keeping your headcount when all about you are...

KEEPING YOUR HEADCOUNT WHEN ALL ABOUT YOU ARELOSING THEIRS: DOWNSIZING, VOLUNTARY TURNOVER

RATES, AND THE MODERATING ROLEOF HR PRACTICES

CHARLIE O. TREVORANTHONY J. NYBERG

University of Wisconsin–Madison

Although both downsizing and voluntary turnover have been topics of great interest inthe organizational literature, little research addresses the topics’ possible relationship.Using organization-level data from multiple industries, we first investigate whetherdownsizing predicts voluntary turnover rates. Second, to support our causal model, weexamine whether aggregated levels of organizational commitment mediate this rela-tionship. Third, we test whether the downsizing–turnover rate relationship is (1)mitigated by HR practices that either embed employees in their organization or conveyprocedural fairness and (2) strengthened by HR practices that enhance career devel-opment. Results support the hypothesized main, mediated, and moderated effects.

Downsizing is commonplace today, with deploy-ment of the strategy constrained by neither com-pany financial health nor employee type (Cascio,2002a). Given the enormous organizational, indi-vidual, and societal implications of downsizing,the phenomenon warrants, and is receiving, sub-stantial research attention. Much of the downsizingresearch involves effects on either employee atti-tudes or organizational performance. Researchersknow very little, however, about downsizing effectson the critical employee behaviors that likely ac-company these attitudes and affect organizationalperformance. Perhaps the most telling of these be-haviors is voluntary employee turnover.

For at least three reasons, assessments of down-sizing viability or success appear to be incompleteif turnover implications are not considered. First,companies undertake downsizing—a planned re-duction in their number of employees (Cascio,1993)—to reduce headcount to some target level.Consequently, unexpected subsequent turnover es-sentially translates into missing that target andleaving the leaner organization understaffed, andthus hindered in its efficiency. Second, turnover-related financial costs, which include employeereplacement, training, and outplacement (Cascio,2000; Sturman, Trevor, Boudreau, & Gerhart, 2003)are quite high, with per-leaver estimates often dou-ble leaver salary (e.g., Johnson, 1995; Solomon,

1988). Given that downsizing is undertaken pri-marily as a cost-cutting measure (Cascio, 1993,2002b), incurring unforeseen expenses of this mag-nitude may jeopardize its success. Third, althoughevidence on the relationship between downsizingand organizational performance is mixed (Cappelli,2000; Madrick, 1995), turnover rates may shed lighton the dynamic. Chadwick, Hunter, and Walston(2004) found that the performance of hospitals thatdownsized appeared to benefit from the high-in-volvement human resource (HR) practices that sev-eral studies (e.g., Arthur, 1994; Batt, 2002; Guthrie,2001; Huselid, 1995) have shown to reduce turn-over rates. Moreover, recent research indicates thatturnover rates are negatively associated with busi-ness unit and organizational success (Glebbeek &Bax, 2004; Kacmar, Andrews, Rooy, Steilberg, &Cerrone, 2006; McElroy, Morrow, & Rude, 2001;Shaw, Gupta, & Delery, 2005). Thus, turnover rates,as well as the HR practices that affect them, may bekey elements in understanding how, and when,downsizing influences organizational performance.

The downsizing-turnover relationship is also ofconsiderable interest from the perspective of vol-untary turnover research itself. Because more valu-able employees are more likely to facilitate organi-zational success, the field is becoming increasinglyfocused on their retention (e.g., Lavelle, 2003; Lee,Mitchell, Sablynski, Burton, & Holtom, 2004; Stur-man et al., 2003; Trevor, 2001; Trevor, Gerhart, &Boudreau, 1997). Hence, studying voluntary turn-over among downsizing survivors, who have beenearmarked as the employees most worth keeping(Martin & Bartol, 1985), is especially relevant. Such

We thank Barry Gerhart for his invaluable assistancewith this project. We also appreciate the insights of SaraRynes and our three anonymous reviewers.

� Academy of Management Journal2008, Vol. 51, No. 2, 259–276.

259

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download or email articles for individual use only.

research also should provide insights into the un-folding model of turnover (Lee, Mitchell, Wise, &Fireman, 1996), a major recent addition to the lit-erature whose key tenets have received little atten-tion in multivariate predictive models.

