The Unfolding Model of Voluntary Turnover: A Replication and Extension

Author(s): Thomas W. Lee, Terence R. Mitchell, Brooks C. Holtom, Linda S. McDaniel and
John W. Hill
Source: The Academy of Management Journal , Aug., 1999, Vol. 42, No. 4 (Aug., 1999),
pp. 450-462
Published by: Academy of Management

Stable URL: https://www.jstor.org/stable/257015

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/257015?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to
The Academy of Management Journal

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
? Academy of Management Journal
1999, Vol. 42, No. 4, 450-462.

THE UNFOLDING MODEL OF VOLUNTARY TURNOVER:

A REPLICATION AND EXTENSION

THOMAS W. LEE
TERENCE R. MITCHELL
University of Washington

BROOKS C. HOLTOM
Vanderbilt University

LINDA S. McDANIEL
University of North Carolina

JOHN W. HILL
Indiana University

This work reports further theoretical development of Lee and Mitchell's (1994) u
folding model of voluntary turnover, which describes different psychological pat
that people take when quitting organizations. Ambiguities in the model were iden
fied, and hypotheses aimed at resolving these ambiguities were tested on a sample of
229 former employees from the "Big 6" public accounting firms. The results provide
theoretical and quantitative extension of an earlier qualitative assessment of th
unfolding model. Implications are discussed.

The intellectual roots for most of the current the-

Furthermore, it is informative to note how turn-
ory and research on voluntary turnover are March over studies are commonly conducted. At time 1,
and Simon's (1958: 84-111) ideas about the ease employees' attitudes and intentions are assessed; at
and desirability of movement job holders see them- time 2, those people who leave are recorded; and
selves as having. Typically, researchers have stud- then, statistical comparisons between stayers and
ied the perceived ease of movement as the per- leavers are made. Certainly, this common design
ceived number and type of job alternatives and yields valuable information about predictive rela-
have defined the perceived desirability of move- tionships; however, far less knowledge is gained
ment as an individual's level of job satisfaction. about the reasons and processes involved with
The empirical evidence indicates a modest rela- leaving. In addition to the common predictive
tionship between levels of satisfaction and turn- study, studies with different research designs are
over but an inconsistent relationship between the needed to explain better why and how people
perceived number and type of alternatives and leave. Perhaps most revealing about academic un-
turnover. More recently, the prevailing theory and derstanding of turnover is O'Reilly's (1991: 442)
research on turnover have focused on quitting in- comment that the empirical studies are "useful in-
duced by lower levels of job satisfaction, with thecrements" and his statement that the most interest-
intention to leave viewed as quitting's immediate ing attribute of this research is its methodological
antecedent. Meta-analyses have indicated that the rather than its substantive contribution. When pre-
proportion of variance shared by levels of satisfac- vailing theory, empirical results, study design, and
tion and turnover is 3.6 percent, and the proportion current incremental knowledge gains are consid-
shared by intention to leave and actual leaving is 12 ered together, new research directions and designs
percent (Hom & Griffeth, 1995: 40, 47). In our view, appear needed.
greater understanding of the turnover process than Several years ago, Lee and Mitchell (1994) argued
that suggested by these empirical findings is that it was time to develop alternative theory about
needed.
how and why people leave organizations. Drawing
on informal interviews with people who had left
We thank Tad Beckwith (Seattle Pacific University),their jobs and a comprehensive review of the extant
Susan Crandall (University of Washington), Dan Daltonturnover research, they proposed the unfolding
(Indiana University), and Rick Mowday (University ofmodel of voluntary turnover. Although individuals
Oregon) for their generous help on this study. experience unique circumstances when they leave
450

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
1999 Lee, Mitchell, Holtom, McDaniel, and Hill 451

organizations, they appear to follow one of four ples include unsolicited job offers, changes in mar-
psychological and behavioral paths when quitting. ital state, transfers, and firm mergers. People com-
In the only published test of the unfolding model, pare shocks and their surrounding circumstances
Lee, Mitchell, Wise, and Fireman (1996) demon- to their own images (that is, their values, goals, and
strated that these four paths completely described plans for goal attainment; see Beach [1997]) and, if
the leaving process for approximately 63 percent of the two are incompatible, thoughts of leaving oc-
their sample; conversely stated, 37 percent did not cur. Second, a script is a preexisting plan of action,
precisely fit the four theorized leaving processes. and it can be based on past experience, observation
The purpose of the study reported here was to of others, reading, or social expectations. Third,
suggest and empirically test several improvements search is the activities involved with looking for
to the unfolding model. alternatives to a current job and the evaluation of
those alternatives. Fourth, image violations occur
THEORY AND HYPOTHESES when an individual's values, goals, and strategies
for goal attainment do not fit with those of the
The Unfolding Model's Major Components and
employing organization or those implied by the
Paths
shock. Fifth, lower levels of job satisfaction occur
First, a shock is a particular, jarring event that when people come to feel over time that their jobs
initiates the psychological analyses involved in no longer provide the intellectual, emotional, or
quitting a job. A shock can be positive, neutral, or financial benefits they desire.
negative; expected or unexpected; and internal or Figure 1 depicts the unfolding model's four the-
external to the person who experiences it. Exam- orized paths. In path 1, a shock triggers the enact-

FIGURE 1

The Unfolding Model of Voluntary Turnover a b

Search and/or
Engaged Image Evaluation of
Shock Script Violation Satisfaction Alternatives Likely Offer Path

_- Yes* Yes*

Yes
m No 10- No --- 1

Yes
Yes -- 3
' Low Yes
' ~ _- NYes _o.
No*
No No* No*
No*
Yes*
Yes*
Irrelevant Ye

No
No 2

Yes*
Yes*

Yes 1 4b
Yes
Yes
IMI,No
No
No No*
K

No* NoYes *
No

No 0- 4a

a This figure includes the changes to the unfolding model

b An asterisk (*) indicates that the route is not classifiabl
could leave an organization that would not be part of one

