A review of job embeddedness: Conceptual, measurement issues, and directions for future research

12
A review of job embeddedness: Conceptual, measurement issues, and directions for future research Mian Zhang a, , David D. Fried b , Rodger W. Griffeth b a School of Economics and Management, Tsinghua University, Beijing, China b Department of Psychology, Ohio University, United States article info abstract Job embeddedness (JE) theory is a relatively new perspective in turnover research. Although the- oretical and empirical implications are encouraging, the construct of JE is still under development. In this paper, we identify five conceptualization and measurement issues warranting additional discussion and research. These are: (1) the composite measure vs. the global measure, (2) concep- tual and measurement problems of community embeddedness, (3) conceptual and measurement problems associated with links, (4) discriminant validity issues and (5) the cultural boundaries of JE. For each issue, we discuss our concerns and suggest future research directions. © 2012 Elsevier Inc. All rights reserved. Keywords: Job embeddedness Turnover Composite measures Global measures Social networks 1. Introduction Voluntary employee turnover is defined as voluntary termination of membership within an organization by an employee who receives monetary compensation for participating in that organization (Mobley, 1982). Voluntary turnover produces tangible costs, such as recruitment and training expense, as well as intangible costs, such as declining employee morale and customer dis- satisfaction (Hom & Griffeth, 1995). Understanding the mechanisms behind voluntary turnover may be helpful in designing in- terventions to reduce turnover and its potential costs (Hom & Griffeth, 1995; Maertz & Campion, 1998; Price, 2001). Following March and Simon's (1958) theory of organization equilibrium, numerous studies have explored the turnover pro- cess (Hom & Griffeth, 1995; Maertz & Campion, 1998). This vein of research has led to many multivariate turnover models (e.g., Hom & Griffeth, 1991; Lee, Mitchell, Sablynski, Burton, & Holtom, 2004; Mobley, 1977; Mobley, Griffeth, Hand, & Meglino, 1979; Mobley, Horner, & Hollingsworth, 1978; Price & Mueller, 1981) with only modest prediction in turnover, typically account- ing for less than 5% to 25% of its variance (Hom & Griffeth, 1995; Maertz & Campion, 1998). The modest predictive strength has inspired recent interest in the development of new theories to gain insight into the turnover pro- cess (Lee & Mitchell, 1994; Lee, Mitchell, & Sablynski, 1999; Lee, Mitchell, Wise, & Fireman, 1996; Lee et al., 2004; Maertz & Campion, 1998; Mitchell, Holtom, Lee, Sablynski, & Erez, 2001). Job embeddedness (JE) theory represents one of these new perspectives (Holtom, Mitchell, Lee, & Eberly, 2008), focusing on factors that encourage an employee to remain with an organization. Several studies reveal that JE predicts incremental variation in turnover after controlling for traditional turnover predictors, such as job satisfaction and quit intentions (Crossley, Bennett, Jex, & Burnfield, 2007; Lee et al., 2004; Mallol, Holtom, & Lee, 2007; Mitchell et al., 2001). Although these results are encouraging, the construct of JE is still under development (Mitchell et al., 2001). With a few exceptions (Crossley et al., 2007; Hom, Rogers, Allen, & Zhang, 2010; Lee et al., 2004; Murphy & Hom, 2008), the scope of most research has been limited to empirical testing of potential outcomes (e.g., employee turnover) of job embeddedness. Few studies have examined con- ceptualization and measurement issues. To date, efforts to develop the construct include the addition and deletion of several items from the original 40-item JE measure (Holtom, Mitchell, Lee, & Tidd, 2006); the integration of ideas from social network theories with embeddedness theory (Hom et al., 2010; Murphy & Hom, 2008); and the development of a seven-item global measure of JE based on a reflective, rather than formative, measurement model (Crossley et al., 2007). Human Resource Management Review 22 (2012) 220231 Corresponding author. E-mail address: [email protected] (M. Zhang). 1053-4822/$ see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.hrmr.2012.02.004 Contents lists available at SciVerse ScienceDirect Human Resource Management Review journal homepage: www.elsevier.com/locate/humres

Transcript of A review of job embeddedness: Conceptual, measurement issues, and directions for future research

A review of job embeddedness: Conceptual, measurement issues, anddirections for future research

Mian Zhang a,⁎, David D. Fried b, Rodger W. Griffeth b

a School of Economics and Management, Tsinghua University, Beijing, Chinab Department of Psychology, Ohio University, United States

a r t i c l e i n f o a b s t r a c t

Job embeddedness (JE) theory is a relatively new perspective in turnover research. Although the-oretical and empirical implications are encouraging, the construct of JE is still under development.In this paper, we identify five conceptualization and measurement issues warranting additionaldiscussion and research. These are: (1) the compositemeasure vs. the globalmeasure, (2) concep-tual andmeasurement problems of community embeddedness, (3) conceptual andmeasurementproblems associatedwith links, (4) discriminant validity issues and (5) the cultural boundaries ofJE. For each issue, we discuss our concerns and suggest future research directions.

© 2012 Elsevier Inc. All rights reserved.

Keywords:Job embeddednessTurnoverComposite measuresGlobal measuresSocial networks

1. Introduction

Voluntary employee turnover is defined as voluntary termination of membership within an organization by an employee whoreceives monetary compensation for participating in that organization (Mobley, 1982). Voluntary turnover produces tangiblecosts, such as recruitment and training expense, as well as intangible costs, such as declining employee morale and customer dis-satisfaction (Hom & Griffeth, 1995). Understanding the mechanisms behind voluntary turnover may be helpful in designing in-terventions to reduce turnover and its potential costs (Hom & Griffeth, 1995; Maertz & Campion, 1998; Price, 2001).

Following March and Simon's (1958) theory of organization equilibrium, numerous studies have explored the turnover pro-cess (Hom & Griffeth, 1995; Maertz & Campion, 1998). This vein of research has led to many multivariate turnover models(e.g., Hom & Griffeth, 1991; Lee, Mitchell, Sablynski, Burton, & Holtom, 2004; Mobley, 1977; Mobley, Griffeth, Hand, & Meglino,1979; Mobley, Horner, & Hollingsworth, 1978; Price & Mueller, 1981) with only modest prediction in turnover, typically account-ing for less than 5% to 25% of its variance (Hom & Griffeth, 1995; Maertz & Campion, 1998).

Themodest predictive strength has inspired recent interest in the development of new theories to gain insight into the turnover pro-cess (Lee & Mitchell, 1994; Lee, Mitchell, & Sablynski, 1999; Lee, Mitchell, Wise, & Fireman, 1996; Lee et al., 2004; Maertz & Campion,1998; Mitchell, Holtom, Lee, Sablynski, & Erez, 2001). Job embeddedness (JE) theory represents one of these new perspectives(Holtom, Mitchell, Lee, & Eberly, 2008), focusing on factors that encourage an employee to remain with an organization. Several studiesreveal that JE predicts incremental variation in turnover after controlling for traditional turnover predictors, such as job satisfaction andquit intentions (Crossley, Bennett, Jex, & Burnfield, 2007; Lee et al., 2004; Mallol, Holtom, & Lee, 2007; Mitchell et al., 2001).

Although these results are encouraging, the construct of JE is still under development (Mitchell et al., 2001).With a few exceptions(Crossley et al., 2007; Hom, Rogers, Allen, & Zhang, 2010; Lee et al., 2004;Murphy &Hom, 2008), the scope of most research has beenlimited to empirical testing of potential outcomes (e.g., employee turnover) of job embeddedness. Few studies have examined con-ceptualization and measurement issues. To date, efforts to develop the construct include the addition and deletion of several itemsfrom the original 40-item JE measure (Holtom, Mitchell, Lee, & Tidd, 2006); the integration of ideas from social network theorieswith embeddedness theory (Hom et al., 2010; Murphy & Hom, 2008); and the development of a seven-item global measure of JEbased on a reflective, rather than formative, measurement model (Crossley et al., 2007).

Human Resource Management Review 22 (2012) 220–231

⁎ Corresponding author.E-mail address: [email protected] (M. Zhang).

