Causes and consequences of collective turnover: a meta-analytic review.

TL;DR: Results generally support expected relationships across the 6 categories of collective turnover antecedents, with somewhat stronger and more consistent results for 2 categories: human resource management inducements/investments and job embeddedness signals.

Abstract: Given growing interest in collective turnover (i.e., employee turnover at unit and organizational levels), the authors propose an organizing framework for its antecedents and consequences and test it using meta-analysis. Based on analysis of 694 effect sizes drawn from 82 studies, results generally support expected relationships across the 6 categories of collective turnover antecedents, with somewhat stronger and more consistent results for 2 categories: human resource management inducements/investments and job embeddedness signals. Turnover was negatively related to numerous performance outcomes, more strongly so for proximal rather than distal outcomes. Several theoretically grounded moderators help to explain average effect-size heterogeneity for both antecedents and consequences of turnover. Relationships generally did not vary according to turnover type (e.g., total or voluntary), although the relative absence of collective-level involuntary turnover studies is noted and remains an important avenue for future research.

Summary

Introduction

• Given growing interest in collective turnover (i.e., employee turnover at unit and organizational levels), the authors propose an organizing framework for its antecedents and consequences and test it using meta-analysis.
• The authors take a broad view of collective turnover to organize its relevant causes and consequences conceptually and then use meta-analysis to determine which factors matter most from an empirical standpoint.
• They classified consequences in terms of productivity, firm performance, and customer outcomes.
• The authors focus on antecedents for two primary reasons.
• The authors also include meta-analyses of consequences to highlight findings that extend their understanding beyond existing research.

Causes of Collective Turnover

• The authors organize the multitude of collective turnover antecedents into six categories: (a) HRM inducements and investments, (b) HRM expectation-enhancing practices, (c) shared attitudes toward the job and organization, (d) quality of work group and supervisory relations, (e) job alternative signals, and (f) job embeddedness signals.
• In light of these arguments, the authors expect the following: Hypothesis 4: Climate, cohesiveness, supervisory relations, and OCBs will be negatively related to collective turnover; age diversity and tenure diversity will be positively related to collective turnover.
• Based on the arguments above, the authors expect the following: Hypothesis 5: Alternative availability, average employee education, and size will be positively related to collective turnover; site quality, unemployment rate, and establishment age will be negatively related to collective turnover.
• Higher job embeddedness is associated with decreased turnover, as individuals perceive a high cost (e.g., sacrifice) of leaving environments to which they feel a higher degree of attachment (e.g., links and fit; Mitchell et al., 2001)—in essence, reducing the desirability of movement (March & Simon, 1958).

Moderators of Antecedent-Turnover Relationships

• Within antecedent-turnover relationships, the authors proposed four moderators based on theory and past research.
• Hypothesis 8 stated that internal mobility-turnover relationships would be stronger when measured in terms of actual promotion rates rather than perceived internal mobility.
• Results did not support Hypothesis 9 or 10.

Consequences of Collective Turnover

• Meta-analytic results for the collective turnover- organizational effectiveness relationships are reported in Table 6.
• As stated in Hypothesis 11, the authors expected that collective turnover would be negatively related to proximal and distal measures of organizational effectiveness (and positively related to measures of counterproductivity, error/loss rates, and absenteeism).
• No significant relationships were found for absenteeism, operating profit, return on assets, return on equity, sales, or sales growth.
• These results indicate partial support for Hypothesis 11.

Moderators of Turnover-Effectiveness Relationships

• Given the diverse array of settings in which turnover’s consequences have been studied, the authors expect that overall effect-size estimates will be heterogeneous, thereby suggesting moderators.
• The magnitudes of collective turnover- effectiveness relationships will be stronger when collective turnover is measured as voluntary turnover rather than when collective turnover is measured as a total rate.
• Immunity effects can be generated by several means, for instance, minimizing the training or educational requirements necessary to perform a job, de-skilling or otherwise dividing components of a job into discrete tasks, increasing routinization or centralization, decreasing autonomy, or, more generally, moving toward a control-based (Arthur, 1992, 1994) HR system.
• Second, a manual search of articles published in Academy of Management Journal, Journal of Applied Psychology, Journal of Management, and Personnel Psychology was conducted from the year 2000 forward.
• Altogether, these procedures resulted in the initial identification of 128 potentially eligible studies.

Coding Procedure

• Discrepancies were resolved by discussion and unanimous consensus over a series of meetings until the final variable groupings were decided.
• With the final variable list in hand, the authors then defined each variable/construct to facilitate valid and consistent organization of effect sizes into appropriate categories for analysis.
• The authors coded turnover type according to voluntariness (i.e., voluntary, involuntary, or total/combined).
• Distal outcomes included sales, sales efficiency, sales growth, operating profit, profit margin, return on assets, and return on equity (proximal = 0, distal =1).

