Why High and Low Performers Leave and What They Find Elsewhere: Job Performance Effects on Employment Transitions

Abstract

Little is known about how high and low performers differ in terms of why they leave their jobs, and no work examines whether pre-quit job performance matters for post-quit new-job outcomes. Working with a sample of approximately 2,500 former employees of an organization in the leisure and hospitality industry, we find that the reported importance of a variety of quit reasons differs both across and within performance levels. Additionally, we use an ease-of-movement perspective to predict how pre-quit performance relates to post-quit employment, new-job pay, and new-job advancement opportunity. Job type, tenure, and race interacted with performance in predicting new-job outcomes, suggesting explanations grounded in motivation, signaling, and discrimination in the external job market.

Full text

Cornell University ILR School
DigitalCommons@ILR
CAHRS Working Paper Series Center for Advanced Human Resource Studies
(CAHRS)
6-26-2007
Why High and Low Performers Leave and What
They Find Elsewhere: Job Performance Effects on
Employment Transitions
Charlie O. Trevor
University of Wisconsin-Madison
John P. Hausknecht
Cornell University, jph42@cornell.edu
Michael J. Howard
Harrah's Entertainment
This Article is brought to you for free and open access by the Center for Advanced Human Resource Studies (CAHRS) at DigitalCommons@ILR. It
has been accepted for inclusion in CAHRS Working Paper Series by an authorized administrator of DigitalCommons@ILR. For more information,
please contact jdd10@cornell.edu.
Please take our short DigitalCommons@ILR user survey.
Trevor , Charlie O. ; Hausknecht , John P. ; and Howard , Michael J., "Why High and Low Performers Leave and What They Find
Elsewhere: Job Performance Effects on Employment Transitions" (2007). CAHRS Working Paper Series. Paper 466.
http://digitalcommons.ilr.cornell.edu/cahrswp/466

CAHRS at Cornell University
615B Ives Hall
Ithaca, NY 14853-3901 USA
Tel. 607 255-9358
www.ilr.cornell.edu/CAHRS
WORKING PAPER SERIES
Why High and Low Performers
Leave and What They Find
Elsewhere: Job Performance
Effects on Employment
Transitions
Charlie O. Trevor
John P. Hausknecht
Michael J. Howard
Working Paper 07 – 11

Why High and Low Performers Leave CAHRS WP07-11
Page 2 of 42
Why High and Low Performers Leave and
What They Find Elsewhere:
Job Performance Effects on Employment
Transitions
Charlie O. Trevor
University of Wisconsin-Madison
John P. Hausknecht
Cornell University
Michael J. Howard
Harrah’s Entertainment
June 2007
http://www.ilr.cornell.edu/cahrs
This paper has not undergone formal review or approval of the faculty of the ILR School. It is
intended to make results of Center research available to others interested in preliminary form to
encourage discussion and suggestions.
Most (if not all) of the CAHRS Working Papers are available for reading at the Catherwood
Library. For information on what’s available link to the Cornell Library Catalog:
http://catalog.library.cornell.edu if you wish.

Why High and Low Performers Leave CAHRS WP07-11
Page 3 of 42
Abstract
Little is known about how high and low performers differ in terms of why they leave their
jobs, and no work examines whether pre-quit job performance matters for post-quit new-job
outcomes. Working with a sample of approximately 2,500 former employees of an organization
in the leisure and hospitality industry, we find that the reported importance of a variety of quit
reasons differs both across and within performance levels. Additionally, we use an ease-of-
movement perspective to predict how pre-quit performance relates to post-quit employment,
new-job pay, and new-job advancement opportunity. Job type, tenure, and race interacted with
performance in predicting new-job outcomes, suggesting explanations grounded in motivation,
signaling, and discrimination in the external job market.
KEYWORDS: employee turnover, job performance, ease of movement

Why High and Low Performers Leave CAHRS WP07-11
Page 4 of 42
Why High and Low Performers Leave and What They Find Elsewhere:
Job Performance Effects on Employment Transitions
Employee job performance is vital to how successfully organizations compete.
Consequently, job performance-specific employment transitions into and out of organizations
are of particular importance. Despite the importance of such resource flows, we still know
relatively little about why high and low performers leave, and virtually nothing about what these
very different types of employees eventually find in the external market.
The criticality of retaining talent is chronicled in the recent popular press (e.g., Lavelle,
2003), as well as in the academic literature (e.g., Allen & Griffeth, 2001; Campion, 1991; Lee,
Mitchell, Sablynski, Burton, & Holtom, 2004; Mossholder, Bedeian, Norris, Giles, & Feild, 1988;
Trevor, 2001; Trevor, Gerhart, & Boudreau, 1997; Williams & Livingstone, 1994). Simply stated,
because high performers are more likely to facilitate organizational success, researchers and
professionals are increasingly focused on their retention. Given the organizational performance
implications, targeting retention efforts toward high performers is a reasonable strategy, yet
requires first that we actually know what is important to high performer retention. Clearly,
research on voluntary turnover has identified an enormous list of factors that, on average, make
it more likely that employees will quit. Yet, several turnover studies of interactions that involve
measures of performance or talent indicate that the conventional wisdom regarding the general
effects of turnover antecedents does not apply across all levels of employee value. Hence,
because we know far less about why high performers leave their organizations than we know
about turnover rationale in general, performance-specific turnover rationale remains an open,
and considerably important, question.
In addition, little is known about what high and low performers find once they have left,
yet this issue has considerable implications for the individuals themselves, for the organizations
that hire these high and low performers, and for understanding how human resources of
disparate value progress through the multi-employer career paths that are so prevalent today.
The vast majority of turnover research ends with the turnover behavior itself or with post-quit

Why High and Low Performers Leave CAHRS WP07-11
Page 5 of 42
investigation of why people left (e.g., Lee, Mitchell, Wise & Fireman, 1996). Of the few studies to
track leavers to new organizations (e.g., Boswell, Boudreau, & Tichy, 2005; Brett & Stroh, 1997;
Dreher & Cox, 2000; Lam & Dreher, 2004; Wilk, Desmarais, & Sackett, 1995), none have
investigated the role of pre-quit job performance as a substantive predictor of post-quit
employment outcomes. Thus, our focus here is twofold: (1) the relationship between job
performance and the reasons why people quit; and (2) the relationship between job
performance and subsequent external market outcomes.
Theory and Hypotheses
To examine job performance’s role in employee transition out of one organization and
into another, we conducted a post-exit study of leavers from a single organization. In the first
half of our study, this allowed us to study retrospective perceptions of several potential quit
reasons. Although there are disadvantages to studying leavers retrospectively (many of which
were negated by our organization’s approach to post-exit data collection), prior retrospective
work on leavers reveals that method’s potential for extensive examination of a variety of factors
in quit behavior (e.g., Lee et al., 1996). Indeed, we were able to assess more potential quit
reasons than predictive studies can usually accommodate; this breadth is important, as
contingent pay is currently the only specific workplace condition researchers have identified as
of particular importance to high performer turnover. For the second half of our study, the sample
of leavers allows us to be the first to investigate (old-job) performance effects on external job
market outcomes; few studies have tracked employees from one organization to another, and
none have studied job performance effects on the post-exit fates of leavers.
Performance-Specific Voluntary Turnover
The most well-documented relationship in the performance-turnover literature is the
negative linear association between the two constructs. Four meta-analyses (Bycio, Hackett, &
Alvares, 1990; Griffeth, Hom, & Gaertner, 2000; McEvoy & Cascio, 1987; Williams &
Livingstone, 1994) support this finding of low performers being more likely than high performers

Why High and Low Performers Leave CAHRS WP07-11
Page 6 of 42
to leave organizations. When analyses are not constrained to linear relationships, evidence
also indicates that performance can be curvilinearly linked to turnover (Salamin & Hom, 2005;
Trevor et al., 1997; Williams & Livingstone, 1994), with both low performers and high performers
being more likely to quit.
However, it is the moderation of the relationship between performance and turnover that
speaks to the value of examining performance-specific quit reasons. Such interdependencies
suggest that there are factors that affect the turnover of high performers differently than they
affect the turnover of low performers (or, equivalently, that performance’s relationship with
voluntary turnover depends on the level of some third variable). Most notably, several
researchers have found that pay contingencies moderate the performance-turnover relationship
(Griffeth et al., 2000; Harrison, Virick, & William, 1996; Salamin & Hom, 2005; Trevor et al.,
1997; Williams & Livingstone, 1994), with contingent pay meaning little to low performer
turnover but substantially reducing the likelihood that high performers will leave. Two additional
studies are consistent with this theme. Lee et al. (2004) found that on-the-job embeddedness
(i.e., attachment to the organization that is driven by on-the-job elements), a construct that
correlated at .73 with job satisfaction, was more likely to be associated with a turnover reduction
when performance was high, rather than low. Similarly, Trevor (2001), who studied human
capital rather than performance, reported that job satisfaction effects on quitting were stronger
for those employees high in cognitive ability, education, and vocation-specific training (who,
presumably, would tend to be better performers). Taken together, these several studies
suggest that the effects on turnover of contingent pay, and of at least some of the factors that
drive job satisfaction and job embeddedness perceptions, are likely to be larger when
employees are more valuable.
This suggestion, however, also indicates how little we actually know about the context of
performance-specific turnover. Because job satisfaction and job embeddedness are broad
multidimensional perceptions, their moderating roles may not tell us about the role of any single

