facets of job satisfaction. This study
suggests that policies aimed to improve satisfaction with “work and supports” and
financial satisfaction will be most effective. Clearly, with the financial pressures many
institutions face today, providing substantially better financial circumstances would be
challenging if not impossible. However, they may be able to free up resources by
investing less in adjunct health and retirement benefits (which are not linked to
retention). If they were to transfer those savings to adjuncts in the form of better
financial compensation, they would probably have more success retaining non-tenure
track labor. Nearly all full-time faculty and about half of part-time faculty currently

42
receive health or retirement benefits, so such an action would likely have a serious
impact.

Of course, administrators may also develop policies aiming to improve “work and
supports.” Indeed, many retention strategies employ these kinds of interventions (Kezar
2013, Eagan et al. 2015, Gappa, Austin, and Trice 2007). The findings in this study
imply that such strategies are both reasonable and will likely lead to success. Such
policies, however, will be unable stem all forms of turnover of course. There will always
be occupational mobility outside an institution’s control. In such circumstances, it seems
sensible that institutions think more carefully how to best manage turnover.
Departments may be able to set up an architecture that helps to deal with higher levels
of churn. This might include securing the long-term employment of core administrative
faculty with greater responsibility orienting, training and supporting regular faculty
additions. They may also want to centralize curriculum so that new faculty members can
focus on teaching—not course-planning. There is some evidence that departments are
already implementing these kinds of changes (Schuster and Finkelstein 2006).

The design of this study had limitations of course. Relying on cross-sectional

data, this study never observed actual turnover behavior. Instead, a cognitive measure,
“turnover intentions,” was utilized. While this is a good proxy for turnover, it is not a
perfect substitute. Future work should pursue actual turnover data (ideally longitudinal),
thus allowing a more complete exploration of the turnover process. As with any
regression framework, omitted variables potentially bias results and compromise test
statistics. While the HERI instrument was advantageous for its diverse set of
satisfaction items, it did not contain every important predictor of turnover (Cotton and
Tuttle 1986). If faculty with different turnover propensities37 drifted into different kinds of
faculty roles, this would clearly bias results.

The timing of this sample also introduces an important limitation. Data collection
for the 2010 wave occurred shortly following the “Great Recession” when labor market
conditions were tight. This probably constrained mobility and influenced the turnover
intentions of adjunct faculty at the time (Steers and Mowday 1981, March and Simon
1958). Results from this time period, thus, may not generalize well to time periods with
better economic conditions. Of course, some researchers have found limited effects of
the labor market (Zhou and Volkwein 2004), so the economic environment may also be
less important than what many think. This would be a useful area of future research.

Conclusion
Despite inevitable limitations, this study undoubtedly advanced this line of
research in important ways. It showed that finer-grained distinctions are absolutely
necessary in order to fully understand turnover patterns in academia. It also provided
important insights into how work experiences are related to turnover among adjunct
faculty. By targeting satisfaction components empirically related to turnover, institutions
can help secure a happy, committed and stable adjunct workforce.

References

37 Propensities not captured by the model.

43
Allison, Paul D., Richard Williams, and Enrique Moral-Benito. 2017. “Maximum
Likelihood for Cross-lagged Panel Models with Fixed Effects.” Socius 3: 1-17.

Antony, James Soto and Ruby A. Hayden. 2011. “Job Satisfaction of American Part-
Time College Faculty: Results from a National Study a Decade Later.”
Community College Journal of Research and Practice 35:689–709.

Asparouhov, T., & Muthen, B.. 2010. Bayesian analysis using Mplus: Technical
implementation. Available from http://www.statmodel.com/download/Bayes3.pdf.

Antony, James Soto and James R. Valadez. 2002. “Exploring the Satisfaction of Part-
Time College Faculty in the United States.” The Review of Higher Education
26(1):41–56.

Barnes, Laura L. B., Menna O. Agago, and William T. Coombs. 1998. “Effects of Job-
Related Stress on Faculty Intention to Leave Academia.” Research in Higher
Education 39(4):457–69.

Biglan, Anthony. 1973. “The Characteristics of Subject Matter in Different Academic

Areas.” Journal of Applied Psychology 57:195–203.

Bluedorn, A. C. 1982. “The Theories of Turnover: Causes, Effects, and Meaning.” in

Research in the Sociology of Organizations, edited by S. B. Bacharach.
Greenwich, CT: JAI Press.

Carnevale, David G. and Jo Marie Rios. 1995. “How Employees Assess the Quality of
Physical Work Settings.” Public Productivity & Management Review 18(3):221–
31.

Conley, Valerie Martin and David W. Leslie. 2002. Part-Time Instructional Faculty and
Staff: Who They Are, What They Do, and What They Think. Retrieved November
7, 2017.

Cotton, John L. and Jeffrey M. Tuttle. 1986. “Employee Turnover: A Meta-Analysis and
Review with Implications for Research.” Academy of Management Review
11(1):55–70.

Daly, Cheryl and Jay Dee. 2006. “Greener Pastures: Faculty Turnover Intent in Urban
Public Universities.” The Journal of Higher Education 77(5):776–803.

Eagan, Kevin, Audrey Jaeger, and Ashley Grantham. 2015. “Supporting the Academic
Majority: Policies and Practices Related to Part-Time Faculty’s Job Satisfaction.”
The Journal of Higher Education 86(3):448–83.

Ehrenberg, Ronald, Hirschel Kasper, and Daniel Rees. 1991. “Faculty Turnover at
American Colleges and Universities: Analyses of AAUP Data.” Economics of
Education Review 10(2):99–110.

44
Gappa, Judith M., Ann E. Austin, and Andrea G. Trice. 2007. Rethinking Faculty Work:
Higher Education’s Strategic Imperative. San Francisco, CA: Jossey-Bass.

Gappa, Judith M. and David W. Leslie. 1993. The Invisible Faculty. Improving the Status
of Part-Timers in Higher Education. San Francisco, CA: Jossey-Bass Inc.

Hackman, JR and GR Oldham. 1980. Work Redesign. Reading, MA.

Hagedorn, Linda Serra. 2000. “Conceptualizing Faculty Job Satisfaction: Components,

Theories, and Outcomes.” New Directions for Institutional Research
2000(105):5–20.

Herzberg, Frederick, Bernard Mausner, and Barbara Bloch Snyderman. 1959. The
Motivation to Work. Transaction Publishers.

Hirschfeld, Gerrit and Ruth Von Brachel. 2014. “Multiple-Group Confirmatory Factor
Analysis in R–A Tutorial in Measurement Invariance with Continuous and Ordinal
Indicators.” Prac Assess Res Eval 19(7).

Holtom, Brooks C., Terence R. Mitchell, Thomas W. Lee, and Marion B. Eberly. 2008.
“Turnover and Retention Research: A Glance at the Past, a Closer Review of the
Present, and a Venture into the Future.” Academy of Management Annals
2(1):231–74.

Hom, Peter W., Thomas W. Lee, Jason D. Shaw, and John P. Hausknecht. 2017. “One
Hundred Years of Employee Turnover Theory and Research.” The Journal of
Applied Psychology 102(3):530–45.

Hurtado, Sylvia, Kevin Eagan, John Pryor, Hannah Whang, and Serge Tran. 2012.
“Undergraduate Teaching Faculty: The 2010-2011 HERI Faculty Survey.”
Monograph Report.

Johnsrud, Linda and Vicki Rosser. 2002. “Faculty Members’ Morale and Their Intention
to Leave: A Multilevel Explanation.” The Journal of Higher Education 73(4):518–
42.

Kalleberg, Arne L. 1977. “Work Values and Job Rewards: A Theory of Job Satisfaction.”
American Sociological Review.

Kezar, Adrianna. 2013. “Departmental Cultures and Non-Tenure-Track Faculty:

Willingness, Capacity, and Opportunity to Perform at Four-Year Institutions.” The
Journal of Higher Education 84(2):153–88.

Kezar, Adrianna and Cecile Sam. 2010. Understanding the New Majority of Non-Tenure-
Track Faculty in Higher Education: Demographics, Experiences, and Plans of
Action: ASHE Higher Education Report. Jossey-Bass, Inc.

45
Klassen, Robert M., Ellen L. Usher, and Mimi Bong. 2010. “Teachers’ Collective
Efficacy, Job Satisfaction, and Job Stress in Cross-Cultural Context.” The Journal
of Experimental Education 78(4):464–86.

Lee, Thomas William, Tyler C. Burch, and Terence R. Mitchell. 2014. “The Story of Why
We Stay: A Review of Job Embeddedness.” Annual Review of Organizational
Psychology and Organizational Behavior 1(1):199–216.

March, J. G. and H. A. Simon. 1958. Organizations. Oxford, England: Wiley.

Maynard, Douglas C. and Todd Allen Joseph. 2008. “Are All Part-Time Faculty
Underemployed? The Influence of Faculty Status Preference on Satisfaction and
Commitment.” Higher Education: The International Journal of Higher Education
and Educational Planning 55(2):139–54.

Meyer, John P. and Natalie J. Allen. 1991. “A Three-Component Conceptualization of

Organizational Commitment.” Human Resource Management Review 1(1):61–
89.

Mitchell, Terence and Thomas Lee. 2001. “The Unfolding Model of Voluntary Turnover
and Job Embeddedness: Foundations for a Comprehensive Theory of
Attachment.” P. 189+ in Research in Organizational Behavior, vol. 23.

Mobley, W. H. 1977. “Intermediate Linkage in the Relationship between Job Satisfaction

and Employee Turnover.” Journal of Applied Psychology 62:237–40.

Mobley, W. H., Rodger W. Griffeth, Herbert H. Hand, and B. M. Meglino. 1979. “Review
and Conceptual Analysis of the Employee Turnover Process.” Psychological
Bulletin 86(3).

Morse, N. C. 1953. Satisfactions in the White-Collar Job. Oxford, England: Survey

Research Center, Institute.

Murphy, Gregory, James Athanasou, and Neville King. 2002. “Job Satisfaction and
Organizational Citizenship Behaviour: A Study of Australian Human‐service
Professionals.” Journal of Managerial Psychology 17(4):287–97.

Muthen, Bengt. 2010. Bayesian analysis in Mplus: A brief introduction. Available from
http://www.statmodel.com/download/introbayesversion%203.pdf.

National Center for Education Statistics. 2015. Characteristics of Postsecondary Faculty.

U.S. Department of Education. Institute of Education Sciences.

Oshagbemi, Titus. 1997. “Job Satisfaction and Dissatisfaction in Higher Education.”