1053-4822/$ – see front matter © 2012 Elsevier Inc. All rights reserved.doi:10.1016/j.hrmr.2012.02.004

Contents lists available at SciVerse ScienceDirect

Human Resource Management Review

j ourna l homepage: www.e lsev ie r .com/ locate /humres

It is important to emphasize the necessity of theoretical development and appropriate measures before exploring the relation-ships between JE and other outcomes. To that end, we have identified five issues warranting additional discussion and researchfor JE. We believe turning attention to these areas will help advance the development of JE research. These five areas are: (1) thecomposite measure vs. the global measure, (2) conceptual andmeasurement problems of community embeddedness, (3) concep-tual and measurement problems associated with “links,” (4) discriminant validity issues and (5) the cultural boundaries of JE.Throughout this discussion we provide suggestions and directions for future research to improve the JE construct.

2. An overview of job embeddedness theory

Job embeddedness (JE) is defined as a broad constellation of psychological, social, and financial influences on employee retention(Mitchell et al., 2001; Yao, Lee,Mitchell, Burton, & Sablynski, 2004). These influences are present on the job, aswell as outside the employ-ee's immediatework environment, and are often likened to strands in a “web” or “net” inwhich a person can become “stuck” (Mitchell etal., 2001). Individuals with a greater number of strands becomemore enmeshed in the web and have greater difficulty leaving their job.

Job embeddedness comprises three dimensions: links, fit, and sacrifice. Each dimension is further categorized into two sub-dimensions (i.e., organization and community) based on whether the influences occur on-the-job, or off-the job. Appendix Alists sample items of the latest scale included in Lee et al. (2004).

2.1. Links

Links are defined as formal or informal connections between a person, institutions, or other people (Lee et al., 2004;Mitchell et al.,2001). JE theory posits that a number of links attach an employee andhis or her family in a social, psychological, and financialweb thatincludes co-workers and non-work friends, groups, and the community in which he or she lives. There are links to entities in the or-ganization, such as a team of work colleagues in one's working team, and there are links to entities in one's community, such as rel-atives, friends and social groups. The greater the number of links between the individual and the web and the more important thoselinks are, the more a worker is bound to the job, the supervisor and entities in the organizations such as teams (Lee et al., 2004;Mitchell et al., 2001).

2.2. Fit

Fit refers to an employee's perceived compatibility with the organization and surrounding community (Lee et al., 2004;Mitchell et al., 2001). Good person–organization fit occurs when an employee's personal values, career aspirations, knowledge,skills, and ability are compatible with the organizational culture, and with the requirements of his or her job. In addition, a personwill consider how well he or she fits with aspects of the community and surrounding environment such as climate, weather con-ditions, religious beliefs, and entertainment activities (Mitchell et al., 2001). JE theory postulates that the better the fit with theorganization and the surrounding community, the stronger the ties to the organization (Lee et al., 2004; Mitchell et al., 2001).

2.3. Sacrifice

Sacrifice is the perceived psychological, social, or material cost of leaving one's organization and one's community (Lee et al.,2004; Mitchell et al., 2001). Leaving an organization incurs job-related losses: e.g., giving up familiar colleagues, interesting pro-jects or desirable benefits. Leaving an organization may also result in community-related losses (e.g., giving up an easy commute,good day care, or local club membership).

In the following sections, we discuss each of the five issues we have identified. For each issue, we first point out our concerns andthen furnish possible suggestions on how to ameliorate the concerns. We organize the sequence of the five issues by the followinglogic. First, we discuss theoretical and statistical issues by comparing the two JEmeasures (i.e., the compositemeasure and the globalmeasure) currently in use.We give this topic top priority because investigators usually need to use the twomeasures for understand-ing the construct of JE and make choice between them when conducting empirical studies. Then, we focus on discussing four issuesrelating to the composite measure because (1) most prior studies employed the composite measure, and (2) it has the advantages oftheoretical richness, as we will depict in the later section. Among the four issues, we first discuss community embeddedness becausethis part reflects a unique contribution of JE. Then, we further discuss the three dimensions of JE: links, fit and sacrifice. Finally, wediscuss the cultural boundaries because there is an emerging trend that studies examine JE theory in a cross-cultural context.

3. The composite measure vs. the global measure

Crossley et al. (2007) suggested that their composite measure of JE, as outlined above, has both theoretical and statistical limita-tions. Therefore, they developed an alternative seven-item “global” JE measure that assesses general attachment to the organizationand includes items such as, “I feel attached to this organization,” “It would be difficult for me to leave this organization,” and “I amtightly connected to this organization.” This globalmeasure does not distinguish betweenwork-related andnon-work-related factors,nor does it distinguish between links, fit, and sacrifice. We will now discuss and compare the theoretical and statistical properties ofthe two JE measures.

221M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

3.1. Theoretical comparisons

Crossley et al. (2007) contended that their global measure has two theoretical advantages. First, it allows participants to incor-porate information from their own judgment because its items assess general attachment to the organization rather than specificaspects. In contrast, composite measure may omit some aspects that may be important to the individual or include some irrele-vant aspects. Second, a global measure could capture the unique weightings that an individual may place on different facets whenforming a summary perception. For instance, some respondents may consider community factors as more important than job-related factors when forming embeddedness perceptions. In this case, when using the global measure, an individual could reporthis or her embeddedness perception by mainly assessing community factors. In contrast, when combining factors additively inthe composite measure, the final score of embeddedness is the average of job-related factors and community factors, which can-not capture the unique weightings. Crossley and colleagues found that the global measure predicted variation in intentions toquit, intention to search, and turnover, after controlling for the composite measure and organizational attitudes. They contendedthat their global measure serves as a succinct companion to the original composite measure.

The global measure does not contain items that explicitly refer to influences occurring outside the immediate work environ-ment (i.e., off-the-job factors), a critical aspect of embeddedness theory (Mitchell et al., 2001). In a clarification to Crossley et al.(2007), Crossley, Bennett, Jex, and Burnfield (2011, p.1316) contended that the component of community embeddedness is indi-rectly included in their global measure because the leading instructions of the measure asked respondents to rate the jobembeddedness as a whole [“After considering both work-related (such as relationships, fit with job, benefits) and non-work relat-ed factors (such as neighbors, hobbies, community perks), please rate your agreement with the statements below”].

Compared to the global measure, the composite measure has the advantage of theoretical richness and contributions becauseit explicitly includes non-attitudinal and off-the-job components. Mitchell et al. (2001) argued the JE construct emphasizes non-attitudinal characteristics. The component of links is the best example: all its items assess objective status or numbers (e.g., “howmany co-workers do you interact with regularly?”). Although fit and sacrifice reflect one's perception, the composite measureemphasizes cognitive content of specific facets. For example, employees have a cognitive match between their human capitaland their jobs, though they may have less emotional attachment to their organizations (“my job utilizes my skills and talentswell”). The global measure, however, assesses a more general reaction, which may mix emotional and attitudinal components.

Although researchers have acknowledged the importance of non-work factors in understanding turnover (Hom & Kinicki,2001; Price, 2001; Steers & Mowday, 1981), the construct of JE represents the first attempt to represent a broad assembling ofoff-the-job influences on employee retention (Mitchell et al., 2001). For example, Price and associates assert that kinship respon-sibility is an important antecedent of turnover; the measure of kinship responsibility contains the items measuring marriage sta-tus, number of children and relatives in the same community (Price, 2001). Lee and colleagues contend kinship responsibility issimilar conceptually and empirically to links of community embeddedness (Lee et al., 2004). However, JE extends the content ofkinship responsibility because off-the-job JE additionally contains fit and sacrifices to community.

3.2. Statistical comparisons

The seven-item global measure has a statistical advantage over the composite measure because it was developed based on areflective rather than formativemeasurement model. A reflective model differs from a formative model in the hypothesized direc-tion of causality between items and their latent constructs (Jarvis, Mackenzie, Podsakoff, Mick, & Bearden, 2003; MacKenzie,Podsakoff, & Jarvis, 2005). In a reflective model the direction of causality flows from the latent construct to the items. Items (com-monly referred to as ‘indicators’ of the construct in reflective models) are hypothesized to be highly-correlated because they rep-resent the same construct. A wide array of common methods can be used to evaluate scale properties (e.g., reliability analysis,exploratory factor analysis, and structural equation modeling) of a reflective model (MacKenzie et al., 2005). However, methodsfor evaluating psychometric properties of formative measurement models are less-developed (Jarvis et al., 2003; MacKenzie et al.,2005). Indeed, coefficient alphas and factor loadings are only useful for descriptive purposes in studies that utilize the compositemeasure (Felps et al., 2009; Lee et al., 2004).