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
452 Academy of Management Journal August

ment of a preexisting action plan or script. The a qualitative study that involved semistructured
person who has experienced the shock leaves with- interviews with 44 nurses who had recently quit
out considering his or her current attachment to the their jobs, and they found that these quits could be
organization and without considering alternatives. reliably classified by the essential features of each
Moreover, levels of job satisfaction are essentially decision path specified by the theory. All the es-
irrelevant in path 1. In path 2, a shock prompts the sential features of paths 1, 2, 3, 4a, and 4b were
person to reconsider her or his organizational at- found for 6, 6, 8, 7, and 6 nurses, respectively (33 of
tachment because image violations have occurred. 44). Thus, some evidence to support the unfolding
After completing these deliberations, the person model was obtained. However, the 11 nurses who
leaves without a search for alternatives. In path 3, a did not exactly fit a path pointed to potential con-
shock produces image violations that, in turn, ini- ceptual ambiguities in the model and became one
tiate the person's evaluation of both the current job basis for further theorizing. In addition, this initial
and various alternatives; thus, in path 3, leaving study highlighted several undefined gaps, which
typically includes search and evaluation. With are those issues and circumstances not addressed
path 4, lower levels of job satisfaction are the pre- or specified by the theory. Thus, the developers of
cipitator, instead of a shock. In path 4a, lower lev- the model deemed it necessary to identify and re-
els of satisfaction become so salient that people solve the conceptual ambiguities relating to (1)
leave without considering alternatives, but in path scripts and (2) alternatives and the theoretical gaps
4b, these lower levels explicitly lead to job search involved in (3) job offers as shocks and (4) individ-
and subsequent evaluation of alternatives. The uals' search and evaluation of alternatives.
other possible routes, identified in Figure 1 by as- Scripts. Originally, scripts about leaving in re-
terisks, do not define paths 1-4b and constitute sponse to a shock were theorized to affect quitting
potential falsifications of the unfolding model; in path 1 but were theorized to not affect quitting in
these routes are ways that people could leave orga-paths 2-4b. More specifically, leaving an organiza-
nizations that would not fall into one of the unfold-
tion under path 1 conditions was viewed as begin-
ing model's paths. ning with an event that prompts an individual to
Path 4b represents the process suggested by most consider quitting. As a result of this shock, the
turnover theories, whereby people leave because of person searches his or her memory for a preexisting
lower levels of job satisfaction, but only after they behavioral plan for leaving under these or very
have engaged in searching for, evaluating, and select- similar circumstances. If such a recollection is
ing alternative jobs; this outline constitutes a very found in memory, a match occurs, and if the co
rational choice process. In the unfolding model, the sponding script is deemed appropriate, it is
other paths suggest ways that people leave that have acted. If such a recollection is not found, a matc
not typically been discussed in the literature. First, does not occur, and another path might be evok
three paths (1, 2, and 3) are initiated by a distinct, That is, a shock can initiate leaving via paths 2
jarring, and recognizable event-a shock. Second, the 3 as well. Whereas the role of scripts appears
processes that describe three of the paths (1, 2, and sonably well specified in path 1, Lee and Mit
4a) do not involve search or comparisons between a (1994) did not unambiguously specify the natur
current and alternative jobs. Third, in one path (1), scripts in the other paths. In particular, they
people leave by following a scripted behavioral se- peared to equate the absence of a direct effec
quence. The descriptions of these paths (and their scripts on quitting in paths 2-4b with its nonex
combinations of psychological events and behaviors) ence.
are new to turnover research. Theoretically, the paths Recent theory and research on fram
suggest that leaving is far more complex than it has 1997: 21, 23) and "sensemaking" (Weic
typically been deemed. In particular, the model sug- 125, 129) have suggested that scripts
gests that variables other than lower levels of satisfac- widely held than was originally prop
tion can prompt leaving a job. The content of various unfolding model. More specifically, w
jarring events (shocks), for instance, may help explain rect effect may be limited to path 1,
who leaves, how quickly they leave, and why they theorized, the existence of scripts may
leave. ited to path 1. Instead, they may coexis
factors while paths 2, 3, 4a, and 4b
instance, an employee might have de
Conceptual Ambiguities, Theoretical Gaps, and
script about leaving in response to a sh
Suggested Improvements
decided not to enact it because the scrip
To date, only Lee and colleagues (1996) have with other scripts or because actual ena
directly tested the unfolding model. They reported too difficult or inappropriate. As Walle

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
1999 Lee, Mitchell, Holtom, McDaniel, and Hill 453

ell (1991) demonstrated, anticipating an event may ing an organization for an alternative meant becom-
be very different from actually experiencing it. ing an employee with another organization. Theory
Moreover, the empirical evidence, albeit quite lim- has failed to address alternative forms of work,
ited, appears to support this point. In Lee and col- such as purchasing a firm or returning to graduate
leagues' (1996) study, six nurses left under path 1 school. We also suggest expanding the meaning of
conditions (scripted quitting), but four other job alternative to include alternative forms of work
nurses, who were classified as quitting via paths 3 and nonwork options; furthermore, pursuing these
or 4, held scripts that did not affect their leaving. In nonjob alternatives implies search and evaluation
order to clarify the nature of scripts, we propose processes that are similar to those that occur with
that a script may exist in paths 2, 3, and 4 but that traditional job alternatives.
that script must not be engaged, or carried out, in
Job offer as shock. Lee and Mitchell (1994: 66-
those paths.
68) initially theorized that the shocking effect of an
Job alternatives. In its original form, the unfold-
unsolicited job offer or inquiry must occur via path
ing model posited that job alternatives are evaluated
3, because some mental comparison of the current
in path 3 (shock-induced search and evaluations)
job and the unsolicited offer was presumed to fol-
and in path 4b (lower job satisfaction-induced
low. No such deliberations were thought to occur
search and evaluations). Unfortunately, Lee and
Mitchell (1994) did not clearly specify the nature of for paths 1 (scripted leaving) and 2 (no search), and
alternatives. Like most turnover theories, the model path 4 is precipitated by lower levels of satisfaction
implied that an alternative was a person's holding instead of by a shock. However, an unsolicited job
some subjective probability, substantially greater offer or inquiry can also initiate a preexisting script
than zero, that another job could be obtained. Two with no other mental deliberations. For example,
problems arise. First, it is not clear whether this an employee could (1) receive an unsolicited offer,
probability must have a value of one or can be less (2) have a preexisting script in place (for example,
than one; in other words, need the person actually "They offered a job before. I turned it down, and I
have an offer in hand? Second, it is not clear what was sorry. If they offer again, I'll accept it"), (3)
constitutes an alternative. Can it be a general labor enact the script, and (4) quit. We suggest that an
market perception, a specific job, or a nonwork unsolicited job offer (or inquiry) can be a prompt-
option? Evaluating a job alternative appears to re- ing shock in path 1 as well as in path 3. It is
semble the process of judging the likelihood of a important to note that an enacted script must not be
unique event (Beach, 1997: 94-95). People first accompanied by an active search or evaluation of
gather information about the event and then sub- alternatives for leaving to be classified as path 1. If
jectively determine the probability of its occur- active search or evaluation occurs, leaving should
rence. In the unfolding model, they would deter- be classified as path 3.
mine whether the subjective probability of Search and evaluation. Lee and Mitchell (1994:
obtaining another job (or some nonwork option) is 66-69) initially theorized that job search and eval-
sufficient for them to take subsequent action (that uation of alternatives must occur together. How-
is, the probability exceeds a threshold value
ever, such coupling presumes that job search pro-
viewed as justifying expending effort on evaluating
duces an alternative to evaluate. It is quite possible,
specific attributes or as justifying quitting). In view
for example, to search for a job and find no alter-
of the available theory and research, it seems un-
natives to evaluate. If the shock itself is a job offer,
likely that a job alternative must have a subjective
there may be no search, but an evaluation of alter-
probability of 100 percent to prompt action. We
natives would likely occur. Moreover, the decision
suggest that a job offer need not be in hand for it to
affect the cognitive deliberations involved with to pursue nonjob options involves the same issues.
quitting; instead, an alternative to a current job Thus, search and evaluation should be decoupled.
need only be perceived as highly likely to prompt Either search or evaluation needs to occur in paths
subsequent action. Moreover, the limited empirical 3 and 4b. In sum, five modifications to the unfold-
evidence appears to support this suggestion. In Lee ing model that add richness and real-world com-
and colleagues (1996), several nurses quit their jobs plexity to our understanding of voluntary turnover
after conducting job searches, with no actual offers are proposed and were tested in the reported study.
in hand. They appeared to view getting new jobs as
quite likely and thus compared their likely alterna- Hypothesis 1. A higher proportion of individu-
tives to their current jobs. als leaving jobs will be classified into the the-
Like virtually all turnover theories, the unfolding orized paths using the revised unfolding model
model initially contained the assumption that leav- of voluntary turnover than the original model.

