Turnover Rates and Organizational Performance: A Meta-Analysis
Tae-Youn Park
Vanderbilt University
Jason D. Shaw
University of Minnesota
The authors conducted a meta-analysis of the relationship between turnover rates and organizational
performance to (a) determine the magnitude of the relationship; (b) test organization-, context-, and
methods-related moderators of the relationship; and (c) suggest future directions for the turnover
literature on the basis of the findings. The results from 300 total correlations (N ⫽ 309,245) and 110
independent correlations (N ⫽ 120,066) show that the relationship between total turnover rates and
organizational performance is significant and negative (⫽–.15). In addition, the relationship is more
negative for voluntary (⫽–.15) and reduction-in-force turnover (⫽–.17) than for involuntary
turnover (⫽–.01). Moreover, the meta-analytic correlation differs significantly across several
organization- and context-related factors (e.g., types of employment system, dimensions of organiza-
tional performance, region, and entity size). Finally, in sample-level regressions, the strength of the
turnover rates– organizational performance relationship significantly varies across different average
levels of total and voluntary turnover rates, which suggests a potential curvilinear relationship. The
authors outline the practical magnitude of the findings and discuss implications for future organizational-
level turnover research.
Keywords: meta-analysis, organizational performance, turnover rates
The relationship between turnover rates and organizational per-
formance has been examined from various disciplinary perspec-
tives, including organizational psychology, sociology, economics,
and human resource management. Perhaps because interest in the
topic is highly dispersed, the research literature has provided little
integration; indeed, some extant results seem conflicting. For
example, some studies have shown a negative relationship between
turnover rates and organizational outcomes such as sales (e.g.,
Baron, Hannan, & Burton, 2001; Batt, 2002; Huselid, 1995),
customer service (e.g., Kacmar, Andrews, Van Rooy, Steilberg, &
Cerrone, 2006; Plomondon et al., 2007), profit (e.g., McElroy,
Morrow, & Rude, 2001; Van Iddekinge et al., 2009), and return on
assets (e.g., Messersmith, Guthrie, & Ji, 2010; Shen & Cannella,
2002). But many studies have failed to find negative relationships
(e.g., Arthur, 1994; Guthrie, 2001; Kesner & Dalton, 1994; Shaw,
Duffy, Johnson, & Lockhart, 2005; Wagner, Pfeffer, & O’Reilly,
1984; Zimmerman et al., 2005), and some have even reported
significantly positive associations (e.g., Keck, 1997; Virany, Tush-
man, & Romanelli, 1992).
As momentum in this area grows, three recent reviews have
described the state of organizational-level turnover literature as
“much less well developed” than individual-level turnover re-
search (Shaw, 2011, p. 187), as an “area of inquiry [that] merits
further investigation” (Holtom, Mitchell, Lee, & Eberly, 2008,
p. 252), and as an area where “much remains to be learned”
(Hausknecht & Trevor, 2011, p. 379). As a starting point for
future research, it is worthwhile to consider and summarize
what the existing empirical literature tells us about turnover
rates and organizational performance. Practitioners may also
benefit from a quantitative summary to judge whether they have
correctly or over-stated their intuitive concerns about turnover’s
potential costs and benefits. Our purpose, therefore, is to per-
form a meta-analysis of the relationship between turnover rates
and organizational performance to (a) determine the magnitude
of the relationship between these variables; (b) test
organization-, context-, and methods-related moderators of the
relationship; and (c) suggest future directions for the turnover
literature on the basis of the meta-analytic findings.
Theoretical Perspectives on the Turnover Rates and
Organizational Performance Relationship
The relationship between turnover rates and organizational per-
formance has been examined, in general, under three alternative
views: (a) turnover rates at any level disrupt organizational per-
formance (e.g., Osterman, 1987); (b) turnover rates are most
disruptive at low- to moderate-levels, but the disruptive effects are
attenuated at high levels (e.g., Price, 1977); and (c) turnover rates
This article was published Online First December 17, 2012.