Meta-Analytic Procedures

• Random-effects meta-analyses were conducted according to procedures recommended by Hunter and Schmidt (1990, 2004) and Borenstein, Hedges, Higgins, and Rothstein (2009).
• In addition, the authors report three statistics to quantify heterogeneity: the 𝑄𝑄 statistic, the weighted sum of squares and its associated 𝑝𝑝 value (a statistically significant 𝑝𝑝 value allows one to reject the null hypothesis that effect sizes are constant across studies);.
• This is a result of efforts to avoid double counting and to use the most detailed information available.
• Finally, the summary category was composed of any kind of measured turnover (i.e., voluntary, involuntary, or total) but, to avoid double counting, excluded total turnover effect sizes from a study if voluntary and/or involuntary effect sizes for that same study were available.
• For turnover type, the authors report meta-analytic results separately for total, voluntary, and involuntary turnover (as well as for the summary category).

Antecedents of Collective Turnover

• In Hypothesis 1, the authors stated that HRM inducement and investment practices would negatively relate to collective turnover.
• This overall pattern is reinforced by statistically significant 𝑄𝑄 values, which indicate that most effect-size estimates are heterogeneous.
• Results for the quality of work group and supervisory relations-collective turnover relationship are reported in Table 3.
• Thus, the authors found partial support for Hypothesis 4.
• Unexpectedly, the authors did not find a significant effect between turnover and size or average employee education.

Moderators of Turnover-Consequence Relationships

• Examining 𝑄𝑄 statistics for turnover- effectiveness relationships revealed that moderators were likely.
• The authors hypothesized that turnovereffectiveness relationships would be weaker for distal (vs. proximal) outcomes (Hypothesis 13), weaker for between-organization (vs. within-organization) studies (Hypothesis 14), weaker in industries characterized by high turnover rates (Hypothesis 15a), and weaker in settings where job complexity is low (Hypothesis 15b).
• Stated differently, consistent with Hypotheses 13 and 14, relationships were weaker for distal outcomes and between-organization studies (i.e., the overall average negative effect shifts closer to zero for distal outcomes and betweenorganization studies).
• Upper quadrants depict the strongest correlations, highlighting the importance of certain collective characteristics (e.g., average employee age, average employee tenure, unionization percentage) and HR practices (e.g., high-commitment HR systems, internal mobility, routinization) in predicting turnover.

Future Research

• Further, at a conceptual level, some conclusions about turnover interventions (e.g., HR practices) or turnover’s impacts on performance depend, in part, on exactly who leaves.
• Another opportunity pertains to the potential moderating effects of unemployment rates.
• In addition, although the authors account for type of turnover measurement (e.g., summary, total, voluntary, involuntary) and discuss differences in findings where possible, in some cases they were unable to determine how primary sources measured turnover.
• Staffing selectivity related relatively strongly and negatively to collective turnover, while the sophistication of selection systems themselves shared only weak and nonsignificant relationships.

Figures

Table 8

Table 1

Full text

Cause and Consequences 1
Causes and Consequences of Collective Turnover: A Meta-Analytic Review
Angela L. Heavey
Florida International University
Jacob A. Holwerda and John P. Hausknecht
Cornell University
Journal of Applied Psychology (2013), 98(3), 412-453

Cause and Consequences 2
Abstract
Given growing interest in collective turnover (i.e., employee turnover at unit and
organizational levels), the authors propose an organizing framework for its antecedents and
consequences and test it using meta-analysis. Based on analysis of 694 effect sizes drawn from
82 studies, results generally support expected relationships across the 6 categories of collective
turnover antecedents, with somewhat stronger and more consistent results for 2 categories:
human resource management inducements/investments and job embeddedness signals. Turnover
was negatively related to numerous performance outcomes, more strongly so for proximal rather
than distal outcomes. Several theoretically grounded moderators help to explain average effect-
size heterogeneity for both antecedents and consequences of turnover. Relationships generally
did not vary according to turnover type (e.g., total or voluntary), although the relative absence of
collective-level involuntary turnover studies is noted and remains an important avenue for future
research.
Keywords: collective turnover, organizational performance, retention, meta-analysis