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
454 Academy of Management Journal August

Theoretical Extensions 4b during the second period because 4b involves

more elaborate mental deliberations in the forms of
Decision path speed. Lee and Mitchell (1994)
search and evaluations.
suggested that two factors affect the speed with
which decision paths unfold. First, the amount of
Hypothesis 2a. The duration between the
mental deliberation varies among paths. Whereasfirst thoughts of quitting and a decision to
paths 1 and 2 involve fewer mental processes (thatleave and the duration between a decision to
is, no comparison of alternatives), path 3 involves
leave and actual quitting will be shorter in
elaborate deliberations involving comparisons paths 1 and 2 than in paths 3, 4a, and 4b,
among alternatives. By inference, paths 1 and 2 and these durations will be shorter in path 3
should occur more quickly than path 3. Further- than in path 4b.
more, the mental deliberations involved with paths
4a and 4b are theorized to unfold gradually over Hypothesis 2b. The duration between a deci-
time. Thus, paths 4a and 4b should unfold more sion to leave and actual quitting will be shorter
slowly than paths 1 and 2. Second, the available in path 4a than in path 4b. No difference is
information on a path's components varies. In expected between paths 1 and 2 for both dura-
paths 1, 2, and 4a, shocks, evaluations, scripts, or tions and between paths 4a and 4b on the first
images are readily available for an individual's di- duration.
rect use in mental deliberations. In contrast, some
of the theorized components from paths 3 and 4b Shocks. In the earlier presentation of the model,
are less readily available for immediate use. These Lee and Mitchell (1994) did not clearly specify the
paths involve the (1) availability of jobs in the attributes of shocks for different paths. In the sub-
external labor market, (2) specific options found sequent presentation, however, Lee and coauthors
during a subsequent job search, and (3) complexity (1996) suggested that shocks could be distin-
in evaluating located alternatives. On the basis of guished as (1) expected or unexpected, (2) positive
availability of information alone, paths 1, 2, and 4a or negative, and (3) personal or organizational.
should occur more quickly than paths 3 or 4b. Those authors explored nondirectional associa-
Although the paths' durations were originally tions between paths and shock attributes. In con-
studied as aggregates, they can be meaningfully trast, the present study offers directional hypothe-
ses that are based on the nature of each shock-
separated into and measured as two sequential pe-
riods that vary in their levels of internal or external
induced decision path. In path 1, shocks oft
control. The first period is the time between an involve the larger, ongoing processes in an indivi
ual's life. For example, Lee and colleagues (199
employee's first thoughts of quitting and the sub-
reported separate cases of path 1 leaving in wh
sequent decision to leave. This period seems
an individual followed a spouse to another cit
largely, though not completely, under the individ-
(after the shock of the spouse's transfer) or beca
ual's cognitive control, as he or she is deciding
a full-time "stay-at-home" parent (after the shock
whether to leave. The second period, the time be-
giving birth). In these cases, quitting was only on
tween a decision to leave and actual leaving, is
part of a mosaic of issues. Thus, shocks that indu
likely affected by external forces; leaving may de-
scripted behavior (path 1) are expected to invo
pend upon the time it takes to enact a script for
issues that are more personal than organization
path 1, employment opportunities for paths 3 and
In path 2, leaving is theorized to be a response to
4b, or felt obligations for ongoing projects and cur-
event, usually an abhorrent one (Lee & Mitch
rent coworkers for path 2. The separation of paths
1994: 65-66). As a result, leaving is expected
into two periods does not affect our reasoning on
involve negative shocks. In path 3, leaving is char
comparisons involving path 1, 2, or 3. When the
acterized by its decidedly rational spirit. Job sea
nature of onset, shocks, amount of mental deliber-
or an unsolicited offer elicits comparisons betwe
ation, and availability of information are consid-
alternatives and the current position. Shocks t
ered together, the following might be expected:
induce such analytical deliberations (path 3) a
First, paths 1 and 2 occur more quickly than 3, 4a,
expected to involve more organizational and j
and 4b. Second, path 3 occurs more quickly than 4b
offer characteristics than personal issues.
because shocks add to the felt pressures to leave.
For paths 4a and 4b, this separation allows for an Hypothesis 3a. Path 1 will be positively related
additional hypothesis. Because there is no clear to the occurrence of personal shocks.
onset of this path, no difference is expected be-
tween paths 4a and 4b during the first period; in Hypothesis 3b. Path 2 will be positively related
contrast, path 4a should occur more quickly than to negative shocks.