Despite the organizational performance and vol-untary turnover implications of downsizing, onlyone study (Batt, Colvin, & Keefe, 2002) has testedwhether downsizing predicts voluntary turnoverrates. Consequently, we used a broad sample toexplore the relationship, the untested contentionthat organizational commitment mediates this as-sociation, and the extent to which the relationshipdepends on how organizations manage their hu-man resources.

CONCEPTUAL DEVELOPMENTAND HYPOTHESES

Three issues require clarification before we de-rive our hypotheses. First, the unfolding model ofturnover (Lee et al., 1996) describes how turnoverdecisions can be functions of “shocks” (i.e., jarringevents for employees), with downsizing cited asone such shock (Holtom, Mitchell, Lee, & Inder-rieden, 2005). We focus on the shock implicationsof both downsizing as an event and the severity(i.e., magnitude) of the downsizing. The event as-pect is evident in Lee and Mitchell’s definition of ashock as a “very distinguishable event that jarsemployees toward deliberate judgments about theirjobs and, perhaps, to voluntarily quit their job”(1994: 60). The severity emphasis evolves from ourassumption that larger downsizings are more jar-ring, shocking, and threatening than are smallerdownsizings; this logic is consistent with downsiz-ing’s operationalization as a percentage or propor-tion in prior downsizing research (e.g., Batt et al.,2002). Second, although downsizing is a multifac-eted process, our treatment focuses on layoffs. Be-cause such job loss is involuntary, layoffs representdownsizing at its most severe and are particularlyconsistent with the notion of downsizing as a “psy-chological contract” violation (Robinson & Rous-seau, 1994) and, thus, a jarring event, or shock, thatmay prompt surviving employees to consider leav-ing. Third, we primarily rely on individual-leveltheories to derive organization-level predictions.Recent work indicates that organization-level turn-over findings (e.g., Batt et al., 2002; McElroy et al.,2001; Shaw, Delery, Jenkins, & Gupta, 1998; Shawet al., 2005) are consistent with expectations in-ferred from individual-level frameworks; similarly,researchers have used individual-level perspec-tives on organizational commitment both to specu-late about organization-level relationships involv-

ing turnover (e.g., Batt et al., 2002; Huselid, 1995;Mishra & Spreitzer, 1998) and, via aggregation, totest such relationships (e.g., Simons & Roberson,2003).

Downsizing and Voluntary Turnover Rates

Main effect. The unfolding model of turnoverdescribes psychological and behavioral paths thatemployees take when quitting (Lee et al., 1996). Inthe model’s most commonly followed turnoverpath (path 3), a shock leads employees to reevalu-ate their current work circumstances, consider jobalternatives, and then quit, usually to take anotherjob offer or to pursue a likely offer. The results ofsubstantial individual-level research addressingthe effects of downsizing on survivor attitudes areconsistent with this path and the presumption thatdownsizing is an unpleasant shock that promptsfear and resentment. Downsizing has been foundto negatively affect organizational commitment(Brockner, Grover, Reed, Dewitt, & O’Malley, 1987;Knudsen, Johnson, Martin, & Roman, 2003), jobsatisfaction (Armstrong-Stassen, 2002; Luthans &Sommer, 1999), job involvement (Brockner, Grover,& Blonder, 1988), and trust (Armstrong-Stassen,2002). Given that such attitudes, particularly organ-izational commitment and job satisfaction, consis-tently predict voluntary turnover (e.g., Tett &Meyer, 1993), even at the business-unit level (e.g.,Simons & Roberson, 2003), a positive downsizing-turnover relationship seems likely. Indeed, in theonly study to examine the relationship, Batt et al.(2002) showed a positive link between the percent-age of core employees displaced and quit rates inthe telecommunications industry.

It is possible, however, that despite the attitudi-nal fallout, downsizing survivors are not morelikely to leave. Surviving job cuts, especially if thedownsizing is attributed to strategic restructuringrather than simple cost cutting, could lead togreater appreciation of what one has, a perceptionof relative safety (i.e., having “dodged a bullet”),expectations of more varied work, and belief thatthe company is positioned to excel. Overall, how-ever, we believe it unlikely that such retention-enhancing impressions will fully counteract thenegative attitudinal effects described above. Hence,we first attempted to replicate the Batt et al. (2002)finding in a broad sample of occupations, firms,and industries.

