Content uploaded by David R. Hekman
Author content
All content in this area was uploaded by David R. Hekman on Nov 10, 2014
Content may be subject to copyright.
Content uploaded by Brooks C. Holtom
Author content
All content in this area was uploaded by Brooks C. Holtom on Oct 28, 2014
Content may be subject to copyright.
TURNOVER CONTAGION: HOW COWORKERS’ JOB
EMBEDDEDNESS AND JOB SEARCH BEHAVIORS
INFLUENCE QUITTING
WILL FELPS
Erasmus University
TERENCE R. MITCHELL
University of Washington, Seattle
DAVID R. HEKMAN
University of Wisconsin–Milwaukee
THOMAS W. LEE
University of Washington, Seattle
BROOKS C. HOLTOM
Georgetown University
WENDY S. HARMAN
Central Washington University
This research developed and tested a model of turnover contagion in which the job
embeddedness and job search behaviors of coworkers influence employees’ decisions
to quit. In a sample of 45 branches of a regional bank and 1,038 departments of a
national hospitality firm, multilevel analysis revealed that coworkers’ job embedded-
ness and job search behaviors explain variance in individual “voluntary turnover”
over and above that explained by other individual and group-level predictors. Broadly
speaking, these results suggest that coworkers’ job embeddedness and job search
behaviors play critical roles in explaining why people quit their jobs. Implications are
discussed.
As the global economy becomes increasingly
knowledge based, organizations that can success-
fully retain their human resources have an advan-
tage over organizations that cannot. Indeed, a num-
ber of studies have shown that turnover negatively
effects performance (e.g., Shaw, Gupta, & Delery,
2005). Hatch and Dyer summarized such findings
with the observation that “firms with high turnover
significantly under-perform their rivals” (2004:
1155). As such, organizational leaders are inter-
ested in understanding why people choose to leave
their jobs and insights that might help with em-
ployee retention (Ulrich & Smallwood, 2006). Ac-
cordingly, researchers have spent considerable ef-
fort developing and testing models to explain why
people quit.
To explain the phenomenon of employee turn-
over, the social sciences have offered both psycho-
logical (i.e., micro) and organizational and eco-
nomic (i.e., macro) explanations. On the micro
side, job satisfaction and organizational commit-
ment have captured most of the research interest.
On the macro side, economic research often looks
at particular industries or localities to explain how
market forces such as unemployment rates or job
supply and demand affect the frequency with
which people leave their jobs (e.g., Banerjee & Gas-
ton, 2004). Sociological research has also looked at
how turnover affects and is affected by institutional
changes within and across industries (e.g., Have-
man, 1995), as well as organizational variables such
as size (Price, 1977).
The unique contribution of management scholar-
ship is not only to investigate the individual or
institutional level, but also what emanates from the
careful exploration of “the space in between”
(Bradbury & Lichtenstein, 2000). For this reason,
organizational researchers are often encouraged to
do “meso-level” research, in which individuals are
studied in their social contexts (e.g., House, Rous-
seau, & Thomas-Hunt, 1995; Johns, 2006). How-
ever, there is surprisingly little work on how social
relationships affect turnover. To quote Pfeffer,
“Turnover has most often been examined as the
娀 Academy of Management Journal
2009, Vol. 52, No. 3, 545–561.
545
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download or email articles for individual use only.
consequence of an individual decision process,
with the individual acting in isolation.... Virtu-
ally all of the dominant models of turnover concep-
tualize it as an individual decision, without con-
sidering the effect of social structure” (1991: 795).
Although Pfeffer’s comment overlooks the work of
economists and sociologists, he is broadly correct
in stating that the bulk of management research on
turnover focuses on individual attitudes as the sole
precursor to leaving. The influence of one’s imme-
diate coworkers on turnover decisions (what Pfeffer
describes as social structure) has been largely
ignored.
This article investigates the social dimensions of
quitting and offers a model of “turnover contagion”
in which the decision to stay at or leave a job is
influenced by coworkers. We provide evidence that
turnover decisions are a domain in which cowork-
ers can influence an actor’s thoughts, judgments,
feelings, and behaviors (Salancik & Pfeffer, 1978).
Two field studies support the predictive validity of
our model, offering new insights into the interper-
sonal precursors of “voluntary turnover” (job leav-
ing). We argue that this type of meso-level research
can widen researchers’ conceptual lenses, increase
our ability to predict turnover, and enhance the
utility of turnover research for practitioners.
TOWARD A THEORY OF
TURNOVER CONTAGION
Turnover Research Heritage
March and Simon’s (1958) seminal book, Orga-
nizations, marks the real beginning of the attempt
to develop an overall theory explaining why people
leave their jobs. According to March and Simon’s
theory, the two factors that determine whether an
employee will leave his or her job are the perceived
desirability of leaving the employing organization
(conceptualized as job satisfaction and organiza-
tional commitment) and the perceived ease of leav-
ing the organization (conceptualized as the quality
of job alternatives). The research focusing on job
satisfaction and organizational commitment, in
particular, has been extensive. Mobley (1977) iden-
tified the sequential and intermediary variables
leading from job dissatisfaction to eventual quit-
ting. In an exemplar of programmatic turnover re-
search, Price and Mueller (1986) added to this
model by cataloging the antecedents of organiza-
tional commitment and job satisfaction, including
pay, social integration, instrumental communica-
tion, formal communication, centralization, routi-
nization, role overload, promotional opportunity,
professionalism, general training, supervisor sup-
port, coworker support, and distributive justice
(Price, 1977; Price & Mueller, 1986). It is important
to note that in Price and Mueller’s model, as in
virtually all other traditional models, various fac-
tors influence turnover through their impact on
organizational commitment and job satisfaction,
which in turn influence intent to leave, which then
leads to voluntary turnover.
The result of subsequent scholarship based on
these ideas is both impressive and troublesome. It
is impressive in that turnover theory and research
have proceeded programmatically in such a way
that researchers can be confident about a pair of
assertions. First, less satisfied and less committed
employees think about leaving, look for alternative
jobs, are more likely to quit, and do each of these to
a greater degree when they believe that desirable
job alternatives exist. Second, many individual-
and macro-level variables are related to turnover
through satisfaction and commitment. However,
the turnover literature is also troublesome in that
even the most inclusive models leave the vast ma-
jority of variance unexplained (e.g., Griffeth, Hom,
& Gaertner, 2000; Maertz & Campion, 1998; Price &
Mueller, 1986). A number of authors have therefore
suggested that scholars need to expand their con-
ceptual lenses to better understand employee turn-
over (e.g., Kammeyer-Mueller, Wanberg, Glomb, &
Ahlburg, 2005; Maertz & Campion, 1998; Mitchell
& Lee, 2001; Mossholder, Settoon, & Henagan,
2005). The framework we describe below, in which
we outline the turnover contagion process, is such
an expansionary attempt.
The Turnover Contagion Process
The central theoretical claim made here is that
when an employee’s coworker engages in behaviors
antecedent to leaving a job, these activities some-
times spill over onto others in such a way that the
affected others are more likely to leave. Put more
precisely, a coworker’s search for job alternatives or
actual quitting can spread, through a process of
social contagion, to affect another employee’s quit-
ting behavior. Like the contagion of illness, the
process involves the transmission of something
from one individual to another. For us, the “some-
thing” is the tendency to leave a job. Others have
used the contagion metaphor to understand the
spread of burnout (Bakker & Schaufeli, 2000), emo-
tions (Barsade, 2002), and long work hours (Brett &
Stroh, 2003).
We believe that the primary mechanism in turn-
over contagion is people’s pervasive tendency to
compare themselves to others. Research on social
comparison has documented that this tendency is
546 JuneAcademy of Management Journal
among the most robust and ubiquitous of psycho-
logical phenomena (Kruglanski & Mayseless, 1990).
“The notion that people rely on others to help
define reality in ambiguous circumstances has long
been a core tenet in social psychology” (Degoey,
2000: 58). Salancik and Pfeffer (1978) extended
Festinger’s (1954) original work on social compar-
ison to organizational behavior and job attitudes,
and Bandura (1977) applied these insights to learn-
ing theory. Social comparisons are especially likely
to be made in novel, risky, or ambiguous situations
(Festinger, 1954; Tesser, Campbell, & Mickler,
1983; Wooten & Reed, 1998). When comparisons
reveal differences with a relevant other’s thoughts,
feelings, or behaviors, an individual’s propensity to
change his or her understanding of a situation so
that thoughts, feelings, and behaviors become con-
sistent with those of the relevant other increases
(Festinger, 1954). Chartrand and Bargh stated this:
“Throughout the history of psychology, many have
argued that the act of perceiving another person’s
behavior creates a tendency to behave similarly
oneself” (1999: 813).