Why High and Low Perform
ers Leave CAHRS WP07-11
Page 7 of 42
contributing source of the perceptions; that is, multi-dimensional measures can mask the effects
of their specific components or facets (e.g., Campbell & Campbell, 2003). Thus, the research
cited above identifies only contingent pay as a specific workplace condition that is more
important to the retention of high performers. Consequently, in this study we attempt to expand
our knowledge of actual workplace elements that should play a larger role in quitting when
performance is high. In contrast, however, we also argue that other workplace factors should be
more important quit reasons when performance is low, which has meaningful implications as
well. To attempt to understand which workplace elements matter more for high performer
turnover and which matter more for low performer turnover, we examine a variety of specific quit
reasons in a single study. Our general approach to performance-specific quit reasons (as well
as to the performance-specific external market outcomes that we address later) is outlined in
Figure 1.
Reasons of Greater Importance to High Performers than to Others
We first examine quit reasons that, like pay contingency, should matter more when job
performance is high. Our general argument is rooted in equity theory: employees with higher
levels of performance inputs should, from a fairness perspective, expect to receive higher levels
of appropriate employment outcomes (Adams, 1965). Disproportionately low levels of these
outcomes will result in perceptions of inequity for high performers; inequity perceptions can then
lead to dissatisfaction and turnover (Aquino, Griffeth, Allen & Hom, 1997). In organizations
where reward distribution rules favor an equity norm (vs. equality or need; Deutsch, 1975), high
performers will be the most deserving of, and expectant of, high levels of employment
outcomes. Thus, they should also tend to be the most sensitive to (i.e., most influenced by)
these outcomes when considering whether to stay or leave. Hence, employment outcomes
commonly associated with performance inputs should be particularly relevant to the turnover
decisions of high performers.

Why High and Low Performers Leave CAHRS WP07-11
Page 8 of 42
Figure 1
Job Performance Effects on Elements of Employment Transition
Job
Performance
Quit Reason Importance External Market Outcomes
Pay
Advancement Opportunity
Supervisor, Misunderstand Job,
Job Demands, Absenteeism Policy
Pulled toward a better job,
rather than pushed away
Post-Quit Employment
New-Job Pay
New-Job Advancement
Note. This represents the main effects in the study; moderation of certain performance effects by tenure, job level, race,
and gender is also examined.
+
-
+
+
+
+

Why High and Low Performers Leave CAHRS WP07-11
Page 9 of 42
Most obvious among these outcomes is pay. Indeed, such pay components as
commissions (Harrison et al., 1996), bonuses (Salamin & Hom, 2005), and pay growth (Trevor
et al., 1997) have been found to have the greatest effect on voluntary turnover when
performance was high. Other aspects of pay also should be expected to be particularly
important for high performers. These employees may believe that they deserve to fare better in
terms of pay level. Their retention may also be especially sensitive to whether group-based pay
adequately rewards their individual contributions (Gerhart & Rynes, 2003). Thus, we predict
that pay issues in general should be more important reasons for leaving when performance is
high.
In addition to pay, the opportunity for advancement is also an equitable outcome for high
performers. Promotions are positively related to job performance (e.g., Salamin & Hom, 2005)
and should thus be quite salient for high performers. Moreover, because promotions tend to
include pay increases that are greater than are available through standard merit processes
(Milkovich & Newman, 2002), high performers have an incentive-based, as well as an equity-
based, stake in caring substantially about advancement opportunity. As a result, we would
expect that high performing leavers are more likely than low performing leavers to cite the
opportunity to advance as important to their quitting behavior.
Hypothesis 1: Job performance will be positively related to the importance of pay
and advancement opportunity as leavers’ quit reasons.
Reasons of Less Importance to High Performers than to Others
While one goal of this study is to identify quit reasons that matter more to high than to
low performers, what we find to be relatively unimportant for high performer turnover may be
equally meaningful. For example, when deciding how to allocate limited resources, investment
in retention-oriented policies that primarily help secure only the least valuable employees may
be unwise. Alternatively, if there are reasonable employment practices that tend to be
unpopular with employees (e.g., disciplinary policies), and this discontent is generally confined

Why High and Low Performers Leave CAHRS WP07-11
Page 10 of 42
to low performers, these practices may be well worth keeping. Because this perspective has
not yet been studied, we examine several different conditions that may be quite influential in
turnover decisions of lower performers, but that may largely lose their relevance as job
performance increases.
We focus on four conditions that appear to be rooted in negative reactions that are more
likely to emerge for low performers. First, lack of a clear understanding about what is expected
of the employee is a key dimension of role ambiguity, which leads to job dissatisfaction and
intent to leave (Ngo, Foley, & Hoi, 2005), as well as turnover (Fisher & Gitelson, 1983; Jackson
& Schuler, 1985). Because role ambiguity also is greater among poor performers (Fisher &
Gitelson, 1983; Jackson & Schuler, 1985), the tendency for this failure to understand the job to
lead to turnover may be greater when job performance is low. Similarly, role overload, as
characterized by too many demands on the employee, leads to emotional exhaustion and intent
to leave (Ngo et al., 2005). Given that role overload also is more prevalent among poor
performers (Cooper, Dewe, & O’Driscoll, 2001), the likelihood that increased job demands are
daunting enough to precipitate turnover may be greater when performance is low. A third
potential source of negative attitudes and quitting that may be particularly meaningful to low
performers is one’s immediate supervisor. Because subordinate performance is one of the
supervisor’s primary concerns, low performance creates a salient opportunity for friction and
discord that is largely absent for high performers. Thus, low performance is likely to be
associated with a poor relationship, and thus dissatisfaction, with one’s supervisor (Gerstner &
Day, 1997). In contrast, high performance should insulate the employee from the primary
potential point of contention with the supervisor. Overall, we contend that, for low performers,
failure to understand the job, increased job demands, and one’s supervisor will contribute to
psychological distress of some type, thereby emerging as relevant reasons for low performer
turnover. These three reasons for leaving should diminish in relevance as performance
increases.

Why High and Low Performers Leave CAHRS WP07-11
Page 11 of 42
Finally, a similar low performer emphasis applies to formal disciplinary policies, such as
one designed to control employee absenteeism. An absenteeism policy provides a disincentive
for missing work, as penalties are incurred as days missed increase. Consequently, because
low performers tend to be absent more often (Stumpf & Dawley, 1981), they also are more likely
to suffer from the policy’s consequences; thus, they may tend to deem the policy to be
problematic and to cite it as a reason for quitting. Such a policy, however, would likely be of
limited relevance in the turnover behavior of high performing (and low absenteeism) employees.
Hypothesis 2: Job performance will be negatively related to the importance of
failure to understand the job, increased job demands, one’s supervisor, and an
absenteeism policy as leavers’ quit reasons.
Reason Differences within Performance Levels
Hypotheses 1 and 2 address how quit reason importance may vary across performance
levels. There are also grounds, however, for being interested in how these reasons’ importance
may vary within performance levels. A quit reason, for example, that was very, and equally,
important to both high and low performers would have considerable implications for
organizations, but would not emerge as meaningful in our across-performance assessments. In
contrast, analysis of such a reason within the high performing subset would, at least partially,
reveal its impact. Given the importance of retaining those who bring the most value to the
organization, we focus on within-group differences among high performers in particular.
Our equity-based arguments that the importance-to-quitting of pay and opportunity for
advancement should increase with performance imply that these reasons will be rather
important to high performers in an absolute sense. High performance is generally deemed to
warrant high levels of such rewards. Similarly, our arguments that the importance-to-quitting of
failure to understand the job, increased job demands, one’s supervisor, and an absenteeism
policy should diminish with performance implies that these reasons will be low in importance for
high performers. Hence, we expect that high performers will rate these latter four quit reasons

Why High and Low Performers Leave CAHRS WP07-11
Page 12 of 42
as less important than pay and advancement opportunity. Additionally, we expect pay and
advancement opportunity to be more important to high performer quitting than are several other
quit reasons that were measured here.
Hypothesis 3: High performing leavers will rate pay and opportunity for
advancement as more important to quitting than other reasons.
Performance Effects on What Leavers Find Elsewhere
In addition to our focus on leavers’ performance-specific quit reasons, our second major
emphasis in studying employment transitions involves examining how high and low performing
leavers vary in terms of post-quit external market outcomes. Our hypotheses are grounded in
March and Simon’s (1958) theory of organizational equilibrium, and in particular, their notion of
“ease of movement”, which refers to the degree to which employees could transition from one
organization to another without difficulty. Trevor (2001) labeled the individual attributes that
enhance ease of movement as “movement capital.” Job performance is one of these movement
capital elements, as performance, even if relatively invisible outside of the organization,
produces promotions, better reference letters, and a variety of success experiences, all of which
are positive signals on the job market (Trevor et al., 1997). Accordingly, high performers should
have the benefit of greater ease of movement in the external market.
To test the implications of these notions, we first address whether high performers are
more likely than low performers to be employed shortly after quitting. If high performers
possess greater ease of movement in the job market, it seems reasonable that they should, all
else equal, more easily acquire a new job before, or just after, leaving the old one. On the other
hand, this mobility advantage may lead high performers to feel more efficacious in their ability to
find suitable employment (Wanberg, Kanfer, & Rotundo, 1999), which could result in being more
likely to quit without having another job lined up. Similarly, such confidence could lead them to
be more hesitant to accept the first viable employment offer, under the belief that they may
acquire others that are more lucrative. On balance, however, we believe these tendencies