Cause and Consequences 3
The issue of collective turnover—that is, “the aggregate levels of employee departures
that occur within groups, work units, or organizations” (Hausknecht & Trevor, 2011, p. 353)—
has a long history in management and applied psychology research. Discussions of
organizational-level turnover rates extend back nearly a century, as seen in early work addressing
“rates of departure” (Greenwood, 1919, p. 187) and the “stability of employment” (Fish, 1917, p.
162). Topical interest further formalized via several influential accounts of collective turnover’s
causes and consequences (March & Simon, 1958; Mobley, 1982; Price, 1977; Staw, 1980). More
recently, this attention has intensified in terms of empirical studies (e.g., Batt & Colvin, 2011;
Shaw, Dineen, Fang, & Vellella, 2009; Siebert & Zubanov, 2009; Trevor & Nyberg,
2008), theoretical contributions (Dess & Shaw, 2001), and narrative reviews (Hausknecht &
Trevor, 2011; Shaw, 2011). Indeed, over 100 studies have been published on the topic—nearly
two thirds in the last decade alone—mostly in leading journals in management and related fields
(Hausknecht & Trevor, 2011).
Alongside the growing interest, the number of constructs that have been postulated as
potentially related to collective turnover has grown substantially. This increase stems from
turnover rates serving as a key predictor or outcome variable across a wide range of both
emerging and established research topics at group, unit, and firm levels (e.g., high-commitment
human resource [HR] practices, collective attitudes, human and social capital, organizational
demography). Moreover, while general understanding of how different constructs relate to
turnover within a given topical domain has appeared (e.g., within strategic human resource
management [HRM]; see Combs, Liu, Hall, & Ketchen, 2006), there has been little systematic
attention to understanding which constructs matter most across these areas. In this study, we take
a broad view of collective turnover to organize its relevant causes and consequences

Cause and Consequences 4
conceptually and then use meta-analysis to determine which factors matter most from an
empirical standpoint.
Our intended contributions are threefold. First, we classify the multitude of variables that
have been studied to date into a more theoretically parsimonious organizing framework. Second,
we leverage this framework and use meta-analysis to understand which factors most strongly
relate to collective turnover. Given vast differences in sample sizes across studies, we examine
whether between-study differences represent sampling error rather than variation in true effects.
Third, we test several study-level moderators that could explain effect-size heterogeneity. In
doing so, we address a number of extant theoretical propositions that can benefit from the cross-
study analysis that meta-analysis allows. We conclude by discussing implications for research
and practice. Throughout the article, we refer to collective turnover or simply turnover,
recognizing that similar terms exist (e.g., unit-level or organizational turnover, turnover rates).
Organizing Framework
In a recent narrative review, Hausknecht and Trevor (2011) organized causes of
collective turnover into three areas: HRM practices, collective attitudes/perceptions, and
collective characteristics. They classified consequences in terms of productivity, firm
performance, and customer outcomes. Although we aim to test their overall framework
empirically, we focus on antecedents for two primary reasons. First, the collective turnover
literature contains substantially more effect sizes for antecedents than for outcomes. Based on
our literature search and inclusion criterion of at least three available independent effect-size
estimates, we identified 40 antecedents and 12 outcomes (Appendix A lists names, definitions,

Cause and Consequences 5
and sample citations for all 52 variables). Indeed, of the 694 effect-size estimates identified here,
526 (75.8%) pertained to turnover antecedents. From a variable standpoint, 40 of 52 variables
(76.9%) captured presumed turnover causes. Thus, the literature on antecedents is much larger—
and more diffuse—and could therefore benefit substantially from systematic consolidation.
Second, given several recent studies that exclusively focused on turnover’s consequences (e.g.,
Hancock, Allen, Bosco, McDaniel, & Pierce, 2013; T. Park & Shaw, 2013; Shaw, 2011),
systematic examination of its determinants remains an important but as of yet untested issue.
Although we emphasize antecedents, we also include meta-analyses of consequences to highlight
findings that extend our understanding beyond existing research.
Causes of Collective Turnover
We organize the multitude of collective turnover antecedents into six categories: (a)
HRM inducements and investments, (b) HRM expectation-enhancing practices, (c) shared
attitudes toward the job and organization, (d) quality of work group and supervisory relations, (e)
job alternative signals, and (f) job embeddedness signals. Rather than provide an exhaustive
account of the theory and findings for all constructs within each category, we focus on general
descriptions of the categories in the text and include a more detailed rationale in Table l.
1
1
We note up front that, although researchers debate whether antecedents and consequences of
turnover vary across different turnover types (e.g., voluntary vs. involuntary; see Batt & Colvin,
2011; T. Park & Shaw, 2013; Shaw et al., 1998), extant collective turnover research almost
uniformly reflects total or voluntary turnover rates, particularly with regard to antecedents. As
such, our theory and hypotheses generally pertain to total or voluntary turnover rates, yet we also
report relationships with involuntary turnover in the tables in those few cases where possible. We
discuss turnover types more explicitly in later sections.