JobEmbeddedness

download JobEmbeddedness

of 13

Transcript of JobEmbeddedness

  • 8/8/2019 JobEmbeddedness

    1/13

    THE EFFECTS OF JOB EMBEDDEDNESS ON ORGANIZATIONALCITIZENSHIP, JOB PERFORMANCE, VOLITIONAL ABSENCES, AND

    VOLUNTARY TURNOVER

    THOMAS W. LEE

    TERENCE R. MITCHELLUniversity of Washington, Seattle

    CHRIS J. SABLYNSKICalifornia State University, Sacramento

    JAMES P. BURTONUniversity of Washington, Bothell

    BROOKS C. HOLTOMGeorgetown University

    This study extends theory and research on job embeddedness, which was disaggre-gated into its two major subdimensions, on-the-job and off-the-job embeddedness. Ashypothesized, regression analyses revealed that off-the-job embeddedness was signif-icantly predictive of subsequent voluntary turnover and volitional absences,whereas on-the-job embeddedness was not. Also as hypothesized, on-the-job embed-dedness was significantly predictive of organizational citizenship and job perfor-mance, whereas off-the-job embeddedness was not. In addition, embeddedness mod-erated the effects of absences, citizenship, and performance on turnover. Implicationsare discussed.

    For over 45 years, management scholars havetheorized about and empirically investigated thecauses of employees voluntarily leaving jobs, or

    voluntary employee turnover (Maertz & Cam-pion, 1998). In their classic book, Organizations,March and Simon (1958) provided much of thetheoretical underpinning for the psychological re-search on voluntary turnover. They conceptualizedemployee turnover as a reflection of an employeesdecision to participate in the activities of his or herorganization. They also outlined how such a deci-sion to participate differs in substantial ways froma decision to perform. As a result of this conceptu-alization, most research on the participation deci-sion has treated the performance decision as a

    largely independent deliberation. A more thoroughreading of March and Simon and of other, recentresearch suggests, however, a closer link between

    the decisions to participate and to perform than hasbeen traditionally thought to exist. Recent theoryand research have suggested new and different

    ways to think about turnover, going beyond a strictfocus on an employee decision to participate. Add-ing considerable richness have been the work ofHulin and associates, on a general withdrawal con-struct (e.g., Hulin, 1991); of Lee and associates, onmultiple paths for leaving, described in their un-folding model (e.g., Lee & Mitchell, 1994); and ofMitchell and associates, on job embeddedness, aconstruct including both on- and off-the job causesof turnover (e.g., Mitchell, Holtom, Lee, Sablynski,& Erez, 2001). Equally important, these new ideashave helped scholars better understand the concep-

    tual and empirical links between employee with-drawal and work performance.

    This study had two specific purposes. First, wesought to extend theory and research on job embed-dedness by demonstrating how its major compo-nents (that is, on- and off-the-job embeddedness)differentially predicted the decision to perform (or-ganizational citizenship and job performance) andthe decision to participate (volitional absences andvoluntary turnover). Second, we sought to showhow these embeddedness components might beprocesses through which the decisions to perform

    We thank John Cotton (Marquette University) and GregBigley (University of Washington) for their suggestionson a draft and Judith Aptaker (University of Washington)and Greg Anderson (University of Washington) for theirhelp with the data.

    Susan Jackson served as action editor for this manu-script.

    Academy of Management Journal

    2004, Vol. 47, No. 5, 711722.

    711

  • 8/8/2019 JobEmbeddedness

    2/13

    and to participate could be conceptually and em-pirically linked.

    CONCEPTUAL BACKGROUND

    March and Simon (1958: Chapters 3 & 4) clearlydifferentiated between the decisions to perform

    and to participate. They explained the performancedecision in terms of motivational concepts such asgoals, expectancies, and social control (for in-stance, norms, group pressure, and rewards). Incontrast, they explained the participation decisionin terms of perceived desirability of movement andperceived ease of movement. Over the years, desir-ability of movement has come to mean work atti-tudes like job satisfaction or organizational com-mitment, whereas ease of movement has come tomean perceived job alternatives or actual unem-ployment rates. More specifically, most turnover

    theory has the premise that people leave if they areunhappy with their jobs and job alternatives areavailable. This focus on dissatisfaction, low com-mitment, and prevalent job alternatives dominatesthe study of voluntary turnover. Although gener-ally valid, the traditional models have had modestsuccess in predicting turnover (e.g., Griffeth, Hom,& Gaertner, 2000), with their variables seldom ex-plaining more than 10 percent of variance.

    New ways to think about turnover may beneeded. In this research, we attempted to integrateMarch and Simons ideas about the links betweenthe decisions to perform and to participate withmore recent research on employee withdrawal byHulin, Lee, and Mitchell. First, March and Simon(1958) suggested that withdrawal occurs over timeand includes more types of participation decisionsthan just turnover. They stated, The motivation towithdraw factor is a general one that holds for bothabsences and voluntary turnover (March & Simon,1958: 93). In other words, both absences and turn-over reflect decisions about participation. Second,they suggested that many off-the-job factors are im-portant determinants of why people stay or leave.For instance, as March and Simon wrote, Families

    often have attitudes about what jobs are appropri-ate for their members (1958: 72) and The integra-tion of individuals into the community has fre-quently been urged by organizations because itoffers advantages for public relations and reducesvoluntary mobility (1958: 72). Thus, March andSimon theorized that severing participation entailsmore than dissatisfaction-induced leaving. It in-volves multiple actions, community, and family.Furthermore, both on- and off-the-job factors areimportant antecedents of employee turnover.

    Hulin and associates proposed a broader concep-

    tualization of withdrawal than is found in mostcontemporary turnover research. They advocatedfor and empirically demonstrated the validity of ageneral withdrawal construct (Hanish & Hulin,1991). More specifically, withdrawal was theorizedto include multiple work behaviors occurring se-quentially over time, such as poor citizenship, de-

    creased job performance, increased absences, andfinally leaving. The withdrawing person demon-strates a progression of withdrawal from the verymild and easy to the difficult and decisive (Hulin,1998: 11). These ideas suggest that the decisionsto perform and participate are related, with thedecision to perform preceding the decision toparticipate.