Education + Training 39(9):354–59.

Oshagbemi, Titus. 1999. “Overall Job Satisfaction: How Good Are Single versus
Multiple‐item Measures?” Journal of Managerial Psychology 14(5):388–403.

46
Peters, Lawrence H., Edward J. O’Connor, and Cathy J. Rudolf. 1980. “The Behavioral
and Affective Consequences of Performance-Relevant Situational Variables.”
Organizational Behavior and Human Performance 25(1):79–96.

Porter, Lyman W., William J. Crampon, and Frank J. Smith. 1976. “Organizational
Commitment and Managerial Turnover: A Longitudinal Study.” Organizational
Behavior and Human Performance 15(1):87–98.

Price, James. 1977. The Study of Turnover. Ames, Iowa: Iowa State University Press.

Price, James L. and Charles W. Mueller. 1981. “A Causal Model of Turnover for Nurses.”
Academy of Management Journal 24(3):543–65.

Rice, Robert W., Michael R. Frone, and Dean B. McFarlin. 1992. “Work—nonwork
Conflict and the Perceived Quality of Life.” Journal of Organizational Behavior
13(2):155–68.

Rosser, Vicki J. 2004. “Faculty Members’ Intentions to Leave: A National Study on Their
Worklife and Satisfaction.” Research in Higher Education 45(3):285–309.

Ryan, John, Richard Healy, and Jason Sullivan. 2012. “Oh, Won’t You Stay? Predictors
of Faculty Intent to Leave a Public Research University.” Higher Education
63:421–437.

Schuster, Jack H. and Martin J. Finkelstein. 2006. The American Faculty: The
Restructuring of Academic Work and Careers. Baltimore, MD: The Johns
Hopkins University Press.

Seifert, Tricia A. and Paul D. Umbach. 2008. “The Effects of Faculty Demographic
Characteristics and Disciplinary Context on Dimensions of Job Satisfaction.”
Research in Higher Education 49(4):357–81.

Smart, John C. 1990. “A Causal Model of Faculty Turnover Intentions.” Research in

Higher Education 31(5):405–24.

Spector, Paul. 1997. Job Satisfaction: Application, Assessment, Causes, and

Consequences. 1st ed. Thousand Oaks, CA: SAGE Publications.

Spector, Paul E. 1992. Summated Rating Scale Construction: An Introduction. 1st ed.
Newbury Park, Calif: SAGE Publications, Inc.

Steers, Richard and Richard Mowday. 1981. “Employee Turnover and Post-Decision
Accommodation Processes.” Research in Organizational Behavior 3:235+.

Tabachnick, T. G. and L. S. Fidell. 2001. Using Multivariate Statistics. Boston: Allyn and
Bacon.

47
Toutkoushian, Robert and Marcia Bellas. 2003. “The Effects of Part-Time Employent and
Gender on Faculty Earnings and Satisfaction: Evidence from the NSOPF:93.”
The Journal of Higher Education 74(2):172–95.

Vroom, Victor. 1964. Work and Motivation. New York, New York: John Wiley & Sons.

Waltman, Jean, Inger Bergom, Carol Hollenshead, Jeanne Miller, and Louise August.
2012. “Factors Contributing to Job Satisfaction and Dissatisfaction among Non-
Tenure-Track Faculty.” The Journal of Higher Education 83(3):411–34.

Weller, Ingo, Brooks C. Holtom, Wenzel Matiaske, and Thomas Mellewigt. 2009. “Level
and Time Effects of Recruitment Sources on Early Voluntary Turnover.” The
Journal of Applied Psychology 94(5):1146–62.

Xu, Yonghong Jade. 2008. “Faculty Turnover: Discipline-Specific Attention Is

Warranted.” Research in Higher Education 49(1):40–61.

Zhou, Ying and James Volkwein. 2004. “Examining the Influences on Faculty Departure
Intentions: A Comparison of Tenured versus Nontenured Faculty at Research
Universities Using NSOPF-99.” AIR Forum Issue 45(2):139–76.

Appendix
Table 7: Descriptive Statistics (n=8418)

Aspiring
Full-time Career-Ender Expert Freelancer
Academic
n 4527 2315 210 715 612
Avg. Age 53 52 65 48 49
Female 41% 68% 47% 53% 56%
Married 84% 83% 82% 75% 77%
White 86% 91% 92% 79% 84%
Ph.D. 14% 24% 30% 23% 37%
Professional 12% 9% 12% 9% 9%
Masters 60% 60% 53% 60% 49%
BA or Less 15% 7% 4% 8% 5%
Teacher 98% 93% 95% 97% 80%
Researcher 1% 1% 1% 1% 4%
Administrator/Other 2% 5% 4% 2% 17%
University 28% 27% 23% 26% 36%
Public 29% 36% 39% 39% 38%
Research I 2% 3% 2% 2% 5%
Research II 20% 15% 12% 15% 21%
Research III/Doctoral 4% 6% 6% 7% 9%
Bachelor's/Master's 73% 75% 79% 76% 64%
Associates/Other 0% 0% 0% 0% 1%
Hard/Applied 55% 32% 33% 37% 44%
Hard/Pure 2% 3% 1% 2% 4%
Soft/Applied 18% 24% 27% 18% 19%
Soft/Pure 26% 41% 39% 44% 33%
Selective 4% 6% 8% 5% 14%

48
 Table 8: Standardized Factor Loadings

Respondent's satisfaction with: axis 1 axis 2 axis 3 axis 4

Salary 0.062 0.269 -0.051 0.411
Health benefits 0.012 0.823 0.005 -0.028
Retirement benefits -0.023 0.87 0.014 0.024
Opportunity for scholarly pursuits 0.141 0.222 0.008 0.454
Teaching load 0.345 0.041 -0.084 0.406
Quality of students 0.351 0.038 0.021 0.139
Office/lab space 0.317 0.199 0.069 0.076
Autonomy and independence 0.550 0.034 0.201 -0.072
Professional relationships with other faculty 0.027 0.018 0.909 -0.026
Social relationships with other faculty -0.042 -0.001 0.755 0.081
Competency of colleagues 0.283 0.014 0.383 0.038
Job security 0.012 0.078 0.088 0.562
Departmental leadership 0.379 -0.081 0.255 0.167
Course assignments 0.609 -0.016 0.003 0.164
Freedom to determine course content 0.689 0.03 -0.014 -0.092
Availability of child care at this institution -0.019 0.023 0.041 0.447
Prospects for career advancement -0.025 -0.026 0.045 0.818
Clerical/administrative support 0.313 -0.042 0.118 0.16
Tuition remission for your children/dependents 0.122 0.291 -0.024 0.215

49
 Table 9: Latent Factors with Satisfaction Items

The measurement model in this study utilized nineteen job satisfaction items
(Table 9). Diagnostic tests38 were used to identify the optimal number of factors. Using
oblique rotation and principal axis factor extraction39, the EFA identified the latent factor
structure in the matrix of satisfaction items. All but one item40 successfully loaded using
traditional criteria (Tabachnick and Fidell 2001). The summated ratings scales were
tested for reliability using Cronbach’s alpha—a calculation of internal consistency. All
factors were associated with a Cronbach alpha above .71, which Spector (1992)
considers “excellent” internal consistency41.

38Tests included scree plots, eigenvalues, parallel analysis and optimal coordinates.
39Oblique rotation was utilized because the study expected latent factors to be correlated.
Principal axis factoring was factoring was employed because this study was most interested in
identifying the underlying constructs in the data.
40 The only item below the common threshold of .3 was satisfaction with tuition remission (.291).

As it was close to the threshold, this variable was not excluded from the measurement model.
41 .71 and higher is excellent, .63 to .7 is very good, .55 to .62 is good, .45 to .54 is fair, and .3 to

.44 are poor Comrey (1992).

50
 CHAPTER 4: ATTRITION: NON-TENURE TRACK FACULTY LEAVING ACADEMIA

Abstract
The work environment and experiences of faculty beginning off the tenure track
are typically inferior to those of traditional tenure-line faculty. However, there is little
research on how these differences may impact career attrition from academia. Given
that most faculty members now work off the tenure line, it is critical to reassess career
attrition in academia. Drawing on panel data from professors in the Survey of Doctorate
Recipients (n=17,129), this study examines the impact of appointment type and work
responsibilities on career permanence. It also assesses how structural characteristics
are related to the risk of occupational turnover.

Introduction
The proportion of faculty working off the tenure track has increased steadily over
the last half century (Schuster and Finkelstein 2006). Today, more than two out of three
postsecondary faculty appointments are made off the tenure line (National Center for
Education Statistics 2015, Schuster and Finkelstein 2006). Largely an institutional cost-
saving measure, non-tenure track faculty earn lower salaries, fewer benefits and
experience generally inferior working conditions (Gappa, Austin, and Trice 2007,
Baldwin and Chronister 2001). For these reasons and others, it is reasonable to assume
that beginning off the tenure-track would make it more difficult to establish a long-term
career in academia. However, this has not been demonstrated empirically. The goal of
this study is to determine the impact of initial appointment type on permanence in the
academic profession. It also examines how structural factors (work activities and
organizational context) affect the risk of academic career attrition.

This study utilizes two decades of panel data from the Survey of Doctorate
Recipients (SDR) (1993-2013). SDR follows a sample of research doctoral degree
recipients who graduated from science, engineering, and health fields. This study
makes use of a specific subset of doctorate recipients—those beginning careers in
academia (n=17,129). This is an important subset. SDR contains a wide range of
demographic, education and career-related information. Importantly, this study is one of
the first to use survival analysis42 in this line of research– a “critical” methodology
repeatedly called for (Morita, Lee, and Mowday 1989, Rhodes and Doering 1983, Hom
et al. 2017). For most attrition-related studies, survival analysis is superior to traditional
regression methods because of how it accounts for censoring (Allison 1982).

This study builds on this line of research in several important ways. For one,
scholars have called for greater consideration of time in turnover research (Hom et al.
2017, Lee, Burch, and Mitchell 2014), but rarely has the timing of turnover been
prioritized in faculty research (Rosser 2005). This study is the first43 to offer robust,
nonparametric estimates of time to faculty career attrition (survival curves). Importantly,
it examines specifically how initial appointment on and off the tenure track impacts

42Also known as event history analysis.