Tae-Youn Park, Owen Graduate School of Management, Vanderbilt
University; Jason D. Shaw, Carlson School of Management, University of
Minnesota.
We thank seminar participants at Cornell University, the Hanken School
of Economics, Tsinghua University, Hong Kong Polytechnic University,
Concordia University, Auburn University, Technical University Munich,
and the London School of Economics for constructive comments on other
versions of this article. We also thank Jacqueline Thompson for editorial
assistance.
Correspondence concerning this article should be addressed to Tae-
Youn Park, Owen Graduate School of Management, Vanderbilt University,
401 21st Avenue South, Nashville, TN 37221. E-mail: TaeYoun.Park@
owen.vanderbilt.edu
Journal of Applied Psychology © 2012 American Psychological Association
2013, Vol. 98, No. 2, 268 –309 0021-9010/13/$12.00 DOI: 10.1037/a0030723
268
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
enhance organizational performance at low- to moderate-levels,
but disrupt performance at high levels (e.g., Abelson & Baysinger,
1984). We summarize the theoretical explanations for these views
below.
Model 1: Linear Negative Relationship View From
Human and Social Capital Theories
Human and social capital theories suggest that turnover rates at
any levels hurt organizational performance. Human capital theory
proposes that more experienced employees perform better because
they accumulate the knowledge and skills (i.e., human capital)
necessary to perform the job (Strober, 1990). From this view, when
experienced employees leave, an organization suffers because it
loses stored/accumulated human capital (Osterman, 1987; Strober,
1990). Organizations may replace employees who leave, but time
must pass before replacements accumulate similar levels of human
capital. Moreover, turnover generates additional human resource
management costs such as recruitment, selection, and training
expenses.
Similar to human capital theory, social capital theory suggests
that turnover is costly because it depletes social capital—“a re-
source reflecting the character of social relations within the orga-
nization, realized through members’ levels of collective goal ori-
entation and shared trust” (Leana & Van Buren, 1999, p. 540).
Increases in turnover rates disrupt an organization’s social fabric
and its operational and collective functions (Dess & Shaw, 2001;
Shaw, Duffy, et al., 2005). In addition, turnover engenders addi-
tional newcomer socialization costs. Human and social capital
theories focus on increases in turnover rates and suggest that
turnover rates are linearly and negatively related to organizational
performance (Shaw, Gupta, & Delery, 2005).
This human and social capital theory perspective appears to
have the most empirical support in the literature. For example,
research has found increases in turnover rates to be negatively
related to customer satisfaction (Morrow & McElroy, 2007), sales
growth (Batt, 2002), return on equity (Cannella & Hambrick,
1993), and profit (Kacmar et al., 2006). In addition, Van Iddekinge
et al. (2009) tested the causal direction of the turnover rates–
performance relationship and showed that retention rates (the
inverse of turnover rates) significantly and positively influenced
the change in unit profitability over time.
In contrast, other theoretical and empirical extensions in the
turnover literature provide a more nuanced picture of the turnover
rates– organizational performance relationship; in particular, they
suggest a curvilinear relationship. From these views, many previ-
ous empirical studies are limited because they fail to address
possible curvilinearity. The form of the curvilinear relationship is
disputed, however, depending on the root of theoretical view-
points, whether from organizational learning and control theories
or from cost-benefit theories. These two alternative models are
discussed below.
Model 2: Attenuated Negative Relationship View From
Organizational Learning and Control Theories
Sharing some common elements of human capital theory, the
organizational learning and control theories suggest an attenuated
negative relationship between turnover rates and organizational
performance. On average, organizations with low turnover rates
have accumulated much human capital. When employees leave,
replacement employees cannot equal the lost human capital until
much time passes. In contrast, organizations with high turnover
rates have workforces that lack accumulated human capital; re-
placements can quickly build equivalent capital and rapidly negate
human capital losses. In addition, continuous workforce replace-
ment becomes routine, so marginal turnover costs are reduced
(Shaw, Duffy, et al., 2005; Shaw, Gupta, & Delery, 2005). From
this viewpoint, an increase in turnover rates from low-to-moderate
levels are more disruptive to organizational performance than an
increase in turnover rates from moderate-to-high levels (Price,
1977; Shaw, Gupta, & Delery, 2005).