There are two problemswith extant compositemeasures that need to be addressed in future research. First, we believe the direc-tion of causality between four items and their latent constructs is ambiguous in Lee et al. (2004) compositemeasure. These four itemsare: “this community I live in is a good match for me”; “I feel like I am a good match for this organization”; “I would incur very fewcosts if I left this organization”; and “I would sacrifice a lot if I left this job”. Although originally conceived as part of a formative mea-surementmodel, it is just as likely that these items are reflective in nature. For example, contrary to a formativemodel (and consistentwith a reflectivemodel), organizational embeddedness (the construct)may cause one to believe that he/shewould sacrifice a lot if he/she left the job (the item). Future research may delete these general items to make the composite measure more consistent with aformative framework.

Second, composite measures assume complete coverage of a construct domain (MacKenzie et al., 2005). However, it seemsthat the items in Lee et al. (2004) JE scale do not fully cover JE's content. For example, leadership may also impact organizationfit because at least one study reveals that high-quality leader–member exchange bolsters staying (Gerstner & Day, 1997). Homet al. (2009) added an item measuring leader–member relationship into their revised scale of job embeddedness.

This problemmay bemore severe in the shortened formof the JE composite scale. Lee et al.'s compositemeasure originally included32 items. Holtom et al. (2006) formed a 21-item shortened version of Lee et al. (2004) measure. Felps et al. (2009) used Holtom et al.(2006) short form of JE scale. However, as MacKenzie et al. (2005, p. 712) argued, “the consequences of dropping a formative indicator

222 M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

from a measurement model are potentially much more damaging than the consequences of dropping a reflective indicator.” Hence,simply reducing scale length may jeopardize content validity of composite measures (Crossley et al., 2007; MacKenzie et al., 2005).

3.3. Conclusions for the comparison

It is hard to conclude which measure is better because both of them have distinct advantages and disadvantages. The compos-ite measure has advantage of theoretical richness. However, as we noted and will further elaborate on, the composite measureneeds improvement to: (1) meet the requirement of a formative measurement model, and (2) test the predictive validity in dif-ferent research settings. The global measure has statistical advantage because it is based on the reflective measurement model.Besides, the compactness of the global measure has an advantage of brevity when restrictions on survey length exist. However,the global measure does not highlight the non-attitudinal and off-the-job factors like the composite measure does.

The research purpose in a specific studymay determine the choice between the compositemeasure and the globalmeasure. If a studyaims to explore the associations between the components of job embeddedness and outcomes, the study should probably use the com-positemeasure. In this case, job embeddedness is likely to be the focal predictor. Ramesh andGelfand's (2010) study is a typical example.The composite measure in their study was needed to test their hypotheses [e.g., “Country will moderate the negative relationship be-tween organization links and turnover such that the relationship is stronger in India than in the United States (Ramesh & Gelfand,2010, p. 810)”] because the measure contains the facet scales. On the other hand, if a study aims to test models using latent constructs,the reflective measure would be a better choice, especially for the studies that use structural equation modeling (SEM).

4. Conceptualization and measurement problems: community embeddedness

According to the JE theory (Mitchell et al., 2001), community factors, such as family ties, marital status, and the number of childrenrequiring parental care, have a significant influence on employee retention. Indeed, the inclusion of non-work factors is one of the keyfactors that distinguish JE from other organizational attachment constructs (e.g., organizational commitment; Mitchell et al., 2001).

Unfortunately, studies have found mixed support for a relationship between community factors and turnover (see Table 1).Mitchell et al. (2001) found that all three dimensions of community embeddedness—community links, fit, and sacrifice—werenegatively-related to turnover in a sample of hospital workers. However, only community links (not fit or sacrifice) were nega-tively associated with turnover in a sample of grocery store workers. Three subsequent studies did not reveal a significant rela-tionship between community embeddedness and turnover when each of the three dimensions was assessed separately(Crossley et al., 2007; Mallol et al., 2007; Ramesh & Gelfand, 2010). Two studies found a significant negative relationship betweentotal community embeddedness (i.e., community embeddedness averaged across the three dimensions) and actual turnover (Leeet al., 2004; Mitchell et al., 2001). However, four other studies did not find a relationship between total community embedded-ness and actual turnover (Allen, 2006; Crossley et al., 2007; Mallol et al., 2007; Ramesh & Gelfand, 2010).

We offer two explanations why community embeddedness is not consistently associated with turnover. First, moderators arelikely to buffer the negative relationship between community embeddedness and turnover. We suggest three constructs mayserve to moderate the community embeddedness—turnover relationship. These are: commute time, job type and financial

Table 1Simple correlations in six job embeddedness studies.

LO LC FO FC SO SC JEO JEC

Mitchell et al. (2001)Grocery store chain −.11/−.14⁎ −.18⁎/−.12 −.18⁎/−.53⁎ −.02/−.09 −.22⁎/−.51⁎ −.11/−.12 −.24⁎/−.57 −.14⁎/−.14⁎

Hospital −.17⁎/−.12 −.14⁎/−.20⁎ −.18⁎/−.41⁎ −.16⁎/−.10 −.13⁎/−.45⁎ −.17⁎/−.15⁎ −.21⁎/−.44⁎ −.20⁎/−.19⁎

Lee et al. (2004)Financial institution −.16⁎/.01 −.16⁎/−.06 −.08⁎/−.56⁎ −.08⁎/−.22⁎ −.02/−.58⁎ −.06/−.17⁎ −.11⁎/– −.13⁎/–

Allen (2006)Financial services −.23⁎/– −.05/–

Crossley et al. (2007)Public organizations −.08/.21⁎ −.10/−.09 −.02/−.43⁎ −.05/−.27⁎ −.08/−.41⁎ .05/−.22⁎ −.08/−.47⁎ −.04/−.28⁎

Mallol et al. (2007)Caucasian −.26⁎/−.27⁎ −.10/−.07 −.12/−.52⁎ −.01/−.22⁎ −.01/−.59⁎ −.06/−.26⁎ −.16⁎/−.60⁎ −.08/−.26⁎

Hispanic −.16⁎/−.11 −.07/−.04 −.14/−.47⁎ .04/.05 −.01/−.50⁎ −.10/−.07 −.13/−.50⁎ −.06/−.05Ramesh and Gelfand (2010)

American call center .00/.12⁎ −.01/−.13⁎ −.02/−.36⁎ .06/.09 −.10/−.67⁎ .02/−.05 −.14⁎/−.51⁎ .03/−.03India call center −.13⁎/.02 −.04/−.22⁎ −.14⁎/−.27⁎ .02/−.12⁎ −.07/−.51⁎ .01/−.19⁎ −.13⁎/−.35⁎ .00/−.23⁎

Total significance for TO/IQ 5/4 3/3 4/8 2/4 2/8 1/5 7/6 3/5

Notes: LO: Organization links; LC: Community links; FO: Organization fit; FC: Community fit; SO: Organization sacrifice; SC: Community sacrifice; JEO:Organization embeddedness; JEC: Community embeddedness; TO: Turnover; IQ: Intentions to quit.The correlations on the left side of slash signify the correlations with turnover behavior while those on the right side indicate the correlations with intentions toquit.We did not include other studies using job embeddedness scale because these studies did not report the components of job embeddedness.We did not conduct meta-analysis because internal coefficients are descriptive for formative models and cannot be used in meta-analysis.⁎ pb .05.

223M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

requirements. Second, when respondents in a sample come from a large area like a big city, the meaning of “community” mayneed specification in measurement items. We will explore these issues in the following section.