    The research by Lee, Mitchell, Holtom, Mc-Daniel, and Hill (1999) expresses related ideas interms of the unfolding model of turnover, accord-ing to which leaving occurs over time and can

    follow various paths. Some turnover happensquickly (for instance, a preexisting script forleaving drives an employee to quit in response tosome event), and some happens more slowly (forinstance, accumulated job dissatisfaction leads to asearch for alternatives). In addition, many peopleleave because of discernable precipitating events,and many of these events occur off the job (a spouserelocates, an unsolicited job offer is received).Thus, specific off-the-job events can precipitateturnover.

    More recently, Mitchell, Holtom, Lee, Sablynski,and Erez (2001) focused on why people stay ratherthan on how they leave. In particular, they drewattention to the reasons people stay through theirjob embeddedness construct. Reflecting the idea ofpeoples being situated or connected in a socialweb, embeddedness has several key aspects: (1)the extent to which people have links to otherpeople or activities, (2) the extent to which theirjobs and communities fit other aspects in their lifespaces, and (3) the ease with which links could be

    brokenwhat they would give up if they left theirpresent settings. Mitchell and his coauthors calledthese three dimensions links, fit, and sacrifice, re-

    spectively, and they are important both on and offthe job. Thus, one can think of a three by twomatrix that shows six dimensions: links, fit, andsacrifice in an organization and in a community.

    Mitchell and his colleagues (2001) provided ini-tial empirical support for job embeddedness. Draw-ing on data from a sample of retail employees andanother sample of hospital employees, they firstreported that job embeddedness was reliably mea-sured as an aggregated score across their six dimen-sions. Second, aggregated job embeddedness corre-lated with intention to leave and predicted

    712 OctoberAcademy of Management Journal

  • 8/8/2019 JobEmbeddedness

    3/13

    subsequent voluntary turnover. Third, job embed-dedness significantly predicted turnover after theeffects of gender, satisfaction, commitment, jobsearch, and perceived alternatives had been con-trolled. Thus, job embeddedness was related to oneof the major decisions about participation, namely,turnover.

    These findings, and the work of March, Simon,and Hulin, suggest three main ideas: First, job em-

    beddedness can be disaggregated into two majorcomponents: on-the-job embeddedness (that is, or-ganizational fit, links, and sacrifice) and off-the-jobembeddedness (that is, community fit, links, andsacrifice). Second, these two components may havedifferent effects on indicators of performance andparticipation (absences and turnover). Third, em-ployee withdrawal occurs over time, with a deci-sion about performing preceding a decision aboutparticipating.

    Hypotheses

    As conceptualized, job embeddedness reflectsemployees decisions to participate broadly and di-rectly, and it moves scholarly attention beyond dis-satisfaction-induced leaving. More aptly, job em-

    beddedness is a retention (or antiwithdrawal)construct. Hulin (1998) never directly measured ageneral withdrawal construct, instead inferring itfrom the occurrence of multiple work behaviors. Ifjob embeddedness is indeed a broad-based reten-tion (antiwithdrawal) construct and if it captures asizable portion of the decision to participate,

    both on- and off-the-job embeddedness should pre-dict not only employee turnover, but also otherwithdrawal behaviors, such as decreasing organiza-tional citizenship behavior, decreasing perfor-mance, and increasing absence. Further, the ex-plained variance in these withdrawal behaviorsshould exceed that explained by job satisfactionand organizational commitment.

    In meta-analyses, Griffith and colleagues (2000)showed that job satisfaction and organizationalcommitment significantly related to absences and

    that absences significantly predicted turnover. Be-cause on-the-job embeddedness correlates to satis-faction, commitment, and turnover (Mitchell et al.,2001), it should predict subsequent absences aswell. However, the effect of on-the-job embedded-ness on absences and turnover may be reduced tozero when researchers control for satisfaction andcommitment. Further, the ideas of Hulin, March,Simon, Lee, and Mitchell about nonwork factorssuggest that off-the-job embeddedness predicts ab-sences and turnover, and it may do so even whensatisfaction and commitment are controlled for.

    That is, leaving a job may have significant effectson an individuals off-the-job life, especially if heor she has to relocate to find new employment.More specifically, people who are embedded intheir communities should want to keep their jobs.Mitchell and colleagues (2001) reported, for exam-ple, that having (1) a working spouse, (2) children

    in a particular school, or (3) involvement in com-munity activities was associated with less turnover.To the extent that absences endanger employmentstatus, they should be lower for people who areembedded on- and off-the-job.

    Extending our reasoning further, off-the-job em-beddedness may be more important to the predic-tion of turnover and absences than on-the-job em-

    beddedness when satisfaction and commitment(which are on-the-job constructs) are controlled.First, at least some of an individuals decisionsabout absence and leaving an organization should

    be associated with thoughts and considerationsabout what would happen if he or she did nothavea job (a hypothetical future or distal state). Thesethoughts (such as job loss owing to being absent toooften [Hulin, 1991]) involve potential disruptionsto the individuals community involvement, espe-cially if relocating were required (March & Simon,1958). In other words, these thoughts do not nec-essarily involve immediate on-the-job consider-ations but do involve off-the-job considerations.

    Second, Mitchell and his colleagues (2001) re-ported higher bivariate correlations between on-the-job embeddedness and satisfaction, commit-ment, and turnover than between off-the-jobembeddedness and satisfaction, commitment, andturnover. Thus, the effects of on-the-job embedded-ness on participation may occur in conjunctionwith work attitudes like satisfaction and commit-ment, whereas the effects of off-the-job embedded-ness on participation may be less shaped by atti-tudes. In other words, on-the-job embeddednessshares more variance with job attitudes than off-the-job embeddedness does; as a result, the highercorrelation between on-the-job embeddedness andturnover may be mostly due to effects shared with

    job attitudes. The lower (but significant) correlation between off-the-job embeddedness and turnovermay reflect different and new information.

    When considered together, the two argumentsmade above(1) people think about the future stateof not having a job and its possible effects on com-munity involvement and (2) the correlations be-tween on- and off-the-job embeddedness and workattitudes differlead to the following expectation:with the attitudes of satisfaction and commitmentcontrolled, the effects of on-the-job embeddednesson the decision to participate at work should not be

    2004 713Lee, Mitchell, Sablynski, Burton, and Holtom

  • 8/8/2019 JobEmbeddedness

    4/13

    significant, but the effects of off-the-job embedded-ness on turnover and absences should remain. Thatis, off-the-job embeddedness adds new informationabout why people are absent or leave.