The application to turnover theory is straightfor-
ward. Given that high levels of risk and uncertainty
often characterize job transition (Steel, 2002), we
expect employees to look to others in evaluating
whether to seek alternative employment. When a
number of coworkers are looking for other jobs, it
may increase the salience and perceived viability of
leaving for a focal employee, especially since im-
mediate coworkers are likely targets for social com-
parison (Kulik & Ambrose, 1992). Conversely, if
few coworkers are looking for other jobs, it is likely
that a focal employee will be less inclined to initi-
ate the turnover process. In either case, social com-
parison helps to answer the question, “Should I
consider leaving?” We posit that the chance that
the answer will be yes increases when many co-
workers are looking for jobs. In this way, the trans-
mission of a tendency to leave occurs as employees
watch and converse with their coworkers. The focal
person may observe such job search behaviors in a
dyadic interaction (e.g., “I am going on a job inter-
view this week”) or in a group interaction (e.g.,
“You all should probably know that I have a job
interview this week”). Moreover, there are a variety
of leaving behaviors that can be observed; the em-
ployee might see a coworker update a re´sume´,
search classified ads, or schedule interviews. In
short, a range of behaviors may indicate that one or
multiple coworkers are in the process of leaving.
Some research has addressed the topic of with-
drawal caused by group-level variables. Mathieu
and Kohler (1990), for example, found that the fre-
quency of absenteeism among work group members
was related to individual employee absenteeism.
And Eder and Eisenberger (2008) demonstrated
that the average tardiness of work group members
is related to individual tardiness. They also
showed, in a second study, that withdrawal behav-
iors carried out at the group level, such as taking
undeserved work breaks or engaging in idle conver-
sation, influence the probability that individuals
do the same. Thus, the idea that withdrawal behav-
iors of group members can influence an individu-
al’s likelihood of engaging in those behaviors
clearly has some precedent. Importantly, we do not
presume here that either job satisfaction or organi-
zational commitment plays a key role in the pro-
cess. The turnover contagion model highlights the
role that simply observing others plays and sug-
gests that a key determinant of whether quitting is
a viable option at any given point in time is
whether coworkers are leaving.
From Theory to Hypotheses
Above we have presented a theory of turnover
contagion whereby the tendency to quit spreads
throughout a work group. We now offer two spe-
cific hypotheses about factors that are central to the
turnover contagion process. First, we hypothesize
that turnover contagion is most likely to occur
when the coworkers around a focal employee are
not “embedded” in their jobs. We choose to focus
on job embeddedness, as opposed to job satisfac-
tion or organizational commitment, because it is a
broader construct that captures a greater range of
factors that provoke leaving. In Mitchell, Holtom,
Lee, Sablynski, and Erez’s (2001) original formula-
tion, the job embeddedness construct addressed
how well people fit in their jobs (e.g., personal
skills are well suited to the work assigned) and
community (e.g., they like the amenities a commu-
nity provides); the interpersonal links they have on
and off the job (e.g., their number of ties to people
and groups); and what they would have to give up
or sacrifice in leaving their place of employment or
community (e.g., what opportunities they would
forego). In sum, job embeddedness includes several
individual-level factors that enmesh employees in
their jobs, and numerous studies have shown it to
be a good predictor of an employee’s tendency to
quit (Allen, 2006; Crossley, Bennett, Jex, & Burn-
field, 2007; Holtom, Mitchell, & Lee, 2006; Holtom
& O’Neill, 2004; Lee, Mitchell, Sablynski, Burton, &
Holtom, 2004; Mitchell et al., 2001; Van Dijk &
Kirk-Brown, 2003; Zatzick & Iverson 2006). In
many of these studies, job embeddedness has gone
beyond job satisfaction and organizational commit-
ment in predicting variance in individual turnover.
2009 547Felps, Mitchell, Hekman, Lee, Holtom, and Harman
When coworkers’ job embeddedness is low, we
believe that the resultant social context will make
individuals more likely to entertain the possibility
of changing jobs. When coworkers are not tethered
to—embedded in—a job, they are likely to be open
to the possibility of leaving. It is this willingness to
leave that transfers from low-embeddedness co-
workers to a focal employee in their work unit.
Thus, we expect the average job embeddedness of
coworkers to predict focal employee turnover. Fur-
ther, since job embeddedness is a broad construct
that includes nonaffective elements such as the
number of links to important others and family ties,
we would expect this effect to be observed even
when a focal employee is satisfied with the work
itself or committed to his/her organization.
Let us briefly take some examples from the
Mitchell et al. (2001) job embeddedness measure to
provide a more grounded understanding of how
turnover contagion might operate. Imagine a work-
place where most people strongly agree with the
following statements: “I feel like I am a good match
for my organization,” “I really love the place I live,”
“I would sacrifice a lot if I left this job,” “My family
roots are in this community,” and “I work closely
with my coworkers.” Interactions among employ-
ees who feel this way are likely to mutually rein-
force each other’s perceptions that “I belong here, I
should be here, and I must remain here.” In such a
setting, people are unlikely to be looking at want
ads, talking about available jobs elsewhere, or say-
ing things that indicate they want to leave. Contrast
this situation with a workplace populated by those
who are less embedded in their jobs and commu-
nities (e.g., people who feel they don’t fit in their
work group or community, or people who have
little to sacrifice in renegotiating their relationships
to their jobs). In this sort of environment, even if
they like their jobs, employees have little to lose by
voicing ideas about leaving or about alternative
avenues of employment (Bartunek, Huang, &
Walsh, 2008). Frequent discussions about leaving
are likely to prime other employees, possibly even
those who are fairly embedded, to consider quit-
ting. Thus, we hypothesize the following:
Hypothesis 1. Coworkers’ job embeddedness is
negatively related to voluntary turnover.
The next question naturally follows: How does a
willingness to quit engendered by low job embed-
dedness influence others? As noted previously, we
hypothesize that the transmission of this leaving
tendency occurs as employees watch and converse
with coworkers searching for alternative employ-
ment. In Study 1 (see below), we gathered data
through a series of focus groups designed to help us
better understand the turnover process. A qualita-
tive analysis of the behaviors discussed by the fo-
cus group members provides some information
about how employees may be influenced by their
coworkers’ comments about leaving. In Study 2, we
sought to specifically measure coworkers’ searches
for alternative employment using the Job Search
Behavior Index (Kopelman, Rovenpor, & Millsap,
1992). These authors reported that this measure
(aggregated to the group level) did an excellent job
of predicting leaving and internal transfer and went
over and above eight affective, perceptional, attitu-
dinal, and intention measures (e.g., organizational
commitment, intent to stay, and general job satis-
faction) in such prediction. For our purposes, the
argument is simple. When an employee sees and
hears about coworkers looking for other jobs, leav-
ing becomes a more salient option for her/him,
which leads to a greater propensity to quit. Figure 1
presents a summary of these ideas. It should be
noted that although both of the studies described
below measured coworkers’ job embeddedness,
only the second study assessed coworker job search
behavior using the Kopelman et al. (1992) measure.
Thus, Hypothesis 2 is only empirically tested in
Study 2.
Hypothesis 2. Coworkers’ job search behaviors
mediate the negative relationship between co-
workers’ job embeddedness and focal em-
ployee turnover.
From Hypotheses to Analytics
Although undisputedly important, pursuing
meso-level research can be challenging (House et
al., 1995). When one defines meso-level research as
research that includes activities and processes that
take place between the micro and macro, this chal-
lenge comes into stark relief. “Micro” and “macro”
are defined relative to each other, and there are a
number of potentially relevant “levels” for both
predictors and criteria, including individuals, dy-
ads, small groups, organizations, industries, and
societies. The number of possible combinations is
extensive; comments by Klein and Kozlowski
(2000) were helpful for our definitional analysis.
We are particularly interested in how the behaviors
that occur in dyads or existing groups (our inde-
pendent variable) influence individual members to
quit (our dependent variable). Rousseau (1985) de-
scribed such an influence process as a cross-level
phenomenon. In our case, the phenomenon of in-
terest is turnover contagion. More specifically, each
person’s job embeddedness reflects an overall
“stuckness” in the job (the inverse of which is a
548 JuneAcademy of Management Journal
willingness to leave), which can be contagiously
transferred through modeling or direct interaction
with coworkers. In actual work situations, individ-
uals may work at various times with just one other
individual (i.e., dyadically) or, as is increasingly
common, in project teams, departments, and inde-
pendent branches (i.e., in a small group). In a small
group, any given individual is likely to send turn-
over contagion stimuli to a number of others as well
as receive this sort of leaving stimuli from a number
of others.
If a researcher wants to capture the cumulative
influence of coworkers on a focal actor’s turnover
decision, it is necessary to somehow combine the
contagious effects of multiple group members
(Klein & Kozlowski, 2000). The simplest way to do
this is to aggregate the scores of coworkers who are
known to work closely with an individual on those
variables that have been theorized to be associated
with the turnover contagion process—in short, job
embeddedness and job search behaviors. Thus, we
propose to test whether the average of coworkers’
job embeddedness scores for a natural group influ-
ences individual employee turnover and, if so,
whether the coworkers’ average level of job search
behaviors mediates this relationship.
It is important to point out two statistical issues
related to this conceptualization. First, given that a
focal individual is nested in a group, it is important
to control for a focal individual’s own level of job
embeddedness or job search behavior in any mul-
tilevel statistical test (Klein & Kozlowski, 2000). All
subsequent analyses do this. Second, this measure-
ment process does not depend on employees com-
ing to some sort of socially agreed upon consensus
about job embeddedness or about job search behav-
iors. For example, one popular type of meso-level
research links group-level consensus about some-
thing (e.g., norms, mood, etc.) to individual behav-
ior. Chan described these as “direct consensus
models” (1998). Our theoretical model is not one of
direct consensus. As such, the methodological
standards used to verify direct consensus effects
(i.e., high degrees of agreement as assessed by in-
traclass correlations, or R
wg
statistics) would be
meaningless for our analysis. Instead, coworkers’
job embeddedness represents what Chan (1998)
called an “additive index model” that does not
hinge upon agreement, but is instead about
whether relevant social comparisons prompt look-
ing for a different job.