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
1999 Lee, Mitchell, Holtom, McDaniel, and Hill 455

Hypothesis 3c. Path 3 will be positively related METHODS

to organizational and unsolicited job offer Procedures
shocks.
Context and sample. During the time our data
Job satisfaction. The effect of lower levels of
were collected, the Big 6 accounting firms, which
satisfaction on turnover may be one of the best-
provided our organizational context, experienced a
documented empirical relationships in the man-
variety of changes that could produce shocks, im-
agement literature. As a result, job satisfaction is a
age violations, and voluntary turnover among part-
major variable in virtually every turnover theory,
ners and managers. Mergers consolidated jobs and
including the unfolding model (Hom & Griffeth,
affected career opportunities. Litigation against
1995: 51-86). Unlike most turnover theories, how-
firms grew substantially and created various per-
ever, the unfolding model posits a varying role for
sonal and financial risks. Nonauditing services,
satisfaction across paths. In path 1, satisfaction is
like consulting, grew substantially, and this growth
seen as largely irrelevant. It has no systematic role affected individuals' career advancement. Thus,
in an individual's scripted behavior. In path 2, the
public accounting firms provided an attractive set-
precipitating shock is theorized to be negative, ting in which to test the modified unfolding model.
which is likely to induce lower levels of job satis-
Each Big 6 public accounting firm had generated a
faction. In path 3, shocks can be positive or nega-
list of all partners and managers who voluntarily
tive. In some cases, for example, a relatively satis-
left its offices in six major U.S. cities in the years
fied employee might leave for a more satisfying job.
1990-92. Approximately 6,000 voluntary leavers
In path 4, low satisfaction is seen as causing were listed. In early 1994, we obtained permission
thoughts and actions associated with leaving. It is
from the Big 6 firms to utilize this list for the
so low that departure occurs without a job search in
present study. From each firm's list, we randomly
4a or with search or evaluation in 4b. Thus, satis-
selected 245 individuals and sent them surveys in
faction should also be low for 4a and 4b. Thus, the
early 1995; a total of 1,470 questionnaires were
theoretical role and level for job satisfaction differ
sent. All surveys were anonymous, and stamped,
across paths.
self-addressed envelopes were provided. Of the
Hypothesis 4a. Job satisfaction will be higher in 1,470 surveys, 301 were returned (20.4%), which
path 3 than in path 2, 4a, or 4b. yielded a return rate comparable to that typically
found in accounting research. Although the firms
In addition to the mean differences across paths had listed all the survey recipients as voluntary
(Hypothesis 4a), a negative relationship between leavers, 63 individuals identified themselves as in-
image violations and levels of job satisfaction in voluntary leavers, and 9 said they were retirees.
paths 2-4b is also theorized. More specifically, im- These 72 people were dropped from the study. The
age violations are specified as having a direct and final and analyzable sample consisted of 229
negative effect on job satisfaction for paths 3, 4a, former Big 6 employees. Their average age was
and 4b. In contrast, image violations are specified 39.93 years (s.d. = 7.19); their average tenure was
as an irrelevant cause of job satisfaction in paths 1 8.08 years (s.d. = 4.93); 69 percent were men; and
and 2; however, image violations are theorized as 44.1 percent had college degrees, with 25.3 percent
negatively related to, though not a cause of, job of the total holding master's of business adminis-
satisfaction in path 2 (that is, the relationship is tration degrees, and all others having at least bach-
concomitant). In path 1, a relationship is not spec- elor's. These demographic values are similar to
ified; it can be negative, null, or positive. those reported in several recent surveys of public
Hypothesis 4b. Job satisfaction is negatively accountants (e.g., Hayes & Hollman, 1996).
correlated with image violation. Categorizing leavers. The key criterion used in
the present study was the investigators' classifica-
In summary, this research had three goals. First, tion of which, if any, decision path each former
we sought to clarify ambiguities in the original employee had followed. Originally, Lee and col-
unfolding model's constructs and paths. Second, leagues (1996) classified nurses into paths on the
we wanted to demonstrate how the constructs and basis of the presence or absence of essential fea-
paths were associated with other issues and con- tures in their leaving scenarios (Figure 1). However,
structs constituting separate and independent our modifications to the unfolding model required
converting those authors' initial classification
ideas. Finally, the results from an initial qualitative
study involving nurses (Lee et al., 1996) were ex- method into a tentative set of decision rules that
tended to a different population (accountants) us- any investigator who might wish to study the un-
ing a far more quantitative method. folding model could apply. Thus, 25 leavers were

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
456 Academy of Management Journal August

randomly selected, and three of the authors applied Some evidence of validity could be inferred from
tentative rules to their survey responses; only mi- empirical support for Hypotheses 2a and 2b, 3a, 3b,
nor inconsistencies were found among the three and 3c, and 4a and 4b, in which the paths are
participating authors. We then drew on these minor associated with other measures. Further evidence
inconsistencies to further modify our decision of validity could be inferred from correlations be-
rules and applied them to another randomly se- tween different items purporting to measure simi-
lected 25 leavers; 100 percent agreement among the lar constructs. Such correlations were calculated
three authors and an additional judge, a doctoral and indicated substantial validity. As illustrations,
student, was reached. These modified rules were some theoretically meaningful and statistically sig-
then applied to the entire sample. To help mini- nificant (p < .001) correlations included the fol-
mize potential for bias, we recruited a volunteer to lowing: (1) "Was the event expected?" had a corre-
replicate the classification judgments made by the lation coefficient of -.91 with "Was the event
four prior judges. This volunteer had no prior con- unexpected?"; (2) "How many acceptable altern
nection to the current study, the researchers, or the tive jobs did your search produce before you le
researchers' institutions, and he was explicitly in- the firm?" and "How many total job offers did y
structed to be skeptical, question broadly, and have before you left your former firm?" had a .5
identify inconsistencies. After reviewing all pub- correlation; and (3) "How compatible were you
lished materials on the unfolding model and hold- personal values/ethics with those of your form
ing discussions with the third author, the volunteer firm?" had a correlation of .64 with "How compat
independently replicated the classification process ible were your professional values/ethics with
and reached 100 percent agreement with the prior those of your former firm?"
judgments.
Classification of leavers into decision paths for
Measures for Hypotheses 2a-4b
Hypothesis 1. Our final classification rules, the
specific survey items that assessed the presence or Volition andpersonal characteristics. Althoug
absence of shocks, scripts, image violations, job the Big 6 firms had listed each respondent as
satisfaction, search, evaluation, and offers-the voluntary leaver, we included the yes/no quest
major features of the decision paths-and the "Did you voluntarily leave the firm?" Tenure (
items' response formats are listed in the Appendix. months), age (in years), gender, and highest degr
We made an explicit effort to write survey items earned and its year were assessed by one quest
that resembled Lee and colleagues' (1996) inter- each.
view questions. However, those researchers relied Speed of decisions. Two fill-in questions were
primarily on single questions in their interviews used to gather data to test Hypotheses 2a and 2b.
and then made multiple judgments based on the The first was "After your first thoughts about leav-
interview data; in contrast, we adopted multiple ing, how long did it take for you to make the final
questionnaire items when feasible. Because we fo- decision to leave the firm? Please specify, in terms
cused more on theory development than on falsifi- of months, weeks or days." The second question
cation, correct classification (or avoiding type II was "After you made the final decision to leave the
errors) was deemed more important than incorrect firm, how long did you stay on the job before you
nonclassification (avoiding type I errors). Thus, an actually left? Please specify, in terms of months,
essential feature for a decision path was deemed weeks or days." Number of days was used in the
present (1) if a respondent answered yes to one or analysis.
more of several questions written to assess the es- Shock characteristics. The specific shock char-
sential feature and (2) if the respondent's other acteristics relevant for Hypotheses 3a-3c were as-
answers did not contradict the prior yes response sessed with the following yes/no questions: (1)
or responses (that is, convergent and discriminant "Was the event expected?" (2) "Was the event un-
validity were demonstrated). If such evidence was expected?" (3) "Would you characterize the event
not found, the essential feature was deemed not as positive?" (4) "Would you characterize the event
present. For example, for survey responses to be as negative?" (5) "Would you characterize the event
classified as constituting a path 1 scenario, evi- as neither positive or negative?" (6) "Did the event
dence of a shock and an active script had to be involve purely personal issues (i.e., unrelated or
present, but no evidence for search, evaluations, orexternal to the job itself)?" (7) "Did the event in-
offers could be present. volve purely firm issues?" (8) "Did the event in-
Validity. Although the agreement among judges volve a combination of personal and firm issues?"
indicated reliable classification, the validity of (9) "Was an unsolicited job offer or inquiry the
these classifications was a more complicated issue. event that first led you to think seriously about