Why High and Low Performers Leave CAHRS WP07-11
Page 13 of 42
would be more than offset by the distinct monetary disincentive associated with being
unemployed for even a short period, by high performers’ advantage in generating new offers
while still working, and by the active recruitment from external employers that is part of that
advantage.
Hypothesis 4: Pre-quit job performance will be positively related to post-quit
employment.
Characteristics of performance-specific leavers’ subsequent employment are
also of interest. That is, a more comprehensive understanding of performance-specific
turnover emerges from knowing not only what may have driven leavers out, but also
what may have enticed them to leave. To address this, we examine performance-
specific impressions of: (1) whether the primary impetus behind quitting was the prior
organization’s shortcoming or the new job’s appeal; and (2) how pay and advancement
opportunity in the new job compared to these two factors in the prior organization.
Turnover models often emphasize how the quit decision depends on the
comparison between current employer utility and future employer utility (e.g., March &
Simon, 1958; Mobley, Griffeth, Hand, & Meglino, 1979). To date, however, nothing is
known regarding performance effects on such a comparison. One implication of the
assumption that high performers enjoy an ease-of-movement advantage is that they will
tend to be offered more attractive employment alternatives than are available to their
lower performing colleagues. Those better alternatives could result from the market’s
recognition of their greater marginal products or via the leverage afforded them by
multiple offers that may result from ease of movement. Moreover, given that job
satisfaction and job performance are positively related in many contexts (Judge,
Thoreson, Bono, & Patton, 2001), we expect that high performers tend, on average, to
view their current situation more positively (or less negatively) than do low performers.
Hence, relative to low performers, high performers should not only tend to acquire more

Why High and Low Performers Leave CAHRS WP07-11
Page 14 of 42
promising alternatives, but also should tend to see their current situation as more
satisfying. Together, this suggests that higher performing leavers would be more likely
than low performing leavers to see themselves as having been pulled into a better job,
rather than pushed from a negative situation.
Hypothesis 5: Pre-quit job performance will be positively related to attributing
quitting to the appeal of a better job elsewhere (as opposed to something
negative about the old job).
We next investigate what leavers of different performance levels actually find in their
new jobs. Specifically, we wish to explore whether pay and advancement opportunity improve
more for high performers than for low performers. It is reasonable to assume that employees
evaluate future employer utility in a manner consistent with how they evaluate current employer
utility. Thus, if, as predicted in Hypothesis 1, pay and advancement opportunity are more
important quit reasons for high performers than for low performers, it then seems probable that
these high performing leavers also would be more motivated to enhance their standing with
regard to these two factors. Additionally, given their ease-of-movement advantage, high
performers should have greater ability to secure the new job that best exemplifies what they
want from the job market.
Hypothesis 6: Pre-quit job performance will be positively related to new-job pay
and new-job advancement opportunity.
Moderators of Performance Links with New-Job Outcomes
The previous hypothesis is potentially conditional upon the manner in which the external
market reacts to additional leaver characteristics. Specifically, we believe that a leaver’s race,
gender, and tenure in the pre-quit organization may affect the extent to which pre-quit
performance predicts new-job pay and advancement opportunity.
Tenure. Two perspectives suggest that the leveraging of pre-quit performance to get
what is wanted from the external market (i.e., better pay and advancement potential,

Why High and Low Performers Leave CAHRS WP07-11
Page 15 of 42
presumably) will be greater when pre-quit tenure is low. First, pre-quit high performance may
serve as a clearer signal of employee ability for employees with low, rather than high, tenure.
That is, whereas high performance at low tenure is particularly dependent on raw ability (e.g.,
Murphy, 1989), and should yield inferences accordingly, high performance at high tenure may
instead at least partially be seen as a function of that tenure (indeed, conceptual models of job
performance often include tenure; Sturman, 2003). Second, the implications of situation
similarity suggest that those making hiring decisions may be most optimistic about near-term
performance when high performing leavers are of low pre-quit tenure. Past performance is
generally considered to be the best predictor of future performance, but the prediction is
stronger when contexts under which past and future performance are measured are similar
(Pulakos & Schmitt, 1995). Thus, hiring the low tenure, high performing leaver will yield the
greatest expectations of high performance in the very near future, as the employee has already
demonstrated the ability to perform well in a new-job situation. In contrast, the expectations for
what the high performing leavers that are of high tenure will do in a low tenure scenario are less
clear.
In sum, we contend that the clearer ability signal and higher expectations for near-term
high performance combine to make high performers particularly attractive job candidates when
their pre-quit tenure is low. Consequently, these low tenure leavers could better exploit the ease
of movement that job performance brings and more effectively land, or negotiate, more lucrative
offers. Less desirable signaling and less optimistic early performance expectations, however,
may somewhat limit the ability to leverage pre-quit performance for high tenure leavers.
Hypothesis 7: The positive relationships between pre-quit job performance and
the new-job’s pay and advancement opportunity will be more evident when
tenure is low.
Gender and race. Among the few studies to track leavers to compare aspects of old jobs
with aspects of new jobs are a set of three that examined gender and race effects on pay
attained via movement in the external job market. This research reports that men fare better

Why High and Low Performers Leave CAHRS WP07-11
Page 16 of 42
than women in terms of gains in pay acquired by such movement (e.g., Brett & Stroh, 1997;
Lam & Dreher, 2004), and that White male movers receive an external market pay premium
relative to all other race/sex groupings (Dreher & Cox, 2000). The authors’ explanations are
grounded in job market discrimination, as these results are deemed inconsistent with traditional
labor-economic perspectives in which it is in the firm’s best interest to allocate pay according to
productivity alone. Specifically, Brett and Stroh (1997), Dreher and Cox (2000), and Lam and
Dreher (2004) cite the following as likely discrimination-related explanations for gender and race
differences in external market outcomes: (1) ties to important informal social networks affect
finding jobs (Burt, 1992; Granovetter, 1995) and success in negotiations (Brodt, 1994), but such
ties are weaker for non-Whites and women (Brass, 1985; Ibarra, 1995); (2) the stereotyping of
women and minorities in the salary negotiation process leads to lower offers and less flexible
bargaining (Ayres, 1995); and (3) search firms tend to disproportionately market males (Judge,
Cable, Boudreau, & Bretz, 1995) and are primarily run by men.
Currently, we know nothing about whether job performance, likely the most important
employee characteristic from the perspective of both the former and the hiring organizations,
matters in this calculus1. Once again, the ease-of-movement advantage is integral to our
prediction. Because discrimination and related factors work against women and minorities in
the marketplace, the ability of women and non-Whites to leverage their performance into new-
job pay may also be constrained. That is, whereas men and Whites should be able to parlay
their high performance into more lucrative offers, the subtle market discrimination that manifests
in decision-maker stereotyping and more limited social networks should diminish the potency of
performance effects for women and non-Whites.
Hypothesis 8: The positive relationship between pre-quit job performance and
new-job pay will be less evident for women and non-Whites.
1 We note that Lam and Dreher (2004) included job performance as a control variable. Although performance was
positively related to pay, the analysis included both movers and stayers, making it unclear whether this effect revealed
internal or external returns to performance.

Why High and Low Performers Leave CAHRS WP07-11
Page 17 of 42
Job level. Finally, although all workers are generally interested in securing more
pay, advancement potential may be particularly important to those who are changing
jobs as part of a career decision. Relative to hourly employees, management and
salaried employees should be more likely to emphasize the long-term career
implications associated with their moves. Hence, these exempt employees should be
more likely than hourly workers to rely upon their performance-driven ease-of-movement
advantages to secure jobs with better advancement potential.
Hypothesis 9: The positive relationship between pre-quit job performance and
new-job advancement opportunity will be more evident for management and
salaried employees.
Method
Sample and Setting
We were allowed access to organizational records and post-exit survey data for a
sample of leavers from a large organization in the leisure and hospitality industry. The sample
consists of 2,510 former employees who voluntarily left their jobs between 2003 and 2005 and
agreed to respond to questions about why they did so. The average age of participants was 38
years, and 52% of the sample was female. The participants were approximately 67% White,
15% African American, 10% Hispanic, 5% Asian, and 2% American Indian. Management
employees constituted 6% of the sample, 16% were (non-management) salaried employees,
and the remaining 78% were hourly. Respondents from 29 different locations across the United
States were included in the study. The organization’s pay system and succession model were
consistent across locations. A merit pay plan was in place for all employees, with the
percentage pay increase determined by performance rating. Promotions were performance-
based and were generally more common for salaried and, in particular, management
employees.