43Kaminski and Geisler (2012) also employ survival analysis to faculty data, however, they did
not examine turnover timing using survival curves.
51
career attrition. This is critical because higher education is dependent on its non-tenure
track labor force but their turnover behaviors have historically been ignored by
researchers (Gappa and Leslie 1993, Kezar and Sam 2010). Second, this study
develops and tests a model using actual turnover data—not “turnover intentions” as
used in most faculty studies. This is a substantial contribution: nearly all theory is based
on models examining a cognitive proxy—a poor substitute when actual turnover data
can be mined from data.

Establishing a better understanding of the academic turnover process is crucial.

Turnover can be either good (e.g., renewing energy, introducing new ideas) or bad44
(e.g., losing talent, eroding morale) for institutions and professions. Whether it is good
or bad is generally a function of how much attrition there is and when it is occurring. By
clarifying the patterns and places of academic attrition, this study provides a useful guide
for administrators, human resource managers and policy makers. Secondly, there are
not enough jobs to go around in academia (Larson, Ghaffarzadegan, and Xue 2014) and
young people may not be aware of the risks of pursing an academic career (Benton
2003). This study provides transparency into the academic career attrition process,
thereby helping young people (and graduate programs) make informed decisions about
their futures. For those who decide to pursue academic careers, the findings imply
strategies for successfully establishing and maintaining a career in academia.

Literature Review
Conceptual Framework
Research on turnover is burdened by extensive conceptual ambiguity and
redundancy45 (Morrow 1983, Maertz and Griffeth 2004), so it is important to establish up
front a clear conceptual framework for the examination of this topic. To be clear, this
study examines career turnover: the event whereby a faculty member changes his or her
occupation to one outside the typical career progression (Lee, Carswell, and Allen 2000,
Blau, Allison, and St. John 1993). By typical, I mean the orderly sequence of
development within an occupational domain and the progressive accumulation of greater
responsibility (Becker and Strauss 1956, Slocum 1966). An occupation is identifiable by
its unique collection of skills, knowledge, and duties. In this study, the terms occupation,
profession, vocation, and career are synonymous.

Occupational turnover is one manifestation of a more general process of

employee withdrawal. Discouraged workers withdraw from their work in many ways.
Some withdraw psychologically, lowering the “potency” of their job involvement. Others
try to improve their work conditions through promotion or unionization (Allport 1962).
Leaving a job and pursing employment elsewhere is a consequential form of withdrawal

44 Attrition requires that organizations recruit and train replacements and it generally results in
lower productivity (Hausknecht, Trevor, and Howard 2009) and performance (Lee, Carswell, and
Allen 2000). Student learning may also be compromised when their instructors have little or no
experience teaching courses (Eagan and Jaeger 2008).
45 Occupational turnover is closely related to occupational retention, career resilience,

occupational commitment, occupational turnover intentions, etc.

52
because of how it disrupts working relationships and often changes work activities
(Hulin, Roznowski, and Hachiya 1985).

In terms of frequency, occupational turnover is somewhat rare. This is because

abandoning a line of work typically entails significant costs like additional training,
severed work relationships, time and resources lost in the transition (Neapolitan 1980,
Blau 2000). It is much more common, for example, for discouraged workers to leave
their organization to work for another organization in the same field. This kind of
turnover, voluntary organizational turnover, merely involves changing one’s employer—
not changing one’s occupation or retiring skill sets (Louis 1980). Perhaps because this
transition is so much more common, organizational turnover has received far greater
research attention and theoretical development (Blau and Lunz 1998).

While distinct from organizational turnover (Blau 2000), the theoretical

architecture behind occupational turnover is nearly always situated in the extensive work
conducted on voluntary organizational attrition (Lee et al. 2000). This study, like others,
draws strongly on insights from the field of organizational theory, as well as the empirical
literature on career attrition. While some have called for greater development of theory
tailored for career turnover (Blau 2007), little research to date has shown these causal
processes (career versus organizational turnover) to differ substantially.

Theory on Turnover
Most theoretical work on turnover is rooted in expectancy theory (Vroom 1964).
Expectancy theory is a general theory of motivation that posits a rational or cognitive
basis for behavior. In this paradigm, an individual identifies a desirable goal, rationally46
evaluates options to accomplish that goal, and subsequently selects the option with the
greatest expectancy of bringing about the desired end. In the context of work, this
means that individuals possess expectations about what their work should be. When a
job meets expectations (or reasonably leads to expectations), employees maintain the
work arrangement. If their work falls short, workers pursue behaviors with a likelihood of
bringing about work expectations. They may be motivated to renegotiate work
responsibilities, seek promotion, consider a reassignment, join a union, or exert less
effort (Hulin et al. 1985). Quitting a career is one possibility and sometimes may be
viewed as the option most likely to bring about one’s work expectations. Importantly,
expectancy theory implies that an individual’s decision to stay or leave is a function of
structural, attitudinal and environmental factors.

March and Simon (1958) were the first to integrate components of expectancy
theory into research on employee withdrawal. (Mobley 1977) and (Price 1977) provided
a sound empirical foundation for many of these ideas, operationalizing job desirability
with measures of job satisfaction. Mobley’s (1977) work was key for framing turnover as
a drawn-out process. His models also elucidated in great detail the cognitive steps that
precede a decision to quit. Mobley et al.’s (1979) content model tied quit decisions to
distal causes (whether some task was disliked or a new job was more attractive). The
specific mechanism was mediated by attitudes and other subjective evaluations of

46 Some debate how truly “rational” quit decisions may be (Lee and Mitchell 1994, Hulin,

Roznowski, and Hachiya 1985), but rationality or bounded rationality is nearly always assumed.
53
external features and circumstances. Price (1977) called turnover the “the degree of
movement across the membership boundary of a social system.” Price (1977) and Price
and Mueller (1981) focused more on the structural and economic context of turnover.
This included the effect of environmental, workplace/organizational features (integration,
pay, professionalism), organizational characteristics, labor market opportunities, and
social relationships. For both Mobley and Price, (dis)satisfaction was specified as a key
driver of turnover.

While work expectations (measured by satisfaction) were essential to Mobley

and Price’s work, another line of research focused more on the forces constraining
mobility. Becker (1960) argued that workers make “side bets” when participating in
social organizations. These are investments that would result in costs if a worker were
to leave. Many side bets are external, like a worker’s choice to invest in a home or
establish “roots” in a community. Some side bets are accumulated through tenure in an
organization, like health and retirement benefits, seniority in an organization, as well as
one’s cultivated workplace image. Often, side bets accrue great value and the cost of
losing organizational membership far exceeds whatever gains could be expected from
changing work arrangements.

Later, Becker’s theories were incorporated into the broader and more formal
construct of job commitment (Porter, Crampon, and Smith 1976, Meyer, Allen, and A.
Smith 1993, Porter et al. 1976). In addition to costs47, workers also tended to develop
affective and normative commitments for organizations (Meyer and Allen 1991).
Affective commitment referred to the employee’s emotional commitment to a group’s
values and goals. Normative commitment recognized the moral obligation many feel for
their organization or profession. Research has shown that workers develop important
commitment attitudes to both their organization and profession (Irving, Coleman, and L.
Cooper 1997, Carson, Carson, and Bedeian 1995). Generally, organizational
commitment is viewed as more stable than job satisfaction (Morrow 1993). This line of
research is important because it recognizes that work choices are always made under
constraints—some rational and some emotional. Workers do not simply change a
career or an organization because they seek to optimize their own job satisfaction and
self-interest. Choices are bounded by costs, obligations, and loyalty to one’s profession,
organization, and colleagues.

Commitment research is closely related to another line of research stressing the

forces that motivate workers to stay (in contrast to forces that constrain)(Mitchell and
Lee 2001). This is the theory of embeddedness, which is less “affect or affect-saturated”
and more contextual than commitment (Lee et al. 2014). Embeddedness recognizes
that workers are enmeshed in larger social systems that facilitate or encourage
individuals to stay. Lee et al. (2014) argued that embeddedness has three critical
components: linkages, perceptions of fit, and desire to avoid sacrifices of leaving.
Feldman and Ng (2007) explored professional embeddedness, finding the important role
of occupation-related groups (e.g., industry contacts), activities in professional societies,
educational investments and occupational status. Some (Hom et al. 2017) have noted

47 Now referred to as continuance commitment.

54
the conceptual overlap between theories of embeddedness and the theory of
organizational commitment48.

While these earlier theories have been important for understanding turnover, they
left much variation in turnover unexplained. The “unfolding model” (Lee and Mitchell
1994) is the latest theoretical perspective to clarify what other features are important to
the turnover process. The unfolding model posits that turnover is principally driven by
“shocks49” that are external to the job itself (Hom et al. 2017). Hagedorn (2000) called
these “triggers.” Parenthood, for example, causes men and women to reconsider work
arrangements and make significant job-related decisions. Sometimes this means
pursuing a job with greater financial rewards and sometimes this means leaving the
workforce to focus on family responsibilities. Unsolicited job offers are another important
shock. Regardless of a worker’s job satisfaction, a secure offer may force a
reconsideration of one’s work and lead to the decision to change employers or
industries. Some have found that external shocks explain more variation in turnover
than any other factors (Holtom et al. 2008)50.

As noted, occupational turnover models are typically situated in organizational

theory; however, researchers have discovered some nuances to the occupational
attrition process. Guided by the theoretical work51 of Rhodes and Doering (1983), Blau
(2007) found work exhaustion52 to be particularly salient to the career change process.
Similarly, Lee and Ashforth (1996) and Barnes, Agago, and Coombs (1998) identified
the related process of “burnout” to be a significant predictor of occupational turnover.
Work exhaustion and burnout are largely channeled through the mediator of job
satisfaction, meaning that job satisfaction may be more important to career attrition than
other socio-psychological variables (Blau 2007, Blau and Lunz 1998, Blau 2000).

Prior Research on Faculty Career Attrition

Few studies have examined academic career attrition. Utilizing a stress-based
model based on Gmelch, Wilke, and Lovrich (1986), Barnes et al. (1998) found that the
two most important predictors of intent to leave academia were stressors related to time
commitment and sense of community. Background characteristics in this study had very
little influence on attrition (e.g., gender, tenure status or academic discipline). Johnsrud
and Rosser (2002) used data from a single institution to develop a model of faculty
turnover intentions. Rosser (2004) later tested a similar model using nationally-
representative data (NSOPF 1999), which included non-tenure track faculty members.

48 However, Mitchell and Lee (2001) and others have found an independent effect of
embeddedness on turnover.
49 Maertz and Griffeth (2004) identified eight reasons for employee turnover.
50 Perhaps sixty percent of turnovers may be driven by shocks described in the unfolding model

(Weller et al. 2009).