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
1999 Lee, Mitchell, Holtom, McDaniel, and Hill 457

leaving?" If a shock had not occurred, these ques- been classified into path 1 and two others, into path
tions were bypassed. 3, but missing data involving a single theoretical
Image violations. Image violations were deter- characteristic prevented their classifications. In
mined with the eight items shown in the Appendix sum, 13 respondents did not report evidence of
and were anchored by five-point Likert scales. image violations, and 4 did not report sufficient
Their alpha reliability was .82. information to classify them.
Job satisfaction. Levels of job satisfaction were The frequency distribution of this study's sample
determined with the 13 items shown in the Appen- into the unfolding model's paths suggests two ad-
dix and anchored by five-point Likert scales. For ditional points. First, we compared the frequency
the classification of persons into paths (Hypothesis distributions of the paths for our sample of accoun-
1), a response of 1 or 2 to any of the 13 questions tants and Lee et al.'s (1996) sample of nurses. Paths
was taken to mean at least some lower level of 1, 2, and 4a occurred significantly more often
satisfaction existed. In contrast, the measurement among nurses than among accountants (X2 = 7.11,
of job satisfaction for Hypotheses 4a and 4b was p < .01, x2 = 6.39, p < .02, and x2 = 9.34, p < .01,
made by averaging these 13 items. Thus, much respectively); in contrast, path 3 occurred signifi-
more information about job satisfaction was ac- cantly more often among accountants than among
cessed and analyzed for Hypotheses 4a and 4b than nurses (X2 = 10.94, p < .001); and no significant
for Hypothesis 1. Factor analysis with "varimax" differences were found for path 4b. These results
rotation indicated a single-factor structure; coeffi- indicate that different occupational groups have
cient alpha was .76, and an averaged composite different patterns of (or paths for) leaving their jobs.
was used in subsequent analyses (x = 3.39, s.d. Second, the sample sizes for paths 1, 2, and 4a
.51). among the accountants were somewhat small and
may have precluded meaningful statistical tests.
The statistical power for these tests ranged from
RESULTS virtually 0 to 20 percent. When significance levels
were reached, substantial effects could be inferred.
Hypothesis 1
Six persons were classified into path 1; 7 persons
Hypotheses 2a and 2b
into path 2; 136 into path 3; 8 into path 4a; and 55
into path 4b. In comparison to the original model The time between the first thoughts of quitting
(defined by the classification method used by Lee and the decision to leave and the time between the
and colleagues [1996]), our modified unfolding decision to leave and actual leaving were hypoth-
model (defined by the current changes to their clas- esized to be shorter for paths 1 and 2 than for paths
sification method) was hypothesized to result in 3, 4a, and 4b and also shorter for path 3 than for 4b.
classification of a higher proportion of individuals The time between the decision to leave and actual
into paths. Application of the original classification leaving was also hypothesized to be shorter for path
method to our current sample resulted in 143 clas- 4a than for path 4b. Because Lee and his coauthors
sifiable quits (62.5%), whereas the modifications (1996) found the time from first thoughts to actual
resulted in 212 classifiable quits (92.6%), a net quitting to be nonnormally distributed, we calcu-
increase of 69 cases (+30.1%, X2 = 59.63, p < .001). lated Q plots and Kolmogorov-Smirnov goodness-
Visual inspection indicated that the modification of-fit tests (KS) for our separate times and found
to scripts alone accounted for a net increase of 47 evidence of nonnormality (KS = 4.40, p < .001, and
cases (+20.5%, X2 = 24.31, p < .001), and the KS = 4.52, p < .001, respectively). Thus, paramet-
modification to job alternatives alone accounted for ric methods were inappropriate for our data.
a net increase of 18 cases (+7.9%, x2 - 3.17, p = To test the hypothesized differences in path
.07). Four other persons became classifiable be- speed, we used logistic regression models to com-
cause of modifications to both scripts and job alter- pare the decision paths' durations. In each regres-
natives. sion, the outcome variable was the occurrence of
Among the 17 individuals who could not be clas- one path (for instance, path 1, coded as 0) as op-
sified by the modified unfolding model, only aposed to that of another path (such as path 2, coded
single theoretical inconsistency in each case pre-as 1). The time between the first thoughts of quit-
cluded classification. In particular, 1 person wouldting and the decision to leave was entered first, and
have been classified into path 2, but no image vio- the time between the decision to leave and actual
lations were reported. Similarly, 12 others would
leaving was entered second. As hypothesized, th
have been classified into path 3, but again no imagefirst duration (between first thoughts of quittin
violations were reported. Two people would haveand the decision to leave) was significantly shorter