Why High and Low Performers Leave CAHRS WP07-11
Page 18 of 42
Procedure
Participants provided information about their reasons for leaving and their new-job
characteristics in questionnaires administered in person (14%), via telephone (66%), or via a
written instrument (20%). Questions were identical across these three types of administration.
Completing the questionnaire took approximately 10 to 15 minutes. We linked the questionnaire
data to organizational records, which we used for all other measures.
To obtain the questionnaire data, members of the human resources department
attempted to contact 6,800 former employees whom the organization had coded as having left
voluntarily. Contact attempts were made, on average, approximately three weeks after the
employee left the organization. Of the 6,800 originally sought, 4,136 completed the
questionnaire, yielding an initial response rate of 60.8%. Of the 2,664 non-respondents, 416
had declined to participate when invited and 2,248 were deemed to be unreachable. The
unreachable designation came only after four separate unsuccessful attempts to contact the
former employee by telephone, followed by non-response to a subsequent mailing of the
questionnaire to the leaver’s last known address. T-tests revealed no significant difference in
performance level, our key predictor variable, between respondents (3.10) and non-respondents
(3.08), despite our considerable statistical power to find any differences. The two groups also
did not differ on gender composition, although respondents and non-respondents did
statistically differ in terms of age (38.8 versus 36.2), tenure (4.1 versus 3.8), being non-White
(36% versus 53%), and exempt status (20% versus 13%). Missing data on either the
questionnaire or on company records reduced the observations available for analysis to 2,510.

Why High and Low Performers Leave CAHRS WP07-11
Page 19 of 42
Data Acquisition
Our questionnaire approach deviates from what has been traditionally described as the
“exit interview” methodology. Strictly speaking, exit interviews are defined as “a discussion
between a representative of an organization and a person whose employment with that
organization has been ended, conducted during one of the employee’s last working days”
(Giacolone & Duhon, 1991, p.83). In the limited empirical research on exit interviews of this
type, researchers have shown that responses obtained were only weakly related to responses
gathered by follow-up questionnaires or third-party interviews (Campion, 1991; Hinrichs, 1975;
Lefkowitz & Katz, 1969), primarily because exit interviews were conducted by untrained
managers who were also the departing employees’ former bosses. Perhaps not surprisingly,
then, when exit interview data were compared to questionnaire responses obtained after the
employees quit, respondents in the exit interviews were less likely to mention their supervisor
(Campion, 1991; Hinrichs, 1975) and were more likely to provide vague responses about why
they left (e.g., “needed at home”; Lefkowitz & Kahn, 1969).
The general consensus from these studies is that the former bosses of departing
employees, especially those who are untrained in conducting exit interviews, are inappropriate
candidates for collecting data on why employees leave. Employees may be reluctant to discuss
poor supervision as one of the primary reasons, and their bosses may selectively interpret their
responses, or use the exit interview as a chance to “re-recruit” the candidate, rather than to elicit
honest information about reasons for leaving. Thus, researchers have concluded that the
validity of exit data can be enhanced by following a number of recommendations: (a) have
representatives from a neutral department such as human resources collect the information; (b)
ensure that these individuals receive proper training; (c) use a methodology that permits
systematic comparisons across respondents; (d) consider supplementing (or even replacing)
the traditional exit interview as defined above with a post-exit questionnaire (Giacalone &
Duhon, 1991; Griffeth & Hom, 2001; Hinrichs, 1975; Lefkowitz & Kahn, 1969).

Why High and Low Performers Leave CAHRS WP07-11
Page 20 of 42
The organization’s approach is consistent with these recommendations. In particular, the
human resources (HR) department was solely responsible for collecting and managing the
information obtained from all exit questionnaires. HR personnel were ideally suited for this
project because they generally had no prior contact with the former employees, which should
promote candid responses about factors that may reflect problems with the supervisor or other
members of the person’s former work group. HR representatives were trained to adhere to the
interview script, ask questions exactly as stated, and avoid asking follow-up questions. The
post-exit questionnaire was structured to ensure that all participants were given the same
instructions, were asked the same set of questions, and were using the same response scales.
In order to alleviate fears of retribution or negative employment references, participants were
promised that responses were business confidential (i.e., within the constraints of the law), and
were told that their responses would be used only to better understand how to improve the work
environment.
In addition to these design elements that address the quality of the data, we also
included measures to control for confounds that potentially remained. Specifically, we coded
each observation for how the questionnaire was administered (i.e., by telephone, in written form,
or in person) and the number of days between the leaver’s termination date and the
questionnaire administration. These variables were then included as controls in our models.
Measures
Dependent variables. The organization’s post-exit survey included twelve items on
reasons for voluntarily leaving. All items were phrased as “What impact did ________ have on
you leaving?” The three possible responses to each item were “little to no impact” (coded as
“1”), “some impact” (coded as “2”), and “strong impact” (coded as “3”). Based on our
hypotheses, we used the six items that addressed the following potential quit reasons: pay,
opportunity for advancement, lack of a clear understanding of the job, the job being too

Why High and Low Performers Leave CAHRS WP07-11
Page 21 of 42
physically demanding or increased job demands, your supervisor, and the absenteeism policy.2
Although the organization’s post-exit survey did not include multi-item measures, or more finely
grained single-item measures, there are indirect indications of quit reason validity. First, face
validity appears to be present, as the quit reason items are simple and straightforward. Second,
the pattern of correlations is consistent with the disparate nature of the quit reasons, as all but
one of the correlations among the reasons is below .14; moreover, the .32 correlation between
pay and advancement opportunity as quit reasons is reasonable given that these two elements
are intertwined in the workplace and that our predictions for their importance are grounded in
near identical rationale.
In addition to quit reasons, the questionnaire tapped into the four dependent variables
associated with the external market. First, individuals were asked whether they were currently
employed at a new job (coded as “1” for yes). Leavers were also asked if they left “due to a
better job” elsewhere, which we coded as “1”, or due to “something negative about working at
[the organization]”, which we coded as “0”. The final two dependent variables are new-job pay
and new-job advancement, which are measures of whether the new job’s pay and advancement
opportunity, respectively, are worse than (“1”), about the same as (“2”), or “better than (“3”) the
old job. Once again, although the organization did not use multi-item measures, the simplicity
of the four items appears to provide evidence of face validity. Also, the pattern of correlations
supports the construct validity of the measures. For example, new-job pay is significantly more
highly correlated with the importance-to-quitting of pay (.34) than it is to any other quit reason;
similarly, new-job advancement opportunity is significantly more highly correlated with the
importance-to-quitting of advancement opportunity (.34) than it is to any other quit reason.
Job performance. All employees were given a performance score by their supervisors
as part of their annual performance review. Ratings were made on a 5-point scale (“1” = Needs
2 The exit interview’s other six potential quit reasons addressed the opportunity to use your skills, your work hours or
schedule, your co-workers, the work environment, benefits, and the job’s distance from home. We did not specifically
hypothesize about these six potential reasons because we saw little a priori foundation for expecting them to vary in
importance according to performance level. To provide context for the importance of the quit reasons that we do

Why High and Low Performers Leave CAHRS WP07-11
Page 22 of 42
Improvement, “2” = Marginal, “3” = Successful, “4” = Highly Successful, “5” = Outstanding
Results). The most recent overall performance rating received (prior to departure) was obtained
from company records. Raters were trained in the performance assessment system and
encouraged to follow a guided distribution close to the following: 5-10% either “Needs
Improvement” or “Marginal”, 65-70% “Successful”, 15% “Highly Successful”, and 10%
“Outstanding”. As comparison, our sample of leavers was rated as follows: 5% either “Needs
Improvement” or “Marginal”, 80% “Successful”, 12% “Highly Successful”, and 3% “Outstanding”.
Apparently, either supervisors provided fewer outstanding ratings than encouraged or
outstanding performers left at a disproportionately low rate (because we were only given access
to leaver performance ratings, we cannot say which).
Moderators. Several of our hypotheses predict performance effects that are moderated
by employee characteristics. These include tenure, which is the number of years employed with
the organization at the time of departure, and dummy variables for non-White (i.e., a race
indicator with Whites as the omitted category) and female. In order to assess job level, dummy
variables were created for salaried and management, with hourly as the omitted category. All of
these moderating variables were obtained from company records.
Control variables. We also needed to account for location-specific explanations for quit
reasons and post-quit outcomes. Thus, to control for such factors as local job market effects
and property-specific climate, we included 28 dummy variables to represent the 29 distinct
properties at which employees worked. We also control for age, so as to better isolate tenure
effects. Finally, we included variables to control for post-exit questionnaire logistics: telephone
and written surveys are dummy variables indicating how the questionnaire was administered,
with face-to-face as the omitted category; lagtime is the number of days between the
termination date and the questionnaire administration (this is coded as “0” for the 9% of the
sample interviewed on or before the termination date).
focus on, however, we do present limited data on these additional six reasons.

Why High and Low Perform
ers Leave CAHRS WP07-11
Page 23 of 42
Analysis Strategy
The post-exit questionnaire contained three-point scales for quit reason importance (i.e.,
little to no impact, some impact, strong impact) and for new-job characteristics (i.e., worse,
about the same, better). These dependent variables represent ordinal rather than interval level
data, thus making the use of ordinary least squares regression problematic. Consequently, we
conducted ordered logit regression analyses, which are specifically designed for examining
independent variable effects on ordinal outcomes. We also conducted logistic regression
analyses when predicting whether leavers were employed after quitting and whether they quit
because of a better job elsewhere, both of which are dichotomous. For both types of analyses,
we used robust standard errors that account for the potential dependence of employee
observations that are clustered within properties. These standard error estimates require only
that observations are independent across, but not necessarily within, the source of the
dependence (Rogers, 1993).
Results
Means, standard deviations, and correlations are presented in Table 1. Of the quit
reasons, pay and advancement are the most interrelated, which is consistent with our similar
rationale for their effects. We next interpret tests of our nine hypotheses; given the number of
hypotheses and associated inferences, we provide a summary of these results in Table 2.