51 Rhodes and Doering 's(1983) career-change model was an attitude-based model firmly
situated in traditional organizational research (Mobley, Horner, and Hollingsworth 1978, Price
1977, and Steers and Mowday 1981). It expressed career withdrawal cognitions (career change
intent) as a function of individual/organizational antecedents and job/career satisfaction.
52 Moore (2000) called work exhaustion “the depletion of emotional and mental energy needed to

meet job demands.”

55
She confirmed that faculty perceptions of their work lives (work activities and supports)
have an important indirect impact on turnover (through job satisfaction). While tenured
faculty viewed their work lives more negatively, she found that tenure status substantially
boosted job satisfaction, resulting in fewer turnover intentions.

Daly and Dee (2006) examined intent to stay among full-time instructional faculty
in urban research settings. Some of these faculty members held tenure and the rest
were on the tenure track. They found that structural antecedents had indirect effects on
intent to stay, channeled through job satisfaction and organizational commitment.
Perceived job opportunities also had important direct effects on intent to stay. Utilizing
NSOPF (1999), Zhou and Volkwein (2004) examined the organizational turnover of
tenured and non-tenured faculty (aggregating all faculty without tenure, including faculty
on the tenure-track, non-tenure track and those at institutions without tenure systems).
They found that the most important components of turnover intentions were education
(whether the faculty member possessed a Ph.D.), seniority/rank, and sense of job
security. Each of these features was important for limiting thoughts about leaving.

Using data from HERI, Ryan, Healy, and Sullivan (2012) examined the
occupational turnover intentions of faculty at a large public research university. They
found that structural antecedents (hard-applied disciplines, fit, support, stress) and
attitudes (satisfaction) were all related to occupational turnover intentions.

Limitations of Earlier Research

While researchers have developed insightful models predicting turnover (Hom et
al. 2017), only on rare instances have they empirically documented the timing of early-
career attrition among postsecondary faculty members. This is especially the case for
non-tenure track faculty, who have historically been disregarded by researchers (Kezar
and Sam 2010, Gappa and Leslie 1993). This study addresses this oversight, producing
robust estimates of faculty attrition that account for censoring.

Secondly, when this line of research has examined the impact of tenure, it has
typically contrasted faculty holding tenure with those on the tenure-track (Barnes et al.
1998, Daly and Dee 2006, Ryan et al. 2012). In the few cases including non-tenure line
faculty, they were confusingly aggregated with tenure-line junior faculty (Rosser 2004).
Only Johnsrud and Rosser (2002) specifically examine the class of non-tenure track
faculty53, but utilizing data from a single university system, it is unclear how
generalizable findings may be. Now that the faculty majority hold non-tenure line
appointments, it is pressing to establish a better understanding of how initial
appointment status impacts career attrition. This study is one of the first to put non-
tenure track faculty at the center of its research agenda.

Third, researchers of turnover make a big assumption when they test theory
using “turnover intentions,” rather than actual turnover. As it is the best predictor of
actual turnover, intentions are admittedly a “good” proxy (Bluedorn 1982, Rosser 2004).
However, institutions are typically interested in turnover behaviors and there is no
substitute for a behavioral measure—especially when the distance between thinking
about quitting and actually quitting is so large. Research has shown that turnover

53 Presumably, “instructors” indicates non-tenure track status.

56
intentions may only share a quarter of its variation with actual observed turnover (Allen,
Weeks, and Moffitt 2005, Griffeth, Hom, and Gaertner 2000) and this should be
concerning. Unlike earlier studies, this study mines faculty panel data to identify actual
turnover, and it uses this information to test a model firmly situated in this line of
research.

Finally, research has long been interested in the structural determinants of

turnover attitudes, cognitions and behaviors (Price 1977). However, the abstract
findings from this line of research are rarely tied to particular structural features of
postsecondary institutions. Daly and Dee's work (2006), for example, examines how
faculty work environments (autonomy, communication openness, distributed justice, role
conflict and workload) impact career turnover. Each of these features, of course,
suggest potential interventions for administrators; however, there is no consideration of
how traditional postsecondary institutional features (e.g., type/level, public/private
control, research intensity, discipline) are related to turnover. Administrators are often
less informed of their human resources than many realize (Cross and Goldenberg 2009),
so identifying the institutional features associated with greater risk of faculty departure is
an important contribution.

Proposed Model
While there are some differences in how researchers have specified turnover
models, this is mostly attributable to authors’ diverse disciplinary orientations—not
fundamental differences in how theorists believe attrition occurs (Smart 1990). In
general, turnover is a function of distal antecedents (e.g., job characteristics,
organization, personality) which affect attitudes (e.g., job satisfaction, commitment).
These attitudes, in the context of external conditions (labor market opportunities54), bring
about turnover cognitions and, for some, actual turnover (Hom et al. 2012, Smart 1990).
This is true both for organizational turnover as well as occupational turnover (Blau 2007,
Rhodes and Doering 1983). As discussed in the review of the literature, research has
shown that this general model extends readily to the case of academia. This study
integrates the Rhodes and Doering (1983) career-change specification with attention to
correlates identified in Cotton and Tuttle's meta-analysis (1986) of employee turnover.

Research Questions
How does initial appointment type affect the timing of career attrition among
tenure-line and non-tenure track faculty? How important is an initial tenure-line
appointment for maintaining a career in academia? Which faculty work activities are
related to risk of career attrition? Are traditional postsecondary institutional features
linked to higher risk of career attrition? What are the implications of these findings for
public policy and the administration of higher education?

Survival analysis is one of the best approaches for answering these kinds of
questions. As Allison (1982) explains, because of censoring and time-varying
explanatory variables, standard regression methods would be inappropriate for this

54 Labor market conditions have a direct effect on turnover (Daly and Dee 2006).
57
study. There would almost surely be serious bias or loss of information. Researchers
have consistently called for greater use of longitudinal methods and survival analysis
(Steel 2002, Lee et al. 2014, Morita et al. 1989) to examine turnover. Rhodes and
Doering (1983) called the use of panel data “critical” to the field.

This study utilizes discrete-time measurements, as this was how the panel data
in this study were collected. The wave intervals of this study are expected to be short
enough to capture variation among different faculty types. Research has shown that
non-tenure track faculty typically remain in their positions for many years, meaning that
job experiences likely span one or more intervals of data collection. According to the
American Federation of Teachers (2010), 32 percent of part-time, non-tenure track
adjuncts have held their positions between six to ten years. A plurality (40 percent) have
worked on their campuses for 11 years or more.

Hypotheses
There are many reasons to suspect that career permanence will be significantly
shorter for faculty beginning their careers off the tenure track. Tenure protects academic
freedom and provides the job security that makes the profession attractive (Xu 2008). It
also reduces stressors linked to turnover (Thorsen 1996). Role conflict is probably also
important as professors (particularly Ph.D. recipients with research backgrounds) are
typically socialized into traditional professional norms that lead one to expect and value
a tenure-line appointment. Beginning off the tenure track, then, probably results in the
kind of role conflict that reduces job satisfaction and drives career attrition (Daly and Dee
2006).

While no national studies map the timing of academic career attrition using
survival curves, considerable research has been conducted on elementary and
secondary school teachers. It is my belief that this body of research can inform our
expectations, although there are admittedly significant differences between teachers and
professors. Ingersoll (2002) found that 14% of school teachers left after one year, 33%
left after three years and 46% left after 5 years55. Estimates from the American
Federation of Teachers (2010) suggest that professors may have similar career attrition
patterns—though slightly less pronounced.

Causal models have tended to find that, net of controls, tenure matters with
regard to career turnover. Tenure status had the single largest total effect on turnover
intentions in Rosser (2004). Bland et al. (2006) found that, when asked whether they
would do it over56, non-tenure track faculty reported less interest in academic careers.
They were also more likely to report expecting to change positions in the near future.
Thus, earlier work suggests that initial appointment off the tenure-track should be
associated with greater risk of career departure, even after controls are introduced. The

55 The National Center for Education Statistics (2014) tabulated that 7.7% of teachers leave their
profession every year (an additional 8.1% change schools).
56 This is also a common measure of occupational commitment.

58
second hypothesis is identical to the first, except that it controls for confounders in a
semi-parametric model (a Cox proportional hazards model).

Hypothesis 1: Faculty with initial appointments off the tenure track will tend to leave their
careers significantly earlier than those beginning with tenure-track appointments.

Hypothesis 2: Net of controls, initial appointments off the tenure track will result in higher
risk of career attrition.

It is well-established that immediate work environments are extremely important

for turnover (Lee et al. 2000, Hom et al. 2017). Thus, the principal activities faculty
members engage in surely have an impact on their career satisfaction, commitment and
turnover. Authors have highlighted the importance of role conflict in job satisfaction.
Employees more aligned with their training and career expectations experience greater
work satisfaction. Blau's findings (2000) on professional context are also suggestive of
role conflict. Academic faculty are generally socialized into research expectations
(particularly for the sample used in this study). So, faculty members appointed to
predominately teaching or administrative roles may experience greater role conflict and
propensity to leave their careers. On the other hand, studies have also shown research
to be more stressful than teaching or administration (Matier 1990, Thorsen 1996, Rosser
2004). Working with students (Xu 2008) and teaching (McGee and Ford 1987, Hoyt
2012) have also been found to be important to retention57. Research experience is also
highly valued (and lucrative) in the private sector, which suggests that “applied” research
faculty may experience a greater “pull” away from academia (Ryan et al. 2012). While
this study acknowledges that role conflict may be important, it expects that research
stress and private sector “pull” will be the dominant forces of career attrition for faculty
with initial research appointments58.

Hypothesis 3: Compared to primary appointments in teaching or administration,

research-intensive positions will have a higher risk of career attrition (controlling for
faculty tenure status).

While some have found that institutional characteristics matter with regard to
turnover (Zhou and Volkwein 2004), little evidence suggests that traditional
postsecondary features impact career turnover in any substantial way. Examining
organizational characteristics, Blau (2000) found little evidence of an effect of
organizational context59 on career turnover. Johnsrud and Rosser (2002) dropped
traditional postsecondary institutional features (e.g., institutional type) from their model
after they failed to show a correlation with turnover in an exploratory analysis. Thus,
research in general suggests that turnover is more of an individual-level process (and

57 For a conflicting opinion, see Smart (1990).

58 On account of findings in Cross and Goldenberg (2009), this study also expects that
administrators will attrite at higher rates than teachers as well.
59 A construct consisting of organizational reduction, shift schedule, and full-time/part-time status.