Several studies have found evidence supporting the attenuated
negative relationship. Shaw, Gupta, and Delery (2005) studied the
trucking and concrete pipe industries and found that the relation-
ship between voluntary turnover rates and organizational perfor-
mance was strongly negative initially but attenuated at higher
turnover levels. Similarly, Ton and Huckman (2008), in a sample
of bookstores, found severe performance decreases as turnover
rates went from low to moderate levels, but the relationship was
attenuated as the rates increased from moderate to high levels.
Interestingly, Alexander, Bloom, and Nuchols (1994) hypothe-
sized an inverted-U-shaped relationship, which we describe below,
but their empirical results provided some support for the attenuated
negative relationship pattern.
Model 3: Inverted-U Relationship View From
Cost-Benefit Theories
In contrast to the attenuated negative relationship, another cur-
vilinear view on the turnover rates–performance relationship pre-
dicts that turnover rates are beneficial at low levels but costly at
high levels. Specifically, these cost-benefit theories propose that
turnover conveys greater benefits than costs at low to moderate
turnover levels, but costs outweigh benefits at moderate to high
levels where the turnover rates–organizational performance rela-
tionship becomes an inverted-U (Abelson & Baysinger, 1984;
Dalton & Todor, 1979; Staw, 1980).
According to this perspective, some turnover benefits organiza-
tions by reducing compensation costs, revitalizing the workforce,
and sorting out poor performers. Turnover reduces compensation
costs related to base pay, vacation, sick leave, and insurance
premiums (Alexander et al., 1994; Jeswald, 1974). In addition,
turnover revitalizes organizations by introducing newcomers who
bring current knowledge and skills (Alexander et al., 1994), re-
ducing employee homogeneity, and increasing the diversity of
ideas (Schneider, Goldstein, & Smith, 1995). Moreover, turnover
can eliminate poor performers and misfits who disrupt the orga-
nization’s culture and values (Abelson & Baysinger, 1984; Dalton
& Todor, 1979). Hence, an optimal turnover rate is found at the
point where benefits maximally exceed the costs. Specifically, at
low to moderate levels where benefits are greater than costs,
increased turnover rates can contribute to organizational perfor-
mance, but as rates rise beyond moderate levels, they have nega-
tive effects.
Several studies have examined the inverted-U-shaped relation-
ship between turnover rates and organizational performance, but
the literature provides very little supportive evidence. Glebbeek
269
TURNOVER RATES AND ORGANIZATIONAL PERFORMANCE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
and Bax (2004) found a curvilinear form, but the relationship
failed to conform to the predicted inverted-U shape; rather perfor-
mance peaked at very high turnover rates. Siebert and Zubanov
(2009) tested the inverted-U hypothesis, but their results failed to
support the curvilinear relationship clearly (Shaw, 2011). The
strongest evidence is found in Meier and Hicklin’s (2007) study;
using a sample of Texas school districts, they found that low levels
of district-level turnover rates were positively related to district
SAT and ACT scores, but the relationship was negative at higher
levels (i.e., an inverted-U-shaped relationship).
Organization- and Context-Related Moderators of the
Relationship Between Turnover Rates and
Organizational Performance
Researchers have identified several factors that influence the
relationship between turnover rates and organizational perfor-
mance (Hausknecht & Trevor, 2011; Shaw, 2011). We examine
three major organization- and context-related factors that possibly
moderate the relationship: (a) turnover rate types, (b) dimensions
of organizational performance, and (c) organizational contexts and
characteristics (e.g., employment system, entity size, industry, and
region).