    Hypothesis 1. After job satisfaction and orga-nizational commitment are statistically con-

    trolled for, off-the-job embeddedness nega-tively relates to voluntary turnover andvolitional absences, whereas on-the-job em-beddedness does not predict these withdrawalbehaviors.

    On the basis of the reviewed theories (Hulin,1991; Lee & Mitchell, 1994; March & Simon, 1958;Mitchell et al., 2001), we also believe that the de-cision to perform should be related to job embed-dedness via motivational effects. Because high on-the-job embeddedness reflects (1) many links, (2) agood fit, and/or (3) consequential things that an

    employee gives up by quitting, the motivation toperform should be high. That is, employees withhigh on-the-job embeddedness will (1) be involvedin and tied to projects and people, (2) feel they fitwell in their jobs and can apply their skills, and (3)sacrifice valued things if they quit. Correspond-ingly, the motivation to perform should be high.(Low motivation should occur when on-the-job em-

    beddedness is low.)The relationship between job embeddedness and

    the decision to perform can be further specified. Inthe last decade, the domain of performance has

    been divided into in-role and extra-role (e.g., Wil-liams & Anderson, 1991). In-role performance issimilar to job-description-based specifications ofperformance, whereas organizational citizenship

    behavior is part of a larger family of extra-role be-haviors (Van Dyne, Cummings, & McLean Parks,1995). Most often, citizenship is seen as an employ-ees actions that help others better perform theirjobs (for instance, training co-workers) and therebyenhance organizational effectiveness.

    Conceptually, the more an individual is job em-bedded (or socially enmeshed) in an organization,the more likely he or she should be to display

    citizenship behaviors. In particular, people may beinterdependent (or linked to one another), andhelpful acts may be consistent with their feelings ofcomfort (or fit) stemming from being part of thatsocial network. The more an employee fits a job,colleagues, and organization, the more natural itshould be to perform citizenship behaviors. In ad-dition, helping others may be perceived as promot-ing others future helpful acts. Foregoing the oppor-tunity to help other interdependent people maywell be seen as a sacrificed opportunity to gain anowed favor. Indeed, the theory and research on

    social exchange (Van Dyne & Ang, 1998), norms ofreciprocity (Gouldner, 1960), perceived organiza-tional support (Rhoades & Eisenberger, 2002), andwork status congruence (Holtom, Lee, & Tidd,2002) suggest that people come to feel obligatedand want to help persons and organizations thathave helped them.

    Much of the above reasoning explicitly involvesthe effect of on-the-job embeddedness on (in-role)job performance and (extra-role) organizational cit-izenship. Most importantly, the attributes of a joband an organization should be significantly moresalient for the immediate motivation (and decision)to perform than are off-the-job factors. On-the jobembeddedness should be more proximal to a deci-sion to perform (as manifested by citizenship be-haviors and job performance) than the more distaldecision to participate (as reflected by turnover andabsences, which involve future states and off-the-

    job considerations). That is, employees have to per-form immediately (or right now), whereas they maybe absent next week or quit next month. Althoughoff-the-job embeddedness should have an effect onperformance, it should be relatively minor. In par-ticular, the saliency of the immediate job and orga-nization supersedes, renders less meaningful, ormakes less potent the more distal effects of off-the-job embeddedness on the decision to perform.

    Hypothesis 2. After job satisfaction and orga-nizational commitment are controlled for, on-the-job embeddedness positively relates to or-

    ganizational citizenship and job performance,whereas off-the-job embeddedness is unrelatedto these performance indicators.

    In their original meta-analysis, Hom and Griffeth(1995) reported an estimated rho of .33 betweenvolitional absences and voluntary turnover, and anestimated rho of.19 between job performance andemployee turnover. In their update, Griffeth et al.(2000) reported a rho of .20 between absences andturnover and a rho as .15 between performanceand turnover. From these summary findings, anenduring conclusion is that increased absence sig-

    nals more turnover and that good performance sig-nals less turnover. We propose that on-the-job em-

    beddedness moderates the effect of absences onturnover and the effect of job performance on turn-over. As suggested above, higher on-the-job embed-dedness reflects more links, better fit, and moreconsequential losses if an employee quits. As such,people with higher on-the-job embeddednessshould to some extent believe and be concernedthat more volitional absences and lower job perfor-mance may endanger the status of being employedand/or attached to their jobs. Conversely, people

    714 OctoberAcademy of Management Journal

  • 8/8/2019 JobEmbeddedness

    5/13

    with lower on-the-job embeddedness should holdthis belief and concern to a lesser extent.

    Hypothesis 3a. On-the-job embeddednessmoderates the positive effect of volitional ab-sences on voluntary turnover in such a waythat these effects are stronger for higher than

    for lower on-the-job embeddedness.Hypothesis 3b. On-the-job embeddednessmoderates the negative effect of job perfor-mance on voluntary turnover in such a waythat these effects are stronger for higher thanfor lower on-the-job embeddedness.

    Less theoretical and empirical evidence exists onthe relationships among organizational citizenshipand participation (absences and turnover) than ex-ists for in-role performance. For example, a review

    by Podsakoff, MacKenzie, Pain, and Bachrach

    (2000) did not report empirical evidence on theserelationships. To our knowledge, only Chen, Hui,and Sego (1998) have reported that supervisor-rated organizational citizenship behaviors (OCBs)predict subordinates subsequent turnover. To theextent that OCBs constitute a form of performance,however, our prior arguments in Hypotheses 3aand 3b should hold for a moderating role of jobembeddedness on the effects of OCBs on turnoverand absences as well.

    Hypothesis 4a. On-the-job and off-the-job em-beddedness moderate the negative effects of

    organizational citizenship behavior on volun-tary employee turnover in such a way thatthese effects are stronger for higher than forlower on-the-job embeddedness.

    Hypothesis 4b. On-the-job and off-the-job em-beddedness moderate the negative effect of or-ganizational citizenship on volitional absencesin such a way that these effects are stronger forhigher than for lower on-the-job embedded-ness.