59
less organizational or institutional). Therefore, while it is important to examine whether
institutional features may impact career turnover, I expect weak or insignificant
relationships with career turnover.

Hypothesis 4: Traditional institutional characteristics will be unrelated to risk of career

attrition.

Method
Sample
This study uses data from the Survey of Doctorate Recipients (SDR) from 1993
to 2013. While SDR goes back to much earlier decades, significant changes were made
to the survey in the early 1990s. Given the complications of interpreting earlier data and
that the focus of this study is on early-career scholars, it was reasonable to exclude the
data of distant waves. SDR consists only of doctoral recipients: every subject has
earned a research doctoral degree from an American postsecondary institution. This
study only used data from individuals who were appointed to a position (as their principal
employment) at a postsecondary institution (at any time after receiving their Ph.D.). This
is a distinct subset of Ph.D.-recipients, as many young people enter graduate school and
the workforce without any intention of ever working in academia. In total, there were
17,129 individuals contributing one or more waves of data. Forty-nine percent were in
tenure-granting institutions. Fifty-one percent were in institutions60 without tenure
systems. Subjects were followed until they were no longer principally employed in
academia (if they ever left).

Nonresponse rates are always an issue—particularly in longitudinal survey

designs. Massey and Tourangeau (2013) report that nonresponse rates over 30% are
quite common in major national surveys today and that is it not uncommon for rates to
exceed 60%. By this standard, nonresponse in the SDR was relatively low: the
unweighted nonresponse rate in 2015 was 32%. A clear concern for this study is
whether this censoring may be informative of the timing of career attrition. If it is,
estimates of model parameters could be significantly biased and this is a difficult
problem to diagnose.

Measures
Dependent variable. The dependent variable in this study was time, in years, from
initial academic appointment to attrition from academia. This value was calculated from
panel data. At each wave, faculty recorded whether or not they were employed at an
educational institution. Their first report of academic employment was viewed as their
start date. Faculty members were followed every wave until they reported employment
outside of academia (or were censored). In addition, one year was added to the attrition
time of each subject to avoid the case of “spontaneous attrition.” Faculty members who
enter and exit in the same year—an important segment—cannot mathematically be
included in a survival analysis. Adding one year to each event time is equivalent to

60 Non-tenure systems include most for-profit and private, 2-year institutions. The number of

public, 2-years institutions and private, 4-year institutions without tenure systems has also been
growing over time (see Figure 3 in Appendix).
60
assuming that academic appointments began midway through the interval that preceded
their “start date.” Most faculty did not leave their careers during the course of this study,
so their event times were censored. Regardless of tenure status, most faculty members
(even non-tenure track) tended to work lengthy periods for academic employers
(American Federation of Teachers 2010).

Independent variables.

Tenure-line appointment was reported by faculty members in each wave. The value
reported at the time of first academic appointment was recorded as initial appointment
type. Faculty were coded as tenure-track, non-tenure track (but at an institution with a
tenure system) or non-tenure track (but without a tenure system at their institution).
Faculty beginning on the tenure-track constituted 31.1% of the sample. 18.4% began off
the tenure-track but in a tenure system institution. 50.5% begin their first academic
appointments at institutions without a tenure system.

Principal work activity was reported in each wave. The value reported at the time of first
academic appointment was recorded as initial work activity. Positions were coded to be
primarily teaching, research or administration/other.

Organizational Characteristics were reported at first wave of academic appointment.

SDR contained information on institutional control (private/public), type/level (two-year,
four-year, medical, or research) and Carnegie research intensity (RI, RII, doctoral
institution, medical/health institution, or other). Ideally, this study would have also
controlled for discipline, however, this information was not provided for all time points
and thus it was excluded from the study.

Covariates. The model developed in this study controls for a range of demographic,
work and organizational variables. Choice of covariates was influenced by Cotton and
Tuttle’s comprehensive meta-analysis of turnover (Cotton and Tuttle 1986).
Demographic controls included age, sex, education, minority, marital, parental and
citizenship status at the onset of their first academic appointment. This study also
controlled for logged entry salary. Entry date (the year of first academic appointment), a
continuous variable, was used to control for cohort effects. Length of interval between
receipt of degree and academic appointment was controlled for as well (Schuster and
Finkelstein 2006). I include a measure of workplace training as this connotes support
and commitment (through social exchange), which are known to be important to turnover
(Ryan et al. 2012). Although some of these covariates varied with time, this study only
utilized measures at the time of academic appointment.

Results
This study utilized two methods of survival analysis. To calculate nonparametric
estimates of survival probabilities (for the first hypothesis), this study utilized the life table
method (actuarial estimator). This method is similar to the Kaplan-Meier estimator but

61
has computational advantages in large samples (Allison 1982)61. To calculate
regression coefficients and test statistics for the remaining hypotheses, this study utilized
a Cox proportional hazard regression.

Life table estimates are reported in the appendix (Table 11). Corresponding
survival probabilities are graphed and presented in Figure 2. For the first hypothesis, it
was reasoned that faculty with initial appointments off the tenure track would leave their
careers at a faster rate than those beginning on the tenure-track. This hypothesis was
confirmed using the score (log-rank) test (p-value<0.001). From Figure 2, we see that
the attrition curve is steepest during the first years of employment. For faculty working
off the tenure line but at an institution with a tenure system, only 72% of them remain in
an academic career after ten years. For faculty working at institutions without a tenure
system, a mere 63% remain in academia ten years later. Furthermore, it is not clear that
the attrition rate ever bottoms out during the course of the study. This suggests career
volatility may be a serious issue at all career stages for non-tenure track faculty. Faculty
with appointments to the tenure line tend to maintain academic careers much longer. In
fact, nearly 90% of tenure-track appointees continue to work in academia a decade later.
Additionally, unlike non-tenure track appointments, the attrition rate of initial tenure-track
appointments levels off after a decade, suggesting the emergence of a degree of
stability among these faculty members.

61 In any case, actuarial-based estimates and the Kaplan-Meier estimates were comparable.
62
 Figure 2: Survival Curves

While the life-table estimates of survival probabilities are important for descriptive
purposes, this method has limitations. Importantly, this approach does not allow for the
control of variables that may be confounding the relationship between tenure-line
appointment and time to attrition. For this reason, we turn to Cox proportional hazards
regression to examine key relationships. The Efron method was employed to handle
ties.

63
 Table 10: Cox Proportional Hazards Models
Model 1 Model 2 Model 3 Model 4
exp(coef) z exp(coef) z exp(coef) z exp(coef) z
Entry Date 0.964 -8.475 ** 0.966 -6.818 ** 0.966 -6.647 ** 0.966 -6.690 **
Tenure status (ref. Tenure-track)
Non-tenure Track (NTT) 2.671 10.360 ** 2.295 8.360 ** 2.280 8.230 ** 2.730 7.166 **
No Tenure System 3.844 17.135 ** 2.711 11.378 ** 2.683 11.078 ** 3.553 10.181 **
Time between Degree & Job 1.004 0.223 1.005 0.270 1.006 0.308
Principal Activity (ref. teacher)
Administrator/Other 1.675 5.051 ** 1.680 5.079 ** 1.647 4.863 **
Researcher 1.550 5.043 ** 1.556 5.077 ** 1.530 4.875 **
Private Control 1.069 1.096 1.072 1.136 0.959 -0.477
Institution Type (ref. Four-year)
Two-year 1.029 0.143 1.019 0.091 0.837 -0.746
Medical 1.187 2.345 * 1.186 2.327 * 1.121 1.175
Research Institute 1.199 2.098 * 1.202 2.126 * 1.080 0.724
Carnegie Status of Grad Inst (ref. RI)
PhD Research II 1.156 1.497 1.155 1.488 1.159 1.521
PhD Doctorate Institution 1.125 1.186 1.123 1.165 1.127 1.221
PhD Other 1.313 1.308 1.302 1.262 1.294 1.229
PhD Medical/Health 1.055 0.397 1.052 0.379 1.055 0.377
Age 0.996 -0.806 0.996 -0.820 0.996 -0.778
Female 0.949 -0.879 0.944 -0.970 0.943 -0.985
Minority 1.073 0.905 1.071 0.879 1.073 0.920
Married 0.917 -1.228 0.917 -1.235 0.913 -1.286
Number of Children 0.981 -0.270 0.980 -0.281 0.983 -0.237
Citizen 1.008 0.110 1.008 0.114 1.012 0.168
Workplace Training 1.036 0.599 1.038 0.636
Logged Salary 0.969 -0.869 0.968 -0.820
Time x NTT 0.831 -1.622
Time x No Tenure System 0.705 -3.553 **
Private x Two-Year 2.772 2.147 *
Private x Medical 1.161 1.078
Private x Research Institute 1.346 1.745

Concordance 0.656 0.669 0.669 0.672

Rsquare 0.047 (max possible 0.935) 0.05 (max possible=0.923) 0.05 (max possible=0.923) 0.053 (max possible 0.923)
* p-value<0.05, ** p-value<0.01

Hypothesis 2 in this study stated that, net of controls, initial appointments off the
tenure track would still result in higher risk of career attrition. Once again, I found
evidence that the risk of career attrition is higher for non-tenure track faculty. From
Model 4 (in Table 10), compared to initial tenure-line appointees, faculty beginning
careers off the tenure track at tenure-awarding institutions are 2.7 times (p-value <0.001)
more likely to leave their academic careers. The hazard of career attrition is even higher
for faculty accepting initial appointments at institutions without a tenure system.
Compared to traditional tenure line faculty, faculty beginning at institutions without tenure
systems have a 3.8 times (p-value <0.001) greater risk of academic career attrition.

For the third hypothesis, it was reasoned that initial appointments to research
positions would result in greater probability of career attrition. Indeed, the regression
model (Model 4) provides support for this hypothesis. Faculty who are first appointed to
primary research positions are more than 1.5 times (p-value <0.001) as likely to leave
their academic careers, compared to those whose principal activity is teaching. Faculty
appointed as administrators, as well, are more likely to leave academia (1.6 times more
likely). However, faculty with initial appointments to administrative roles are rare and
there are probably unique circumstances surrounding such an appointment.

Finally, this study had reason to suspect that traditional institutional

characteristics would be unrelated to the risk of career attrition. Robust standard errors
were used to handle the dependency among institutional-level predictors. Indeed, this
study found little evidence of institutional effects on career attrition (Model 4). This was
64
true for institutional control, type/level and Carnegie research status.