4.1. Possible moderating effects

4.1.1. Commute timeCommute time refers to how long an employee spends on commuting in a working day. If one is highly embedded in his or her im-

mediate community, yet possesses a job outside the community, traveling away from his or her beloved community to the distant jobmay be an aversive experience because time spent travelingmeans less time spent doing community activities. For such people, commu-nity embeddedness may have a neutral or positive, rather than negative, relationshipwith turnover. Zax and Kain (1991) found that lon-ger commutes encouraged Caucasians to quit their jobs but reduced their intentions to move out of their community. A recent studyreveals that a worker's quit positively relates to the worker's commuting distance (Deding, Filges, & Van Ommeren, 2009). Further re-search is needed to study if/how commute time moderates the relationship between community embeddedness and turnover.

4.1.2. Availability of nearby replacement jobsThe availability and proximity of potential replacement jobs may also moderate the relationship between community

embeddedness and turnover. For some jobs (e.g., university faculty), quitting means one must travel distances away from hisor her community to find another job (Mitchell et al., 2001). For these job types, someone with great community attachmentmay find it difficult to quit his or her job.

On the other hand, other job types (e.g., food service) may be plentiful enough to allow relocation within the same commu-nity. For these jobs, community embeddedness is likely to have little effect on turnover behavior. Future studies should testwhether the perception of nearby replacement jobs moderates the relationship between community embeddedness and turn-over. An example item that could be used to measure the perceptions of nearby replacement jobs is: “There are a number of or-ganizations nearby where I could find comparable work.”

4.1.3. Financial requirementsFinancial requirements are defined as factors that influence people's economic need to work (Brett, Cron, & Slocum, 1995). Brief

and associates found thatwhenworkers havemore dependents, earn lessmoney from their jobs, and have less income in their familyunits, they are more likely to viewwork as a tool to support their life (Brief, Brett, Raskas, & Stein, 1997). Employees that do not havetheir financial needs met are likely to focus on on-the-job rather than off-the-job (i.e., community) aspects because they rely heavilyon jobs to support their life. For such employees, organizational embeddedness may have more influence on turnover than commu-nity embeddedness. On the other hand, workers that have their financial needs met may be more likely to pay attention to commu-nity issues, such as the quality of medical care, education, and social services. For such employees, community embeddedness mayhave greater importance than organizational embeddedness in the decision whether to remain with an organization. Future studiesmay utilize survey samples with great variation in financial requirements to test whether financial requirements moderate the rela-tionship between community embeddedness and turnover.

4.1.4. Conclusion and suggestionCompared with organization embeddedness, community embeddedness is not a stable predictor of turnover. We have con-

cern about the fact because fundamentally, the value of a formative measure is the predictive validity it demonstrates. The lackof predictive validity may be because the facets scale of community embeddedness includes factors that do not make peoplefeel “stuck” in their jobs for a particular sample. For example, as we argued, for people who have to spend many commutinghours, their community embeddedness may be not beneficial to staying at their current organizations. We suggest that future re-search test the moderators that reflect the importance of off-the-job embeddedness.

4.2. “Community” needs specification

We argue that “community” needs specification in measurement items. Many items measuring community embeddedness con-tain theword “community” (Lee et al., 2004). In sociology, the concept of community has caused infinite debate, and sociologists haveyet to reach agreement on a definition of the term. Berger (1978) noticed at least 100 different definitions of community. However,extant embeddedness research does not specify the meaning and measurement of community. This misspecification, as describedbelow, may affect generalizability of findings across studies, as well as construct validity.

Consistent with Mitchell and colleagues' argument (Lee et al., 2004; Mitchell et al., 2001), “community” probably refers tocommunities of location (i.e., geographical community). The physical boundaries of a community can range from the local neigh-borhood to the suburb, village, town, city, region, or even an entire nation. However, the measurement items of communityembeddedness do not specify the range of a shared geographical location, and this could lead to conflicting research findings.

4.2.1. Conclusion and suggestionWhen surveys are conducted in small towns, it may not be important to specify “community” because “neighborhood,” “where I

live,” and “community” could be used interchangeablywithout bias. However, when respondents are from big cities and areas, it maybe necessary to give a clear definition of “community.” For example, with respect to the statement, “I really love the place where I

224 M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

live,” community is likely to refer to the local town or city. However, the statement, “Myneighborhood is safe” implies amore narrowdefinition of “community.”Wesuggest that further studies specify themeaning of “community”. An example of definition is that com-munity refers to “a deep psychological and emotional relationship to a group and/or a particular space” (Gottdiener & Budd, 2005).This definition can be used in the instruction part leading measurement items of community embeddedness.

5. Conceptualization and measurement problems: links

Asmentioned, links are defined as “formal or informal connections between a person and institutions or other people” (Mitchell etal., 2001, p. 1104). One of the tenets of JE is that, as the number of links increase job embeddedness becomes stronger, which in turn,decreases turnover (Mitchell et al., 2001; Yao et al., 2004). However, links have not been consistently associated with turnover (seeTable 1). Two studies found a negative relationship between links and turnover (Lee et al., 2004; Mitchell et al., 2001). Yet, two otherstudies did not find an association between links (community and organizational) and turnover (Crossley et al., 2007; Ramesh &Gelfand, 2010). Mallol et al. (2007) found an association between organizational links (but not community links) and turnover. Im-portantly, network research suggests that other relational properties influence turnover besides the number of workplace associatesor associations (Hom & Xiao, 2011; Hom et al., 2010; Murphy & Hom, 2008).

In line with prior research (Holtom et al., 2008; Hom et al., 2010), we contend that the number of connections may not adequatelycover relational constraints on leaving. In the discussion that follows, we identified four conceptual andmeasurement aspects associatedwith links.

5.1. More links do not necessarily translate into higher embeddedness

We propose that inconsistencies in the relationship between turnover and links may be attributed to the way links are hypoth-esized to relate to embeddedness and turnover. Specifically, we speculate that Mitchell and colleague's (Mitchell et al., 2001)tenet—more links produce lower turnover—is not always true.

A high number of organizational links, such as the connections a worker has with peers, teams, and committees, may result inoverload and even conflicting requests from bosses and co-workers. These requests could cause work stress whichmay decrease or-ganizational attachment (Kim, Price, Mueller, & Watson, 1996). Additionally, such connections may increase a sense of obligation toparticipate in social activities with co-workers during off-hours, whichmay reduce quality timewith family members. Studies foundthat greater work interference with family is associated with a higher propensity to quit (Hom & Kinicki, 2001). As Mitchell et al.(2001) noted, “Being highly embedded at workmight lead to work–family role conflicts, and such conflicts might result in turnover”(p. 117). Feeley, Hwang, and Barnett (2008) found that neither the number of in-degree links (i.e., links nominated by others) withfriends, nor the number of network links with peers, were significantly associatedwith turnover. Future research should examine theconditions where organizational links result in greater work load, more work pressures, and higher work–family conflict.

Community links are implied by marital status, whether or not a spouse works, whether a respondent owns his or her home, orwhether one has family roots near home (Lee et al., 2004). According to Mitchell et al. (2001), more community links are usually as-sumed to result in greater embeddedness, which in turn results in lower turnover. However, some empirical evidence suggests thatcommunity links may increase turnover in certain situations. Social network studies have found that the number of social ties is pos-itively correlatedwith alternative job opportunities (Bian, 1997; Granovetter, 1973; Lin, 1990).Mitchell et al. (2001) noticed the pos-sibility that more social ties may increase unsolicited job offers and knowledge about other positions. This knowledge may increasethe number of available alternatives, which in turn may increase turnover. One avenue of future research is to determine whethercommunity links are positively related to perceived job alternatives. Assuming such a relationship exists, the next step might be toexamine whether perceived job alternatives partially mediates the relationship between community links and turnover.

5.2. Consider the quality—not just the quantity of links

Another reason for the inconsistent relationships between links and turnover may be that previous research attention has notgiven adequate attention to the quality—rather than the quantity—of links. Indeed, Mitchell et al. (2001) stated that “certain linksmay be more important than others” (p. 1104). Yet, inadequate efforts have beenmade to examine how the quality of links affectsembeddedness and turnover decisions.