STUDY 1
Methods
Sample. Our first research site was a large recre-
ation and hospitality organization, hereafter re-
ferred to as Funcorp. This organization operates
roughly 200 golf courses, country clubs, private
FIGURE 1
The Turnover Contagion Model
Individual
Voluntary
Turnover
Coworkers’ Job Embeddedness
(average job embeddedness
score of employees in
department or branch)
Coworkers’ Job Search
Behavior (qualitative and
quantitative)
Individual Job
Embeddedness
Level 2 Variables
Level 1 Variables
Sacrifice–Community
Sacrifice–Organization
Fit–Community
Fit–Organization
Links–Community
Links–Organization
2009 549Felps, Mitchell, Hekman, Lee, Holtom, and Harman
business and sports clubs, and resorts. Funcorp
provides services to about 200,000 member fami-
lies throughout the United States. Our initial sam-
ple consisted of 14,981 Funcorp employees who
serve its members. Nine thousand seventy-nine em-
ployees completed our survey, for a response rate
of 60.6 percent. Missing values reduced the number
of usable observations to 8,663, or 57.8 percent of
the initial sample. Within our usable sample were
1,037 club departments. Overall, 39.3 percent of
the respondents were women; the average age was
39.0 years; the average tenure with the organization
was 6.2 years; and 32.6 percent were nonwhites.
The average department size was 14.4, and the
average number of survey respondents per depart-
ment was 8.35. The firm provided demographic
data for all employees, allowing us to statistically
compare respondents and nonrespondents. These
comparisons yielded no significant differences in
gender, age, tenure, race, or turnover rate, provid-
ing some confidence that nonresponse bias was not
a concern.
Measures. Voluntary turnover was measured in
Study 1 as whether an employee voluntarily left the
organization in the 18 months immediately follow-
ing the survey. An 18-month period was reasonable
because it allowed enough time for the indepen-
dent variables to influence employees’ turnover de-
cisions and provided us with a large enough sam-
ple to reliably run statistical tests. Specifically,
2,001 of the employees surveyed choose to leave
Funcorp. This number corresponds to an 18-month
voluntary turnover rate of 23.1 percent (or 15.4
percent annually).
In prior studies of job embeddedness, researchers
have relied on a 40-item measure to capture the six
subdimensions that were then aggregated to create
composite measures (e.g., Mitchell et al., 2001). In
defining the construct, Mitchell et al. (2001) char-
acterized job embeddedness as a formative indica-
tor construct, in that multiple variables are asso-
ciated with the embeddedness construct and
predictive validity represents the major mechanism
for validation of its conceptual meaning (Edwards,
2001). In other words, job embeddedness captures a
large set of things that enmesh people in their jobs
and that predict voluntary turnover.
In the present study, we assessed the degree to
which an employee’s coworkers were enmeshed in
the organization and community (coworkers’ job
embeddedness) using a 21-item measure of job em-
beddedness developed and validated by Holtom,
Mitchell, Lee, and Tidd (2006). In their measure
development study, the product-moment correla-
tion showed a strong relationship between the orig-
inal long form and the revised short form (r ⫽ .92)
used to measure job embeddedness. This measure
was developed using data collected from 769 cor-
rections officers. Given the fact that the short-form
items are also represented in the long form, we
would expect this correlation to be very high. More
importantly, after job satisfaction was controlled
for, the long-form measure of individual job embed-
dedness significantly predicted voluntary turnover
(p ⬍ .001), as did the short-form measure (p ⬍
.001), which provides evidence of predictive valid-
ity for this shorter measure. Further, there was no
difference in the amount of variance in turnover
explained by the two forms of the instrument.
In both samples, the respondents indicated on a
five-point scale the extent to which they agreed
with 18 of the 21 items. The other 3 items involved
yes or no answers. We standardized and averaged
each individual’s scores for each item to create an
individual-level job embeddedness score. These in-
dividual job embeddedness scores were then aver-
aged across employees in each department to create
an aggregate of departmental job embeddedness
(i.e., departmental coworkers’ job embeddedness).
The Appendix reports the survey’s items. Because
individual job embeddedness is a formative (or in-
dicator) construct, high internal consistency (e.g.,
as measured by coefficient alpha) and unidimen-
sionality (e.g., as shown by one-factor-model supe-
riority) are not the standards by which construct
validity should be judged (Diamantopoulos & Win-
klhofer, 2001). However, for descriptive purposes,
we note that coefficient alpha was high (
␣
⫽ .88).
Control variables. Given that we wanted to test
coworkers’ job embeddedness as a predictor of fo-
cal employee turnover, we sought to control for
other variables that might provide alternative ex-
planations. These control variables included both
the individual (level 1) factors of job embedded-
ness, job satisfaction, organizational commitment,
part-time versus full-time status, age, gender, race,
and tenure, as well as the group (level 2) factors of
coworkers’jobsatisfaction, coworkers’organization-
al commitment, department size, and local unem-
ployment rate. Job satisfaction assessed the degree
to which employees expressed satisfaction with ten
dimensions of their jobs (e.g., pay, coworkers, pro-
motion, etc.) using a shortened version of Spector’s
(1985) job satisfaction measure. Spector’s original
scale includes 36 items, but because of survey
length constraints our shortened measure included
only the 2 best-loading items for each subscale (as
based on Spector [1985]). Thus, the respondents
indicated on a five-point scale the extent to which
they agreed with 20 items assessing satisfaction
with various aspects of their jobs. Coefficient alpha
for job satisfaction was .93. We measured organi-
550 JuneAcademy of Management Journal
zational commitment using four items from Meyer,
Allen, and Smith’s measure of affective organiza-
tional commitment. Respondents indicated on a
five-point scale the extent to which they agreed
with the items. Coefficient alpha for this measure
was .85. The employees’ full or part-time work sta-
tus was determined from organizational records at
the time the employee completed the survey (0 ⫽
“full time,” 1 ⫽ “part time”). Part-time employees
worked a maximum of 32 hours per week and did
not receive benefits, whereas full-time employees
were expected to work at least 40 hours per week
and received benefits. We obtained the demo-
graphic variables age, gender, race, and tenure from
the organizations’ records and entered them as con-
trols. We included these employee demographic
variables in the model because we wanted to have
confidence that effects were not based on employ-
ee’s life experiences, social categories, or career
position.
In addition, the analysis contained several group
(level 2) controls because they could also constitute
potential alternative explanations. These include
coworkers’ job satisfaction and coworkers’ organi-
zational commitment, which are the individual-
level variables of job satisfaction and organization-
al commitment averaged over department. We
should reiterate that we were adding group-level
job satisfaction and group-level organizational
commitment simply as conservative controls. Since
they are major predictors of turnover at the indi-
vidual level, they may also control variance in turn-
over when assessed at the group level. However,
we are not postulating that they necessarily op-
erate through a contagion process similar to co-
workers’ job embeddedness (although they
could). Department size was assessed as the num-
ber of employees in each branch or department.
Local unemployment rate was obtained from the
Bureau of Labor Statistics for each zip code
where a club was located.
Analysis
Employees who share a department have the
same coworkers’ satisfaction, coworkers’ commit-
ment, and coworkers’ embeddedness scores. To ig-
nore this dependence by using normal logistic re-
gression would violate a core assumption of
regression analysis. Even excluding the focal actor
from each aggregated score would leave highly in-
terdependent aggregated scores. In fact, aggregated
scores with the focal actor excluded are almost
identical to aggregated scores with the focal actor
included (i.e., the average correlation is .95). There-
fore, the data were analyzed with multilevel logis-
tic regression software called hierarchical general-
ized linear modeling (HGLM; Guo & Zhao, 2000).
The main difference between hierarchical linear
modeling (HLM) and HGLM is that the latter allows
for binary outcome variables (e.g., stay/quit).
HGLM was ideal for our tests because it is designed
to account for nonindependence between group-
level predictor variables. Given that HGLM is sim-
ply a type of multilevel logistic regression analysis,
normally distributed outcome variables and error
terms are not necessary. It has been used to study
multilevel predictors of a wide range of binary out-
comes, including whether a person drops out of
high school, completes college, marries, divorces,
or goes bankrupt (see Guo and Zhao, 2000). HGLM
helps one to disentangle individual-level effects
from social effects by statistically disaggregating
individual (level 1) and group (level 2) effects. In
sum, we used this form of analysis because it pro-
vided the least biased and most informative
method of hypothesis testing in this context.
Results
Tables 1 and 2 report the means, standard devi-
ations, and correlation coefficients between the de-
pendent, independent, and control variables for the
level 1 and level 2 variables. Table 3 presents the
results of the HGLM analysis.