One interesting finding however relates to the interaction between institutional

control (public/private) and institutional type (2-year, 4-year, etc.). This study found that
faculty members working in private, 2-year institutions (interaction effect) had a
significantly higher risk of career attrition. One argument is that private, two-year
colleges may be turbulent places and they create an atmosphere that drives away their
faculty members. Another argument may be that these institutions plug in PhDs who are
not entirely serious about the profession (i.e., there may be a selection bias). However,
these findings may also be the consequence of data limitations. Few faculty members in
this study were primarily employed by private, two-year institutions (despite the fact that
these kinds of institutions are widespread and growing in the United States). In fact, the
test data contain only 20 such faculty members (<1% of the sample). More research is
needed to ascertain whether, indeed, private, two-year institutions may be truly linked to
higher career turnover.

It is surprising, however, that so few of the other covariates in the model were
linked to the timing of attrition. In addition to institutional-level characteristics,
demographic features like age, gender, minority status and number of children were
found to be statistically insignificant. On the other hand, earlier research has found
individual-level, workplace characteristics to be the main drivers of attrition—not
demographic features (Cotton and Tuttle 1980). So perhaps this should not be entirely
surprising. However, even the job characteristic of workplace training was insignificant
in this model, thereby raising concern. Given the data limitations mentioned earlier, the
model may be underspecified and this is probably responsible for the low R-square
(0.053).

Discussion
Non-tenure track faculty earn less, receive fewer benefits and generally
experience inferior working conditions compared to tenure-track peers. They report less
satisfaction with many aspects of their work and perceive less institutional support.
Given our theory on career change (Rhodes and Doering 1983), it is not surprising that
faculty beginning off the tenure track leave their careers at higher rates than tenure-line
faculty. However, the specific timing (attrition pattern) has never been examined in a
robust way. This work also showed that faculty with initial research appointments are
more likely to experience career attrition (than principal teaching appointments). Finally,
while structural factors often impact career attrition, this study was unable to produce
any evidence of an effect tied to traditional postsecondary institutional categories (i.e.,
public/private control, type/level, Carnegie research status).

These findings generally fit in with what has been established in this line of
research. Earlier work has shown that tenured faculty are less likely to leave their
organization or career than tenure-line junior faculty. This study extended our
understanding to faculty beginning off the tenure-track as well. More than tenure-track
appointments, faculty appointed off the tenure-line are even more likely to leave their
academic careers than tenure-line faculty. Presumably, the promise of tenure (tenure-
line) is a powerful motivator of career permanence in academia. The rate at which non-
tenure track faculty leave is striking. This study also presented survival curves showing
the rapid rate of early-career attrition and the fact that attrition never flattens among non-
65
tenure track laborers. This decline was particularly notable at institutions without tenure
systems. By their eighth year, a quarter of non-tenure system faculty members leave
their academic careers. Coupled with the slow, continuous decline of tenure systems at
most institutions62, it is clear that this issue is of growing importance.

The findings with regard to research status and institutional effects generally fit in
with earlier work as well. Career attrition is most closely tied to individual factors of the
immediate job context (i.e., person-job-fit, work experiences). Whether faculty serve
primarily as researchers or teachers probably matters then, to some extent, with regard
to career permanence. In earlier work, teaching was reported as one of the most
satisfying aspects of adjunct work (Waltman et al. 2012). Meanwhile, research activities
have been linked to stress and there is clearly a demand for research experience in the
private sector. Thus, the hypothesis of research-related attrition was largely borne out in
the results. Research appointees leave academic careers at much higher rates. As
career attrition is mostly an individual-level phenomenon, this study did not expect
institution-level effects (and it did not find any independent effects). However,
postsecondary features more closely linked to work experiences (e.g. academic
discipline) may be more closely related to career attrition. It is likely that these
organizational features would be related to organizational turnover. Professional
features (role orientation and professional commitment) tend to be more closely linked to
professional turnover (Blau 2000).

The findings of this study have important implications—and for many actors.
From the point of view of administrators, policy-makers and taxpayers, it should be a
goal to establish the “right amount63” of attrition in the postsecondary labor force.
Attrition should be high enough that “poor-fit” and unproductive workers are reallocated
to other areas of the economy but low enough that critical knowledge, structure, and
know-how are not lost in the process. This study cannot say whether or not attrition may
be too high or too low. But it does provide valuable information that can be used to help
answer that question. By year four, 25% of non-tenure track appointments will have left
academia (compared to only 7% of tenure-track appointees). That volatility implies costs
potentially outweighing the savings associated with faculty off the tenure line.
Administrators should reevaluate their hiring and retention strategies to be sure that their
activities are rational. Policy-makers should reconsider how postsecondary subsidies
are allocated. The more that subsidization is aligned with viable career training, the
more efficient public investments will be. Knowledge about academic career turnover
helps see that public dollars go to support the careers of young people who will put their
advanced training to work.

Finally, while there is an element of risk with any career pursuit, young people
should be as informed as possible before deciding to pursue an academic career. This
study found that, independent of tenure status, most Ph.D. recipients who begin work in
academia end up staying there for lengthy periods. This is particularly the case for
tenure-line appointees and for faculty with initial appointments as teachers. However, of

62 Figure 3 in the appendix shows that the decline of tenure systems is notable in all but public,

four-year colleges.
63 Some turnover is “functional”, removing surplus and unproductive labor (Dalton, Krackhardt,

and Porter 1981).

66
the institutional features tested in this study, none of them were associated with career
permanence.

Of the two analytic methods utilized in this study, the findings from the life table
model are more robust. This is because this methodology makes no assumptions about
the distribution of event times. The hazards model is far more sensitive to the issues
common to linear regression. Including mediators like job satisfaction and commitment
would have improved estimates, however those variables were not collected at every
wave and thus were excluded from the analysis. This problem—omitted variable bias—
however is not unique to this study and is ubiquitous to this line of research. Even the
most extensive models in this line of research (e.g., Meyer & Allen, 1991; Mobley,
Griffeth, Hand & Meglino, 1979) have “neglected or underestimated some important
antecedents” of turnover (Maertz and Griffeth 2004). So, the findings of this study must
be interpreted in consideration of this inevitable issue.

There are other important limitations worth mentioning as well. This study took
as its focus initial appointments and examined the career attrition that followed. Of those
who left an academic career, however, certainly some of them returned to academia at
some point in the future. Jacobs (1989) for example described the “revolving door” for
women in academia. Wolfinger, Mason, and Goulden (2009) described the academic
life course as “both complex and permeable.” Mobility between “careers” has received
very little research consideration and “boomerang faculty” would be a useful avenue for
future work (Lee and Mitchell 1994). Turnover destinations would be another fruitful
area of exploration (Hom et al. 2012, Hom et al. 2017). Where do faculty land after they
leave academic careers? Are the circumstances of non-tenure track faculty superior in
the private sector or in government? More work in this area is merited.

Conclusion
While not without limitations, this study clearly advances the field in important
ways. This line of work has typically assumed that turnover intentions are a reasonably
good proxy for actual attrition. This study utilized the actual behavioral measure of
attrition and confirmed that the processes are generally the same. That is, the causal
model to actual turnover helps extend this body of knowledge to the behavioral act of
quitting. This study also showed how initial employment off the tenure track, as well as
research appointments, are associated with a greater risk of career attrition.

References

Allen, David G., Kelly P. Weeks, and Karen R. Moffitt. 2005. “Turnover Intentions and
Voluntary Turnover: The Moderating Roles of Self-Monitoring, Locus of Control,
Proactive Personality, and Risk Aversion.” The Journal of Applied Psychology
90(5):980–90.

Allison, Paul. 1982. “Discrete-Time Methods for the Analysis of Event Histories.” 13.1.

67
Allport, F. H. 1962. “A Structuronomic Conception of Behavior: Individual and Collective.
I. Structural Theory and the Master Problem of Social Psychology.” Journal of
Abnormal and Social Psychology 64:3–30.

American Federation of Teachers. 2010. “American Academic: A National Survey of

Part-Time/Adjunct Faculty.” 2.

Baldwin, Roger G. and Jay L. Chronister. 2001. Teaching without Tenure: Policies and
Practices for a New Era. Baltimore, MD: Johns Hopkins University Press.
Retrieved (http://eric.ed.gov/?id=ED449758).

Barnes, Laura L. B., Menna O. Agago, and William T. Coombs. 1998. “Effects of Job-
Related Stress on Faculty Intention to Leave Academia.” Research in Higher
Education 39(4):457–69.

Becker, Howard S. 1960. “Notes on the Concept of Commitment.” American Journal of

Sociology 66(1):32–40.

Benton, Thomas H. 2003. “So You Want to Go to Grad School?” The Chronicle of
Higher Education, June 3.

Bland, Carole J., Bruce A. Center, Deborah A. Finstad, Kelly R. Risbey, and Justin
Staples. 2006. “The Impact of Appointment Type on the Productivity and
Commitment of Full-Time Faculty in Research and Doctoral Institutions.” The
Journal of Higher Education 77(1):89–123.

Blau, Gary. 2007. “Does a Corresponding Set of Variables for Explaining Voluntary
Organizational Turnover Transfer to Explaining Voluntary Occupational
Turnover?” Journal of Vocational Behavior 70(1):135–48.

Blau, Gary. 2000. “Job, Organizational, and Professional Context Antecedents as

Predictors of Intent for Interrole Work Transitions.” Journal of Vocational
Behavior 56(3):330–45.

Blau, Gary, Paul Allison, and Natalie St. John. 1993. “On Developing a General Index of
Work Commitment.” Journal of Vocational Behavior 42(3).

Blau, Gary and Mary Lunz. 1998. “Testing the Incremental Effect of Professional
Commitment on Intent to Leave One’s Profession beyond the Effects of External,
Personal, and Work-Related Variables.” Journal of Vocational Behavior
52(2):260–69.

Bluedorn, A. C. 1982. “The Theories of Turnover: Causes, Effects, and Meaning.” in

Research in the Sociology of Organizations, edited by S. B. Bacharach.
Greenwich, CT: JAI Press.

Carson, Kerry D., Paula Phillips Carson, and Arthur G. Bedeian. 1995. “Development
and Construct Validation of a Career Entrenchment Measure.” Journal of
Occupational and Organizational Psychology 68(4):301–20.
68
Cotton, John L. and Jeffrey M. Tuttle. 1986. “Employee Turnover: A Meta-Analysis and
Review with Implications for Research.” Academy of Management Review
11(1):55–70.