Hypothesis 1. Downsizing is positively relatedto voluntary turnover rates.

Mediated relationship. Because attitudinal ef-fects on voluntary turnover are central to so many

260 AprilAcademy of Management Journal

turnover models (e.g., Gerhart, 1990; Hom, Carani-kas-Walker, & Prussia, 1992; March & Simon, 1958;Mobley, 1977; Trevor, 2001), support for Hypothe-sis 1 will be enhanced if a relevant (aggregated)employee attitude emerges as a mediator. Mishraand Spreitzer’s (1998) conceptual model positionsorganizational commitment as such a mediator,and Batt et al. (2002) speculated that their down-sizing–quit rate finding stemmed from employees’perceiving a threat to employment security andsubsequently having reduced organizational com-mitment. Further, empirical work identifies organ-izational commitment as both a downsizing conse-quence (e.g., Brockner et al., 1987; Knudsen et al.,2003) and a voluntary turnover antecedent (Tett &Meyer, 1993 [meta-analysis]).

Hypothesis 2. Organizational commitment me-diates the positive relationship between down-sizing and voluntary turnover rates.

General Unfolding Model Contingency

Paralleling individual-level turnover research(e.g., Gerhart, 1990; Schwab, 1991; Trevor, 2001),

turnover prediction at the organizational levelshould be highly context-dependent. Our proposedinteractions are consistent with a general contin-gency framework that stems from the unfoldingmodel of turnover: a shock such as downsizing willoften result in employee scrutiny of a current jobsituation, which is then followed by job search,evaluation of alternative offers, and, depending onthe evaluation results, quitting or staying (Holtomet al., 2005; Mitchell, Holtom, & Lee, 2001). Thus, adownsizing shock that jars an employee onto thismost common of turnover pathways leads to a quitor stay decision based on an analysis of the utilityof (i.e., the value of and satisfaction with) one’scurrent employment relative to the expected utilityof alternative employment. We argue that HR prac-tice dimensions moderate downsizing effects byaffecting these postdownsizing utility levels. Fig-ure 1 depicts these relationships.

HR Practices That Buffer Downsizing Effects

Certain sets of HR practices should become moresalient as a downsizing shock leads to closer scru-

FIGURE 1HR Practices as Moderators of the Relationship between Downsizing and Voluntary Turnover Ratesa

a Both examples of moderation are consistent with the general contention of the unfolding model of turnover, that shocks lead to greaterscrutiny of the employment context. This scrutiny involves a comparison of the utilities of current and alternative employment. We arguethat HR practices largely determine the results of this comparison, and thus the extent to which downsizing affects voluntary turnoverrates.

2008 261Trevor and Nyberg

tiny of an employment situation. Two of these HRpractice dimensions may buffer (i.e., limit) thenegative impact of downsizing on current em-ployment utility and subsequent turnover. Ex-actly how this buffering effect occurs, however,depends on whether the HR practice dimensionpromotes perceptions of procedural justice or per-ceptions of job embeddedness.

Procedural justice. Procedural justice is the fair-ness of procedures used in decisions that affectemployees (Folger & Greenberg, 1985). The poten-tial for HR practices associated with perceived pro-cedural justice to buffer downsizing effects stemsfrom the fact that employment outcomes (e.g., pay)received over the long term are—given their largeimpact on material gains—seen as more importantthan a single short-term outcome. Procedural jus-tice perceptions fuel optimism regarding these crit-ical long-term outcomes, since fair processes sug-gest, even in light of short-term disappointment,that favorable outcomes lay ahead. This convictionleads people to assign less weight to, and be lessaffected by, short-term outcomes; indeed, a varietyof empirical work indicates that a single employ-ment outcome has a smaller effect in the presenceof procedurally just conditions associated with theoutcome’s allocation (Brockner & Wiesenfeld,1996). This contingency suggests that, given HRpractices that promote procedural justice percep-tions (e.g., a grievance or appeals process, a confi-dential problem-solving avenue, an ombudspersondesignated for complaint resolution), the unfavor-able short-term downsizing outcome will yield asmaller decrement in current employment utility;hence, high levels of such practices should reducethe effect of the downsizing shock on voluntaryturnover.1

Hypothesis 3. Downsizing’s positive relation-ship with voluntary turnover rates diminish athigh levels of HR practices likely to promoteprocedural justice perceptions.