Hypothesis 1 posited a negative relationship be-
tween coworkers’ job embeddedness and voluntary
turnover. Table 3 (model 1) shows a negative and
significant relationship between coworkers’ job
embeddedness and individual voluntary turnover
(

⫽⫺.19, p ⬍ .001). We further suggested that
coworkers’ job embeddedness would predict turn-
over even when coworkers’ job satisfaction and
organizational commitment are controlled. As
shown in Table 3 (model 2), coworkers’ job embed-
dedness remains significantly predictive of turn-
over (

⫽⫺.16, p ⬍ .001), and neither coworkers’
TABLE 1
Study 1 Means, Standard Deviations, and
Correlations of Level 2 Variables
a
Variables Mean s.d. 1 2 3 4
1. Group size 8.35 1.50
2. Local unemployment rate 4.55 1.01 ⫺.08
3. Coworkers’ organizational
commitment
4.04 0.48 ⫺.13 .01
4. Coworkers’ job satisfaction 3.90 0.42 ⫺.02 .04 .63
5. Coworkers’ job
embeddedness
3.79 0.31 ⫺.09 .00 .64 .59
a
k ⫽ 1,037 departments; all correlations greater than .03 are
significant at p ⬍ .01.
2009 551Felps, Mitchell, Hekman, Lee, Holtom, and Harman
job satisfaction nor coworkers’ organizational com-
mitment remains as a significant predictor of turn-
over in this “competitive test” model. Thus, Hy-
pothesis 1 is supported. Although replication was
not the focus of this research, Table 3 does show
results that replicate prior research. Specifically,
individual-level job satisfaction, organizational
commitment, and job embeddedness are signifi-
cant, negative predictors of voluntary turnover
(model 2: satisfaction,

⫽⫺.07, p ⬍ .05; commit-
ment,

⫽⫺.16, p ⬍ .01; embeddedness,

⫽⫺.09,
p ⬍ .01).
Supplementary Qualitative Analysis
Our model suggests that aggregate job embedded-
ness influences individual turnover through conta-
gion of job search behaviors. In Study 1, we did not
quantitatively measure job search behaviors. In-
stead, we conducted content analyses of a deduc-
tive nature based on 11 focus groups at both of our
research sites. To get a wide variety of responses,
we selected Funcorp and Cashcorp
1
sites where
employee turnover from prior years was high (six
focus groups) and other sites where turnover was
low (five focus groups). Eisenhardt and Graebner
(2007) called this method “contrasting polar
types.” Focus groups were conducted before the
survey was sent out in both samples. We asked
focus group participants to tell us about their jobs
and why people stayed or left. The focus groups
lasted 90–120 minutes each and were attended by
an average of eight employees each. All interviews
were audiotaped and later transcribed verbatim.
One of the leading measures of job search behav-
ior (Kopelman et al., 1992) asks respondents to note
which, if any, of ten different search behaviors they
have engaged in during the prior year. The behav-
iors include revising one’s re´sume´, going on a job
interview, and talking with coworkers about getting
a new job. In the focus groups, we were careful not
to put any of the participants under pressure by
asking questions about revising re´sume´s or going
on job interviews. However, when we asked about
the reasons why people stay, we noted that many
1
This is our pseudonym for the bank providing Study
2 data.
TABLE 3
Study 1 HGLM Logistic Regression Results Predicting
Individual Voluntary Turnover
a
Variables
Individual Turnover
Model 1 Model 2
Level 2
Group size ⫺.01 ⫺.01
Local unemployment rate ⫺.03 ⫺.03
Coworkers’ job embeddedness ⫺.19*** ⫺.16***
Coworkers’ organizational
commitment
⫺.06
Coworkers’ job satisfaction .01
Level 1
Age ⫺.47*** ⫺.48***
Tenure ⫺.62*** ⫺.61***
Gender ⫺.05* ⫺.05*
Race ⫺.14*** ⫺.15***
Work status .09** .09**
Job embeddedness ⫺.10** ⫺.09**
Organizational commitment ⫺.17*** ⫺.16***
Job satisfaction ⫺.05* ⫺.07*
Log-likelihood ⫺794.45 ⫺802.81
⫺2[L(

reduced
) ⫺ L(

full
)]
b
16.72***
a
To enhance ease of interpretation, we report standardized
coefficients. n ⫽ 8,663 individuals; k ⫽ 1,037 departments.
b
The significant change in log-likelihood indicates that
model 2 is significantly worse at predicting turnover than model
1. The loss of two degrees of freedom may be the cause.
* p ⬍ .05
** p ⬍ .01
*** p ⬍ .001
TABLE 2
Study 1 Means, Standard Deviations, and Correlations of Level 1 Variables
a
Variables Mean s.d. 12345678
1. Voluntary turnover 0.23 0.42
2. Age 39.98 14.45 ⫺.21
3. Tenure 7.20 6.79 ⫺.17 .41
4. Gender 0.39 0.49 .01 ⫺.07 ⫺.03
5. Race 0.33 0.47 ⫺.08 ⫺.04 .05 ⫺.12
6. Work status 0.26 0.44 .07 ⫺.05 ⫺.24 .05 ⫺.17
7. Organizational commitment 3.97 0.87 ⫺.19 .21 .14 ⫺.06 .12 ⫺.13
8. Job satisfaction 3.89 0.72 ⫺.08 .00 ⫺.02 ⫺.06 .12 .04 .62
9. Job embeddedness 3.76 0.55 ⫺.17 .20 .15 ⫺.01 .06 ⫺.01 .63 .57
a
n ⫽ 8,663 individuals; all correlations greater than .02 are significant at p ⬍ .01.
552 JuneAcademy of Management Journal
spontaneous comments about leaving emerged.
Moreover, it seemed to be much more acceptable to
discuss leaving in the high-turnover locations.
Consequently, we asked two of the authors who
were not involved with conducting the focus
groups to use Atlas.ti qualitative software (a quali-
tative analysis tool that helps users organize, lo-
cate, code, and annotate findings from large vol-
umes of qualitative documents) to independently
count the comments about leaving (e.g., reasons for
leaving, alternative job options, people who had
left or were considering leaving). To ensure the
coding process was blind, all focus-group-identify-
ing information was removed from the transcripts.
Before coding, the two judges discussed how they
would count leaving reasons. For example, they
agreed that when a single focus group participant
listed several leaving reasons, they would count
each reason as unique. One coder counted 158 leav-
ing reasons in the 11 focus groups, and the other
coder counted 163 leaving reasons. Together, the
two judges identified 168 leaving reasons, of which
156 were the same, for 93 percent agreement. All
disagreements were resolved in a discussion be-
tween the two judges, and the judges ultimately
agreed on a final count of 158 leaving reasons.
To assess spontaneous discussions about leaving,
the coders counted the number of reasons for leav-
ing that employees publicly stated in each group.
Employees mentioned that their coworkers left to
obtain more pay, better opportunities or benefits,
and less physically demanding jobs, or to go back to
school. For example, one Cashcorp employee at a
branch where the coworkers’ job embeddedness
score was low made the comment, “Did you know
that at [alternative company], the pay starts at $9 or
$10 and they reimburse 100% of tuition? If I saw
that they were hiring, I could see myself leaving.”
After conducting the focus groups, we surveyed the
employees as part of the broader quantitative por-
tion of our study, as we describe in the Methods
sections pertaining to Studies 1 and 2. From each
focus group participant’s individual job embedded-
ness, commitment, and satisfaction scores, we cal-
culated each focus group’s average level of job em-
beddedness, commitment, and satisfaction. We
imputed the organizational average score to each of
the 5 focus group participants (out of 88) who did
not fill out a survey. The survey data gathered from
focus group participants were also included in the
broader HGLM analysis. The number of coded com-
ments about leaving (a proxy for job search behav-
ior) was then correlated with the group’s average
level of job embeddedness, commitment, and
satisfaction.
Our findings were consistent with our hypothe-
ses about what causes people to search and leave.
The group’s average levels of satisfaction (r ⫽⫺.10,
n.s.) and commitment (r ⫽⫺.27, n.s.) were not
significantly correlated with the number of com-
ments about leaving. However, and consistently
with the turnover contagion model, the group’s
level of job embeddedness was significantly, nega-
tively correlated with the number of comments
about leaving (r ⫽⫺.64, p ⬍ .05). This qualitative
finding regarding coworkers’ job embeddedness is
considerably more speculative than our subsequent
quantitative findings reported for Study 2. As men-
tioned by Lee (1999), qualitative research is not
suited to discussions of prevalence, generalizabil-
ity, or calibration. Qualitative research, however, is
well suited to discussions of description, interpre-
tation, and explanation. Thus, these findings in-
creased our confidence in our conceptual under-
standing and encouraged us to further test whether
coworker job search mediates the relationship be-
tween coworkers’ job embeddedness and focal em-
ployee turnover.
STUDY 2
Hypotheses
In Study 2, we sought not only to replicate the
results of Study 1 but also to gain greater under-
standing of the coworker behaviors that explain the
effect of coworkers’ job embeddedness on individ-
ual employee turnover. Recall that Hypothesis 2
holds that coworkers’ job search behavior mediates
the effect of coworkers’ job embeddedness on indi-
vidual turnover. As described in the theory devel-
opment section, contagion is a process by which
turnover propensity spreads from coworkers to a
focal actor. This process is hypothesized to occur
when employees model each other’s leaving-re-
lated behaviors (i.e. resume revision, reading the
classifieds, going on job interviews, etc.). However,
in Study 1 we did not directly measure job search
behavior. In Study 2 we attempted to assess di-
rectly if job search behavior was more common
where employees were not embedded and if co-
workers’ search behavior mediated the relationship
between coworkers’ job embeddedness and focal
employee turnover.