    METHODSIn early 1998, we contacted, visited, and gained

    access to data at a regional operations center of alarge international financial institution. The locallabor market for this operations center was excep-tionally tight, with unemployment below 3 per-cent. In September 1998, the two senior authorsconducted a focus group with ten randomly se-lected employees, who discussed how this studysmajor variables might embed them in their jobs andcommunity. From this focus groups information,our surveys were tailored to this particular research

    site. In December 1998, an employee survey wasconducted. Individuals names, employee num-

    bers, personal characteristics, job satisfaction, orga-nizational commitment, and on- and off-the-jobembeddedness were assessed with voluntary self-reported measures. In January 1999, unit supervi-sors rated their subordinates organizational citi-

    zenship behavior and job performance. Absencesand turnover for calendar year 1999 were collectedfrom company records.

    Surveys were distributed to 1,650 employees infive separate organizational units. Of the 1,650 sur-veys, 839 surveys (51%) were returned. Ten sur-veys were not included in the analyses becausethey: (1) lacked a signed consent form, (2) wereillegible, (3) had no identifying information, or (4)were identifiable but blank. Thus, our samplesdata come from 829 employees and represent a 50percent response rate. Next, the immediate super-

    visors (the unit managers) of these 829 subordi-nates rated their subordinates in-role performanceand extra-role behavior. Of the 829 surveys, match-ing unit manager surveys were returned for 636individuals (76.7%). Within our sample, 75.3 per-cent were women; the overall average age was 34.2years (s.d. 9.9), and the average tenure with theorganization was 6.6 years (s.d. 5.1). The major-ity of respondents had some college (48.3%) or aB.A. or B.S. degree (25.1%). Statistical comparisons

    between the sample and overall population (all em-ployees within the operations center) yielded nosignificant differences in age, gender, and tenure.In addition, no significant differences were foundon turnover, job performance, and organizationalcitizenship across our five organizational units.Significant differences were found in absences inone of our units; the other four units did not differin absences. Moreover, the ratings of the 20 super-visors who rated only a single subordinate werecompared to a random sample of another 20 super-visors who rated multiple subordinates. No signif-icant differences were found between supervisorsratings for one versus more subordinates on citizen-ship or job performance. These comparisons sug-

    gest that sampling bias, although not completelydiscounted, was not a major problem.

    Measures

    Voluntary turnover. Each month of the year fol-lowing administration of our survey, the host orga-nization provided a list of individuals who hadvoluntarily or involuntarily left the organization.One hundred thirty-six individuals were classifiedas voluntary leavers (16.4%), and 12 others wereinvoluntarily terminated. To verify these lists, we

    2004 715Lee, Mitchell, Sablynski, Burton, and Holtom

  • 8/8/2019 JobEmbeddedness

    6/13

    tried to contact each person who was classified as avoluntary leaver. Seventy-two of the 136 voluntaryleavers were telephoned during the month follow-ing their quitting and confirmed their voluntaryleaving; the other 64 individuals could not bereached. In the analysis, stayers were coded as 0and leavers as 1.

    Volitional absences. The host organization pro-vided absenteeism records for the year followingadministration of the survey. The organization clas-sified absences as paid (excused) or unpaid (unex-cused). Because we were concerned with volitionalabsences, our analysis focused on unpaid absences.The total number of monthly unpaid hours absentper employee was observed for the 12 months fol-lowing the administration of the survey. We wereable to obtain absentee data for 761 employees.Because some employees left the organization priorto the end of the 12-month observation period, an

    average monthly absenteeism figure was calculatedfor all persons. Our unpaid absence data also ex-hibited a positive skew and a kurtosis and lackednormality. In order to achieve better fitting andmore normal distributional properties, a squareroot transformation was applied.

    Organizational citizenship behavior. The im-mediate supervisor of each survey respondent ratedthe latters citizenship behavior on eight items thatwere adapted from the Van Dyne and LePine (1998)organizational citizenship scale. Response optionswere 1 (never) to 5 (always), and a sample itemis volunteers to do things that are not required. Atotal of 632 (76.2%) employees were rated. A factoranalysis indicated unidimensionality, and an aver-aged composite was used in the analysis ( .93).

    Job performance. The participants unit manag-ers assessed their job performance with the six-itemscale developed by Williams and Anderson (1991).Its response options were 1 (never) to 5 (al-ways), and a sample item is performs all tasksthat are expected of him or her. Job performanceratings were completed for 632 of the employees(76.2%). A factor analysis indicated unidimension-ality, and an averaged composite was used in the

    analysis ( .92). Job embeddedness. Although most items corre-

    sponded directly to Mitchell and associates mea-sure of job embeddedness, a few minor edits wererequired to fit the measure to the current samplessetting. These changes incorporated unique en-meshing opportunities available to the employeeswithin the host organization and its local commu-nity. In addition, additional items emerged fromthe focus group and meetings with representativeemployees, managers, and upper management. TheAppendix shows all our embeddedness items. As

    did Mitchell and colleagues (2001), we averageditems for on- and off-the-job embeddedness overtheir three subdimensions into composite scores(s .84 and .82, respectively).

    Organizational commitment was assessed witheight items from Meyer and Allens (1997) measureof affective commitment. A sample item is I enjoy

    discussing my organization with people outside ofit. Job satisfaction was measured with three items.A sample item is All in all, I am satisfied with myjob. Five-point Likert scales were used, and factoranalyses indicated unidimensionality. Averagedcomposites were used in the analysis (respectives .85 and .86).

    An exploratory factor analysis with varimax ro-tation was conducted on all self-reported items forjob embeddedness, organizational commitment,and job satisfaction. Visual inspection of the screeplot suggested a three-factor solution. Items for job

    satisfaction, organizational commitment, and thefit and sacrifice dimensions of on-the-job embed-dedness loaded on factor 1 (however, one sacri-fice item did not load at all). All items for the fitand sacrifice dimensions of off-the-job embedded-ness loaded on factor 2. The links items for on- andoff-the-job embeddedness loaded on factor 3, ex-cept for two that did not load at all. Given theirconceptual overlap, the loading of all items forsatisfaction, commitment, and on-the-job embed-dedness, fit and sacrifice, onto a single factor was to

    be expected and suggested some evidence for con-vergent validity. The separate factors for off-the-jobembeddedness, fit and sacrifice, and for links sug-gest some evidence of discriminant validity. (Thefactor pattern matrix is available upon request tothe senior author.)