Job embeddedness. The second HR practice di-mension of interest from the perspective of buffer-ing downsizing spillover is job embeddedness,which is the level to which employees’ are boundin a social web (both on and off the job) that keepsthem attached to their organization (Lee et al.,2004; Mitchell, Holtom, Lee, Sablynski, & Erez,2001). To examine whether employment outcomesthat tie employees’ lives to their current employerreduce downsizing effects on turnover, we createda job embeddedness index from five practices thatresearchers (Mitchell, Holtom, & Lee, 2001; Mitch-ell, Holtom, Lee, Sablynski, & Erez, 2001) havespecifically identified as contributing to job embed-dedness perceptions (i.e., defined benefits plans,sabbaticals, on-site childcare, hiring for organiza-tional fit, and flextime).

Although Mitchell, Holtom, and Lee stated thatjob embeddedness “seems to buffer the individualfrom the effects of shocks that might otherwiseprompt one to leave” (2001: 104), empirical re-search has yet to explicitly test whether shock ef-fects depend on either job embeddedness or thepractices that lead to it. A specific rationale forexpecting this interdependence is evident in a dis-cussion of multiple employment outcomes thatBrockner and Wiesenfeld (1996) provided as anextension of their procedural justice moderationlogic that we described above. Those authors ar-gued that, because people prefer positive outcomes,an unfavorable outcome in an exchange relation-ship may lead to greater attention to other relevantoutcomes in the relationship. This premise, re-cently supported in a multiple pay outcome con-text (Trevor & Wazeter, 2006), thus suggests that theunfavorable downsizing outcome should lead togreater scrutiny of other relevant employment out-comes, such as those associated with the HR prac-tices that yield job embeddedness perceptions (e.g.,a defined benefit plan, on-site childcare, and fitwith the organization). This greater attentionshould, at high levels of these embeddedness-rele-vant HR practices, make heightened job embedded-ness perceptions very salient; the heightened senseof attachment, in turn, would constrain downsiz-ing-driven reduction in current employment utilityand reduce downsizing effects on turnover rates. Incontrast, the unfavorable downsizing effect on em-ployees should be greater at lower levels of embed-dedness-relevant practices, where less favorableoutcomes mean that there is little sense of embed-dedness and attachment to constrain downsizing-driven reduction in the utility of the current em-ployment (i.e., little buffering of the downsizingshock).

1 Our position implies that procedural justice percep-tions other than those solely associated with a downsiz-ing event itself will reduce downsizing impact. As de-scribed above, procedural justice in one domain canreduce the effects of distributive outcomes because it canlead people to infer that they will be treated justly, andthus likely will fare well, in the future (Brockner &Wiesenfeld, 1996). Given this stipulation of proceduraljustice generalizing from one domain to another, HRpractices likely to yield procedural justice perceptionsnot specific to downsizing should also be likely to miti-gate downsizing effects by yielding optimism regardingfuture treatment.


Hypothesis 4. Downsizing’s positive relation-ship with voluntary turnover rates diminishesat high levels of HR practices likely to embedemployees in their organization.

HR Practices That Enhance AlternativeEmployment Utility

Whereas our focus in the previous two hypothe-ses was on HR practice dimensions that bufferdownsizing-driven reduction in the utility of cur-rent employment, here we address practices that,following a downsizing, should enhance the utilityof alternative employment. Because alternative em-ployment options are functions of both job search(e.g., Schwab, Rynes, & Aldag, 1987) and individ-ual attributes that signal employee quality to themarket (e.g., Spence, 1973), we focus on HR prac-tices that positively affect search and signaling ca-pabilities. Career development practices, in partic-ular, should foster these capacities.