Methods
Sample. Our second site was a retail bank in the
U.S. Midwest. Cashcorp owns and operates 45
branch offices in two states and has roughly two
billion dollars in assets. We sent a survey to all 486
employees. Three-hundred and twenty employees
2009 553Felps, Mitchell, Hekman, Lee, Holtom, and Harman
completed the survey, for a response rate of 66
percent. Missing values reduced the number of us-
able observations to 234 and the final response rate
to 48 percent. In our usable sample, respondents
were from 45 branches; 77.1 percent were women;
the average age was 37.8 years; the average tenure
with the organization was 6.1 years; and 8.2 per-
cent were nonwhites. The average branch size was
10.8 (5.2 survey respondents per branch). In the
two years following the survey, 60 employees who
completed our survey voluntarily left Cashcorp,
which equates to a two-year voluntary turnover rate
of 25.7 percent (or 12.9 percent per year). Finally,
nonresponse bias was unlikely because employees
who completed our survey and those who did not
were not significantly different in terms of gender,
age, tenure, race, or turnover rate.
Measures. The measures and methods were only
slightly different from those for Study 1. The most
important addition was that Study 2 included the
ten-item Job Search Behavior Index (Kopelman et
al., 1992;
␣
⫽ .83). We aggregated this measure
to the unit level in order to assess the amount of
job search activity occurring in a particular bank
branch. This index seeks to tap the actual behaviors
involved in looking for a new job and includes
items such as, “[During the past year, have you]
revised your re´sume´?” “. . . read the classified/
help wanted advertisements in the newspaper?”
“. . . sent copies of your re´sume´ to a prospective
employer?” “. . . talked to coworkers about getting a
new job?” “. . . gone on a job interview?” The re-
sults of Kopelman and colleagues (1992) suggest
that this index may be a better and more behavior-
ally grounded predictor of employee turnover than
are intention to leave and attitudinal variables.
Study 2 also employs slightly better measures of
job satisfaction and organizational commitment.
Whereas the first study used a shortened 20-item
version of Spector’s Job Satisfaction Index (1985),
Study 2 used the full 36-item scale. In addition,
Study 1 only assessed affective commitment, and
Study 2 used a more comprehensive 18-item mea-
sure of organizational commitment that includes
affective, normative, and continuance commitment
(Meyer, Allen, & Smith, 1993). Both studies used
the same short version of Holtom and colleagues’
(2006) job embeddedness scale. Once again, the
reliability scores for these scales were high (
␣
⫽
.93, job satisfaction;
␣
⫽ .89, organizational com-
mitment; and
␣
⫽ .82, job embeddedness). Finally,
we used Griffeth and Hom’s (1988) 5-item Index of
Perceived Job Alternatives at the individual level to
control for the effect of employee perceptions on
the job alternative–turnover relationship. Finally,
as in Study 1, we aggregated five variables (i.e., job
search, job embeddedness, job satisfaction, organi-
zational commitment, and job alternatives) to the
unit level. Our rationale for aggregating job satis-
faction, organizational commitment, and job alter-
natives was the same as for Study 1. They were
seen as conservative controls.
Analysis. The analytic technique also remained
the same (i.e., HGLM), with the exception that we
tested mediation using both the traditional Baron
and Kenny (1986) standard and the Sobel test,
which directly assesses the statistical significance
of the change in regression coefficients when the
mediator is added to the equation. As Baron and
Kenny (1986) suggested, the Sobel test offers a con-
firmatory and rigorous test of mediation.
Results
Tables 4 and 5 present the level 1 and level 2
descriptive statistics. Table 6 presents results of
the regression of the job attitude variables on the
mediator (coworkers job search behavior) as well
as results of the HGLM analysis for actual quit-
ting. The main purpose of this study, though, was
to test whether coworkers’ job search behavior
would mediate the relationship between cowork-
TABLE 4
Study 2 Means, Standard Deviations, and Correlations of Level 2 Variables
a, b
Variables Mean s.d. 123456
1. Group size 5.20 6.07
2. Local unemployment rate 4.08 2.46 ⫺.08
3. Coworkers’ job alternatives 3.34 0.41 .08 ⫺.14
4. Coworkers’ organizational commitment 2.97 0.35 .04 .03 ⫺.53
5. Coworkers’ job satisfaction 3.27 0.31 .01 ⫺.09 ⫺.52 .66
6. Coworkers’ job embeddedness 2.36 0.27 .19 ⫺.02 ⫺.19 .52 .46
7. Coworkers’ search behavior 0.45 0.19 ⫺.12 .13 .38 ⫺.60 ⫺.55 ⫺.60
a
The correlation between coworkers’ job embeddedness and comments about leaving is ⫺.64, p ⬍ .05. The relationships between
coworkers’ job satisfaction and coworkers’ organizational commitment and comments about leaving are not significant.
b
k ⫽ 45 departments; all correlations greater than .19 are significant at p ⬍ .05.
554 JuneAcademy of Management Journal
ers’ job embeddedness and focal employee vol-
untary turnover. Results indicate that coworkers’
job embeddedness was significantly and nega-
tively related to coworkers’ job search behavior
(

⫽⫺.41, p ⬍ .001) even after we controlled for
bank branch size, local unemployment rate, co-
workers’ organizational commitment, coworkers’
job satisfaction, and coworkers’ perceptions of
job alternatives. Next, we regressed voluntary
turnover on coworkers’ job embeddedness (mod-
els 1 and 2 in Table 6). Replicating our first
study’s findings, coworkers’ job embeddedness
TABLE 6
Study 2 OLS and HGLM Regression Results Examining the Influence of Coworkers’ Job Embeddedness
a
Variables Coworker Job Search
Individual Turnover
Model 1 Model 2 Model 3
Level 2
Group size ⫺.09 ⫺.04 ⫺.10 ⫺.21* ⫺.19*
Local unemployment rate .20 .20 .16 .06 ⫺.04
Coworkers’ job alternatives .13 .18 ⫺.17 ⫺.26
Coworkers’ organizational commitment ⫺.32* ⫺.24 ⫺.11 ⫺.04
Coworkers’ job satisfaction ⫺.18 ⫺.08 ⫺.27 ⫺.29
Coworkers’ job embeddedness ⫺.41*** ⫺.92*** ⫺.68** ⫺.46
Coworkers’ job search .59**
Level 1
Age ⫺.58** ⫺.52* ⫺.54*
Tenure ⫺.26 ⫺.18 ⫺.18
Gender .08 .16 .15
Race .29 .30 .33
Work status .29 .27 .28
Job embeddedness .12 .10 .12
Organizational commitment ⫺.15 ⫺.15 ⫺.15
Job satisfaction ⫺.10 ⫺.07 ⫺.10
Job alternatives ⫺.10 ⫺.11 ⫺.10
Job search behavior index .87*** .90*** .87***
R
2
.44 .55
⌬R
2
from previous model
.11***
Log-likelihood ⫺330.25 ⫺330.55 ⫺326.63
⫺2[L(

reduced
) ⫺ L(

full
)]
7.84**
a
To enhance ease of interpretation, we report standardized coefficients. n ⫽ 234 individuals; k ⫽ 45 branches.
* p ⬍ .05
** p ⬍ .01
*** p ⬍ .001
TABLE 5
Study 2 Means, Standard Deviations, and Correlations of Level 1 Variables
a
Variables Mean s.d. 12345678910
1. Voluntary turnover 0.26 0.44
2. Age 37.51 12.15 ⫺.37
3. Tenure 5.89 7.05 ⫺.26 .47
4. Gender 0.73 0.45 .10 ⫺.15 ⫺.03
5. Race 0.09 0.28 .20 ⫺.09 ⫺.04 .06
6. Work status 0.13 0.33 .16 ⫺.22 ⫺.13 .13 .09
7. Job alternatives 3.36 0.95 .07 .02 .03 .07 ⫺.13 ⫺.06
8. Organizational commitment 2.97 0.62 ⫺.26 .06 .13 .05 .04 .01 ⫺.33
9. Job satisfaction 3.28 0.53 ⫺.24 .04 ⫺.04 ⫺.07 ⫺.02 ⫺.02 ⫺.34 .55
10. Job embeddedness 2.41 0.47 ⫺.28 .30 .27 ⫺.04 ⫺.14 ⫺.17 ⫺.23 .56 .47
11. Job search behavior index 0.43 0.31 .45 ⫺.37 ⫺.33 .03 .10 .06 .19 ⫺.44 ⫺.37 ⫺.49
a
n ⫽ 234 for all variables; all correlations greater than .13 are significant at p ⬍ .05.
2009 555Felps, Mitchell, Hekman, Lee, Holtom, and Harman
significantly predicted turnover (

⫽⫺.92, p ⬍
.001). Then we regressed voluntary turnover si-
multaneously on coworkers’ job search behavior
and coworkers’ job embeddedness (model 3 in
Table 6). Coworkers’ job embeddedness becomes
nonsignificant when the mediator is added.