Turnover Rate Types
Turnover researchers have often operationalized turnover rates
as the number of departing employees divided by the total number
of employees (e.g., Arthur, 1994; Guthrie, 2001). This operation-
alization, which we call total turnover rates, omits employees’
reasons for leaving. A more refined approach is distinguishing
voluntary and involuntary turnover rates based on reasons for
leaving; voluntary turnover rates refer to the proportion of em-
ployee departure initiated by employees (e.g., resignations), and
involuntary turnover rates refer to the proportion of departure
initiated by organizations (e.g., firings, discharges, dismissals,
terminations; Shaw, Delery, Jenkins, & Gupta, 1998). Voluntary
turnover rates include resignations for higher wages, career oppor-
tunities, further education, and job dissatisfaction for example
(Campion, 1991) and exclude discharges, retirements, transfers,
and promotions (Batt, 2002). In contrast, involuntary turnover
rates include resignations caused, for example, by failure to meet
expectations and expired employment contracts (Campion, 1991;
McElroy et al., 2001). Reduction-in-force (RIF) turnover (down-
sizing) is a separate category because “no replacement employees
are planned and the departing employees are presumed to have
been at least minimally competent” (McElroy et al., 2001, p.
1295).
1
Researchers have often suggested that voluntary and involuntary
turnover have different consequences (e.g., Hausknecht & Trevor,
2011; Holtom et al., 2008; Shaw, 2011). Highly skilled, high-
performing employees may be more likely to leave voluntarily
because they have external employment opportunities (Trevor,
2001). For an organization, voluntary quits are often surprising and
unmanageable (Shaw et al., 1998). Thus, voluntary turnover rates
are likely to be negatively related to organizational performance.
In contrast, the relationship between involuntary turnover rates
and organizational performance has long been assumed to be
positive because organizations choose to discharge employees for
individual performance deficiencies or other behavioral problems
(Holtom et al., 2008). Assuming that poor performers are properly
replaced with better performers, the removal of poor performers
should be associated with better organizational performance (Dal-
ton, Todor, & Krackhardt, 1982; Hollenbeck & Williams, 1986).
In addition, this sorting effect may help remedy poor hiring deci-
sions (Shaw et al., 1998), and maintain performance-oriented
norms among remaining employees (Trevino, 1992). Some re-
searchers have, however, recently questioned the presumed posi-
tive relationship and have proposed that the involuntary turnover
rates and organizational performance have a negative relationship
instead. Hausknecht and Trevor (2011) argued that high involun-
tary turnover rates “may have little to do with the employee
movement per se (which is the foundation for the voluntary
turnover rate hypothesis) but may instead simply reflect a low-
quality workforce and the subsequent poor performance that this
group is expected to provide” (p. 369). From a somewhat different
view, Batt and Colvin (2011) suggested that both voluntary and
involuntary turnover disrupt organizational performance because
both incur recruitment and training costs and disrupt social con-
nections. Although their data failed to fully support the argument,
the relationship was in a direction consistent with their expecta-
tion: in the customer satisfaction regression model the coefficient
for involuntary turnover rates was negative although not statisti-
cally significant. In sum, the literature has predominantly focused
on a positive relationship between involuntary turnover rates and
organizational performance, but recent attention reports a negative
relationship.
Views on the relationship between RIF turnover rates and or-
ganizational performance have also been equivocal. RIF objectives
are often to enhance productivity and profitability by eliminating
redundant or unnecessary jobs and employees (Dewitt, 1998; Free-
man & Cameron, 1993). Thus, RIF proponents argue that RIF
reduces organizational slack and operating costs, and enhances
efficiency and profitability (e.g., Brookman, Chang, & Rennie,
2007; Cascio & Young, 2003; Chalos & Chen, 2002; Palmon, Sun,
& Tang, 1997; Yu & Park, 2006). Yet, opponents argue that RIF
hurts organizational performance because it increases employment
instability and voluntary turnover rates among those remaining
(Trevor & Nyberg, 2008). In addition, RIF disrupts social capital
(Pfeffer, 1998) and engenders behavioral rigidity and risk aversion
(Cameron, Whetton, & Kim, 1987; Cascio, 1993), which overturn
the temporal benefits (Hallock, 1998). Recently, Datta, Guthrie,
Basuil, and Pandey (2010) conducted a thorough qualitative re-
view of the RIF turnover rates and organizational performance
relationship and suggested that, despite somewhat equivocal em-
pirical findings, the overall relationship was likely negative.