At higher levels of career development practicesthat enhance search (e.g., a career resource center,formal employee assessment to identify develop-ment opportunities) and/or signaling (e.g., formalsuccession planning, a transfer policy to supportlearning), employees should enjoy the ease ofmovement in the job market that alternative em-ployment options bring. In previous research usingemployee quality signals (Trevor, 2001) and mar-ket-level job availability (Carsten & Spector, 1987;Gerhart, 1990; Trevor, 2001) as indications of highlevels of such ease of movement, job dissatisfactionhas been more likely to lead to voluntary turnover.Thus, it also seems likely that downsizing-drivenattitudinal fallout and scrutiny of the employmentsituation will be more likely to lead to quitting athigher levels of career development practices thatenhance search, signaling, and subsequent alterna-tive employment options (i.e., when ease of move-ment is high). In other words, shocked employeeswill be more likely to leave when HR practicesenhance alternative employment utility by helpingto provide better external job opportunities (seeFigure 1).

We recognize, however, that career developmentpractices also prepare employees for internal op-portunities. As a downsizing shock drives employ-ees to evaluate their current employment situation,the perception of internal movement potentialcould enhance the utility of the current employ-ment. On balance, however, we expect that careerdevelopment practices should do more, in the post-downsizing environment of decaying commitmentand satisfaction, to enhance the utility of alterna-

tive employment (via improved search and signal-ing) than to improve current employment utility.

Hypothesis 5. Downsizing’s positive relation-ship with voluntary turnover rates strengthensat high levels of HR practices likely to enhancecareer development.

METHODS

Sample

Our sample consisted of employee-friendly com-panies that were invited to submit detailed infor-mation in applications to be included in the 1998and 1999 Fortune magazine articles on “the 100best companies to work for in America” (Branch,1999; Levering & Moskowitz, 2000). Levering andMoskowitz (2000) chose the invitees from theirown database of over 1,000 companies; criteria forbeing invited to compete for the “100 Best” listincluded minimum existence (ten years) and em-ployment (500 people) levels, as well as simplybeing considered by the authors to be a viable can-didate. Of the 1998 invitees, 206 companies agreedto complete the application process (Branch, 1999),in 1999, 236 did so (Levering & Moskowitz, 2000).Hence, we began with a sample of 442 company-years. Missing data limited the sample for our mul-tivariate turnover rate analyses to 106 companies in1998 and 161 companies in 1999, for a total sampleof 267 company-years.

The organization-level data used here comefrom an extensive company-level questionnaire de-signed by Hewitt Associates, a management con-sulting firm. Each company applying to be on theFortune “100 best” list was asked to complete thisquestionnaire on management practices. The sur-vey was typically filled out by a team of the com-pany’s managers and human resource profession-als. In 1998, the mean reported person-hoursdevoted to the survey was 47.3, suggesting consid-erable effort on the part of the respondents. Surveyitems were in reference to United States–basedfull-time exempt (i.e., salaried) and nonexempt(i.e., hourly) employees.

Each company applying to be on the “100 best”list also was required to administer an employeeattitude survey created by the Great Place to Work®

Institute. Although the attitudinal items used inselecting companies for the list were proprietaryand unavailable to us, we did have access to twoother items collected from the applicant compa-nies’ employees. In 1998 and 1999, the applyingcompanies were instructed to randomly survey 250and 275 employees, respectively. Explicit direc-tions for randomly sampling employees and ad-


ministering the attitude survey were provided toeach applicant. Companies distributed the ques-tionnaires with preaddressed and stamped enve-lopes. Employees were directed to return the ques-tionnaires directly to the data processor at the GreatPlace to Work® Institute, thus ensuring anonymityand confidentiality. Companies applying to be onthe “100 best” list were also required to share theprecise sampling mechanism followed in distribut-ing employee questionnaires. The authors of theFortune articles on the “100 best” companies (Le-vering & Moskowitz, 2000) also checked for sam-pling irregularities by comparing survey respon-dent demographic characteristics with eachcompany’s overall demographic characteristics,which were reported in the company-level survey.Response rates for companies averaged 53 percentin 1998 and 52 percent in 1999.

Company and employee surveys were adminis-tered in the summers of 1998 and 1999. For addi-tional detail on the data acquisition, see Fulmer,Gerhart, and Scott (2003).