Moreover, the Sobel test shows that the change in
the regression coefficient for coworkers’ job em-
beddedness is itself significant (Sobel t ⫽⫺8.6,
p ⬍ .001). Thus, Hypothesis 2 is supported. Cowork-
ers’ job search behavior appears to mediate the rela-
tionship between coworkers’ job embeddedness and
turnover. As in our first study, we found that when
we included coworkers’ satisfaction and commitment
variables in the model, only coworkers’ job embed-
dedness was still significant.
The implications of our results are apparent in
the effect sizes and statistical ramifications
within our sample. On average, using a simple
log-odds transformation, we found that a one
standard deviation increase in coworkers’ job em-
beddedness decreased the probability of an indi-
vidual voluntarily leaving from 15.4 percent per
year to 8.5 percent per year at Funcorp, and from
12.9 percent per year to 4.2 percent per year at
Cashcorp. This equates to a decrease in voluntary
turnover of 45 percent at Funcorp and 67 percent
at Cashcorp, given controls for other variables in
the model. We speculate that the Cashcorp re-
sults are stronger because the units are smaller
and more exclusive. In the confined space of a
bank branch, people saw their unit members
more frequently and were exposed to only their
fellow branch members’ leaving behaviors.
Across the two samples, a one standard deviation
decrease in coworkers’ job search behavior de-
creased the probability of an individual “turning
over” by 35 percent. Moreover, in comparing the
effect sizes of the individual and coworker vari-
ables, we found them to be roughly equal predic-
tors of focal actor quitting. Thus, the job embed-
dedness and job search behavior of coworkers
had a sizable influence on focal actor turnover
decisions.
DISCUSSION
Extending social comparison theory to the do-
main of turnover, we investigated the role of co-
workers’ attitudes and behaviors on individual
employee turnover propensity. In two separate
samples, we found that aggregated coworkers’ job
embeddedness was a valid predictor of individ-
ual voluntary turnover. There are a variety of
ways to demonstrate validity: (1) controlling for
alternative explanations, (2) predicting a crite-
rion, (3) explaining additional variance over and
above that explained by competitive constructs,
and (4) showing the process by which something
has an effect (i.e., establishing mediation). It
should be noted that all four of these criteria have
been demonstrated for coworkers’ job embedded-
ness. In analyses (1) controlled for demographic
characteristics (e.g., tenure, age, work status, and
gender), perceived and objective measures of job
alternatives, and department size, coworkers’ job
embeddedness (2) predicts turnover in two dis-
tinct samples, (3) exceeds the prediction of sim-
ilarly aggregated variables such as coworkers’ job
satisfaction and coworkers’ organizational com-
mitment, (4) and can plausibly be seen to operate
through the observation of coworkers’ job search
behaviors. Thus, we can have some confidence
that coworkers’ job embeddedness is one impor-
tant driver of turnover contagion.
But going beyond job embeddedness, we also
provide evidence through both qualitative and
quantitative analyses that coworkers’ job search
may act as a critical mechanism in the turnover
contagion process. These findings are conservative
in that job search behaviors have more recently
evolved to include Internet job search, job clearing-
house websites, and e-mail correspondence about
positions. The Job Search Behavior Index (Kopel-
man et al., 1992) does not account for these new
ways of searching for a job.
A particular strength of the study is the replica-
tion of findings for two large samples in two very
different settings. Cashcorp employees worked in
self-contained branches with relatively few people.
The employees at Funcorp, in contrast, were
grouped according to department within a larger
organizational unit (a club). These department
members were not isolated from other employees
in different departments. We suspect that this fac-
tor dilutes the influence of coworkers’ job embed-
dedness. As such, finding that coworkers’ job em-
beddedness influences voluntary turnover at
Funcorp represents a more rigorous test of our hy-
potheses. Taking the results of these samples to-
gether enhances our confidence in the robustness
of our inferences.
Limitations and Future Research
We have argued that taking an average of cowork-
ers’ job embeddedness and coworkers’ job search
behaviors makes sense as a way to capture the
turnover contagion stimuli to which a focal indi-
vidual is exposed. However, we note that the as-
sessment of turnover contagion was indirect. Nei-
ther the more speculative qualitative data gathered
556 JuneAcademy of Management Journal
from the focus groups nor the more rigorous job
search behaviors gathered in Study 2 measured
what a focal person actually heard or saw cowork-
ers do. Such behavioral data are difficult to gather
but would seem to be a necessary component of
future research on this topic. Moreover, there are
other variables that might signal how likely it is for
contagion to occur—for instance, how close desks
are situated to each other, how often or effectively
coworkers communicate with each other, friend-
ship level, and status similarity. Given that such
data were not available in our samples, the current
research employed a simpler (and more conserva-
tive) measurement of turnover contagion. Future
research could productively build on these find-
ings to identify alternative operationalizations of
the turnover contagion process, as well as modera-
tors of these effects.
Further, our research has not included all vari-
ables known to be related to turnover. In partic-
ular, a valuable contribution to future research
would be to include more macro variables, such
as organizational support, leadership quality,
and compensation policies, that might be fruit-
fully integrated as antecedent, moderator, or al-
ternative mechanisms for the turnover contagion
model developed here. In particular, it is possible
that norms about the legitimacy of leaving might
develop and could affect turnover (Abelson,
1993). Qualitative work by Bartunek et al. (2008)
and Rumery (2003) has suggested that such col-
lective norms can develop and that they may
affect turnover attitudes and behaviors. Unfortu-
nately, our data cannot speak to this issue. More-
over, it should be pointed out that if such norms
were to exist, they would be predicated on exten-
sive social comparison (Bartunek et al, 2008) and
thus would act as a complementary rather than
substitute mechanism for turnover contagion.
Thus, although a normative factor could add to
the prediction of individual turnover, we do not
believe that effect will replace or be as strong as
the contagion effect captured here. Said differ-
ently, we have attempted to control for the vari-
ables most likely to provide alternative explana-
tions for our findings, yet we have not controlled
for all of them, nor have we included all the
variables that may be involved in the process.
The inclusion of these additional variables may
both clarify and extend the current research.
Finally, another potential limitation concerns the
issue of weights for the subdimensions of job em-
beddedness. As a robustness check, we ran all the
analyses using the weighted approach suggested by
Law, Wong, and Mobley (1998). Specifically, we
ran a logistic regression in which we regressed
turnover on the six job embeddedness items to get
the weights of each of the dimensions. We then
multiplied each individual’s score on each dimen-
sion by the weight for each dimension and added
the six resulting products together. This created the
weighted job embeddedness score. We then aggre-
gated these weighted individual job embeddedness
scores among the members of each department to
create our measure of coworkers’ job embedded-
ness. The results of this analysis are virtually iden-
tical to those obtained from the straight aggregation
approach but with slightly improved predictive va-
lidity, which we would expect (Edwards, 2001;
Howell, Breivik, & Wilcox, 2007; Law et al., 1998).
We chose not to report or base our conclusions on
the results using these weights for three reasons.
First, such weights capitalize on sample-specific
variance and error. Second, findings reported in
other studies suggest that the contributions to turn-
over of the dimensions vary by samples (Allen,
2006; Lee et al., 2004, Zatzick & Iverson, 2006).
Thus, using weights means that the construct is
essentially different with every sample, which
makes it difficult to meaningfully compare results
across studies (Howell et al., 2007) and thus re-
duces generalizability and complicates theory
building. Third, it was more conservative (e.g., less
likely to capitalize on chance) and in line with
previous research to use the aggregate job embed-
dedness score. However, we should add that this
variation in weights points to the need for future
research into the potential moderators of the rela-
tionship between the subdimensions of job embed-
dedness and turnover or performance. Better infor-
mation is still needed about how and under what
conditions job embeddedness subdimensions in-
fluence turnover. Finally, though we did not use
the weights in the research reported here, we rec-
ognize that such sample-specific information may
be what is most valuable in making prescriptions
for any given organization.
Managerial Implications
Organizations can use the results of this study to
design specific interventions aimed at reducing
voluntary turnover. A primary implication is that,
at the group level, job embeddedness is an impor-
tant antecedent to eventual turnover. Beyond just
affecting individual decision making, it also influ-
ences whether the social environment incites leav-
ing. Of particular interest in the context of this
research is a study by Allen (2006). He found that
collective socialization tactics—where newcomers
experience common learning experiences with a
group or cohort—increase embeddedness in an or-
2009 557Felps, Mitchell, Hekman, Lee, Holtom, and Harman
ganization. Such socialization tactics provide a
common message about the organization, roles, and
appropriate responses. This common message may
shape how groups of people interpret organization-
al events such as the loss of a respected coworker or
a large number of coworkers simultaneously. In
short, organizations could actively manage the con-
tent of collective socialization experiences as well
as attend to influential individuals in social net-
works (Bartunek et al., 2008; Mossholder et al.,
2005).