Cross, John and Edie Goldenberg. 2009. Off-Track Profs: Nontenured Teachers in
Higher Education.

Dalton, Dan, David M. Krackhardt, and Lyman W. Porter. 1981. “Functional Turnover: An
Empirical Assessment.” Journal of Applied Psychology 66:716–21.

Daly, Cheryl and Jay Dee. 2006. “Greener Pastures: Faculty Turnover Intent in Urban
Public Universities.” The Journal of Higher Education 77(5):776–803.

Eagan, Kevin and Audrey J. Jaeger. 2008. “Effects of Exposure to Part-Time Faculty on
Community College Transfer.” Research in Higher Education 50(2):168–88.

Gappa, Judith M., Ann E. Austin, and Andrea G. Trice. 2007. Rethinking Faculty Work:
Higher Education’s Strategic Imperative. San Francisco, CA: Jossey-Bass.

Gappa, Judith M. and David W. Leslie. 1993. The Invisible Faculty. Improving the Status
of Part-Timers in Higher Education. San Francisco, CA: Jossey-Bass Inc.

Gmelch, Walter H., Phyllis K. Wilke, and Nicholas P. Lovrich. 1986. “Dimensions of
Stress among University Faculty: Factor-Analytic Results from a National Study.”
Research in Higher Education 24(3):266–86.

Griffeth, Rodger W., Peter W. Hom, and Stefan Gaertner. 2000. “A Meta-Analysis of
Antecedents and Correlates of Employee Turnover: Update, Moderator Tests,
and Research Implications for the next Millennium.” Journal of Management
26(3):463–88.

Hagedorn, Linda Serra. 2000. “Conceptualizing Faculty Job Satisfaction: Components,

Theories, and Outcomes.” New Directions for Institutional Research
2000(105):5–20.

Hausknecht, John P., Charlie O. Trevor, and Michael J. Howard. 2009. “Unit-Level
Voluntary Turnover Rates and Customer Service Quality: Implications of Group
Cohesiveness, Newcomer Concentration, and Size.” The Journal of Applied
Psychology 94(4):1068–75.

Holtom, Brooks C., Terence R. Mitchell, Thomas W. Lee, and Marion B. Eberly. 2008.
“Turnover and Retention Research: A Glance at the Past, a Closer Review of the
Present, and a Venture into the Future.” Academy of Management Annals
2(1):231–74.

Hom, Peter W., Thomas W. Lee, Jason D. Shaw, and John P. Hausknecht. 2017. “One
Hundred Years of Employee Turnover Theory and Research.” The Journal of
Applied Psychology 102(3):530–45.

69
Hom, Peter W., Terence R. Mitchell, Thomas W. Lee, and Rodger W. Griffeth. 2012.
“Reviewing Employee Turnover: Focusing on Proximal Withdrawal States and an
Expanded Criterion.” Psychological Bulletin 138(5):831–58.

Hoyt, Jeff E. 2012. “Predicting the Satisfaction and Loyalty of Adjunct Faculty.” The
Journal of Continuing Higher Education 60(3):132–42.

Hulin, Charles, Mary Roznowski, and Donna Hachiya. 1985. “Alternative Opportunities
and Withdrawal Decisions. Empirical and Theoretical Discrepancies and an
Integration.” Psychological Bulletin 97:233–50.

Ingersoll, Richard M. 2002. “The Teacher Shortage: A Case of Wrong Diagnosis and
Wrong Prescription.” NASSP Bulletin 86(631):16–31.

Irving, P. Gregory, Daniel Coleman, and Christine L. Cooper. 1997. “Further

Assessments of a Three-Component Model of Occupational Commitment:
Generalizability and Differences across Occupations.” Journal of Applied
Psychology 82.

Jacobs, Jerry A. 1989. Revolving Doors: Sex Segregation and Women’s Careers.
Stanford University Press.

Johnsrud, Linda and Vicki Rosser. 2002. “Faculty Members’ Morale and Their Intention
to Leave: A Multilevel Explanation.” The Journal of Higher Education 73(4):518–
42.

Kaminski, Deborah and Cheryl Geisler. 2012. “Survival Analysis of Faculty Retention in
Science and Engineering by Gender.” Science 335(864).

Kezar, Adrianna and Cecile Sam. 2010. Understanding the New Majority of Non-Tenure-
Track Faculty in Higher Education: Demographics, Experiences, and Plans of
Action: ASHE Higher Education Report. Jossey-Bass, Inc.

Larson, Richard C., Navid Ghaffarzadegan, and Yi Xue. 2014. “Too Many PhD
Graduates or Too Few Academic Job Openings: The Basic Reproductive
Number R 0 in Academia.” Systems Research and Behavioral Science
31(6):745–50.

Lee, Kibeom, Julie Carswell, and Natalie Allen. 2000. “A Meta-Analytic Review of
Occupational Commitment: Relations with Person- and Work-Related Variables.”
Journal of Applied Psychology 85(5):799–811.

Lee, R. T. and B. E. Ashforth. 1996. “A Meta-Analytic Examination of the Correlates of

the Three Dimensions of Job Burnout.” The Journal of Applied Psychology
81(2):123–33.

Lee, Thomas W. and Terence R. Mitchell. 1994. “An Alternative Approach: The
Unfolding Model of Voluntary Employee Turnover.” The Academy of
Management Review 19(1):51–89.
70
Lee, Thomas William, Tyler C. Burch, and Terence R. Mitchell. 2014. “The Story of Why
We Stay: A Review of Job Embeddedness.” Annual Review of Organizational
Psychology and Organizational Behavior 1(1):199–216.

Louis, Meryl Reis. 1980. “Career Transitions: Varieties and Commonalities.” The
Academy of Management Review 5(3):329–40.

Maertz, Carl and Rodger Griffeth. 2004. “Eight Motivational Forces and Voluntary
Turnover: A Theoretical Synthesis with Implications for Research.” Journal of
Management 30(5):667–83.

March, J. G. and H. A. Simon. 1958. Organizations. Oxford, England: Wiley.

Massey, Douglas S. and Roger Tourangeau. 2013. "Where do We Go from Here?

Nonresponse and Social Measurement." The Annals of the American Academy
of Political and Social Science 645(1):222-236.

Matier, Michael W. 1990. “Retaining Faculty: A Tale of Two Campuses.” Research in

Higher Education 31(1):39–60.

McGee, Gail W. and Robert C. Ford. 1987. “Faculty Research Productivity and Intention
to Change Positions.” The Review of Higher Education 11(1):1–16.

Meyer, John, Natalie Allen, and Catherine A. Smith. 1993. “Commitment to

Organizations and Occupations: Extension and Test of a Three-Component
Conceptualization.” Journal of Applied Psychology 78:538–51.

Meyer, John P. and Natalie J. Allen. 1991. “A Three-Component Conceptualization of

Organizational Commitment.” Human Resource Management Review 1(1):61–
89.

Mitchell, Terence and Thomas Lee. 2001. “The Unfolding Model of Voluntary Turnover
and Job Embeddedness: Foundations for a Comprehensive Theory of
Attachment.” P. 189+ in Research in Organizational Behavior, vol. 23.

Mobley, W. H. 1977. “Intermediate Linkage in the Relationship between Job Satisfaction

and Employee Turnover.” Journal of Applied Psychology 62:237–40.

Mobley, W. H., Rodger W. Griffeth, Herbert H. Hand, and B. M. Meglino. 1979. “Review
and Conceptual Analysis of the Employee Turnover Process.” Psychological
Bulletin 86(3).

Mobley, W. H., S. O. Horner, and A. T. Hollingsworth. 1978. “An Evaluation of

Precursors of Hospital Employee Turnover.” The Journal of Applied Psychology
63(4):408–14.

Moore, Jo Ellen. 2000. “Why Is This Happening? A Causal Attribution Approach to Work
Exhaustion Consequences.” The Academy of Management Review 25(2):335–
49.

71
Morita, June, Thomas Lee, and Richard Mowday. 1989. “Introducing Survival Analysis
To Organizational Researchers: A Selected Application To Turnover Research.”
Journal of Applied Psychology 74:280–92.

Morrow, Paula. 1983. “Concept Redundancy in Organizational Research: The Case of

Work Commitment.” Academy of Management Review 8.

Morrow, Paula C. 1993. The Theory and Measurement of Work Commitment.

Greenwich, Conn: JAI Press.

National Center for Education Statistics. 2015. Characteristics of Postsecondary Faculty.

U.S. Department of Education. Institute of Education Sciences.

National Center for Education Statistics. 2014. Table 1: Number and Percentage
Distribution of Public School Teacher Stayers, Movers, and Leavers: 1988-89
through 2012-13. U.S. Department of Education. Institute of Education Sciences.

Neapolitan, Jerome. 1980. “Occupational Change in Mid-Career: An Exploratory

Investigation.” Journal of Vocational Behavior 16(2):212–25.

Porter, Lyman W., William J. Crampon, and Frank J. Smith. 1976. “Organizational
Commitment and Managerial Turnover: A Longitudinal Study.” Organizational
Behavior and Human Performance 15(1):87–98.

Price, James. 1977. The Study of Turnover. Ames, Iowa: Iowa State University Press.

Price, James L. and Charles W. Mueller. 1981. “A Causal Model of Turnover for Nurses.”
Academy of Management Journal 24(3):543–65.

Rhodes, Susan R. and Mildred Doering. 1983. “An Integrated Model of Career Change.”
Academy of Management Review 8(4):631–39.

Rosser, Vicki J. 2004. “Faculty Members’ Intentions to Leave: A National Study on Their
Worklife and Satisfaction.” Research in Higher Education 45(3):285–309.

Rosser, Vicki J. 2005. “Measuring the Change in Faculty Perceptions Over Time: An
Examination of Their Worklife and Satisfaction.” Research in Higher Education
46(1).

Ryan, John, Richard Healy, and Jason Sullivan. 2012. “Oh, Won’t You Stay? Predictors
of Faculty Intent to Leave a Public Research University.” Higher Education
63:421–437.

Schuster, Jack H. and Martin J. Finkelstein. 2006. The American Faculty: The
Restructuring of Academic Work and Careers. Baltimore, MD: The Johns
Hopkins University Press.

Smart, John C. 1990. “A Causal Model of Faculty Turnover Intentions.” Research in

Higher Education 31(5):405–24.

72
Steel, Robert. 2002. “Turnover Theory at the Empirical Interface: Problems of Fit and
Function.” Academy of Management Review 27:346–60.