In sum, based on the existing literature, we can reasonably
expect that voluntary turnover rates will be negatively related to
organizational performance. Views are contradictory about invol-
untary and RIF turnover effects, but recent qualitative reviews
1
Although many organizations classify turnover as voluntary, involun-
tary, and RIF turnover, some turnover types are not clearly voluntary or
involuntary, such as retirement, health problems, pregnancy, and separa-
tion by mutual agreement. Accordingly, the turnover literature may benefit
from the development and use of alternative classifications turnover rate
types.
270
PARK AND SHAW
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
suggest that RIF turnover rates and organizational performance
will also be negatively related.
Dimensions of Organizational Performance
The broad concept of organizational performance comprises
many operationalizations (e.g., P. J. Richard, Devinney, Yip, &
Johnson, 2009). Turnover researchers have often categorized
performance into proximal (workforce-related outcomes) and
distal (financial, market, and shareholder return) outcomes.
Turnover research has most often examined workforce-related
performance such as productivity, partly because human and
social capital theory foundations can be most directly applied to
those proximal outcomes (Dess & Shaw, 2001; Osterman, 1987;
Shaw, Gupta, & Delery, 2005). Financial and market-oriented
organizational performances have been regarded as distal out-
comes because several other factors, such as general economic
conditions, may dilute the direct turnover effects. For example,
Kacmar et al. (2006) proposed a turnover-efficiency-profit
model showing that turnover reduces restaurant profits by
lengthening customer wait-time. As such, the most proximal
measures of organizational performance might be those related
to employee interactions and attitudes such as customer satis-
faction and absenteeism. Time must pass before the cycle of
customer service and employee attitude changes affect cus-
tomer spending, unit-level workforce productivity, and eventu-
ally unit profits. Indeed, the literature has generally assumed,
and some evidence has found, that turnover rates will be more
strongly related to workforce-related measures than financial
measures (e.g., Huselid, 1995; Kacmar et al., 2006; Shaw,
Gupta, & Delery, 2005). In this meta-analysis, we categorize
the organizational performance dimensions into three broad
categories—the most proximal, moderately proximal, and dis-
tal—and we expect that the turnover rates–performance rela-
tionship will be strongest for the most proximal measures (e.g.,
customer satisfaction, employee work attitudes, absenteeism),
modest for moderately proximal measures (e.g., quality, safety,
workforce productivity), and weak for distal ones (e.g., finan-
cial performance).
Organizational Context and Characteristics
The relationship between turnover rates and organizational per-
formance may be different depending on the context or environ-
ment in which turnover occurs (e.g., Arthur, 1994; Batt & Colvin,
2011; Shaw, Gupta, & Delery, 2005). The organizational literature
frequently mentions several contextual factors as potentially im-
portant to the turnover rates–performance relationship. Next, we
briefly discuss these factors: employment systems, entity size,
industries, and region.
Organizations use different employment systems in their ap-
proaches to human resource management. The strategic human
resource management literature (e.g., Arthur, 1992, 1994;
Shaw, Gupta, & Delery, 2005) suggests that organizations
shape employee behaviors and work attitudes using two dis-
tinctive employment systems: (a) primary employment systems
that forge psychological links between organizational and em-
ployee goals (also called commitment systems), and (b) sec-
ondary employment systems that emphasize labor cost reduc-
tion, efficiency improvement, and employee compliance with
specified rules and procedures (also called control systems).