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
458 Academy of Management Journal August

in path 1 than in paths 4a and 4b (X2 = 2.77, p < Hypotheses 4a and 4b

.05, and X2 = 4.74, p < .01, respectively); it was
Job satisfaction was hypothesized to be higher in
significantly shorter in path 2 than in paths 4a and
path 3 (x = 3.39, s.d. = .52) than in paths 2 ( =
4b (X2 = 2.79, p < .05, and X2 = 5.37, p < .01,
3.34, s.d. = .44), 4a (x = 3.39, s.d. = .47), and 4b (x
respectively); and it was also significantly shorter
= 3.23, s.d. = .49). No hypothesis was offered for
in path 3 than in path 4b (X2 = 4.00, p < .05).
path 1 (k = 3.5, s.d. = .51). As hypothesized, job
Unexpectedly, paths 1 and 2 did not differ from
satisfaction was significantly higher in path 3 than
path 3 on duration. And the first duration was also
in path 4b (t = 1.95, p = .03, one-tailed test).
shorter for path 3 than for 4a (X2 = 2.71, p < .05), an
Contrary to hypothesis, job satisfaction was not
unhypothesized finding that may suggest the grad-
significantly higher in path 3 than in path 2 or 4a.
ual onset of path 4 leaving.
With the effect of the duration between the first In addition, the mean job satisfaction of the accoun-
tants who had outside offers (in path 3) was signif-
thoughts of quitting and the decision to leave held
icantly higher than the job satisfaction of those who
constant, as hypothesized, in path 3 the second
did not have offers (t = 2.60, p < .01, two-tailed
duration (between the decision to leave and actual
test). Job satisfaction was also hypothesized to cor-
leaving) was significantly shorter than it was in
relate negatively with image violation (Hypothesis
path 4b (X2 = 2.94, p < .05). Unexpectedly, paths 4a
4b). As expected, the bivariate correlations between
and 4b did not differ for the second duration and,
job satisfaction and the eight image items were all
although the finding was not hypothesized, path 3
statistically significant (p < .001) and negative in
unfolded more quickly than 4a (X2 = 3.79, p < .05),
direction, and they ranged from -.42 to -.24. The
which may also suggest the gradual onset of path 4
adjusted coefficient for the multiple correlation be-
leaving; but, as expected, differences between
tween job satisfaction and these eight items (R2)
paths 1 and 2 and 4a and 4b were not significant.
was .27 (p < .001).

Hypothesis 3a-3c
DISCUSSION
It was hypothesized (1) that path 1 would be
positively related to personal shocks, (2) that path 2 In the present study, we sought further con
would be positively related to negative shocks, and tual refinement of the unfolding model of vol
(3) that path 3 would be positively related to orga- turnover. Specific modifications to the unf
nizational and job offer shocks. Using two-by-two model were proposed and tested via Hypoth
contingency tables, we cross-tabulated the absence These modifications resulted in a significa
or presence of each decision path (coded 0 and 1, increase of 30.1 percent in the classification o
respectively) with the absence or presence (coded 0 leavers. In particular, changes concerning
and 1) of each of the hypothesized shock character- alone accounted for a net increase of 20.5 p
istics. To gain additional insights about shocks, we and changes concerning job alternatives alo
content-analyzed respondents' written comments counted for another net increase of 7.9 perce
using an open coding scheme (Lee, 1999: 48). As addition, the present study provided theo
extensions to the unfolding model. As impl
hypothesized, path 1 was positively related to per-
sonal shocks (X2 = 3.11, p = .08, 4 = .17). More- not specified by the unfolding model, paths 1
over, path 1 respondents reported experiencing one occurred significantly more quickly than path
merger and five personal (family) shocks. Contrary 4b, and path 3 occurred more quickly than 4b
to hypothesis, path 2 was not significantly related potheses 2a and 2b); different shock character
to negative shocks. Path 2 respondents reported were associated with each path (Hypothe
five organizational shocks (such as mergers, trans- 3c); and, finally, job satisfaction was higher i
fers, and colleagues' dismissals) and two personal 3 than in path 4b and negatively correlated
shocks, such as family problems or liability suits. image violations (Hypotheses 4a and 4b). O
As hypothesized, path 3 was positively related to whole, these results appear to extend our u
organizational shocks (X2 = 7.67, p = .006, 4 = .27) standing of the unfolding model beyond that
and unsolicited offer shocks (X2 = 3.42, p = .06, and colleagues (1996). The two studies sugge
4 = .18). Path 3 respondents reported experiencing people use different, distinct, and systematic
58 organizational shocks, 65 job offer shocks, and chological processes, or paths, when leaving o
13 personal shocks. izations.

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
1999 Lee, Mitchell, Holtom, McDaniel, and Hill 459