Why High and Low Performers Leave CAHRS WP07-11
Page 24 of 42
Table 1
Means, Standard Deviations, and Correlations
Variable Mean SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
1. Pa y 1. 687.8
2. Advance 1.65 .86 .32
3. Understand 1.08 .35 .03 .09
4. Demands 1.23 .58 .04 .03 .12
5. Supervisor 1.75 .91 -.11 .04 .09 .07
6. Absenteeism 1.14 .47 .03 .02 .08 .11 .05
7. Employed .67 .47 -.03 .23 .00 -.05 -.12 -.02
8. Better job .44 .50 .27 .10 -.07 -.16 -.45 -.07 .48
9. New-job pay 2.66 .65 .34 .08 .03 -.02 -.13 -.01 -.02 .23
10. New-job adv. 2.67 .57 .15 .34 .01 .03 .04 .00 -.02 .01 .27
11. Performance 3.10 .60 .03 .05 -.03 -.01 -.07 -.09 .06 .07 .03 .04
12. Female .52 .50 -.13 -.08 .05 .08 .07 .06 -.13 -.16 -.13 -.04 .00
13. Non-White .33 .47 .01 -.03 -.01 .01 -.03 .02 -.01 .06 .04 -.01 -.04 .01
14. Salaried .16 .37 .01 .10 .04 .02 .06 -.09 .02 .04 .00 .04 .06 -.01 -.07
15. Management .05 .23 -.04 .03 .03 .02 .01 -.07 .03 .02 .01 .00 .11 -.08 -.11 -.11
16. Tenure 3.92 4.36 -.09 -.02 -.01 .09 .01 -.06 .00 .00 -.11 .06 .14 -.02 -.08 .15 .13
17. Age 38.17 12.23 -.07 -.08 .02 .11 .02 -.02 -.10 -.10 -.01 -.01 .05 -.02 -.17 .06 .05 .28
18. Lagtime 21.21 30.26 .01 .05 .05 .07 .09 .05 .05 -.10 -.08 -.04 .02 .03 .00 .04 .01 .03 .05
19. Telephone .66 .47 -.05 -.06 -.04 -.04 -.03 -.05 -.17 -.14 .07 .11 -.01 -.02 .00 -.03 .02 -.01 -.02 -.30
20. Written .20 .40 .06 .07 .06 .08 .09 .07 .06 -.07 -.08 -.08 .04 .05 .00 .05 .00 .03 .05 .61 -.69
Note. N = 1,148-2,510; correlations with absolute values above .05 are significant at p < .05.

Why High and Low Performers Leave CAHRS WP07-11
Page 25 of 42
Table 2
Summary of Hypotheses and Empirical Support
Predicted Direction/
Hypothesis Coefficient Sign Support
Hypothesis 1: Performance increases will be associated with increases in the importance-to-
quitting of pay and advancement opportunity + yes
Hypothesis 2: Performance increases will be associated with decreases in the importance-
to-quitting of failure to understand the job, inability to meet job demands, one’s
supervisor, and an absenteeism policy - yes
Hypothesis 3: For best performers, the most important quit reasons will be pay and
advancement opportunity N/A
partial
(supervisor and opportunity
to use skills were as
important as pay)
Hypothesis 4: Pre-quit job performance will be positively related to new-job employment + yes
Hypothesis 5: Pre-quit job performance will be positively related to reporting being pulled to
a better job rather than pushed away from something negative + yes
Hypothesis 6: Pre-quit job performance will be positively related to new-job pay and new-job
advancement + no
Hypothesis 7: Hypothesis 6 relationships will be stronger when tenure is low - yes
Hypothesis 8: Hypothesis 6 new-job pay relationship will be weaker for women and non-
Whites - yes (race); no (gender)
Hypothesis 9: Hypothesis 6 new-job advancement relationship will be stronger for
management and salaried + yes

Why High and Low Performers Leave CAHRS WP07-11
Performance-Specific Quit Reasons (Hypotheses 1-3)
We first predicted that job performance would be positively related to the importance of
pay and advancement opportunity in quitting. The significant positive performance coefficients
in the first two models in Table 3 support Hypothesis 1. Interpretation of the effects for ordered
logit models parallels logistic regression interpretation: the raw coefficient, which represents the
change in the log-odds of an outcome predicted by a one unit change in the independent
variable, is made more interpretable by exponentiating, subtracting one, and multiplying by 100.
Hence, the .17 and .18 coefficients in the first and second models indicate that a one unit
increase in performance increases the odds of a higher rating of the importance-to-quitting of
pay and advancement opportunity by 19% (i.e., [e.17 - 1] x 100=19%) and 20%, respectively.
Hypothesis 2 predicted that job performance would be negatively related to the
importance of failure to understand the job, increased job demands, one’s supervisor, and an
absenteeism policy as quit reasons. The final four models in Table 3 indicate strong support for
Hypothesis 2, as low performing leavers, compared to high performing leavers, rated these
reasons as substantially more important to turnover. For example, a one unit increase in
performance increases the odds of a lower importance-to-quitting rating of an absenteeism
policy and of not clearly understanding the job by 27% and 36%, respectively.
In addition to such tests across performance levels, we also investigated differences in
quit reasons within the high performer group. Hypothesis 3 predicted that high performing
leavers would report pay and opportunity for advancement as more important than other quit
reasons. We test this in Table 4, where we present, for high performing leavers (i.e., those
rated as “5”), mean levels of quit reason importance. As comparison, we also list the means for
average (rated as “3”) and low performers (rated as “1”). T-tests reveal that high performing
leavers rated advancement opportunity as significantly more important to quitting than all other
(non-pay) reasons; these high performers also rated pay as significantly more important to
quitting than all other (non-advancement) reasons except supervisor and opportunity to use
skills. Thus, Hypothesis 3 was largely, though not completely, supported.
Page 26 of 42

Why High and Low Performers Leave CAHRS WP07-11
Given that we had found that supervisory importance diminished as performance
increased (see Table 3), the importance of the supervisor was unexpectedly strong for high
performers. Additionally, the supervisor was the highest rated quit reason for average and low
performers (see the Discussion for more on the supervisor issue). A final point of interest in
Table 4 is the enduring importance of pay and advancement opportunity; while they clearly
matter more to turnover for high performers (as we tested in Table 3), they also retain
meaningful impact at average and low performance levels, as described in the Table 4 note.
Table 3
Ordered Logit Regressions of Quit Reason Importance on Employee Job Performance
Dependent Variable
Predictors Pay Advance Understand Demands Supervisor Absenteeism
Property dummies yes yes yes yes yes yes
Lagtime -.30 .15 .47 .25 .37* .24
(.18) (.20) (.32) (.19) (.17) (.27)
Telephone .00 -.41* -.60 -.38 .05 -.49
(.18) (.20) (.34) (.25) (.18) (.36)
Written .51* -.10 -.05 .20 .17 .16
(.22) (.25) (.35) (.26) (.20) (.37)
Female -.54*** -.29** .61** .46*** .27*** .35**
(.06) (.06) (.21) (.09) (.08) (.13)
Non-White -.11 -.19* -.05 .15 -.07 .01
(.07) (.09) (.22) (.16) (.09) (.12)
Age -.02** -.02*** .01 .02*** .00 .00
(.00) (.00) (.01) (.00) (.00) (.01)
Tenure -.04*** -.01 -.04 .03* .00 -.03
(.01) (.02) (.03) (.02) (.01) (.02)
Salaried .07 .47*** .56** -.02 .28*** -1.04***
(.16) (.12) (.21) (.15) (.08) (.30)
Management -.25 .17 .92* .17 .29 -33.68***
(.20) (.23) (.46) (.33) (.16) (.29)
Performance .17* .18* -.44*** -.18* -.26*** -.32***
(.08) (.09) (.12) (.10) (.05) (.13)
Model log-likelihood -2256*** -2269*** -565*** -1245*** -2295*** -873***
N 2464 2478 2462 2475 2491 2467
Note. Standard errors in parentheses; raw coefficients are reported for all models. The lagtime coefficient and standard error
were multiplied by 100 to allow interpretation.
* p < .05; ** p < .01; *** p < .001
Page 27 of 42

Why High and Low Performers Leave CAHRS WP07-11
Table 4
Mean Values of Quit Reason Importance
for Highest, Average, and Lowest Performers
Quit Reason Importance Means
Lowest Average Highest
Reason Performers Performers Performers
Pay 1.51 1.70 1.79
Advance opportunity 1.56 1.65 1.86
Opportunities to use skills 1.32 1.46 1.64b
Failure to understand job 1.19 1.08 1.07ab
Inability to meet job demands 1.24 1.23 1.20ab
Supervisor 1.96 1.76 1.60b
Absenteeism policy 1.33 1.15 1.00 ab
Co-workers 1.17 1.20 1.11 ab
Benefits 1.16 1.25 1.27 ab
Work hours 1.68 1.58 1.23 ab
Environment
Page 28 of 42
1.15 1.24 1.26 ab
Distance 1.19 1.21 1.28 ab
Note. The superscripts indicate that, for the highest performers, the quit reason’s importance is
statistically smaller than the importance of pay (“a”) and advancement opportunity (“b”). Because
Hypothesis 3 involves within-performance differences only for high performers, t-tests are presented
only within the highest performers’ column. We note, however, that, for average performers, supervisor
importance is statistically greater than all other reasons’ importance; for lowest performers, the
supervisor mattered more than all reasons except work hours. Also, the importance levels of pay and
advancement opportunity are statistically greater than the importance of 18 of 20 possibilities for
average performers, and 14 of 20 possibilities for lowest performers.