Measures

Voluntary turnover rate. Company survey re-spondents were asked to provide the number offull-time voluntary leavers during the most recent12 months and their number of full-time employeesfrom 12 months prior to the survey. We divided theformer by the latter to create the voluntary turnoverrate. We then took the natural logarithm of theresult to transform the highly skewed dependentvariable data to a more normal distribution.

Downsizing rate. The company survey asked forthe calendar year of the most recent reduction inforce. We considered an organization to havedownsized if its respondent reported a reduction inforce in the survey’s calendar year or in the preced-ing two calendar years (e.g., a 1999 survey respon-dent with a reported downsizing in 1999, 1998, or1997). Although there is no research on how longdownsizing effects on survivors may persist, weused the two preceding years because a shock thateventually leads to quitting often necessitates aconsiderable period of evaluation and search beforesuitable alternative employment is found. On themore recent end of the time frames, we includedcurrent year (year t) downsizing in the measure forthree reasons. First, news and rumors of downsiz-ing often precede the actual event, potentiallyprompting pre-event voluntary turnover. Consis-tent with this concern is Cascio’s (2002a) conten-tion that employees tend not to be surprised bylayoffs and the finding of Swaen, Bultmann, Kant,and Van Amelsvoort (2004), in which psychologi-

cal distress significantly increases both one monthafter a downsizing announcement and before theevent itself. Second, including year t downsizingwas necessary to ensure that we captured the im-mediate (within-year) effects on voluntary turn-over. This was a particular concern given that“turnover intentions” appear to peak, relative topredownsizing levels, when measured closer to theevent itself (Allen, Freeman, Russell, Reizenstein,& Rentz, 2001). Third, simultaneity concerns asso-ciated with same-year downsizing and turnovermeasures are minimal, as companies presumablydo not downsize in response to increased turnover.

The company survey also asked what percentageof employees was downsized.2 The response,which we first converted to a proportion, allowedus to model the severity of the downsizing, underthe assumption that shock is greater followinglarger downsizings. We then logged the highly(positively) skewed downsizing rates (after addinga constant of .0001 because the natural logarithm ofzero is undefined; see Cohen, Cohen, West, andAiken [2003]). Logging an independent variable en-ables examination of underlying nonlinearity inwhich effects are considerably larger at the lowestlevels of the raw values of that variable (Cohen etal., 2003). This process allowed us to infer whetherthe transition from nondownsizer to downsizer(i.e., the event) carried shock effects in addition tothose carried by downsizing severity (we explorethis distinction further in the Results and Discus-sion sections).

All company respondents reporting a downsiz-ing also reported implications for “terminated em-ployees in the most recent downsizing,” indicatingthat all reported downsizings involved termina-tions. Downsizings were reported in 102 cases (38percent of the sample), with mean downsized pro-portions of .014 overall and .036 among thedownsizers.

HR practices as moderators. We used 12 mea-sures to create three theoretically driven additiveindexes of HR practices. Working at the index,rather than practice, level was consistent not onlywith our theory, but also with considerable prior

2 Unfortunately, the item wording differed in the twoyears, as the 1998 survey asked, “What percentage ofemployees were dismissed due to the most recent down-sizing?” and the 1999 survey asked, “What percentage ofemployees left due to the most recent downsizing?”Analysis of companies that were present in both years,however, indicates that the two items were answeredfrom similar perspectives, as the correlation between re-sponses to the 1998 item and the 1999 item (when refer-ring to the same downsizing event) was .90.


research. Batt (2002) and Delery (1998), for exam-ple, described theoretically driven additive indexesin which higher scores meant greater investment ina range of applicable practices, and similar resultsemerged as a function of practices substituting forone another in an index. An index of this type iscomposed of multiple elements whose values arenot caused by an underlying construct but, whensummed, yield a value whose magnitude is rele-vant to the construct (Batt & Valcour, 2003). Suchindexes are the preferred method for creating asingle measure from a series of underlying dimen-sions that can be cumulated to determine the levelof a construct of interest (Shaw et al., 2005).