Second, individual-level factors that increase in-
dividual job embeddedness should also be consid-
ered (Mitchell et al., 2001). Prior research has iden-
tified a number of antecedents to on-the-job
embeddedness. For example, personality variables
such as conscientiousness, extraversion, and agree-
ableness have demonstrated a strong, positive rela-
tionship with on-the-job embeddedness (Giosan,
Holtom, & Watson, 2005). Thus, reducing voluntary
turnover through selection is one clearly actionable
approach (Barrick & Zimmerman, 2005). Further,
both perceived supervisor support and perceived
organizational support have been demonstrated to
positively predict levels of on-the-job embedded-
ness (Giosan et al., 2005) and reduced voluntary
turnover (Maertz, Griffeth, Campbell, & Allen,
2007). Other suggestions include developing
schedules that fit employee needs (Holtom, Lee, &
Tidd, 2002), providing creative benefit alternatives
or cafeteria plans, tailoring benefits to meet indi-
vidual needs and enhance work-life balance, allow-
ing employees input into designing work environ-
ments, and providing incentives or perks based on
tenure (Giosan et al., 2005).
Off-the-job embeddedness can also be increased
in a number of ways. For example, one firm was
able to increase community embeddedness and
subsequent retention by recruiting and hiring from
communities close to their facilities and avoiding
relocating employees whenever possible (Holtom
et al., 2006). Similarly, another firm increased links
in the community by supporting community ser-
vice by employees (e.g., giving two days off per year
for community service), allowing them to volunteer
in local student programs as mentors, and encour-
aging professional involvement in community-
based professional organizations (Holtom et al.,
2006). Finally, one organization augmented em-
ployee-perceived community-related sacrifice and
subsequent retention by providing home-buying as-
sistance (Holtom et al., 2006). In sum, there are
many ways that organizations can systematically
seek to reduce the rate of voluntary, avoidable turn-
over by enacting programs designed to increase job
embeddedness at the meso and micro levels.
Job search behavior may also have managerial
implications, and these are potentially more con-
troversial. Specifically, managers could prohibit
gossiping about people who are looking for other
jobs, especially “on company time.” Such a prohi-
bition might inhibit the spread of contagious infor-
mation, but bald attempts at concertive control may
provoke reactance, ill-will, and even sabotage. Per-
haps a more realistic alternative is for managers to
track job embeddedness and turnover at the team
level. Where embeddedness is low and/or turnover
is high, they might actively try to raise embedded-
ness scores or reconstitute a group with some peo-
ple who have high embeddedness. Such changes
might reduce the job search behaviors demon-
strated by group members.
Concluding Thoughts
Tackling turnover theory at the meso level is not
new; it has been advocated at the organization cul-
ture level (Abelson, 1993) and even empirically
examined on occasion (cf. Feeley & Burnett, 1997.
However, it is our belief that these approaches have
not focused enough on social factors—specifically,
the attitudes, characteristics, and behaviors of focal
employees’ coworkers. Although researchers per-
haps do not naturally think of quitting as a social
phenomenon, our research suggests that it is and
that additional research regarding the social predic-
tors of turnover is warranted.
REFERENCES
Abelson, M. A. 1993. Turnover cultures. In G. Ferris & G.
Rowland (Eds.), Research in personnel and human
resource management, vol. 11: 339–376. Green-
wich, CT: JAI Press.
Allen, D. G. 2006. Do organizational socialization tactics
influence newcomer embeddedness and turnover?
Journal of Management, 32: 237–256.
Bakker, A. B., & Schaufeli, W. B. 2000. Burnout contagion
processes among teachers. Journal of Applied So-
cial Psychology, 30: 2289–2308.
Bandura, A. 1977. Social learning theory. New York:
General Learning Press.
Banerjee, D., & Gaston, N. 2004. Labour market signaling
and job turnover revisited. Labour Economics, 11:
599– 622.
Baron, R. M., & Kenny, D. A. 1986. The moderator-medi-
ator distinction in social psychological research:
Conceptual, strategic, and statistical considerations.
Journal of Personality and Social Psychology, 51:
1173–1182.
Barrick, M. R., & Zimmerman, R. D. 2005. Reducing vol-
558 JuneAcademy of Management Journal
untary turnover, avoidable turnover through selec-
tion. Journal of Applied Psychology, 90: 159–166.
Barsade, S. G. 2002. The ripple effect: Emotional conta-
gion and its influence on group behavior. Adminis-
trative Science Quarterly, 47: 644 –675.
Bartunek, J. M., Huang, Z., & Walsh, I. J. 2008. The
development of a process model of collective turn-
over. Human Relations, 61: 5–38.
Bradbury, H., & Lichtenstein, B. M. B. 2000. Relationality
in organizational research: Exploring the space be-
tween. Organization Science, 11: 551–564.
Brett, J. M., & Stroh, L. K. 2003. Working 61 plus hours a
week: Why do managers do it? Journal of Applied
Psychology, 88: 67–78.
Chan, D. 1998. Functional relations among constructs in
the same content domain at different levels of anal-
ysis: A typology of composition models. Journal of
Applied Psychology, 83: 234–246.
Chartrand, T. L., & Bargh, J. A. J. 1999. The chameleon
effect: The perception-behavior link and social inter-
action. Journal of Personality and Social Psychol-
ogy, 76: 893–910.
Crossley, C. D., Bennett, R. J., Jex, S. M., & Burnfield, J. L.
2007. Development of a global measure of job em-
beddedness and integration into a traditional model
of voluntary turnover. Journal of Applied Psychol-
ogy, 92: 1031–1042.
Degoey, P. 2000. Contagious justice: Exploring the social
construction of justice in organizations. In B. M.
Staw & R. I. Sutton (Eds.), Research in organization-
al behavior, vol. 22: 51–102. Greenwich, CT: JAI
Press.
Diamantopoulos, A., & Winklhofer, H. M. 2001. Index
construction with formative indicators: An alterna-
tive to scale development. Journal of Marketing
Research, 38: 269–277.
Eder, P., & Eisenberger, R. 2008. Perceived organizational
support: Reducing the negative influence of co-
worker withdrawal behavior. Journal of Manage-
ment, 34(1): 55–68.
Edwards, J. R. 2001. Multidimensional constructs in
organizational behavior research: An integrative
analytical framework. Organizational Research
Methods, 4: 144–192.
Eisenhardt, K. M., & Graebner, M. E. 2007. Theory build-
ing from cases: Opportunities and challenges. Acad-
emy of Management Journal, 50: 25–32.
Feeley, T. J., & Barnett, G. A. 1997. Predicting employee
turnover from communication networks. Human
Communication Research, 23: 370–387.
Festinger, L. 1954. A theory of social comparison pro-
cesses. Human Relations, 7: 117–140.
Giosan, C., Holtom, B., & Watson, M. 2005. Antecedents
to job embeddedness: The role of individual, organ-
izational and market factors, Journal of Organiza-
tional Psychology, 5: 31–44.
Griffeth, R. W., & Hom, P. W. 1988. A comparison of
different conceptualizations of perceived alterna-
tives in turnover research. Journal of Organiza-
tional Behavior, 9: 103–111.
Griffeth, R. W., Hom, P. W., & Gaertner, S. 2000. A meta-
analysis of antecedents and correlates of employee
turnover: Update, moderator tests, and research im-
plications for the millennium. Journal of Manage-
ment, 26: 463–488.
Guo, G., & Zhao, H. 2000. Multilevel modeling for binary
data. In K. Cook & J. Hagan (Eds.), Annual review of
sociology, vol. 26: 441– 462. Stanford, CA: Annual
Reviews.
Hatch, N. W., & Dyer, J. H. 2004. Human capital and
learning as a source of sustainable competitive ad-
vantage. Strategic Management Journal, 25: 1155–
1178.
Haveman, H. A. 1995. The demographic metabolism of
organizations: Industry dynamics, turnover, and ten-
ure distributions. Administrative Science Quar-
terly, 40: 586– 618.
Holtom, B. C., Lee, T. W., & Tidd, S. 2002. The relation-
ship between work status congruence and work-re-
lated attitudes and behaviors. Journal of Applied
Psychology, 87: 903–915.
Holtom, B. C., Mitchell, T. R., & Lee, T. W. 2006. Increas-
ing human and social capital by applying job embed-
dedness theory. Organizational Dynamics, 35(4):
316–331.
Holtom, B. C., Mitchell, T. R., Lee, T., & Tidd, S. 2006.
Less is more: Validation of a short form of the job
embeddedness measure and theoretical exten-
sions. Paper presented at the annual meeting of the
Academy of Management, Atlanta.
Holtom, B. C., & O’Neill, B. S. 2004. Job embeddedness: A
theoretical foundation for developing a comprehen-
sive nurse retention plan. Journal of Nursing Ad-
ministration, 34: 216–227.
House, R. J., Rousseau, D. M., & Thomas-Hunt, M. 1995.
The meso-paradigm: A framework for the integration
of micro and macro organizational behavior. In B. M.
Staw & L. L. Cummings (Eds.), Research in organi-
zational behavior, vol. 17: 71–114. Greenwich, CT:
JAI Press.
Howell, R. D., Breivik, E., & Wilcox, J. B. 2007. Recon-
sidering formative measurement. Psychological
Methods, 12: 205–218.
Johns, G. 2006. The essential impact of context on organ-
izational behavior. Academy of Management Re-
view, 31: 386– 408.
Kammeyer-Mueller, J. D., Wanberg, C. R., Glomb, T. M., &
Ahlburg, D. 2005. The role of temporal shifts in
turnover processes: It’s about time. Journal of Ap-
plied Psychology, 90: 644– 658.