The two employment systems often coexist in an organization
(Lepak & Snell, 1999) depending on the employees (Bamberger
& Meshoulam, 2000; Delery & Shaw, 2001; Lepak & Shaw,
2008; Lepak, Taylor, Tekleab, Marrone, & Cohen, 2007;
Siebert & Zubanov, 2009). For example, full-time managers are
more appropriately managed under primary or commitment-
based employment systems because they need less supervision
and have more discretion in their job tasks. Part-time employees
are typically managed under secondary or control-based em-
ployment systems because they perform routine tasks with
clearly specified rules and procedures.
Researchers have often suggested that turnover rates more
strongly and negatively affect organizational performance under
primary employment systems than under secondary employment
systems (Arthur, 1994; Guthrie, 2001). Because organizations
invest more in pay, training, benefits, and socialization programs
for employees under primary systems, their turnover is more costly
in terms of lost investments and human and social capital deple-
tion. In contrast, organizations select secondary system employees
less carefully and invest less in their services, so their departure
depletes less human and social capital (Shaw, Dineen, Fang, &
Vellella, 2009). For example, in a sample of retail chain employ-
ees, Siebert and Zubanov (2009) compared full-time employees
under a commitment system and part-time employees under a
control system and found that turnover rates were more strongly
and negatively related to sales when the turnover occurred in
commitment systems.
Executives strongly influence organizational performance be-
cause they make important strategic decisions (viz., upper echelon
theory; Hambrick & Mason, 1984). Departures among executive
team members may be the most strongly related to organizational
performance because of lost information necessary for strategic
decisions and altered executive team composition (Virany et al.,
1992; Wagner et al., 1984). Executive turnover also incurs signif-
icant human resource management costs because executives are
managed under distinctive and elaborate employment systems
designed to carefully select, motivate, and retain them (e.g., Ger-
hart & Rynes, 2003). Thus, in our analyses, we separate executive-
or top-management-team turnover rates from turnover rates of
employees in primary and secondary systems. To summarize, we
expect that turnover rates and organizational performance will be
more strongly and negatively related in samples managed by
primary and executive employment systems than in those managed
by secondary employment systems.
The literature holds two contrasting views about the moderating
effects of entity size on turnover rates–performance relationships.
Some have proposed that larger organizations will show a weak-
ened negative relationship because larger groups can buffer turn-
over’s disruptions (e.g., Green, Anderson, & Shivers, 1996; Koz-
lowski & Bell, 2003). In addition, equivalent turnover rates will
inflict less damage on larger organizations because they can better
withstand the same proportional information losses (Carley, 1992).
Others, in contrast, have argued that larger organizations will show
stronger negative turnover rates–performance relationships be-
cause smaller entities can handle socialization and adjustment
processes more efficiently (Hausknecht, Trevor, & Howard, 2009).
271
TURNOVER RATES AND ORGANIZATIONAL PERFORMANCE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Thus, entity size is an important moderator, but the direction and
magnitude of the effect remains unknown.
Moreover, we expect industry to moderate the turnover rates
and organizational performance relationship. Strategic human
resource management and human capital theory literature sug-
gest that the importance of human capital varies across indus-
tries because organizations adopt different technology and work
structures depending on the characteristics of their industries
(e.g., Datta, Guthrie, & Wright, 2005; Dess & Shaw, 2001). For
example, Datta et al. (2005) argued that in industries with high
levels of capital intensity (e.g., manufacturing), organizational
decision makers place greater emphasis on leveraging invest-
ments in technology, equipment, and physical resources and
place relatively less emphasis on human capital development.
In other industries (e.g., health care, hospitality) employees
play a central role in the functioning of the organization and
therefore human capital losses through high turnover rates may
have substantial negative effects on performance. As such, it is
reasonable to expect that the relationship between turnover
rates and organizational performance will be stronger in indus-
tries where the leveraging of human capital is more important to
organizational performance than in industries with high capital
intensity. In support of this line of reasoning, Shaw, Park, and
Kim (2012) found that the negative relationship between turn-
over rates on organizational performance were exacerbated
among organizations that invested heavily in human capital (see
also, Arthur, 1994; Guthrie & Datta, 2008).