An examination of the organizational attachment and social networking literatures reveals qualitative differences among links. Wepropose that the relationships between links and turnover depend on type of link. The links developed for the purposes of informationseeking and advicenetworkingmayhave aneutral relationshipwith embeddedness because this type of linksmaynot be associatedwithemployee retention. Other links, such as those developed for the purposes of friendship networking and trust networking,may positivelyaffect organizational attachment (Morrison, 2002). Some links, such as a “hindrance network” that identifies people who impede one'swork (Cross & Parker, 2004), may even have a negative effect on attachment. In support, Jehn (1995) found that relationship conflictin groups decreased organizational attachment. Labianca and Brass (2006) argued that dysfunctional relationships may create poortask-related and socio-emotional outcomes. Future studies should examine how different types of links and networks relate to jobembeddedness.

225M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

5.3. Consider structural characteristics of links

Structural characteristics of links—such as size (Morrison, 2002), strength (Feeley & Barnett, 1997; Morrison, 2002; Mossholder,Setton, & Henagan, 2005), range (Morrison, 2002), network density (Morrison, 2002), centrality (Feeley et al., 2008; Hom et al.,2010;Mossholder et al., 2005) , network constraint (Hom et al., 2010), and structural equivalence (Feeley & Barnett, 1997)—have dif-ferential effects on organizational attachment. For example, Krackhardt and Porter (1986) found that turnover did not occur stochas-tically, but in structurally equivalent clusters in a perceived interpersonal communication network. One study revealed thatindividuals located in the periphery of a social network were more likely to leave their jobs (Feeley & Barnett, 1997). In a study ofnewcomers' socialization in organizations, Morrison (2002) found that network size, strength, range and status were associatedwith greater organizational commitment. Using survival analysis over a five-year period,Mossholder et al. (2005) found that networkcentrality decreased turnover behavior. Flap andVölker (2001) found that a networkwith a bow-tie structure, inwhich a focal actor isthe link between two or more mutually exclusive cliques, had strong negative effects on satisfaction with social aspects of a job. Onestudy revealed that employees who reported a greater number of out-degree links (i.e., the links nominated by the respondent) withfriends (i.e., out-degree centrality) were less likely to leave (Feeley et al., 2008).

In two recent studies (Hom et al., 2010; Murphy & Hom, 2008), Hom and coauthors utilized a social network approach to elu-cidate the role of structural characteristics of links in job embeddedness. Murphy and Hom (2008) created a measure of embedd-edness that included two types of network characteristics: the affective strength of links (i.e., the quality of links) and networkclosure (i.e., the degree to which everyone knows everyone else in a network; Burt, 2005). Murphy and Hom (2008) revealedthat network closure was positively associated with college persistence because more durable relationships embedded in thirdparties prevent people from leaving their organizations (Feeley & Barnett, 1997; Krackhardt & Porter, 1986).

Hom and associates refined the conceptualization of links by incorporating more social network constructs (Hom et al., 2010).Using an ego-centered social network approach (Morrison, 2002) and a whole-network approach (Scott, 2000), Hom and col-leagues integrated a comprehensive set of network constructs: strength of links, network closure, centrality, and other potentiallyembedding social forces such as normative prescription (Hom et al., 2010). Regression analyses revealed that normative pressuresto quit, defecting links, and sparse networks among external professional contacts explained an additional 14% of the variance inwithdrawal cognitions beyond that explained by job embeddedness, job satisfaction, and perceived alternatives (Hom et al.,2010). Future studies should replicate Hom et al.'s (2010) study with more diverse samples. In addition, future studies shouldcontinue to refine the conceptualization of links by exploring additional network characteristics, such as structural equivalence.

5.3.1. Conclusion and suggestionsWe conclude that the number of links is not sufficient in measuring the objective connections that make people stuck in their jobs.

Consistent with Hom and associates' work (Hom et al., 2010; Murphy & Hom, 2008), we propose that links should be measured withsocial network indices that depict theway people are connected. Further, we posit that four indices (number ofweak ties, extent of struc-tural holes, network range and network status) deserve prior attention because the indices suggest that people own social capital in theirorganizations (Fang, Duffy, & Shaw, 2011). The number ofweak ties refers to theweak connectionswith different people. Comparedwiththosewhoestablish strong connectionswithothers, peoplewhohave variousweak tiesmayhave the advantage of achieving information(Granovetter, 1973). This index is similar toMitchell et al. (2001) construct of links. Extent of structural holes indicates that people differin the extent their contacts are unconnected (Burt, 1992). Burt's structural hole theory posits that people could benefit from their broker-age roles by controlling the flowof information and exerting influence (Burt, 1992, 2005). Network range indicates towhat extent peoplehave contacts from different departments or teams (Ibarra, 1995;Morrison, 2002). Peoplewho havewidespread connections could haveadvantages of achieving information (Ibarra, 1995) and reinforcing social identity (Morrison, 2002). Network status assesses towhat de-gree people have contacts positioned at high level of the organizational hierarchy (Lin, 1990; Seibert, Kraimer, & Liden, 2001). Peoplewhohave contacts with high ranks could mobilize more resources to achieve desired outcomes (Lin, 1990; Seibert et al., 2001). We proposethat an individual's social capitalwhich is formedby connectionswith other contacts inhis or her organization (i.e., internal social capital)is negatively associatedwith turnover whereas an individual's social capital which is formed by connectionswith other contacts outsideshis or her organization (i.e., external social capital) is positively related to turnover.

5.4. Certain items may reduce predictive strength of links

Wehave concernswith the construct validity of extantmeasures of organization links (Lee et al., 2004;Mitchell et al., 2001). Linksaremeasured by three items: organizational tenure, the length of the current job position, and industrial experience (i.e., the length ofworking time in an industry). According to JE theory, all three items should be negatively related to turnover. Consistentwith the the-ory, a meta-analysis reveals that organizational tenure is negatively-related to turnover (Griffeth, Hom, & Gaertner, 2000; Maertz &Campion, 1998). However, the nature of the relationship between turnover and the two other items thatmeasure links—length of thecurrent position and industrial experience—may not coincide with JE theory.

We suggest that, under certain circumstances, industrial experience may increase turnover. Because experienced employeespossess greater skills and knowledge, they are more marketable to employers and therefore possess higher quality and more nu-merous job alternatives than their inexperienced counterparts. It is well-established that employees with more alternatives aremore likely to quit (Allen & Griffeth, 2001). Thus, it is conceivable that, contrary to predictions made by embeddedness theory,industrial experience is positively (rather than negatively) related to turnover.

226 M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

In addition,we speculate that the length of the current positionmayhave aneutral (orweak) relationshipwith turnover. In particular,for people who are in their career plateau, occupying a position for many years may indicate that they are stuck in a dead-end job andhave little chance for promotion. As Burt (1992) noted, “spending a long time in the same rank is a signal of low achievement” (p. 130).

5.4.1. SuggestionThus, although studies reveal that organizational tenure is consistent with JE theory's conceptualization of links, in that orga-

nizational tenure is negatively-related to turnover (Hom, Roberson, & Ellis, 2008), industrial experience and the length of the cur-rent positions are two measures of links that may not be consistent with JE theory. Future studies may eliminate these two itemsand use tenure plus network indices such as number of weak ties, extent of structural holes, network range and network status asthe measure of organizational links.

6. Discriminant validity issues

In this section, we discuss discriminant validity issues regarding two components of the JE composite measure: fit and sacri-fice. It should be noted that the discriminate validity issues discussed below have no substantial influence if studies use the ag-gregated score of the JE's components. However, because prior studies put forward the hypotheses that include thecomponents of fit and sacrifice (for example, Mallol et al., 2007; Ramesh & Gelfand, 2010), it is necessary to examine the discrim-inant validity issues.

As mentioned, fit refers to an employee's perceived compatibility with the organization or surrounding community (Mitchellet al., 2001); sacrifice is the perceived cost of material or psychological benefits that may be forfeited by leaving one's job, or one'scommunity (Mitchell et al., 2001). Mitchell et al. (2001) implied that fit and sacrifice are distinct constructs; yet some evidenceindicates that they are, in fact, quite similar.