Why High and Low Perform
ers Leave CAHRS WP07-11
Page 29 of 42
Performance-Specific New-Job Outcomes (Hypotheses 4-9)
Results from our focus on the external market aspect of performance-specific
employment transitions are presented in Table 5. Hypotheses 4 and 5 involved predictions
about pre-quit job performance relationships with new-job employment and the perception that
quitting was due to being pulled to a better job. As is evident in Model 1, Hypothesis 4 was
supported, as, controlling for time between quitting and the questionnaire administration, a one
unit increase in pre-quit performance increases the odds of being employed at a new job at the
time of the questionnaire by 23%. Hypothesis 5 also was supported, as Model 2 in Table 5
reveals that a one unit increase in pre-quit performance increases the odds of leavers reporting
that they were pulled toward a better job (rather than pushed away by something negative) by
28%.
The final four hypotheses all address pre-quit job performance relationships with pay
and advancement opportunity in the new job. Hypothesis 6 is the main effect prediction that
leavers who were higher performers were more likely to find improved pay and advancement
opportunity in their new job. Hypothesis 6 was not supported (see Models 3 and 5 in Table 5).
Some interactions, however, did qualify this lack of support, as is indicated in Models 4 and 6.
First, in support of Hypothesis 7 and our signaling and ease-of-movement rationale, we found
that pre-quit performance’s positive relationships with new-job pay and new-job advancement
opportunity were more evident at lower tenure. Specifically, a one unit increase in pre-quit
performance increases the odds of leavers reporting a higher rating on the new-job pay scale by
26% when tenure is low (the 25th percentile; 1.02 years), but only by 7% when tenure is high
(the 75th percentile; 5.18 years).3 Also, a one unit increase in pre-quit performance increases
the odds of leavers reporting a higher rating on the new-job advancement scale by 38% when
tenure is low, but increases the odds of reporting a lower rating by 4% at high tenure (see
Figure 2, top).
3 We use percentiles here because positive skew in tenure’s distribution makes using the moderator’s mean plus and minus one standard deviation
inappropriate (e.g., the mean minus one standard deviation yields “negative tenure”). To be consistent, we also set tenure at its median when
evaluating the other interactions in the study.

Why High and Low Performers Leave CAHRS WP07-11
Page 30 of 42
Table 5
Ordered Logit and Logistic Regressions of External Market Elements
on Pre-Quit Job Performance and its Moderators
Dependent Variable
New-Job New-job New-job New-job
Predictors Employed Better job
Advancement Pay Pay Advancement
Property dummies yes yes yes yes yes yes
Lagtime .59* .60* -.59* -.62* .10 .05
(.24) (.25) (.25) (.25) (.26) (.26)
Telephone -.89*** -.22 .68 .71 -.06 -.11
(.22) (.24) (.38) (.38) (.23) (.24)
Written -.58* -1.70* .51 .57 -.67** -.67**
(.26) (.86) (.35) (.37) (.23) (.24)
Female -.56*** -.65*** -.64*** -.67 -.22 -.22
(.09) (.12) (.17) (.71) (.11) (.12)
Non-white -.20* -.13 .10 1.51 -.06 -.09
(.10) (.07) (.16) (.89) (.15) (.15)
Age -.02*** -.02** .01 .00 .00 .00
(.00) (.01) (.01) (.01) (.00) (.01)
Tenure .01 .00 -.09*** .04 .01 .29**
(.02) (.02) (.02) (.07) (.02) (.10)
Salaried .10 .23* .05 .03 .15 -2.06
(.12) (.12) (.10) (.11) (.15) (1.07)
Management .26 .27 -.13 -.09 -.29 -.3.78***
(.27) (.20) (.40) (.39) (.28) (1.05)
Performance .21** .25* .12 .41 .10 .24
(.08) (.11) (.11) (.28) (.13) (.16)
Performance-X-tenure -.04* -.09***
(.02) (.03)
Performance-X-female .01
(.22)
Performance-X-non-White -.46*
(.27)
Performance-X-salaried .69**
(.35)
Performance-X-management 1.07**
(.30)
Model log-likelihood -1331*** -1061*** -955*** -952*** -980*** -968***
N 2208 1723 1465 1465 1433 1433
Note. Standard errors in parentheses; raw coefficients are reported for all models. The lagtime coefficient and standard error were multiplied by 100 to allow interpretation.
* p < .05; ** p < .01; *** p < .001

Why High and Low Performers Leave CAHRS WP07-11
Figure 2
Moderated Job Performance Effects on New-Job Pay
and New-Job Advancement Opportunity
0.5
0.6
0.7
0.8
0.9
12345
Job Performance
New-Job Advancement
Low Tenure
High Tenure
0.6
0.7
0.8
0.9
12345
Job Performance
New-Job Pay
Whit es
Non-Whit es
0
0.2
0.4
0.6
0.8
1
12345
Job Performance
New-Job Advancement
Salaried
Management
Hourly
Note. To aid in interpretation, all y-axis units are in terms of probabilities.
Page 31 of 42

Why High and Low Performers Leave CAHRS WP07-11
Hypothesis 8 predicted that pre-quit performance effects on new-job pay would depend
on gender and race. We found support only for the latter. Model 4 reveals that performance’s
positive effect on new-job pay emerged for Whites but not for non-Whites. A one unit increase
in pre-quit performance increases the odds of leavers reporting a higher rating on the new-job
pay scale by 39% for Whites, but increases the odds of reporting a lower rating by 12% for non-
Whites (see Figure 2, middle). Finally, we found support for Hypothesis 9, in that pre-quit
performance was more important to new-job advancement opportunity for management and
salaried employees than for hourly employees. Specifically, Model 6 indicates that a one unit
increase in pre-quit performance increases the odds of reporting a higher rating on the new-job
advancement scale by 203% and 108% for management and salaried leavers, respectively, but
increases the odds by only 4% for hourly leavers (see Figure 2, bottom).
Discussion
Our purpose here was to examine performance-specific employee transitions out of and
into organizations. Our hypotheses addressing job performance relationships with specific quit
reasons and with new-job outcomes were largely supported (see Table 2), although with some
notable exceptions.
Performance-Specific Quit Reasons (Hypotheses 1-3)
There are a number of implications that arise from our findings that quit reason
importance is often related to job performance level. First, the relationships themselves suggest
the importance of a focused, targeted retention policy (Steel, Griffeth, & Hom, 2002), as the use
of more general retention approaches without regard to performance-specific differences may
lead to investments that facilitate low performer retention but have little effect on high
performers. Other than contingent pay (e.g., Harrison et al., 1996; Salamin & Hom, 2005;
Trevor et al., 1997), exactly what specific conditions organizations might manipulate to target
performance-specific retention efforts has received little attention. Our work indicates that pay
in general is a more important quit reason when performance is high. We were unable,
however, to assess whether it was contingent pay specifically or a more holistic impression of
Page 32 of 42

Why High and Low Performers Leave CAHRS WP07-11
pay driving this relationship; thus, future research that teases apart the performance-specific
importance-to-quitting of various aspects of the general pay construct would be very helpful.
Advancement opportunity was also a particularly important quit reason for high performers. It is
possible, then, that publicizing existing career ladders or restructuring so as to enhance
advancement opportunity may help to retain top performers in spite of less than favorable pay
conditions.
Quit reasons that are more compelling for low performers are also meaningful. An
apparently poor fit between job requirements and employee knowledge, skills, and abilities
(KSA’s) seemed more likely to drive out low performers, as failure to understand the job and
increased job demands were much more relevant to low performer quits. This finding highlights
the importance of an effective staffing function, as hires with the appropriate KSA’s would be
both less likely to experience such lack of fit and more likely to perform well. It also indicates
that complex and challenging jobs appear to pose greater threats to the retention of low
performers, arguably adding to the endorsement of such types of job design. Importantly,
however, we note that the absolute importance levels of failure to understand the job and high
job demands for low performers are below those of pay and advancement opportunity (see
Table 4). This suggests that, for these low performing employees, problems with the nature of
the work itself are less likely to prompt quitting than are problems with pay and advancement
opportunity, which matter most for the best employees but still matter substantially for all.
The absenteeism policy was also of greater importance-to-quitting for low performers
than for high performers. This reflects an almost ideal result for the organization. The presence
of such a disciplinary policy is designed to punish and thus constrain absenteeism, an
expensive, unproductive behavior. While such a policy may not be well-liked, reactions to the
policy only appear to manifest into turnover for low performers. Hence, the policy, while
relatively inexpensive, appears to precipitate leaving for the least valuable employees,
presumably reduces absenteeism, and produces no apparent fallout among the best
performers.
Page 33 of 42