Last, we anticipate that region may moderate the relationship
between turnover rates and organizational performance. Labor
market policies, regulations, and human resource management
practices vary dramatically across regions of the world (Ahmad
& Schroeder, 2003; Pfeffer, 1998). In particular, there are
considerable differences in the rigidity of labor markets across
regions. European labor markets tend to be less flexible than
those in North America and Asia because of strict employment
policies, heavy regulation, and emphasis on collective bargain-
ing agreements. These characteristics likely not only reduce the
frequency and the variance in voluntary and involuntary turn-
over rates—serving to reduce the bivariate relationship— but
may also increase the predictability of turnover. The ability to
plan and prepare for turnover events may lessen negative effects
on organizational performance.
Methods-Related Moderators of the Relationship
Between Turnover Rates and Organizational
Performance
We also explore possible differences in the turnover rates and
organizational performance relationship by using methods-related
moderators. We identify potential upward/downward biases on the
turnover rates-performance correlations caused by variance in
methods rather than true theoretical variance. Specifically, we
examine three research design factors (unit of analysis, data struc-
ture, and source of turnover rates information) and three publica-
tion factors (role of turnover rates, hypothesized relationship, and
publication status) that possibly moderate the turnover rates-
performance relationship.
Research Design
Research design-related factors may moderate the relationship
between turnover rates and performance, including (a) unit of
analysis (unit-level vs. organization-level), (b) data structure
(cross-sectional vs. lagged vs. panel), and (c) source of turnover
rates information (organizational record vs. key information).
Shaw (2011) and Hausknecht and Trevor (2011) suggested that
considering distinctions between cross-organization samples (with
different policies, practices, and organizational forms) and cross-
unit samples (with similar policies, practices, and organizational
forms) could potentially provide better understanding of the rela-
tionship between turnover rates and organizational performance.
Cross-organization samples offer some advantages because the
variation in turnover rates and organizational performance can be
large, and such samples allow researchers to explore potential
contextual moderators including industry dynamics (e.g., Guthrie
& Datta, 2008) and staffing and employment policy differences
(e.g., Bamberger & Philips, 1991; Lepak et al., 2007). In contrast
to the cross-organization samples, cross-unit studies can be better
for addressing causality issues by holding certain threats to internal
validity constant (Shadish, Cook, & Campbell, 2001), and by
ensuring consistent definition and measurement of turnover rates.
We make no specific prediction about which unit of analysis
produces stronger turnover rates-performance correlations, but it
would be informative to examine whether and how the unit of
analysis moderates the relationship.
Another possible design-related moderator is data structure:
cross-sectional, lagged, and panel data structures. Meta analytic
summary using zero-order correlations fails to ensure causality.
Reverse causality concerns might be relatively more serious when
turnover rates and organizational performance are measured con-
currently than when a time-lagged performance variable is used.
Panel data are another possible data structure. For regression-based
data analysis, panel data might advantageously address reverse
causality because panel data allow researchers to control for po-
tential confounding factors. In terms of correlations for meta-
analysis, however, correlations from panel data are similar to the
correlations from cross-sectional data because the convention in
the literature is for researchers to report a single between-
organization correlation. For example, Siebert and Zubanov (2009)
analyzed data from 325 retail stores over a 5-year window (1,625
store years), but reported a single between-store correlation (N ⫽
325) of –.24 between the full-time turnover rates and store perfor-
mance. This correlation—the association between turnover rates
averaged across the years of the study and organizational perfor-
mance averaged across the years of the study—is similar to a
cross-sectional correlation, albeit over a longer time window.
Last, the source of turnover rates information is a potential
research design-related moderator. Unit-level research often relies
on archival sources and/or key informants. A concern about using
key informants is that few people can accurately report organiza-
tional information such as turnover rates and organizational per-
formance. In addition, informants with inadequate knowledge and
low motivation to provide accurate data will damage the accuracy
and reliability of organizational information (Delery & Shaw,
2001). To examine the potential bias of using key information data
(versus archival data), we examine the source of turnover rates
information for its possible moderation effects.
272
PARK AND SHAW
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