Several empirical studies reveal that fit and sacrifice are highly-correlated (Cunningham, Fink, & Sagas, 2005; Mallol et al.,2007; Mitchell et al., 2001). For a sample of hospital employees, Mitchell et al. (2001) revealed that the correlation between com-munity fit and community sacrifice was .73. In a subsequent study with two samples (Cunningham et al., 2005), results revealedcorrelations of .70 and .71. Mallol et al. (2007) found that the correlations between organization fit and organizational sacrificewere .75 for Caucasian workers and .67 for Hispanic workers, and the correlations between community fit and community sacri-fice were .76 for Caucasian workers and .57 for Hispanic workers.

One explanation for the high associations between fit and sacrifice is that, in some cases, sacrifice and fit may be mixed. The fitbetween aspects of the organization (e.g., organizational culture) may also be viewed as sacrifices associated with leaving. For in-stance, organizational fit items, such as “I fit with this organization's culture” (Lee et al., 2004, p. 721), could be thought of as apsychological cost of leaving the organization (i.e., organizational sacrifice), if framed as such: If I leave I will give up my senseof identity and relationship with members that share common values. In addition, one organizational fit item, “I like the authorityand responsibility I have at this company,” appears similar to one particular organizational sacrifice item: “I have a lot of freedomon this job to decide how to pursue my goals” (Lee et al., 2004, p. 721–722). Having authority on the job is a necessary conditionof being able to have the freedom to choose how to pursue one's goals.

Similarly, organizational sacrifice may capture the concept of organizational fit. For instance, one organizational sacrifice itemreads, “I feel that people at work respect me a great deal” (Lee et al., 2004, p. 722). One may argue that respect from other co-workers is a part of organizational culture, and thus represents fit.

Some community fit and community sacrifice items may overlap. For example, the community sacrifice item, “My neighbor-hood is safe” (Lee et al., 2004, p. 722), could be framed in terms of community fit: A safe neighborhood is compatible with my values,goals, and personality. These examples suggest that the conceptual overlap may account for the high correlations between fit andsacrifice in the extant literature.

6.1. Suggestions

To reduce the possible overlaps between some fit items and sacrifice items, we suggest that the definition of fit and sacrifice beoperationalized more explicitly. Specifically, we suggest that the definition of sacrifice focuses on perceived material cost and ex-cludes the component of psychological cost because: (1) when people consider quitting their jobs, fit could be regarded as psy-chological cost, and (2) the items that measure psychological cost could also be framed as fit indices.

There are two possible avenues for future research. One avenue could involve the use of subject matter experts to examine thecontent of fit and sacrifice (Anastasi & Urbina, 1997). A group of experts would be given a list of items that belong to the two di-mensions; however, they would not have a priori knowledge of the dimension each item belongs to. Then, they would review thedefinitions of fit and sacrifice and assign items to one of the two dimensions, or omit items that do not belong to either dimension.Finally, the experts would resolve any classification disagreements.

Following MacKenzie et al. (2005) recommendation, the second path involves utilizing composite confirmatory factor analysisto verify the discriminant validity of embeddedness dimensions with a formative rather than reflective measurement model. Pre-vious research that has tested the discriminant validity with confirmatory factor analysis has done so with a reflective measure-ment model: i.e., mathematical relationships were represented as causal arrows pointing from the constructs to the items(Cunningham et al., 2005; Halbesleben & Wheeler, 2008). The validity of findings of these studies is questionable, because a re-flective measurement model is inconsistent with the formative framework of job embeddedness theory. Future studies could test

227M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

the discriminant validity of dimensions of the composite measure using a formative approach recommended by Brown (2006).We expect that the items which are used to measure fit and sacrifice could be loaded on two distinct dimensions in the compositeconfirmatory factor analysis.

7. Cultural boundaries of job embeddedness

The potential role of cultural influences on job embeddedness deserves further scrutiny. To our knowledge, only three studies(Hom et al., 2009; Ramesh & Gelfand, 2010; Tanova & Holtom, 2008) have been conducted on samples of non-U.S. workers.Tanova and Holtom (2008) revealed that job embeddedness explained a significant amount of variation in turnover beyondthe effects of demographic and traditional predictors such as job satisfaction for a sample of employees from four European coun-tries. However, Tanova and Holtom (2008) did not employ Lee et al. (2004) composite scale. Thus, to what extent the compositescale of job embeddedness can be generalized to other cultural settings remains unclear.

The findings from other two studies suggest that job embeddedness theory be tailored to particular cultural contexts. Hom etal. (2009) found that the relationships between job embeddedness and organizational constructs could be generalized to a Chi-nese sample of middle-level managers only after substantial item revision. Hom et al. (2009) added more items assessing per-son–organization and person–job fit. In addition, interviews and open-ended surveys with part-time Chinese MBA studentswere conducted to generate additional items [e.g., loss of “Guanxi (Relationship)” connections as a sacrifice, for a detailed discus-sion about Guanxi, see Farh, Tsui, Xin, and Cheng (1998)] or revisions (e.g., more explicit examples of fringe benefits for sacrifice).In a recent study, Ramesh and Gelfand (2010) conducted a cross-culture study with samples from American and India call centers.The findings reveal the dimensions of job embeddedness influence turnover with cultural differences. Person–job fit was found tobe more important to reducing turnover in the United States than in India, whereas organization links and community links werefound to be more important to decreasing turnover in India than in the United States (Ramesh & Gelfand, 2010).

Cultural differences may affect the generalizability of JE theory. In this paper, we take China as an example to illustrate thatthere is cultural boundary to the theory of JE. We choose China because the country has been found to own a distinct culturewhich is different with the culture in the United States. According to Hofstede's cultural theory (Hofstede, 2001), national cultureis composed of five main dimensions: power distance index, individualism, masculinity, uncertainty avoidance index, and long-term orientation. Studies have shown that China's national culture has major differences with the United States' national culturein two aspects: individualism (the following numbers depict how the countries score on individualism: China, 20; the UnitedStates, 91) (Hofstede & Hofstede, 2004, p. 78–79) and long-term orientation (China, 118; the United States, 29) (Hofstede &Hofstede, 2004, p. 211). Thus, China provides a good example to test if the theory of JE can be generalized to other cultural set-tings. Because Hom et al. (2009) revised the measure of on-the-job embeddedness and found the revised measure worked ade-quately in China, we focus on two issues of generalization of community embeddedness.

7.1. The less important role of community embeddedness

Aryee and coauthors argue that in Chinese society, work may be viewed as a means of enhancing the family's well-being (Aryee,Field, & Luk, 1999). Yang and associates used the theory of individualism/collectivism to explain work–family priority in a comparativestudy between American and Chinese (Yang, Chen, Choi, & Zou, 2000). She and coauthors argued that Chinese aremore inclined to givepriority to work than are Americans because collectivism encourages Chinese to have family-based work ethic which means workingfor the long-term welfare of the family (Yang et al., 2000). According to this family-based work ethic, work overtime is a self-sacrificemade for the benefit of the family rather than a sacrifice of the family for the selfish pursuit of one's own career development.Moreover,Chinese are likely to view sacrificing family time for work as a short-term cost incurred to gain long-term benefits (Yang et al., 2000).

Apart from cultural values, Chinese firms are implementing “expectations-enhancing” human resourcesmanagement (HRM) sys-tems (moving away from state-owned bureaucracies) as they strive to be more competitive in global markets (Hom et al., 2009;Wang, Tsui, Zhang, & Ma, 2003). Unlike traditional HRM systems (furnishing broad benefits, such as housing, and lifetime jobs formeager employee contribution), contemporary systems demand greater workforce productivity in exchange for inducements, andjob security. Thus, Chinese employees are likely to bemore on-the-job embedded than ever before—by having tomeetmore stringentperformance standards and perform duties for teams and superiors than go beyond their job duties (Hom et al., 2009).

Wang and colleagues found that work interference with family was not associated with turnover intentions among Chineseemployees, though this stressor often induced withdrawal among American employees (Wang, Lawler, Walumbwa, & Kan,2004). Apparently, Chinese employees assign higher priority to work over family, as they do not form withdrawal cognitionswhen their job interferes with family activities and time (Wang et al., 2004). Following the logic of work priority, we speculatethat some items in the composite measure of community embeddedness would have negligible effect on Chinese employees'quit decision (e.g., “the weather where I live is suitable for me”). Future studies could test whether nationality moderates the re-lationship between community embeddedness and turnover. We hypothesize that, for Chinese people, community embedded-ness is likely to be less negatively associated with turnover.