    Analyses. Logistic regression equations were cal-culated for Hypotheses 1, 3a, 3b, and 4a, and ordi-nary least squares (OLS) regressions were calcu-lated for Hypotheses 2 and 4b. We examined themain underlying assumptions of all the statisticaltests of hypotheses and found no major violations(such as outliers, major deviations from normality,or multicolinearity). In particular, the variance in-

    flation factors for the regressions that containedonly main effects were all below 3. As expected,however, multicollinearity did emerge when in-teraction terms were entered into the regressionanalyses.

    RESULTS

    Table 1 shows the means, standard deviations,and correlations for all variables in this study. Off-and on-the-job embeddedness significantly relatedto turnover, citizenship, performance, satisfaction,

    716 OctoberAcademy of Management Journal

  • 8/8/2019 JobEmbeddedness

    7/13

    and commitment. Whereas off-the-job embedded-ness did, on-the-job embeddedness did not relatesignificantly to volitional absences.

    Tests of Hypotheses

    Hypothesis 1 holds that, when satisfaction andcommitment are statistically controlled, off-the-jobembeddedness remains negatively related withturnover and absences, whereas on-the-job embed-dedness is unrelated. Equation 1 in Table 2, which

    reports the results of analyses of the hypothesizeddirect effects, shows the regression of turnover ontosatisfaction, commitment, and on- and off-the-jobembeddedness. As hypothesized, the coefficient foroff-the-job embeddedness is significant and showsa negative effect, whereas the coefficient for on-the-job embeddedness is nonsignificant. Equation 2shows the regression of absences onto satisfaction,commitment, and off- and on-the-job embedded-ness. The coefficient for off-the-job embeddednessis significant and negative, whereas the coefficient

    TABLE 1Means, Standard Deviations, and Correlationsa, b

    Variable Mean s.d. 1 2 3 4 5 6 7

    1. Voluntary turnover 0.16 0.372. Performance (in-role) 4.08 0.65 .12**3. OCB (extra-role) 3.07 0.88 .08* .62***

    4. Volitional absences 3.49 7.43 .22*** .17*** .17***5. On-the-job embeddedness 2.67 0.49 .11** .11** .19*** .016. Off-the-job embeddedness 2.88 0.54 .13*** .10** .11** .16*** .33***7. Job satisfaction 3.60 0.84 .10** .02 .07 .03 .73*** .23***8. Organizational commitment 2.91 0.71 .09** .02 .06 .07 .71*** .22*** .69***

    a n 805 for column 1 (turnover); n 809 for all other variables (n 623 for performance and OCB).b Column 1 reports point-biserial correlations; all other columns report product-moment correlations (two-tailed tests of significance).

    The correlations for volitional absences are based on square-root transformations.* p .05

    ** p .01*** p .001

    TABLE 2Effects of Job Embeddedness

    Predictors

    Dependent Variables

    Equation 1:VoluntaryTurnovera

    Equation 2:VoluntaryAbsencesb

    Equation 3:Job Performanceb

    Equation 4:Organizational

    Citizenship Behaviorb

    Job satisfaction .91 .01 .19** .10Organizational commitment .92 .11* .06 .11On-the-job embeddedness .80 .02 .27*** .32***Off-the-job embeddedness .60** .18*** .07 .05

    F or 2 log-likelihood 706.78 6.78*** 6.24*** 8.52***R2 .04 .04 .05R2 or 2

    On-the-job embeddedness .02*** .04***Off-the-job embeddedness 7.75** .03***

    n 805 739 620 620

    a Logistic regression. The entries are exponentiated bs. Entries above 1.00 indicate positive effects, and entries below 1.00 indicatenegative effects.

    b The entries are standardized regression coefficients when all variables are entered into the equation.* p .05

    ** p .01*** p .001Two-tailed tests.

    2004 717Lee, Mitchell, Sablynski, Burton, and Holtom

  • 8/8/2019 JobEmbeddedness

    8/13

    for on-the-job embeddedness is nonsignificant.Thus, Hypothesis 1 is supported.

    Hypothesis 2 holds that, when satisfaction andcommitment are controlled, on-the-job embedded-ness remains positively related with citizenshipand performance, whereas off-the-job embedded-ness is unrelated. Equation 3 in Table 2 shows the

    regression of performance onto satisfaction, com-mitment, and off- and on-the-job embeddedness.As hypothesized, only the coefficient for on-the-jobembeddedness is significant and positive, whereasthe coefficient for off-the-job embeddedness is notsignificant. Equation 4 shows the regression of cit-izenship onto satisfaction, commitment, and off-and on-the-job embeddedness. The coefficient foron-the-job embeddedness is significant and posi-tive, whereas the coefficient for off-the-job embed-dedness is not significant. Thus, Hypothesis 2 issupported.

    Hypothesis 3a predicts that on-the-job embed-dedness moderates the positive effect of volitionalabsences on quitting, and Hypothesis 3b predictsthat on-the-job embeddedness moderates the nega-tive effect of job performance on quitting. In eachcase the moderation is such that the effects arestronger for higher than for lower on-the-job em-

    beddedness. Equation 1 in Table 3, which reportsthe results of analyses of the hypothesized moder-ation effects, shows the regression of turnover ontosatisfaction, commitment, on- and off-the-job em-

    beddedness, absences, and the interactions be-tween on-the-job embeddedness and absences and

    between off-the-job embeddedness and absences.As hypothesized, the coefficient for the interaction

    between on-the-job embeddedness and absences isstatistically significant and shows a positive effect,whereas the interaction between off-the-job embed-dedness and absences is nonsignificant. To de-scribe this interaction, we calculated separate re-gressions for high and low groups based on amedian split of on-the-job embeddedness. The highgroup has a steeper positive slope than the lowgroup for absences regressed on turnover (exp b 1.63, p .001, vs. 1.27, p .01). Equation 2 in

    Table 3 shows the regression of turnover onto sat-isfaction, commitment, on- and off-the-job embed-dedness, performance, and the interactions be-tween on-the-job embeddedness and performanceand between off-the-job embeddedness and perfor-mance. The coefficient for the interaction betweenon-the-job embeddedness and performance is sig-nificant and shows a negative effect, whereas thecoefficient for the interaction between off-the-jobembeddedness and performance is nonsignificant.To describe the significant interaction, we calcu-lated separate regression equations for high and

    low groups based on a median split of on-the-jobembeddedness. The high group has a negative andsignificant slope (exp b 0.41, p .001), whereasthe low group has a negative but nonsignificant

    slope for performance regressed on turnover. Thus,Hypotheses 3a and 3b are supported.