Why High and Low Performers Leave CAHRS WP07-11
Two unforeseen findings in the performance-specific data were particularly interesting.
First, as is evident in Table 4, work hours were cited as low performers’ second most important
reason for quitting. The table also indicates that the reason appears to be less important for
high performers. To test this with the appropriate control variables, we replicated the Table 3
ordered logit regression approach, but with work hours importance as the dependent variable.
Consistent with the mean levels in Table 4, the regression indicated that job performance was
negatively associated with work hours’ importance-to-quitting (p < .001). In retrospect, this
finding has intuitive appeal. At the hourly level, which comprises 78% of our sample,
supervisors have an incentive to reward high performers with more favorable schedules,
thereby saddling lower performers with less desirable schedules that may contribute to
dissatisfaction. At the exempt level, additional hours for high performers may be a vehicle to
either display or attain higher performance, and thus would not tend to be a major reason for
quitting; exempt low performers, however, may be more likely to see more work hours as an
uncompensated input and a source of frustration. Thus, at both exempt and hourly levels, it
seems reasonable that work hours are a bigger factor in quitting when performance is low.
A second unexpected but noteworthy finding here involved the importance of one’s
supervisor. Performance increases were associated with less importance of the supervisor as a
quit reason, as predicted in our Hypothesis 2 rationale that high performance would insulate
employees from perhaps the primary source of friction in the supervisor-subordinate relationship
(i.e., performance problems). The supervisor, however, was still the fourth most important
factor in high performer turnover (see Table 4). This surprising finding underscores the
criticality of training, development, and hiring practices that will improve the supervisor-
employee relationship, given that it is important to quitting across the performance continuum.
The fact that high performance did not appear to insulate high performers from contention with
the supervisor to the extent that we had envisioned suggests potentially important contextual
factors. For example, future research might explore factors that constrain the “insulation effect”
and keep supervisor importance-to-quitting high for top performers; candidates for this
Page 34 of 42

Why High and Low Performers Leave CAHRS WP07-11
moderation might include supervisor ability, supervisor personality, and the demographics of the
supervisor-employee dyad.
Performance-Specific New-Job Outcomes (Hypotheses 4-9)
Because no research to date has chronicled the job performance effects on new-job
outcomes, the second focus in our study of performance-specific employee transitions was what
high and low performers find after leaving. The external market outcomes were largely
consistent with our emphasis on ease-of-movement advantages that accrue to high performers.
We first found that high performers were more likely to be employed at the time of the exit
questionnaire and to see themselves as pulled into a better situation. Given their ease-of-
movement advantage, high performers presumably were more able to acquire superior offers,
and to do so more quickly.
Somewhat in contrast to this result and interpretation, we did not find support for our
prediction that high performers would be more successful in leveraging their ease-of-movement
advantages so as to garner better pay and advancement opportunity in the new job. Two
explanations seem particularly plausible. First, our dependent variables were somewhat coarse
and may not have been sensitive enough to allow us to pick up the effect. Second, it may be
that people only tend to leave for or accept new jobs with better pay and advancement
opportunity, somewhat regardless of prior performance and ease of movement. That is, it may
be that pay and advancement are so critical to most people’s new-job decisions that, regardless
of performance level, they are reluctant to leave the old job or accept an offer for a new one
without the assurance, or at least the belief, that such an improvement is likely. The range
restriction on the new-job pay and new-job advancement opportunity measures is consistent
with both explanations, as 77% and 72% of leavers reported improvement in new-job pay and
new-job advancement opportunity, respectively.
This lack of support for Hypothesis 6, however, was qualified by the interactions
involving leaver tenure, race, and job type. The tenure moderation indicated that, when pre-quit
tenure was relatively low, high pre-quit performers did in fact fare better than their low
Page 35 of 42

Why High and Low Performers Leave CAHRS WP07-11
performing comparisons in new-job pay and advancement. This reinforces the ease-of-
movement and signaling perspectives, as high performance appears to have signaled ability
and or near-future performance more strongly when tenure was low than when it was high.
Whereas the tenure moderation argument is grounded in signaling, the race moderation
explanation involves job market discrimination. We found that, for White leavers only, high pre-
quit performers did in fact fare better than their low performing comparisons in new-job pay and
new-job advancement opportunity. Because minority groups do suffer from disadvantages in
such factors as social networks, negotiations, and stereotyping that are likely to play a role in
external market outcomes (Dreher & Cox, 2000), it also seems likely that these factors my have
made it more difficult for non-Whites to leverage their performance in the external market.
Research with data to directly test this race-based explanation of the degradation of ease-of-
movement advantages would be of considerable interest.
Our failure to find a gender by performance interaction was surprising. Given that men
fare better than women in terms of gains in pay acquired by external market movement (e.g.,
Brett & Stroh, 1997; Lam & Dreher, 2004), we expected that, as in the race case, performance
signals might be marginalized for women. One interpretation that fits our data is that the
discriminatory forces at work are strong enough in the non-White case to discount high
performance, but are not strong enough in the female case.
Acquiring Leaver Data
The exit interview was originally conceived as an informal opportunity for supervisors to
gather from leavers, just prior to actual departure, the primary reasons for quitting. It appears
that practitioner usage of the exit interview remains high (88% of companies surveyed; cited in
Steel et al., 2002), yet given the dearth of published research on the exit interview methodology
relative to its widespread organizational usage, there is a pressing need to evaluate the quality
of these programs. The limited empirical data that is available on the exit interview (as
compared to the selection interview, for example) suggests that departing employees will be
reluctant to describe specific reasons for leaving when former supervisors conduct the
Page 36 of 42

Why High and Low Performers Leave CAHRS WP07-11
interview, especially when the primary drivers of quitting are supervisor-related. Nevertheless, it
seems premature to discount the value of exit interview programs as a potentially useful
approach to understanding why people leave. Careful design of post-exit questionnaires can
enhance the quality of the data obtained (e.g., by using structured interviews, promising
confidentiality, training interviewers, and avoiding the use of former managers as interviewers),
as was done here. An additional design consideration relates to the appropriate use of control
variables to account for extraneous influences on the relationships observed. In this study, we
controlled for the time lag between the date of employee termination and the questionnaire
administration, as well as the data collection method, as both were related to our dependent
variables at times.
Given that we believe that many of the reported problems with exit questionnaires can
be prevented, or at least substantially reduced, it is important to consider the tradeoffs
associated with design choice. Clearly, a predictive study of voluntary turnover has certain
advantages. Ideally, we wish to be able to predict voluntary turnover, so as to eventually take
steps to control it. Predictive designs enhance causality inferences that are integral to prediction
and control. On the other hand, there are advantages to the exit, and particularly the post-exit,
timing of data acquisition. In most predictive studies of turnover, where antecedents are
measured and are then related to turnover at some future date, it is possible that important
reasons for leaving will be missed because attitudes and other perceptions can change over the
course of the study window (Steel, 2002). In contrast, collecting exit data shortly after those who
quit left provides researchers with a method of assessing attitude-behavior (and event-behavior)
linkages that is less dependent on the timing and sequencing of data collection.
Additionally, asking former employees about exactly why they left at times enables more
comprehensive inquiry into the relevant context than predictive studies can provide. An example
of the effectiveness of this approach can be seen in the recent development of the unfolding
model of turnover, an influential framework that emphasizes multiple turnover pathways and
shocks (i.e., jarring events) that can prompt turnover. In a series of studies of the unfolding
Page 37 of 42

Why High and Low Performers Leave CAHRS WP07-11
model (e.g., Lee, Mitchell, Holtom, McDaniel, & Hill, 1999; Lee et al., 1996), the authors relied
on interviewing and surveying leavers to gain the data necessary to isolate the different paths to
(and reasons for) quitting. It is likely that only through accessing leaver retrospection could such
data have been obtained. The principle is similar here: acquiring detailed retrospective data
from performance-specific leavers is the best first step to teasing apart the many potential
explanations for why they left, thereby providing a conceptual and practical roadmap for future
work. Finally, an additional benefit is that, should the exit data collection take place a
reasonable amount of time after the final work day, as in the present study, it can also be used
to inquire about the new job. This tracking of people across organizations provides a means for
addressing rich questions, such as the performance-specific new-job issues investigated here.
Limitations
The dependent variable measures used here may not have been sensitive enough to
share as much variation with predictors as would measures with multiple items and finer
response scales. The organization’s retrospective approach, however, did allow us to obtain
information on a broad array of turnover antecedents and new-job outcomes while maximizing
the likelihood that participants would complete the questionnaire. Moreover, the dependent
variable measures are straightforward and appear to have reasonable face validity.
Additionally, generalizability, as is so often the case, may be a concern here. Our
sample was confined to the leisure and hospitality industry. This has the advantage of avoiding
unmeasured industry-specific sources of variation in employment transitions that might
complicate a multi-industry study. At the same time, however, other industries may well exhibit
different turnover dynamics than we observed here. For example, our ease-of-movement
construct, which was a foundation for several of our hypotheses, could be at a substantially
different level and thus function differently in, say, manufacturing or high-tech industries. Yet,
because our sample encompassed a broad array of exempt and nonexempt jobs, many of
which are not industry-specific, our findings at least have the potential to extend beyond this
sample’s industry.
Page 38 of 42

Why High and Low Performers Leave CAHRS WP07-11
Conclusion
Our purpose here was to examine two consequential but largely unstudied aspects of
employment transitions by studying (1) the various reasons why low and high performers leave,
and (2) what each group finds in the external job market. We found that job performance was
important to a variety of reasons for quitting, to viewing the departure as being driven by a better
job elsewhere, to finding a job in the external market, and, under certain contingencies, to the
level of pay and advancement opportunity in this new job. As such, we believe our study
contributes to the important research base aimed at understanding the flow of critical human
resources across organizational boundaries.
Page 39 of 42