7.2. Revision of community embeddedness in cultural analysis

We argue that the measurement items of community embeddedness are likely to need revision in other cultural settings. For ex-ample, many citizens of China are likely to have difficulty understanding “community” because the concept is not indigenous.

228 M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

Sociologists argue that Chinese society has a tradition of family rather than community orientation (Fei, 1992; Fukuyama, 1995),while cross-cultural scholars often note that the Chinese—a collectivist people—differentiate between in-group (often family mem-bers) and out-group members (acquaintances) (Hofstede, 2001). Hence, Chinese people are unlikely to feel much embeddednessin the broader community than to their family, form fewer links and less fit with those who are not in-group members.

7.2.1. SuggestionBecause the measure of community links contains items assessing people's connections to their family and friends, the mea-

sure is likely to make sense among Chinese people. The measures of community fit and sacrifice, however, exclude items refer-ring to family and friends and some other important factors such as living place in the school district with excellent educationfor children. Hence, community fit and sacrifice may have less effect on Chinese people's turnover behavior. To keep the pre-dictive validity of community embeddedness, researchers should find the relevant factors in the specific cultural settings byconducting qualitative studies before the measurement items are used in empirical studies. Future research may consider re-vising the scale of community fit and sacrifice for research with the Chinese. For example, community fit may contain itemsassessing the relationships with relatives and close friends. Community sacrifice may include the cost of leaving relativesand close friends.

8. Conclusion

The creation of JE has made a significant contribution to the extant turnover literature. Most of the prevailing theory and research onvoluntary turnover came from the ideas of March and Simon (1958), which is related to the perceived ease and desirability of leavingone's job. The traditional wisdom was that people become dissatisfied with their jobs, search for alternatives, compare those optionswith their present jobs using a rational cost–benefit decision process, and leave if any of the alternatives are judged to be better thantheir current situation (Mobley, 1977). The construct of JE, however, emphasizes the important roles of non-attitudinal and off-the-jobfactors in understanding employees' attachment to their organizations. To truly capture JE, we believe it is crucial that future researchersuse instruments that include both non-attitudinal and off-the-job factors above and beyond attachment to the organization.

Responding to Mitchell and colleagues' proposition of improving JE's conceptualization and measurement, we summarizedfive possible areas of deficiency with regard to JE. We focused on conceptual and measurement problems associated with thescale developed by Mitchell, Lee and their colleagues. We then provide avenues and suggestions for future research for thetopic of JE. It should be acknowledged that the work of Mitchell, Lee and their colleagues points to a promising new direction.JE contributes to understanding turnover beyond intra-organizational attitudes and extra-organizational alternatives. These vir-tues notwithstanding, there is a need to improve the measurement of JE. We hope this paper will inspire research efforts to im-prove the conceptualization and measurement of JE.

Acknowledgments

The authors thank two anonymous reviewers for their helpful and constructive comments. We also thank Sean Robinson andJustin Weinhardt for their helpful comments on an earlier draft of this article. The authors gratefully acknowledge financial sup-port for this research from the grants by the National Natural Science Foundation of China (nos. 70972023 and 71121001).

Appendix A. The sample items of job embeddedness composite measurea

Links, communityAre you currently married?Do you own the home you live in? (mortgaged or outright)

Links, organizationHow long have been in your present position? (years)How many co-workers do you interact with regularly?

Fit, communityI really love the place where I live.I like the family-oriented environment of my community.

Fit, organizationMy job utilizes my skills and talents well.I feel like I am a good match for this organization.

Sacrifice, communityLeaving this community would be very hard.My neighborhood is safe.

Sacrifice, organizationI have a lot of freedom on this job to decide how to pursue my goals.The perks on this job are outstanding.

aNote: The full version can be obtained in Lee et al. (2004, p.721–722).

229M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

References

Allen, D. G. (2006). Do organizational socialization tactics influence newcomer embeddedness and turnover? Journal of Management, 32, 237–256.Allen, D. G., & Griffeth, R. W. (2001). Test of a mediating performance–turnover relationship highlighting the moderating roles of visibility and reward contingen-

cy. The Journal of Applied Psychology, 86, 1014–1021.Anastasi, A., & Urbina, S. (1997). Psychological testing (7th ed.). Upper Saddle River, NJ: Prentice-Hall.Aryee, S., Field, D., & Luk, V. (1999). A cross-cultural test of a model of the work–family interface. Journal of Management, 25, 491–511.Berger, A. (1978). The city: Urban communities and their problems. Dubuque, Iowa: William C. Brown.Bian, Y. (1997). Bringing strong ties back in: Indirect ties, network bridges, and job searches in China. American Sociological Review, 62, 366–385.Brett, J. F., Cron, W. L., & Slocum, J. W., Jr. (1995). Economic dependency on work: A moderator of the relationship between organizational commitment and per-

formance. Academy of Management Journal, 38, 261–271.Brief, A. P., Brett, J. F., Raskas, D., & Stein, E. (1997). Feeling economically dependent on one's job: Its origins and functions with regard to worker well-being. Jour-

nal of Applied Social Psychology, 27, 1303–1315.Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York: Guilford.Burt, R. S. (1992). Structural holes: The social structure of competition. Cambridge, MA: Harvard University Press.Burt, R. S. (2005). Brokerage and closure: An introduction to social capital. Oxford, UK: Oxford University Press.Cross, R., & Parker, A. (2004). The hidden power of social networks: Understanding how work really gets done in organizations. Boston: Harvard Business School Press.Crossley, C. D., Bennett, R. J., Jex, S. M., & Burnfield, J. L. (2007). Development of a global measure of job embeddedness and integration into a traditional model of

voluntary turnover. The Journal of Applied Psychology, 92, 1031–1042.Crossley, C. D., Bennett, R. J., Jex, S. M., & Burnfield, J. L. (2011). Development of a global measure of job embeddedness and integration into a traditional model of

voluntary turnover: Clarification to Crossley et al. (2007). The Journal of Applied Psychology, 96, 1316-1316.Cunningham, G. B., Fink, J. S., & Sagas, M. (2005). Extensions and further examination of the job embeddedness construct. Journal of Sport Management, 19,

319–337.Deding, M., Filges, T., & Van Ommeren, J. (2009). Spatial mobility and commuting: The case of two-earner households. Journal of Regional Science, 49, 113–147.Fang, R. L., Duffy, M. K., & Shaw, J. D. (2011). The organizational socialization process: Review and development of a social capital model. Journal of Management,

37, 127–152.Farh, J. L., Tsui, A. S., Xin, K., & Cheng, B. S. (1998). The influence of relational demography and guanxi: The Chinese case. Organization Science, 9, 471–488.Feeley, T. H., & Barnett, G. A. (1997). Predicting employee turnover from communication networks. Human Communication Research, 23, 370–387.Feeley, T. H., Hwang, J., & Barnett, G. A. (2008). Predicting employee turnover from friendship networks. Journal of Applied Communication Research, 36, 56–73.Fei, X. (1992). From the soil: The foundations of Chinese society. Berkeley, CA: University of California Press.Felps, W., Hekman, R. D., Mitchell, R. T., Lee, W. T., Harman, S. W., & Holtom, C. B. (2009). Turnover contagion: How coworkers' job embeddedness and coworkers'

job search behaviors influence quitting. Academy of Management Journal, 52, 545–561.Flap, H., & Völker, B. (2001). Goal specific social capital and job satisfaction: Effects of different types of networks on instrumental and social aspects of work. Social

Networks, 23, 297–320.Fukuyama, Y. F. (1995). Trust: The social virtues and the creation of prosperity. New York: Free Press.Gerstner, C. R., & Day, D. V. (1997). Meta-analytic review of leader–member exchange theory: Correlates and construct issues. The Journal of Applied Psychology,

82, 827–844.Gottdiener, M., & Budd, L. (2005). Key concepts in urban studies. Thousand Oaks, CA: Sage.Granovetter, M. S. (1973). The strength of weak ties. The American Journal of Sociology, 78, 1360–1380.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.Halbesleben, J. R. B., & Wheeler, A. R. (2008). The relative roles of engagement and embeddedness in predicting job performance and intention to leave. Work and