Steers, Richard and Richard Mowday. 1981. “Employee Turnover and Post-Decision
Accommodation Processes.” Research in Organizational Behavior 3:235+.

Thorsen, Elizabeth J. 1996. “Stress in Academe: What Bothers Professors?” Higher

Education 31(4):471–89.

Vroom, Victor. 1964. Work and Motivation. New York, New York: John Wiley & Sons.

Waltman, Jean, Inger Bergom, Carol Hollenshead, Jeanne Miller, and Louise August.
2012. “Factors Contributing to Job Satisfaction and Dissatisfaction among Non-
Tenure-Track Faculty.” Journal of Higher Education 83(3).

Weller, Ingo, Brooks C. Holtom, Wenzel Matiaske, and Thomas Mellewigt. 2009. “Level
and Time Effects of Recruitment Sources on Early Voluntary Turnover.” The
Journal of Applied Psychology 94(5):1146–62.

Wolfinger, Nicholas H., Mary Ann Mason, and Marc Goulden. 2009. “Stay in the Game:
Gender, Family Formation and Alternative Trajectories in the Academic Life
Course.” Social Forces 87(3):1591–1621.

Xu, Yonghong Jade. 2008. “Faculty Turnover: Discipline-Specific Attention Is

Warranted.” Research in Higher Education 49(1):40–61.

Zhou, Ying and James Volkwein. 2004. “Examining the Influences on Faculty Departure
Intentions: A Comparison of Tenured versus Nontenured Faculty at Research
Universities Using NSOPF-99.” AIR Forum Issue 45(2):139–76.

73
Appendix

Figure 3: Changes in Tenure Systems (1993-2015)

Table 11: Life Table

Tenure/Track Non-Ten/Track No Tenure system

Wave Surv. Prob. Std. Err Surv. Prob. Std. Err Surv. Prob. Std. Err
1 1.000 0.000 1.000 0.000 1.000 0.000
2 0.969 0.002 0.878 0.006 0.830 0.004
3 0.935 0.004 0.807 0.008 0.731 0.005
4 0.908 0.005 0.752 0.010 0.675 0.006
5 0.889 0.006 0.725 0.011 0.641 0.007
6 0.878 0.006 0.706 0.011 0.617 0.008
7 0.868 0.007 0.689 0.012 0.599 0.008
8 0.862 0.007 0.677 0.013 0.590 0.009
9 0.857 0.008 0.665 0.014 0.582 0.009
10 0.857 0.008 0.660 0.015 0.574 0.010
11 0.857 0.008 0.650 0.017 0.565 0.012

Table 12: Study Entrants by Year

Year
1993 1995 1997 1999 2001 2003 2006 2008 2010 2013
Number of
2428 1967 3340 1850 1936 1764 2309 1826 2099 1917
Study Entrants

74
 CHAPTER 5: CONCLUSION

While the numbers of tenure-line and fixed term faculty increase every year, the
pace of hiring among the fixed term segment is much higher. As a result, the proportion
of faculty working under contingent contracts has been steadily climbing for about half a
century now. The best evidence shows that more than two out of three postsecondary
faculty members work off the tenure line today (National Center for Education Statistics
2015, Schuster and Finkelstein 2006). Non-tenure track faculty are no longer the
ancillary supports or short-term fixes they once were: they are now the professoriate.

Researchers have investigated many themes related to this important change;

however, the scope of this dissertation focused on two important topics: classification
and the job satisfaction and turnover process of postsecondary adjuncts. In Chapter 2,
this study utilized a cluster analysis to extract natural structure in a dataset containing
extensive information on non-tenure track faculty. Three distinct clusters of full-time
adjuncts emerged from the analysis. Administrative adjuncts—highly integrated in their
departments—often worked principally as departmental administrators. Core teaching
faculty were also highly integrated, often holding higher rank, earning more and
receiving greater opportunities for professional development. Peripheral teaching faculty
worked full-time but tended to lack traditional status and ties in their departments.

This study also identified five types of part-time adjunct faculty. Similar to the
full-timers, Adjunct Administrators commonly perform administrative tasks. Adjunct
Experts often hold full-time careers outside of academia and probably teach in order to
share their expertise with young people. There were also three types of aspiring
academics. With advanced credentials and a greater focus on academic work,
Academic Aspirants often resembled traditional faculty. Journeyman Aspirants
commonly worked outside of academia, but nevertheless hoped to acquire a full-time
position in academia. Freelancing Aspirants were similar to the Journeymen but
typically work less in formal economic channels.

The other important topic of this dissertation related to the job satisfaction and
turnover process of postsecondary adjuncts. In Chapter 3, cross-sectional data was
used to investigate the job satisfaction and turnover intentions of different subclasses of
adjunct faculty. Importantly, by modeling relationships using a structural equation
model, this study was able to measure job satisfaction more comprehensively. The
analysis showed expert and freelancer part-time adjuncts to be quite similar with regard
to job satisfaction and turnover intentions. However, career-enders were somewhat
unique with regard to the financial and benefits satisfaction they experience. This study
also showed that the main way satisfaction impacts the turnover intentions of these
faculty members is through financial satisfaction and “work and supports.” Satisfaction
with co-workers and benefits were unrelated to the turnover intentions.

Chapter 4 continued this examination of turnover but extended it to the question

of career attrition (instead of institutional turnover). Using a longitudinal dataset and a
behavioral measure of career attrition, this chapter provided strong empirical evidence
that beginning off the tenure track was linked to higher career attrition. This was true
even after controlling for background characteristics. The study also showed that faculty
with initial research and administrative appointments were more likely to leave
75
academia. Traditional ways of describing postsecondary institutions (i.e., public/private
control, type/level, Carnegie research status), however, were not shown to have
independent effects on faculty career attrition.

Contribution
This dissertation made several important contributions. First, while there existed
a line of research on adjunct typologies, there was great redundancy among adjunct
classes and an unjustifiable omission of potentially useful information in the creation of
adjunct types. This study was the first to leverage information from a high-dimensional
dataset and classify utilizing a computational method. The resultant typology that
emerged was both innovative and confirmatory. The foundational insights of Gappa and
Leslie (1993) and Baldwin and Chronister (2001) were both evident. What the analysis
accomplished, however, was to suggest important nuances that had been overlooked in
earlier classification systems. It produced evidence for a distinction to be drawn
between core and peripheral teaching faculty. It also suggested a nuanced
conceptualization of aspiring academics—one that factored in the adjunct’s level of
engagement in the private sector.

There were also valuable contributions made in relation to job satisfaction and
turnover. As noted earlier, traditional tenure-line faculty have been the focus of earlier
research. When non-tenure line faculty were included in these analyses, they were
typically an afterthought and never disaggregated to the level of this study. The results
of Chapter 3 showed that models could be improved by factoring in the unique patterns
of career-enders. The structural equation model also permitted a nuanced investigation
of how job satisfaction fits into the turnover process. Importantly, this study showed
turnover intentions to operate principally through financial satisfaction channels and
though “work and supports’’—not through benefits or co-worker satisfaction as some
would expect.

Finally, only rarely has the timing of academic turnover been a topic of inquiry in
faculty research. By extracting information in panel data, this study was able to produce
robust, nonparametric estimates of time to career attrition for both tenure-line and fixed-
term faculty. Perhaps it is not surprising that adjuncts leave their academic careers at
higher rates. However, the timing of this attrition was important to document and is an
important point of future deliberation. It is also important to note that this study tested
these hypotheses using actual turnover data—not “turnover intentions” as used in most
faculty studies.

Practical implications
As argued earlier, a better typology means better science and better policy.
Scientists, administrators, policy-makers and others rely on having honed conceptual
tools so that their ideas are well-defined and clearly articulated. The most effective
policy is one that correctly understands a problem at hand and can target, with great
precision, the subjects of the intervention. By honing our understanding of adjunct
types, this study has incrementally improved the conceptual tools necessary to do good
academic and professional work.

The findings with regard to job satisfaction and turnover also have important
practical implications. Most agree that employees have the right to agreeable work
conditions whenever possible. However, it is important to stress the fact that creating
76
agreeable work conditions is also in the interest of most employers as well—certainly
postsecondary institutions. Consistent faculty turnover is potentially a very large
institutional cost as new teachers need to be hired, oriented, trained and managed. It
would be wise, then, for administrators to pursue personnel strategies that minimize
these costs. Chapter 4 demonstrated that, by the end of year four, 25% of faculty
members with non-tenure track appointments leave their careers in academia
(compared to only 7% of tenure-track appointees). This is clearly the kind of volatility
that needs to be understood and dealt with by postsecondary administrators and human
resource managers.

To curb or limit turnover, institutions should design policies that target faculty
financial satisfaction and satisfaction with “work and supports.” There may be many
reasons to pursue strategies impacting benefits or co-worker relationships, but not for
the purposes of retention. Institutions may have their hands tied with regard to financial
satisfaction, but there are probably creative ways to reallocate budgets to incentivize
faculty commitment. In Chapter 3, I proposed channeling resources away from benefits
(which are not linked to turnover) and towards financial support (which has the greatest
impact on turnover). There may also be more cost-effective ways to improve adjunct
faculty satisfaction with “work and supports.” Merely providing them with office space or
access to office materials may be welcomed and appreciated by these faculty members.
In any case, career-enders have unique circumstances and administrators should
probably view them as short-term fixes—not long-term faculty contributors.

Future Work
The work of this dissertation has set the stage for a range of projects to further
elucidate the work lives of non-tenure track faculty. The typology devised in this study
can be fully employed across a range of important adjunct outcomes. How do different
kinds of aspiring academics affect student learning? With extensive efforts outside of
academia, are Journeyman Aspirants able to dedicate themselves to their classes as
well as Academic Aspirants? Do full-time peripheral teaching faculty attrite from their
positions at higher rates than core teaching faculty? These are empirical questions with
promise of extending our knowledge to new domains.

Future work should also consider more complicated forms of institutional turnover
and career attrition. This study simply examined institutional departure (Chapter 3) and
career attrition (Chapter 4). However, mobility back and forth (“boomerang faculty”) is
an important topic that has received very little research consideration. Research should
also begin considering the destinations of adjuncts leaving their positions. How often do
they leave for other institutions? For faculty leaving academic careers, do they land in
more desirable circumstances? It is certainly a great possibility that their circumstances
improve, given how highly skilled these professionals are and how valued their skill-sets
are in the broader economy. Understanding where they land is an important component
to fully evaluating the costs and benefits of our graduate programs and our system of
contingent academic labor.

77