    Hypothesis 4a predicts that on- and off-the-jobembeddedness moderate the negative effect of or-ganizational citizenship on quitting (turnover), andHypothesis 4b makes a similar prediction for voli-tional absences. In both cases, the moderation issuch that these effects are stronger for higher thanfor lower on-the-job embeddedness. In Table 3,equation 3 shows the regression of turnover ontosatisfaction, commitment, on- and off-the-job em-

    beddedness, citizenship, and the interactions be-

    TABLE 3Moderating Effects of Job Embeddedness and

    Work Behaviors on Voluntary Turnovera

    Predictors Equation 1 Equation 2 Equation 3

    Job satisfaction 0.89 0.76 0.76Organizational

    commitment

    0.83 0.85 0.89

    On-the-job embeddedness 0.49 14.27 7.26*Off-the-job embeddedness 0.68 0.09 0.09**Voluntary absences 0.60 Job performance 0.89Organizational

    citizenship behavior0.78

    On-the-job embeddedness voluntary absences

    1.37*

    Off-the-job embeddedness voluntary absences

    1.01

    On-the-job embeddedness job performance

    0.50*

    Off-the-job embeddedness job performance

    1.65

    On-the-job embeddedness organizationalcitizenship behavior

    0.49**

    Off-the-job embeddedness organizationalcitizenship behavior

    1.92**

    2 log-likelihood 622.92 505.40 504.802 b 6.05* 4.82 11.15***n 740 621 621

    a

    Logistic regressions. The entries are exponentiated bs. En-tries above 1.00 indicate positive effects, and entries below 1.00indicate negative effects.

    b For interactions. p .10* p .05

    ** p .01*** p .001Two-tailed tests.

    718 OctoberAcademy of Management Journal

  • 8/8/2019 JobEmbeddedness

    9/13

    tween on-the-job embeddedness and citizenshipand between off-the-job embeddedness and citizen-ship. As hypothesized, the coefficients for the in-teractions between on-the-job embeddedness andcitizenship and between off-the-job embeddednessand citizenship are statistically significant. To de-scribe these interactions, we calculated separate

    regressions for high and low groups based on amedian split of both on- and off-the-job embedded-ness. For the high on-the-job embeddedness group,the slope for citizenship regressed on turnover isnegative and significant (exp b 0.55, p .01),whereas the slope for the low group is nonsignifi-cant. For the high off-the-job embeddedness group,the slope for citizenship regressed on turnover isnonsignificant, whereas the slope for the low groupis negative and significant (exp b 0.65, p .05).Thus, Hypothesis 4a is supported.

    For Hypothesis 4b, absences were regressed onto

    satisfaction, commitment, on- and off-the-job em-beddedness, and the interactions between on-the-job embeddedness and citizenship and betweenoff-the-job embeddedness and citizenship. (This re-gression is not shown but is available upon requestto the senior author.) Only the interaction for on-the-job embeddedness and citizenship is statisti-cally significant ( 0.56, p .05). To describethis interaction, we calculated separate regressionequations for high and low groups based on a me-dian split of on-the-job embeddedness. The highgroup showed a negative and significant slope thatwas steeper than the low groups ( 0.19, p .001, vs. 0.12, p .05). Thus, Hypothesis 4b isonly partially supported.

    Post Hoc Analyses

    Table 2 shows different patterns of significantcontribution toward prediction of turnover, ab-sences, performance, and citizenship for on- andoff-the-job embeddedness. Post hoc, we tested thestronger inferences that off-the-job embeddednessis a better predictor than on-the-job embeddedness

    for turnover and absences and that on-the-job em- beddedness is a better predictor than off-the-jobembeddedness for performance and citizenship.First, by adding scores for on- and off-the-job em-

    beddedness we created an overall job embedded-ness score. Second, we regressed each dependentvariable onto the predictors of satisfaction, commit-ment, and overall embeddedness. Third, we com-pared the difference in variance explained (R2) inthe three OLS regressions shown in Table 2 (that is,predicting absences, performance, and citizenshipwith satisfaction, commitment, on-the-job embed-

    dedness, and off-the-job embeddedness) with thevariance explained in the corresponding regres-sions containing satisfaction, commitment, andoverall embeddedness. For the corresponding lo-gistic regression analyses, we compared deviancesfrom these two equations with a G-test (Hosmer &Lemeshow, 2000). To reiterate, the new regressions

    (both the OLS and logistic versions), with theircoefficients for overall embeddedness, allowed forstronger inferences on the strength of predictions.For absences, the four-variable model explained 4percent of variance (R2 .04), whereas the three-variable model explained 3 percent (R2 .03; bothp .001). The difference between the two R2s wassignificant (F 3.20, p .05, with a one-tailed test,

    but p .10, with a two-tailed test). For perfor-mance, the four- and three-variable models had thesame respective explained variances as the modelsfor absences (both p .001). The difference in R2s

    was significant (F 6.01, p .05 with one- andtwo-tailed tests). For citizenship, the four-variablemodel explained 5 percent of variance, and thethree-variable model, 4 percent (both p .001).The difference in R2s was significant (F 11.51,p .001, with one- and two-tailed tests). For turn-over, the four-variable model had 2 log-likelihoodof 706.78, whereas the three-variable model had a2 log-likelihood of 707.31. The difference be-tween them (G) was nonsignificant. In sum, thesedata generally support the stronger inferences.

    DISCUSSIONThis study expands understanding of job embed-

    dedness. First, off-the-job embeddedness predictedturnover and absences, whereas on-the-job embed-dedness did not (Hypothesis 1). In contrast, on-the-job embeddedness predicted organizational citi-zenship and job performance, whereas off-the-jobembeddedness did not (Hypothesis 2). Second, thetwo components of job embeddedness may be pro-cesses through which the decisions to perform andto participate can be conceptually and empiricallylinked. On-the-job embeddedness moderated the

    positive effect of volitional absences on turnover(Hypothesis 3a), the negative effect of job perfor-mance on turnover (Hypothesis 3b), and the nega-tive effect of citizenship on absences; the modera-tion was such that these effects were stronger forhigher than for lower on-the-job embeddedness(Hypothesis 4a).