Theoretical Implications our classification rate significantly. Furthermore,

inspection of Figure 1 shows 17 possible paths.
At this point, a pause to consider some broader
Five of these constitute the unfolding model (29%),
theoretical issues would be timely. Particular ques-
but 12 (71%) describe processes that constitute fal-
tions to address are, how does the unfolding model
sification. As far as we could tell, none of these 12
contribute to turnover theory, what do scholars
paths occurred in the accountant data. Although
gain in understanding from it, and how useful is it
falsification was actively sought, no such evidence
to managers? Five criteria for judging a theory are
emerged.
widely accepted. First, a good theory's statements
To date, the two empirical studies on the unfold-
can be judged for internal consistency and parsi-
ing model have involved retrospective self-reports.
mony. Second, the theory must be falsifiable.
In particular, our study's turnover events occurred
Third, scholarly understanding should be en-
between 1990 and 1992, and our surveys were sent
hanced. That is, the theory must help make sense of
in 1995. Recall bias could have operated. However,
observations of the empirical world. Fourth, the three different research streams offer indirect evi-
theory should help in the control and management dence that the likelihood of recall bias in turnover
of behavior. Finally, the theory should help inves-
studies is relatively low. First, research strongly
tigators predict when and where the theorized be-
indicates that events held in episodic memory
haviors or phenomena will occur. mental structures are accurately remembered
One might argue that this study added too many (Wheeler, Stuss, & Tulving, 1997). To the extent
questions and that the unfolding model has become that leaving an organization (particularly, shock-
overly complex. However, a review of the model's induced leaving) can be considered a major and
development through its original specification and vivid personal event, the memory of that leaving
initial test (Lee & Mitchell, 1994; Lee et al., 1996)
should reside in episodic memory and be accu-
suggests that such is not the case. First, four of the rately recalled. Second, the research on self-based
constructs stayed the same: shocks, scripts, image referencing structures clearly demonstrates both
violation, and satisfaction. Search and evaluation
that such memories are accurate and that they grow
were decoupled, but the ideas underlying them increasingly accurate over time (Symons & John-
remained essentially the same. In both the qualita- son, 1997). To the extent that an individual's leav-
tive and quantitative tests of the model, no new ing an organization is self-based, or voluntary, the
constructs emerged. Second, the overall number of memory of that leaving should be accurate. Third,
questions (items) asked in the two empirical stud- recent reanalysis of data that assessed the accuracy
ies were comparable. More specifically, the actual of recall-based studies in organizational research
numbers of items used for each construct were as
suggests that such retrospective designs are not
follows (the first number is for nurses, the second necessarily
is biased and remain a viable research
for auditors): shock, 8 versus 13; scripts, 4 versus strategy
3; (Miller, Cardinal, & Glick, 1997). Albeit
image violations, 11 versus 8; job satisfaction, not indirectly, the accumulating evidence from three
measured versus 13; search, 5 versus 3; evaluation, independent research streams on recall bias leads
2 versus 3; and offers, 2 versus 6. More questions to the conclusion that individuals' own memories
about shocks, evaluation, and offers, but fewer of their quitting organizations are most often accu-
questions about scripts, image violations, and rate.
search, were asked. Recall that many of the ques- Although recall bias appears of little concern, the
tions about a given construct correlated with one two studies on the unfolding model were retrospec-
another as expected. Given no new constructs, rel- tive and not predictive. In the future, more tradi-
atively few additional questions, and evidence of tional predictive studies are needed for additional
convergent validity for multiple questions about a validation of the unfolding model. For instance,
construct, this study's representation of the unfold- researchers could collect data on the unfolding
ing model shows substantial fidelity to its original model's constructs at time 1 and then statistically
specification and, although complex, the extended compare their predictive ability to discriminate be-
unfolding model is parsimonious. tween stayers and leavers at time 2.
The model is also falsifiable. As mentioned, no
new constructs were added that changed the paths
Implications for Management
or underlying theory. In addition, although Lee and
colleagues' (1996) nursing study uncovered a num- The unfolding model enhances understanding
ber of anomalies and ambiguities, no such prob- and prediction of turnover in at least three ways.
lems emerged with the accounting study. By asking First, relatively few constructs from the unfold-
more precise questions, we were able to increase ing model are needed to make accurate classifi-

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
460 Academy of Management Journal August

cations. For example, shock (versus no shock) quired to leave firms, and the different roles of job
places people into path 1, 2, or 3, ruling out path satisfaction in quitting. Although substantial re-
4; search (including evaluation and offers) places search remains to be done, the unfolding model's
people into path 1 or 2, ruling out path 3, or into basic foundation seems sound.
path 4a or 4b; and script (versus no script) places
people into path 1 or 2. Second, the empirical
tests of the unfolding model inform judgments REFERENCES
about how and why people leave. Among accoun-
tants and nurses, for instance, more people re- Beach, L. R. 1997. The psychology of decision m
ported leaving because of a shocking event than Thousand Oaks, CA: Sage.
because of lower levels of satisfaction. Third, the Hayes, R. D., & Hollman, K. W. 1996. Managing dive
current study confirmed that different paths take Accounting firms and female employees. CPA J
different lengths of time to unfold. For example, nal, 66: 36-39.
leaving by paths initiated by a shock took less Horn, P. W., & Griffeth, R. W. 1995. Employee turnover.
time than leaving by paths initiated by lower Cincinnati: South/Western.
levels of satisfaction.
Lee, T. W. 1999. Using qualitative methods in organi-
Although managers may not be able to control zational research. Thousand Oaks, CA: Sage.
many of the external shocks employees sustain, it
Lee, T. W., & Maurer, S. D. 1997. The retention of knowl-
may be advisable for organizations to develop
edge workers with the unfolding model of voluntary
systems that allow quick responses, such as of-
turnover. Human Resource Management Review,
fers of new internal jobs, telecommuting, or job 7: 247-275.
sharing. Theoretically, the unfolding model sug-
gests multiple and independent ways in whichLee,
a T. W., & Mitchell, T. R. 1994. An alterna
proach: The unfolding model of voluntary em
researcher can assess which path a person takes
turnover. Academy of Management Review
and when that path will be initiated and com- 51-89.
pleted. Pragmatically, if certain paths take longer
Lee, T. W., Mitchell, T. R., Wise, L., & Fireman, S. 1
to occur than others, a manager has more time to
An unfolding model of voluntary employee
respond to those leaving processes. In addition,
over. Academy of Management Journal, 39:
knowledge about common scripts may offer in-
formation about who will leave quickly, perhaps March, J. G., & Simon, H. A. 1958. Organization
requiring quicker organizational response, and York: Wiley.
who will leave more slowly, perhaps permittingMiller,
a C. C., Cardinal, L. B., & Glick, W. H. 1997. Re
slower response. For example, managers might spective reports in organizational research: A r
focus on anticipated events before they happen. amination of recent evidence. Academy of Man
ment Journal, 40: 189-204.
When unemployment is low and jobs are plenti-
O'Reilly, C. A. 1991. Organizational behavior: Where
ful, for instance, more people may leave via path
3 (that is, unsolicited job offers or inquiries). we've been, where we're going. In M. R. Rosenzweig
When negative events happen (such as downsiz- & L. W. Porter (Eds.), Annual review of psychology,
vol. 17: 427-458. Palo Alto, CA: Annual Reviews.
ing or a lack of annual bonuses), people may
leave via path 2. If most employees are young and
Porter, L. W., & Steers, R. M. 1973. Organizational, work,
married, some may leave via path 1 (for instance, and personal factors in employee turnover and ab-
after the birth of a new child). Alternatively senteeism. Psychological Bulletin, 80: 151-176.
stated, managers might identify employees who Symons, C. S., & Johnson, B. T. 1997. The self-reference
are at high risk of leaving and then provide feed- effect in memory: A meta-analysis. Psychological
back, career counseling, or counteroffers when Bulletin, 121: 371-394.
needed. Lee and Maurer (1997) described several Waller, W. S., & Mitchell, T. R. 1991. Conditional prob-
common interventions that managers and organi- ability judgments: Effects of anticipating vs. experi-
zations might consider to discourage or encour- encing the conditioning event. Organizational Be-
age employees' leaving based on the unfolding havior and Human Decision Processes, 49: 302-
model. 325.