2009 559Felps, Mitchell, Hekman, Lee, Holtom, and Harman
Klein, K. J., & Kozlowski, S. W. J. 2000. From micro to
meso: Critical steps in conceptualizing and conduct-
ing multilevel research. Organizational Research
Methods, 3: 211–236.
Kopelman, R. E., Rovenpor, J. L., & Millsap, R. E. 1992.
Rationale and construct validity evidence for the Job
Search Behavior Index: Because intentions and New
Year’s resolutions often come to naught. Journal of
Vocational Behavior, 40: 269–287.
Krackhardt, D., & Porter, L. W. 1985. When friends leave:
A structural analysis of the relationship between
turnover and stayers’ attitudes. Administrative Sci-
ence Quarterly, 30: 242–261.
Krackhardt, D., & Porter, L. W. 1986. The snowball effect:
Turnover embedded in communication networks.
Journal of Applied Psychology, 71: 50–55.
Kruglanski, A. W., & Mayseless, O. 1990. Classic and
current social comparison research: Expanding the
perspective. Psychological Bulletin, 108: 195–208.
Kulik, C. T., & Ambrose, M. L. 1992. Personal and situa-
tional determinants of referent choice. Academy of
Management Review, 41: 55–67.
Law, K. S., Wong, C., & Mobley, W. H. 1998. Toward a
taxonomy of multidimensional constructs. Acad-
emy of Management Review, 23: 741–755.
Lee, T. W. 1999. Using qualitative methods in organi-
zational research. Thousand Oaks, CA: Sage.
Lee, T. W., Mitchell, T. R., Sablynski, C. J., Burton, J. P.,
& Holtom, B. C. 2004. The effects of job embedded-
ness on organizational citizenship, job performance,
volitional absences, and voluntary turnover. Acad-
emy of Management Journal, 47: 711–722.
Maertz, C. P., & Campion, M. A. 1998. 25 years of volun-
tary turnover research: A review and critique. Inter-
national Review of Industrial and Organizational
Psychology, 13: 49– 81.
Maertz, C. P., Griffeth, R. W., Campbell, N., & Allen, D. G.
2007. The effects of perceived organizational sup-
port and perceived supervisor support on employee
turnover. Journal of Organizational Behavior, 28:
1059–1075.
March, J. G., & Simon, H. A. 1958. Organizations. New
York: Wiley.
Mathieu, J., & Kohler, S. 1990. A cross-level examination
of group absence influences on individual absence.
Journal of Applied Psychology, 75: 217–220.
Meyer, J. P., Allen, N. J., & Smith, C. A. 1993. Commit-
ment to organizations and occupations: Extension
and test of a three-component conceptualization.
Journal of Applied Psychology, 78: 538–551.
Mitchell, T. R., Holtom, B. C., Lee, T. W., Sablynski, C. J.,
& Erez, M. 2001. Why people stay: Using job embed-
dedness to predict voluntary turnover. Academy of
Management Journal, 44: 1102–1121.
Mitchell, T. R., & Lee, T. W. 2001. The unfolding model
of voluntary turnover and embeddedness: Founda-
tions for a comprehensive theory of attachment. In
B. M. Staw & R. I. Sutton (Eds.), Research in organ-
izational behavior, vol. 23: 189–246. Greenwich,
CT: JAI Press.
Mobley, W. H. 1977. Intermediate linkages in the rela-
tionship between job satisfaction and employee
turnover. Journal of Applied Psychology, 62: 237–
240.
Mossholder, K. W., Settoon, R. P., & Henagan, S. C. 2005.
A relational perspective on turnover: Examining
structural, attitudinal, and behavioral predictors.
Academy of Management Journal, 48: 607–618.
Pfeffer, J. 1991. Organization theory and structural per-
spectives on management. Journal of Management,
17: 789 – 803.
Price, J. L. 1977. The study of turnover. Ames: Iowa State
University Press.
Price, J. L., & Mueller, C. W. 1986. Absenteeism and
turnover of hospital employees. Greenwich, CT: JAI
Press.
Rousseau, D. M. 1985. Issues of level in organizational
research. In L. L. Cummings & B. M. Staw (Eds.),
Research in organizational behavior, vol. 7: 1–37,
Greenwich, CT: JAI Press.
Rumery, S. M. 2003. The influence of work group cul-
ture on employee turnover. Unpublished doctoral
dissertation, University of Connecticut, Storrs.
Salancik, G. R., & Pfeffer, J. 1978. A social information
processing approach to job attitudes and task de-
sign. Administrative Science Quarterly, 23: 224–
252.
Shaw, J. D., Gupta, N., & Delery, J. E. 2005. Alternative
conceptualizations of the relationship between
voluntary turnover and organizational perfor-
mance. Academy of Management Journal, 48:
50– 68.
Spector, P. E. 1985. Measurement of human service job
satisfaction: Development of the Job Satisfaction
Survey. American Journal of Community Psychol-
ogy, 13: 693–712.
Steel, R. P. 2002. Turnover theory at the empirical inter-
face: Problems of fit and function. Academy of Man-
agement Review, 27: 346–360.
Tesser, A., Campbell, J., & Mickler, S. 1983. The role of
social pressure, attention to the stimulus, and self-
doubt in conformity. European Journal of Social
Psychology, 13: 217–223.
Ulrich, D., & Smallwood, N. 2006. How leaders build
value: Using people, organization, and other in-
tangibles to get bottom-line results. Englewood
Cliffs, NJ: Wiley.
Van Dijk, P., & Kirk-Brown, A. 2003. The relationship
between job embeddedness and the intention to
leave an organization. Paper presented at the APS
560 JuneAcademy of Management Journal
Psychology of Relationships Interest Group, 3rd An-
nual Conference, Deakin University, Melbourne.
Wooten, D. B., & Reed, A. 1998. Informational influence
and the ambiguity of product experience: Order ef-
fects on the weighting of evidence. Journal of Con-
sumer Research, 7(1): 79–99.
Zatzick, C. D., & Iverson, R. D. 2006. Job embeddedness,
gender, and voluntary turnover: Further exten-
sions of a new construct. Paper presented at the
annual meeting of the Academy of Management, At-
lanta.
Will Felps (wfelps@rsm.nl) is an assistant professor at the
Rotterdam School of Management, Erasmus University.
He received his Ph.D. in management from the Univer-
sity of Washington, Seattle. His diverse topics of interests
are linked together by a desire to create more effective
and humane organizations through more valid and co-
herent theories.
Terence Mitchell (trm@u.washington.edu) is a professor
of management and organization and of psychology at
the Foster School of Business at the University of Wash-
ington, Seattle. He received his Ph.D. from the University
of Illinois. His current research interests are organization-
al attachment, motivation, and leadership.
David R. Hekman (hekman@u.washington.edu) is an as-
sistant professor of health care management at the Lubar
School of Business, University of Wisconsin–Milwaukee.
He earned his Ph.D. in management from the University
of Washington, Seattle. His research interests include
employee attachment to the organization, courageous
employee actions, subtle discrimination, and under-
standing why organizations fail.
Thomas W. Lee (orcas@u.washington.edu) is the Hughes
M. Blake Professor of Management and associate dean for
academic and faculty affairs at the Foster School of Busi-
ness, University of Washington, Seattle. He earned his
Ph.D. in management from the University of Oregon. His
primary research interests include employee loyalty, re-
tention and turnover, and work motivation.
Brooks C. Holtom (bch6@msb.edu) is an assistant profes-
sor of management in the McDonough School of Business
at Georgetown University. He received his Ph.D. in man-
agement from the University of Washington, Seattle. His
research interests include the attraction, development,
and retention of human and social capital.
Wendy S. Harman (wendysue@u.washington.edu) is vis-
iting assistant professor of business administration at Cen-
tral Washington University. She received her Ph.D. in man-
agement from the University of Washington Business
School. Her primary research interest focuses on using var-
ious methods of inquiry to positively impact work life.
APPENDIX
Job Embeddedness Scale, Short Form
a
1. My job utilizes my skills and talents well.
2. I feel like I am a good match for my organization.
3. If I stay with my organization, I will be able to achieve
most of my goals.
4. I really love the place where I live.
5. The place where I live is a good match for me.
6. The area where I live offers the leisure activities that I like
(sports, outdoor activities, cultural events & arts).
7. I have a lot of freedom on this job to pursue my goals.
8. I would sacrifice a lot if I left this job.
9. I believe the prospects for continuing employment with
my organization are excellent.
10. Leaving the community where I live would be very hard.
11. If I were to leave the community, I would miss my non-
work friends.
12. If I were to leave the area where I live, I would miss my
neighborhood.
13. I am a member of an effective work group.
14. I work closely with my coworkers.
15. On the job, I interact frequently with my work group
members.
16. My family roots are in this community.
17. I am active in one or more community organizations (e.g.,
churches, sports teams, schools, etc.).
18. I participate in cultural and recreational activities in my
local area.
19. Are you currently married?
20. If you are currently married, does your spouse work
outside the home?
21. Do you own a home (with or without a mortgage)?
a
Source: Holtom et al., 2006.
2009 561Felps, Mitchell, Hekman, Lee, Holtom, and Harman