    Three particular limitations of this study shouldbe noted. First, the timing for some of our measuresprovides only limited support for causal infer-ences. Although our four behavioral outcome vari-ables were measured independently from and after

    2004 719Lee, Mitchell, Sablynski, Burton, and Holtom

  • 8/8/2019 JobEmbeddedness

    10/13

    our respondents self-reports of satisfaction, com-mitment, and embeddedness, organizational citi-zenship and job performance were assessed only amonth after the self-reported survey. Although it islikely that job embeddedness was a cause of ouroutcome variables, the reverse direction may hold.Second, the measures of on- and off-the-job embed-

    dedness are still preliminary and evolving. Al-though our data on factor structures and internalconsistencies produce empirical findings similar toearlier work, these measures are not yet establishedand standard research instruments. Third, we dis-aggregated embeddedness into on- and off-the jobcomponents. In the future, profiles of the six em-

    beddedness dimensions may be useful for predic-tion and understanding.

    It should be mentioned that Meyer and Allens(1997) dimension of continuance commitment andthe embeddedness subdimension of organization-

    related sacrifice (sacrificeorganization) are simi-lar. Both dimensions share notions of sunk costand reluctance to give things up by leaving. How-ever, the original continuance commitment con-struct combined reluctance with the availability ofalternatives, which organization-related sacrificedoes not. Even if items for alternatives are omittedfrom the continuous commitment measure, itemsfor organization-related sacrifice have much morespecific and targeted referents such as perks, re-spect, compensation, benefits (retirement andhealth care), and promotional opportunities. Thus,our measure omits the alternatives idea and of-fers more detail in terms of the specific topics thancontinuance commitment. (See Yao, Lee, Mitchell,Burton, and Sablynski [2004] for a comprehensivecomparison between job embeddedness and relatedconstructs, including continuance commitment.)

    Overall, our results indicate the appropriatenessof studying retention and performance as tandemjob behaviors and viewing job embeddedness as ameaningful mechanism through which to under-stand this linkage. In particular, Mitchell and co-authors (2001) reported significant predictive asso-ciations between aggregated job embeddedness and

    turnover in two samples. In this study, disaggre-gated job embeddedness predicted turnover, ab-sences, in-role performance, and organizational cit-izenship. Thus, meaningful statistical effects werefound over three diverse and sizable samples. Inthe future, it may be timely for researchers to move

    beyond simple prediction and predictive validitydesigns. Given the existing studies, research de-signs that allow for stronger causal inferences arenow needed. Field or quasi field experiments thatinclude interventions aimed at altering job embed-dedness are suggested. Alternatively, field studies

    that measure job embeddedness immediately be-fore and after acts of organizational citizenship andformal appraisals of in-role performance, absencespells, or individual quitting might yield valuableevidence on causal effects.

    In our view, the managerial implications of thisstudy are clear. Job embeddedness (which can be

    established through building community, develop-ing a sense of belonging, establishing deep tiesamong employees, and deepening social capital)may increase retention, attendance, citizenship,and job performance. Furthermore, organizationscan be proactive about job embeddedness: links can

    be increased through teams and long-term projects;sacrifice can be increased by connecting job andorganizational rewards to longevity; and fit can beincreased by matching employees knowledge,skills, abilities, and attitudes with a jobs require-ments. Equally important, managers can increase

    off-the-job embeddedness by providing peoplewith information about the community surround-ing their workplace and by providing social sup-port for local activities and events (Mitchell, Hol-tom, & Lee, 2001).

    In closing, our results suggest that studying em-ployees reasons for both staying and leaving mayenrich knowledge of retention, increasing it beyondwhat the current focus on leaving permits. This

    broader perspective suggests an interesting and po-tentially fruitful direction for future research.

    REFERENCES

    Chen, X. P., Hui, C., & Sego, D. J. 1998. The role oforganizational citizenship behavior in turnover:Conceptualization and preliminary test of key hy-potheses. Journal of Applied Psychology, 83: 922931.

    Gouldner, A. W. 1960. The norm of reciprocity: A pre-liminary statement. American Sociological Re-view, 25: 161178.

    Griffeth, R. W., Hom, P. W., & Gaertner, S. 2000. A meta-analysis of antecedents and correlates of employeeturnover: Update, moderator tests, and research im-

    plications for the millennium. Journal of Manage-ment, 26: 463488.

    Hanish, K. A., & Hulin, C. L. 1991. General attitudes andorganizational withdrawal: an evaluation of a causalmodel. Journal of Vocational Behavior, 39: 110128.

    Holtom, B. C., Lee, T. W., & Tidd, S. T. 2002. The rela-tionship between work status congruence and work-related attitudes and behaviors. Journal of Applied

    Psychology, 87: 903915.

    Hom, P. W., & Griffeth, R. W. 1995. Employee turnover.Cincinnati: South-Western.

    720 OctoberAcademy of Management Journal

  • 8/8/2019 JobEmbeddedness

    11/13

    Hosmer, D. W., & Lemeshow, S. 2000. Applied logisticregression (2nd ed.). New York: Wiley.

    Hulin, C. L. 1998. Behaviors, constructs and time: Pot- holes on the road well traveled. Invited address,annual meeting of the Society for Industrial andOrganizational Psychology, Dallas.

    Hulin, C. L. 1991. Adaptation, persistence and commit-

    ment in organizations. In M. D. Dunnette & L. M.Hough (Eds.), Handbook of industrial and organi-

    zational psychology(2nd ed.): 445507. Palo Alto,CA: Consulting Psychologists Press.

    Hulin, C. L. 2002. Lessons from industrial and organiza-tional psychology. In J. Brett & F. Drasgow (Eds.),The psychology of work: Theoretically based em-

    pirical evidence. Mahwah, NJ: Erlbaum.

    Lee, T. W., & Mitchell, T. R. 1994. An alternative ap-proach: The unfolding model of voluntary employeeturnover. Academy of Management Review, 19:5189.