This study demonstrates that the unfolding Weick, K. E. 1995. Sensemaking in organizations.
model is a viable and an important approach to Thousand Oaks, CA: Sage.
understanding and studying voluntary turnover. It Wheeler, M. A., Stuss, D. T., & Tulving, E. 1997. Toward
can expand scholars' and managers' views of how a theory of episodic memory: The frontal loges and
and why people leave organizations by adding in- autonoetic consciousness. Psychological Bulletin,
formation on the nature of shocks, the time re- 121: 331-354.

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
1999 Lee, Mitchell, Holtom, McDaniel, and Hill 461

APPENDIX surance coverage, retirement plans, sick leave, family

leave)? 2. At your former firm, how satisfied were you
Items, Classification Rules, and Response Formats
with the work environment related to: (9) generating new
Shock client business, (10) competitive pressures, (11) auton-
omy of the work, (12) pressures at work, (13) time flexi-
An appropriate answer to at least one of the following bility?
indicated shock. Responses were open ended, except for
those for items 1, 3, and 4, which were yes/no.
1. Was there a single particular event that caused you Search
to think about leaving? 2. Please describe that event. 3. If
you accepted a job offer you had in hand, was it origi- An appropriate answer to at least one of the following
nally an unsolicited offer or inquiry? (Please answer onlyindicated search. Items 1 and 2 had yes/no responses,
if you had a job offer in hand.) 4. Was there a particular and item 3 was answered on a Likert scale (1, no search,
event or series of particular events related to litigation to 5, very comprehensive search).
that influenced your decision to leave? If yes, please 1. Did you have at least one job offer in hand when you
describe briefly. decided to leave? 2. If you didn't have a job offer in hand
when you actually left, did you believe that getting an
offer was very likely? 3. Before you left the firm, how
Script comprehensive was your job search for another job (e.g.,
did you gather lots of information on other job opportu-
An appropriate answer to at least one of the following
nities or search on a daily basis)?
indicated an engaged script. Responses were yes/no for
item 1 and on a five-point Likert scale for items 2 and 3
(1, strongly disagree; 5, strongly agree). Evaluation
1. If you accepted a job offer you had in hand, was it
originally an unsolicited offer or inquiry? (Please answer An answer of yes to at least one of the following
only if you had a job offer in hand.) 2. I have left a job indicated evaluation of job alternatives. 1. After your fir
before for essentially the same reasons (i.e., very similar thoughts about leaving, did you evaluate any specific job
circumstances). 3. At the time I left my job, I had already alternatives before deciding to leave? 2. After your firs
determined that I would leave the firm IF a certain event thoughts about leaving, did general job availability affec
were to occur (e.g., being accepted to graduate school). your decision to leave (e.g., you were pretty sure yo
could get another job, though you didn't have a specific
job in mind)? 3. In making your final decision to leave,
Image Violation did you seriously consider nonwork options (e.g., stayin
at home, returning to school, taking a sabbatical)? If yo
An answer of 1 or 2 to at least one of the following
responded yes, please indicate the type of nonwork op-
indicated violation. Responses for items 1-4 ranged from
tion you actually pursued.
1, not compatible, to 5, compatible, and those for items
5-8 were from 1, strongly disagree, to 5, strongly agree.
For Hypothesis 3, these items were reverse-coded.
Job Offers
1. How compatible were your personal values/ethics
with those of your former firm? 2. How compatible were An appropriate answer to at least one of the following
your professional values/ethics with those of your former indicated offers. Items 1-5 were answered yes/no, an
firm? 3. How compatible were your personal goals with items 6 and 7 were filled in.
those of your former firm? 4. How compatible were your 1. Was an unsolicited job offer or inquiry the event that
professional goals with those of your former firm? 5. If I first lead you to think seriously about leaving? 2. Did you
had stayed, I would have been able to achieve most of my have at least one job offer in hand when you decided to
career goals. 6. If I had stayed, I would have been able to leave? 3. Did you ultimately accept a job offer that you
achieve most of my personal goals. 7. At my former firm, had in hand? (Please answer only if you had a job offer in
my career was progressing as I expected. 8. At my former hand.) 4. If you accepted a job offer you had in hand, was
firm, my personal goals were progressing as I expected. it originally an unsolicited offer or inquiry? (Please an-
swer only if you had a job offer in hand.) 5. If you didn't
have a job offer in hand when you actually left, did you
Job Satisfaction
believe that getting an offer was very likely? 6. How many
A 1 or 2 answer to at least one of the following indi- acceptable alternative jobs did your search produce be-
cated dissatisfaction. Responses for all items ranged from fore you left your former firm? How many total job offers
1, very dissatisfied, to 5, very satisfied (a = .76). did you have before you left your former firm?
1. At your former firm, how satisfied were you with:
(1) the supervision you received, (2) firm as an employer, Thomas W. Lee is a professor of human resource man-
(3) career opportunities, (4) financial rewards, (5) your agement and organizational behavior at the University of
coworkers, (6) nature of the work, (7) recreational activ- Washington. He earned his Ph.D. in organizational stud-
ities, (8) fringe benefits (e.g., vacation, holiday time, in- ies at the University of Oregon. His current research

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms
462 Academy of Management Journal August

interests include employee turnover, staffing, and quali- clude employee attitudes, voluntary turnover, and alter-
tative and quantitative methods. native employment arrangements.
Terence R. Mitchell is the Edward E. Carlson Professor of Linda S. McDaniel is an associate professor of account-
Business Administration and a professor of psychology ing at the University of North Carolina at Chapel Hill.
at the University of Washington. He earned his Ph.D. in She earned her Ph.D. in accounting at the University of
social psychology at the University of Illinois. His cur- Michigan. Her current research interests include the ef-
rent research interests include employee turnover, em- fects of environmental factors on auditors' judgments
ployee motivation, and decision making. and the effects of accounting policies on analysts' deci-
sions.
Brooks C. Holtom is a visiting assistant professor at the
Owen Graduate School of Management, Vanderbilt Uni- John W. Hill is an associate professor of accounting at
versity. He received his doctorate in human resource Indiana University. He earned his Ph.D. in accounting at
management and organizational behavior from the Uni- the University of Iowa. His current research interests
versity of Washington. His current research interests in- include auditor liability and auditor turnover.

This content downloaded from

136.50.62.222 on Wed, 20 Jan 2021 02:38:53 UTC
All use subject to https://about.jstor.org/terms