Why High and Low Performers Leave CAHRS WP07-11
References
Adams, J. S. (1965). Inequity in social exchange. In L. Berkowitz (Ed.), Advances in
experimental social psychology (pp. 267-299). New York: Academic Press.
Allen, D. G., & Griffeth, R. W. (2001). Test of a mediated performance-turnover relationship
highlighting the moderating roles of visibility and reward contingency. Journal of Applied
Psychology, 86, 1011-1021.
Aquino, K., Griffeth, R. W., Allen, D. G., & Hom, P. W. (1997). Integrating justice constructs into
the turnover process: A test of a referent cognitions model. Academy of Management
Journal, 40, 1208-1227.
Ayres. I.(1995). Further evidence of discrimination in new car negotiations and estimates of its
cause. Michigan Law Review, 94, 109-146.
Boswell, W. R., Boudreau, J. W., & Tichy, J. (2005). The relationship between employee job
change and job satisfaction: The honeymoon-hangover effect. Journal of Applied
Psychology, 90, 882-892.
Brass, D. J. (1985). Men's and women's networks: A study of interaction patterns and influence
in an organization. Academy of Management Journal, 28, 327-343.
Brett, J. M., & Stroh, L. K. (1997). Jumping ship: Who benefits from an external labor market
career strategy? Journal of Applied Psychology, 82, 331-341.
Brodt, S. E. (1994). Inside information and negotiation decision behavior. Organizational
Behavior and Human Decision Processes, 58, 172-202.
Burt. R. S. (1992). Structural holes: The social structure of competition. Cambridge, MA:
Harvard University Press.
Bycio, P., Hackett, R. D., & Alvares, K. M. (1990). Job performance and turnover: A review
and meta-analysis. Applied Psychology: An International Review, 61, 468-472.
Campbell, D.J., & Campbell, K.M. 2003. Global versus facet predictors of intention to quit:
differences in a sample of male and female Singaporean managers and non-managers.
International Journal of Human Resource Management, 14, 1152-1177.
Campion, M.A. (1991). Meaning and measurement of turnover: Comparison of alternative
measures and recommendations for research. Journal of Applied Psychology, 76, 199-
212.
Cooper, G.L., Dewe, P.J. & O’Driscoll, M.P. (2001). Organizational stress: A review and critique
of theory, research, and applications. Thousand Oaks, CA: Sage.
Deutsch, M. (1975). Equity, equality, and need: What determines which value will be used as
the basis of distributive justice? Journal of Social Issues, 31, 137-149.
Dreher, G. F., & Cox, Jr., T. H. (2000). Labor market mobility and cash compensation: The
moderating effects of race and gender. Academy of Management Journal, 43, 890-900.
Fisher, C.D. & Gitelson, R. (1983). A meta-analysis of the correlates of role conflict and
ambiguity. Journal of Applied Psychology, 68, 320–333.
Gerhart, B., & Rynes, S. (2003). Compensation: Theory, evidence, and strategic implications.
Thousand Oaks, CA: Sage.
Gerstner, C. R. & Day, D. V. (1997). Meta-analytic review of leader-member exchange theory:
Correlates and construct issues. Journal of Applied Psychology, 82, 827-844.
Giacalone, R. A., & Duhon, D. (1991). Assessing intended employee behavior in exit interviews.
The Journal of Psychology, 125, 83-90.
Granovetter, M.S. (1995). Getting a job: A study of contracts and careers. Chicago: University of
Chicago Press.
Griffeth, R. W., & Hom, P. W. (2001). Retaining valued employees. Thousand Oaks, CA: Sage.
Griffeth, R. W., Hom, P. W., & Gaertner, S. (2000). A meta-analysis of antecedents and
correlates of employee turnover: Update, moderator tests, and research implications for
the next millennium. Journal of Management, 26, 463-488.
Page 40 of 42

Why High and Low Performers Leave CAHRS WP07-11
Harrison, D.A., Virick, M., & William, S. (1996). Working without a net: Time, performance, and
turnover under maximally contingent rewards. Journal of Applied Psychology, 81, 331-
345.
Hinrichs, J.R. (1975). Measurement of reasons for resignation of professionals: Questionnaire
versus company and consultant exit interviews. Journal of Applied Psychology, 60, 530-
532.
Ibarra, H. (1995). Race, opportunity, and diversity of social circles in managerial networks.
Academy of Management Journal, 38, 673-703.
Jackson, S.E. & Schuler, R.S. (1985). A meta-analysis and conceptual critique of research on
role ambiguity and role conflict in work settings. Organizational Behavior and Human
Decision Processes, 36, 16–78.
Judge, T.A.. Cable, D.M.. Boudreau, J.W., & Bretz, R.D. (1995). An empirical investigation of
the prediction of executive career success. Personnel Psychology, 48, 485-519.
Judge, T. A., Thoresen, C. J., Bono, J. E., & Patton, G. K. (2001). The job satisfaction-job
performance relationship: A qualitative and quantitative review. Psychological Bulletin,
127, 376-407.
Lam, S. S. K., & Dreher, G. F. (2004). Gender, extra-firm mobility, and compensation attainment
in the United States and Hong Kong. Journal of Organizational Behavior, 25, 791-805.
Lavelle, L. (2003). After the jobless recovery, a war for talent. Business Week: 92.
Lee, T. W., Mitchell, T. R., Holtom, B. C., McDaniel, L. S., & Hill, J. W. (1999). The unfolding
model of employee turnover: A replication and extension. Academy of Management
Journal, 42, 450-462.
Lee, T. W., Mitchell, T. R., Sablynski, C. J., Burton, J. P., & Holtom, B. C. (2004). The effects of
job embeddedness on organizational citizenship, job performance, volitional absences,
and voluntary turnover. Academy of Management Journal, 47, 711-722.
Lee, T. W., Mitchell, T. R., Wise, L., & Fireman, S. (1996). An unfolding model of voluntary
employee turnover. Academy of Management Journal, 39, 5-36
Lefkowitz, J., & Kahn, M.L. (1969). Validity of exit interviews. Personnel Psychology, 69, 445-
455.
March, J.G., & Simon, H.E. (1958). Organizations. New York: John Wiley.
McEvoy, G. M., & Cascio, W. F. (1987). Do good or poor performers leave? A meta-analysis of
the relationship between performance and turnover. Academy of Management Journal,
30, 744-762.
Milkovich, G. T., & Newman, J. M. (2002). Compensation (7th ed.). Homewood, IL: Irwin.
Mobley, W.H., Griffeth, R.W., Hand, H.H., & Meglino, B.M. (1979). Review and conceptual
analysis of the employee turnover process. Psychological Bulletin, 86, 493-522.
Mossholder, K.W., Bedeian, A.G., Norris, D.R., Giles, W.F., & Feild, H.S. (1988). Job
performance and turnover decisions: Two field studies. Journal of Management, 14, 403-
414.
Murphy, K.R. 1989. Is the relationship between cognitive ability and job performance stable over
time? Human Performance, 2, 183-200.
Ngo, H.Y., Foley, S., & Hoi, R. (2005). Work role stressors and turnover intentions: A study of
professional clergy in Hong Kong. International Journal of Human Resource
Management, 16, 2133–2146.
Pulakos, E.D., & Schmitt, N. 1995. Experience-based and situational interview questions:
Studies of validity. Personnel Psychology, 48, 289-308.
Rogers, W.H. (1993). sg17: Regression standard errors in clustered samples. Stata Technical
Bulletin, 56, 34-40.
Salamin, A., & Hom, P.W. (2005). In search of the elusive U-shaped performance-turnover
relationship: Are high performing Swiss bankers more liable to quit? Journal of Applied
Psychology, 90, 1204-1216.
Steel, R.P., Griffeth, R.W., & Hom, P.W. (2002). Practical retention policy for the practical
manager. Academy of Management Executive, 16, 149-161.
Page 41 of 42

Why High and Low Performers Leave CAHRS WP07-11
Stumpf, S. A., & Dawley, P. K. (1981). Predicting voluntary and involuntary turnover using
absenteeism and performance indices. Academy of Management Journal, 24, 148-163.
Sturman, M.C. 2003. Searching for the inverted U-shaped relationship between time and
performance: Meta-analyses of the experience/performance, tenure/performance, and
age/performance relationships. Journal of Management, 29, 609-640.
Trevor, C.O. (2001). Interactions among actual ease-of-movement determinants and job
satisfaction in the prediction of voluntary turnover. Academy of Management Journal, 44,
621-638.
Trevor, C. O., Gerhart, B., & Boudreau, J. W. (1997). Voluntary turnover and job performance:
Curvilinearity and the moderating influences of salary growth and promotions. Journal of
Applied Psychology, 82, 44-61.
Wanberg, C. R., Kanfer, R., & Rotundo, M. (1999). Unemployed individuals: Motives, job-search
competencies, and job-search constraints as predictors of job seeking and
reemployment. Journal of Applied Psychology, 84, 897-910.
Wilk, S.L., Desmarais, L.B., Sackett, P.R. (1995). Gravitation to jobs commensurate with ability:
Longitudinal and cross-sectional tests. Journal of Applied Psychology, 80, 79-85.
Williams, C. R., & Livingstone, L. P. (1994). Another look at the relationship between
performance and voluntary turnover. Academy of Management Journal, 37, 269-298.
Page 42 of 42