    Lee, T. W., Mitchell, T. R., Holtom, B. C., McDaniel, L., &Hill, J. W. 1999. Theoretical development and exten-sion of the unfolding model of voluntary turnover.

    Academy of Management Journal, 42: 450462.

    Maertz, C. P., & Campion, M. A. 1998. 25 years of volun-tary turnover research: A review and critique. Inter-

    national Review of Industrial and OrganizationalPsychology, 13: 49 81.

    March, J. G., & Simon, H. A. 1958. Organizations. NewYork: Wiley.

    Meyer, J. P., & Allen, N. J. 1997. Commitment in theworkplace: Theory, research, and application.

    Thousand Oaks, CA: Sage.Mitchell, T. R., Holtom, B. C., & Lee, T. W. 2001. How to

    keep your best employees: The development of aneffective attachment policy. Academy of Manage-

    ment Executive, 15(4): 96108.

    Mitchell, T. R., Holtom, B. C., Lee, T. W., Sablynski, C. J.,& Erez, M. 2001. Why people stay: Using job embed-dedness to predict voluntary turnover. Academy of

    Management Journal, 44: 11021121.

    Podsakoff, P. M., MacKenzie, S. B., Pain, J. B., & Bach-rach, D. G. 2000. Organizational citizenship behav-iors: A critical review of the theoretical and empiri-

    cal literature and suggestions for future research. Journal of Management, 26: 513564.

    Rhoades, L., & Eisenberger, R. 2002. Perceived organiza-tional support: A review of the literature. Journal of

    Applied Psychology, 87: 698714.

    Van Dyne, L., & Ang, S. 1998. Organizational citizenship behavior of contingent workers in Singapore. Acad-emy of Management Journal, 41: 692703.

    Van Dyne, L., & LePine, J. A. 1998. Helping and voiceextra-role behaviors: Evidence of construct and pre-dictive validity. Academy of Management Journal,41: 108119.

    Van Dyne, L., Cummings, L. L., & McLean Parks, J. 1995.Extra-role behaviors: In pursuit of construct and def-initional clarity (a bridge over muddied waters). InL. L. Cummings & B. M. Staw (Eds.), Research inorganizational behavior, vol. 17: 215285. Green-wich, CT: JAI Press.

    Williams, L. J., & Anderson, S. E. 1991. Job satisfaction

    and organizational commitment as predictors of cit-izenship and in-role behaviors. Journal of Manage-

    ment, 17: 601617.

    Yao, X., Lee, T. W., Mitchell, T. R., Burton, J. P., &Sablynski, C. S. 2004. Job embeddedness: Currentresearch and future directions. In R. Griffeth & P.Hom (Eds.), Understanding employee retentionand turnover: 153-187. Greenwich, CT: InformationAge.

    APPENDIXJob Embeddedness Itemsa

    Fit, communityI really love the place where I live.b

    I like the family-oriented environment of mycommunity.This community I live in is a good match for me.I think of the community where I live as home.b

    The area where I live offers the leisure activities thatI like (e.g., sports, outdoors, cultural, arts).

    Fit, organizationMy job utilizes my skills and talents well.b

    I feel like I am a good match for this organization.b

    I feel personally valued by (name of the

    organization).I like my work schedule (e.g., flextime, shift).I fit with this organizations culture.b

    I like the authority and responsibility I have at thiscompany.b

    Links, communityAre you currently married?b

    If you are married, does your spouse work outsidethe home?b

    Do you own the home you live in? (mortgaged oroutright)My family roots are in the community where I live.

    Links, organizationHow long have you been in your present position?(years)b

    How long have you worked for this organization?(years).b

    How long have you worked in the (banking)industry? (years).b

    How many coworkers do you interact withregularly?b

    How many coworkers are highly dependent on you?b

    How many work teams are you on?b

    How many work committees are you on?b

    Continued

    2004 721Lee, Mitchell, Sablynski, Burton, and Holtom

  • 8/8/2019 JobEmbeddedness

    12/13

    APPENDIX Continued

    Sacrifice, communityLeaving this community would be very hard.b

    People respect me a lot in my community.b

    My neighborhood is safe.b

    Sacrifice, organization

    I have a lot of freedom on this job to decide how topursue my goals.b

    The perks on this job are outstanding.b

    I feel that people at work respect me a great deal.b

    I would incur very few costs if I left thisorganization.c

    I would sacrifice a lot if I left this job.b

    My promotional opportunities are excellent here.b

    I am well compensated for my level of performance.b

    The benefits are good on this job.b

    I believe the prospects for continuing employmentwith this company are excellent.b

    a Items 13 for links, community, and links, organization,

    were standardized before being analyzed or included in any

    composites.b Item used by Mitchell, Holtom, Lee, Sablynski, and Erez

    (2001).c Reverse coded.

    Thomas W. Lee ([email protected]) is a professorof human resource management and organizational be-havior and the Evert McCabe Faculty Fellow at the Uni-versity of Washington Business School. He earned hisPh.D. in organizational studies at the University of Ore-

    gon. His current research interests include employee re-tention, staffing, work motivation, and research methods.

    Terence R. Mitchell ([email protected]) is the Ed-ward E. Carlson Professor of Business Administrationand a professor of psychology at the University of Wash-ington Business School. He earned his Ph.D. in socialpsychology at the University of Illinois. His current re-search interests include employee retention, work moti-vation, and decision making.

    Chris J. Sablynski ([email protected]) is an assistantprofessor of organizational behavior and environment atCalifornia State University, Sacramento. He earned hisPh.D. in organizational behavior and human resourcemanagement at the University of Washington BusinessSchool. His current research interests include employeeretention and workplace deviance.

    James P. Burton ([email protected]) is an assistant pro-fessor of management at the University of Washington,Bothell. He earned his Ph.D. in organizational behaviorand human resource management at the University ofWashington Business School. His current research inter-ests include workplace fairness, employee retention, andteaching effectiveness in university settings.

    Brooks C. Holtom ([email protected]) is an assistant profes-sor of management in the McDonough School of Business

    at Georgetown University. He earned his Ph.D. in orga-nizational behavior and human resource management atthe University of Washington Business School. His cur-rent research interests include human and social capitaldevelopment and employee retention.

    722 OctoberAcademy of Management Journal

  • 8/8/2019 JobEmbeddedness

    13/13