Stress, 22, 242–256.Hofstede, G. H. (2001). Culture's consequences: Comparing values, behaviors, institutions and organizations across nations (2nd ed.). Thousand Oaks, CA: Sage.Hofstede, G. H., & Hofstede, G. J. (2004). Cultures and organizations: Software of the mind. New York: McGraw-Hill.Holtom, B. C., Mitchell, T. R., Lee, T. W., & Eberly, M. (2008). Turnover and retention research: A glance at the past, a closer review of the present, and a venture into

the future. Academy of Management Annals, 2, 231–274.Holtom, B. C., Mitchell, T. R., Lee, T. W., & Tidd, S. (2006, August). Less is more: Validation of a short form of the job embeddedness measure and theoretical ex-

tensions. Paper presented at Academy of Management Conference, Atlanta, GA.Hom, P. W., & Griffeth, R. W. (1991). Structural equations modeling test of a turnover theory: Cross-sectional and longitudinal analyses. The Journal of Applied Psy-

chology, 76, 350–366.Hom, P. W., & Griffeth, R. W. (1995). Employee turnover. Cincinnati, OH: South-Western Press.Hom, P. W., & Kinicki, A. J. (2001). Toward a greater understanding of how dissatisfaction drives employee turnover. Academy of Management Journal, 44, 975–987.Hom, P. W., Roberson, L., & Ellis, A. D. (2008). Challenging conventional wisdom about two quits: Revelations from corporate America. The Journal of Applied Psy-

chology, 93, 1–34.Hom, P., Rogers, K., Allen, D., & Zhang, M. (2010, August). An extended model of social embeddedness: Applying social network theory to enrich job embedded-

ness theory. Paper presented at the Academy of Management Conference, Montreal, Canada.Hom, P. W., Tsui, A. S., Lee, T. W., Ping, P. F., Wu, J. B., & Zhang, A. Y. (2009). Explaining employment relationships with social exchange and job embeddedness. The

Journal of Applied Psychology, 94, 277–297.Hom, P. W., & Xiao, Z. (2011). Embedding social networks: How guanxi ties reinforce Chinese employees' retention. Organizational Behavior and Human Decision

Processes, 116, 188–202.Ibarra, H. (1995). Race, opportunity, and diversity of social circles in managerial networks. Academy of Management Journal, 38, 673–703.Jarvis, C. B., Mackenzie, S. B., Podsakoff, P. M., Mick, D. G., & Bearden, W. O. (2003). A critical review of construct indicators and measurement model misspecifica-

tion in marketing and consumer research. Journal of Consumer Research, 30, 199–218.Jehn, K. A. (1995). A multimethod examination of the benefits and detriments of intragroup conflict. Administrative Science Quarterly, 40, 256–282.Kim, S. W., Price, J. L., Mueller, C. W., & Watson, T. W. (1996). The determinants of career intent among physicians at a US air force hospital. Human Relations, 49,

947–976.Krackhardt, D., & Porter, L. W. (1986). The snowball effect: Turnover embedded in communication networks. The Journal of Applied Psychology, 71, 50–55.Labianca, G., & Brass, D. J. (2006). Exploring the social ledger: Negative relationships and negative asymmetry in social networks in organizations. Academy of

Management Review, 31, 596–614.Lee, T. W., & Mitchell, T. R. (1994). An alternative approach: The unfolding model of voluntary employee turnover. Academy of Management Review, 19, 51–89.Lee, T. W., Mitchell, T. R., & Sablynski, C. J. (1999). Qualitative research in organizational and vocational psychology, 1979–1999. Journal of Vocational Behavior, 55, 161–187.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, vo-

litional 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.Lin, N. (1990). Social resources and social mobility. In R. Burt (Ed.), Social mobility and social structure. New York: Cambridge University Press.MacKenzie, S. B., Podsakoff, P. M., & Jarvis, C. B. (2005). The problem of measurement model misspecification in behavioral and organizational research and some

recommended solutions. The Journal of Applied Psychology, 90, 710–730.Maertz, C. P., & Campion, M. A. (1998). 25 years of voluntary turnover research: A review and critique. In C. L. Cooper, & I. T. Robertson (Eds.), International review

of industrial and organizational psychology, Vol. 13. (pp. 49–83)Chichester, England: Wiley & Sons.

230 M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231

Mallol, C. M., Holtom, B. C., & Lee, T. W. (2007). Job embeddedness in a culturally diverse environment. Journal of Business and Psychology, 22, 35–44.March, J., & Simon, H. A. (1958). Organizations. New York: Wiley Press.Mitchell, T. R., Holtom, B. C., Lee, T. W., Sablynski, C. J., & Erez, M. (2001). Why people stay: Using job embeddedness to predict voluntary turnover. Academy of

Management Journal, 44, 1102–1121.Mobley, W. H. (1977). Intermediate linkages in the relationship between job satisfaction and employee turnover. The Journal of Applied Psychology, 62, 237–240.Mobley, W. H. (1982). Employee turnover: Causes, consequences, and control. Reading, Mass.: Addition-Wesley.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.Mobley, W. H., Horner, S. O., & Hollingsworth, A. T. (1978). An evaluation of precursors of hospital employee turnover. The Journal of Applied Psychology, 63,

408–414.Morrison, E. W. (2002). Newcomers' relationships: The role of social network ties during socialization. Academy of Management Journal, 45, 1149–1160.Mossholder, K. W., Setton, R. P., & Henagan, S. C. (2005). A relational perspective on turnover: Examining structural, attitudinal, and behavioral predictors. Acad-

emy of Management Journal, 48, 607–618.Murphy, K., & Hom, P. (2008, April). Extending and enriching job embeddedness theory: Predicting college persistence. Paper presented at the SIOP Annual Con-

ference, San Francisco, CA.Price, J. L. (2001). Reflections on the determinants of voluntary turnover. International Journal of Manpower, 22, 600–624.Price, J. L., & Mueller, C. W. (1981). A causal model of turnover for nurses. Academy of Management Journal, 24, 543–565.Ramesh, A., & Gelfand, M. J. (2010). Will they stay or will they go? The role of job embeddedness in predicting turnover in individualistic and collectivistic cul-

tures. The Journal of Applied Psychology, 20, 807–823.Scott, J. (2000). Social network analysis: A handbook (2nd ed.). London: Sage.Seibert, S. E., Kraimer, M. L., & Liden, R. C. (2001). A social capital theory of career success. Academy of Management Journal, 44, 219–237.Steers, R. M., & Mowday, R. T. (1981). Employee turnover and post-decision accommodation processes. In L. L. Cummings, & B. M. Staw (Eds.), Research in orga-

nizational behavior, Vol. 3. (pp. 235–282)Greenwich, CT: JAI Press.Tanova, C., & Holtom, B. C. (2008). Using job embeddedness factors to explain voluntary turnover in four European countries. International Journal of Human Re-

source Management, 19, 1553–1568.Wang, P., Lawler, J. J., Walumbwa, F. O., & Kan, S. (2004). Work–family conflict and job withdrawal intentions: The moderating effect of cultural differences. In-

ternational Journal of Stress Management, 11, 392–414.Wang, D. X., Tsui, A. S., Zhang, Y., & Ma, L. (2003). Employment relationship and firm performance: A contingency analysis in an emerging economy. Journal of

Organizational Behavior, 24, 511–535.Yang, N. N., Chen, C. C., Choi, J., & Zou, Y. M. (2000). Sources of work–family conflict: A Sino-U.S. comparison of the effects of work and family demands. Academy of

Management Journal, 43, 113–123.Yao, X., Lee, T. W., Mitchell, T. R., Burton, J. P., & Sablynski, C. J. (2004). Job embeddedness: Current research and future directions. In R. Griffeth, & P. Hom (Eds.),

Innovative theory and empirical research on employee turnover (pp. 153–187). Greenwich, CT: Information Age Publishing.Zax, J. S., & Kain, J. F. (1991). Commutes, quits, and moves. Journal of Urban Economics, 29, 153–165.

231M. Zhang et al. / Human Resource Management Review 22 (2012) 220–231