Six of the 12 HR practices used to create theindexes were assessed separately for exempt andnonexempt employees. Because organizations man-age exempt and nonexempt employees differently(Gerhart & Trevor, 1996; Lepak & Snell, 1999), thisseparation allowed for greater precision in identify-ing the extent to which a practice covered the overallworkforce. For example, since we knew the exempt/nonexempt proportions for each organization, a com-pany with 70 percent nonexempt employees that of-fered a certain practice only for exempts would becoded .30 (rather than a less representative 1). Thus,our indexes reveal the presence and, to some extent,the coverage of HR practices (though not the intensityof employee use).

We created a procedural justice index to repre-sent an organization’s commitment to providingemployees with a formal process for addressingperceived injustices. Company survey respondentswere asked about the presence of “an ombudsmanwho is designated to address any employee com-plaints,” “a confidential hotline or other confiden-tial problem/resolution/complaint center,” and “agrievance or appeal process available to nonunionemployees.” Items were coded 1 for “yes” and thevalues were then summed. The three items evolvedirectly from the formal procedures section ofMoorman’s (1991) procedural justice scale, whichassesses the presence of procedures that “provideopportunities to appeal or challenge,” “have allsides affected by the decision represented,” and“hear the concerns of all those affected.” Moor-man’s formal procedures items originated in partfrom Leventhal’s (1980) procedural justice rule ofcorrectability (i.e., correcting past wrongs via griev-ance or appeals), which is consistent with ouritems and is frequently the basis for proceduraljustice measures in management research (e.g.,Tepper & Taylor, 2003; Thau, Aquino, & Wittek,2007).

Two of the three major dimensions of job embed-dedness are fit (i.e., compatibility with one’s com-

pany and off-the-job environment) and sacrifice(i.e., the cost of what would be given up by leaving)(Mitchell, Holtom, & Lee, 2001). Thus, we created ajob embeddedness index of five HR practices thatshould yield perceptions along these dimensions(we did not have reliable HR practice measures oflinks to coworkers, the third dimension of job em-beddedness). Although the HR practices have neverbeen tested as predictors of job embeddedness per-ceptions, Mitchell and his colleagues (Mitchell,Holtom, & Lee, 2001; Mitchell, Holtom, Lee, Sa-blynski, & Erez, 2001) explicitly emphasized cer-tain practices as fostering job embeddedness. Assubstantial sacrifices associated with leaving a job,these authors cited defined benefits plans and sab-baticals (which become increasingly valuableand/or available with increasing employee longev-ity), and on-site childcare, which is related to crit-ical off-the-job concerns. Regarding practices thatpromote fit, hiring for organizational fit has beendescribed as increasing employee compatibilitywith the culture and values of an organization(Mitchell, Holtom, & Lee, 2001), and flextime hasbeen described as important to fit with off-the-jobconsiderations (Mitchell, Holtom, Lee, Sablynski, &Erez, 2001). These practices also parallel itemsused to assess job embeddedness perceptions (e.g.,“I fit with the company’s culture” and “The retire-ment benefits provided by this organization areexcellent”) from Mitchell, Holtom, Lee, Sablynski,and Erez (2001). To assess the job embeddednesspractices, the company survey inquired about thepresence of a defined benefits plan, paid sabbati-cals, on-site childcare, and “flexible or nonstand-ard arrival and departure times,” and the impor-tance of organizational fit in hiring decisions.Respondents ranked the importance of fit on a 1–14scale on which 1 equaled “the most important hir-ing criterion”; we reversed these fit scores by sub-tracting from 15. Because the paid sabbatical andhiring-for-fit items separately addressed the ex-empt and nonexempt populations, we weighted theresponses by exempt and nonexempt proportionsbefore combining. We then standardized and addedthe scores for the five practices to create the jobembeddedness index.

Similarly, we created a career development in-dex tapping four practices likely to improve abilityto compete in the external job market via enhancedsignaling or search capabilities. The presence of acareer resource center demonstrates a company’swillingness to commit resources to employee de-velopment and should provide employees withtools and strategies for enhancing their careers.Companies were also asked about the presence of“assessment of skills, knowledge, and abilities to


identify employees’ development opportunities”;the existence of this practice should make employ-ees more aware of their marketable characteristicsand possible career paths. Together, these two prac-tices (career resource center and employee assess-ment) appear likely to enhance search efficiency.The presence of formal succession planning ofteninvolves a variety of developmental activities, suchas education, rotation through various jobs and lo-cations, and special (visible) assignments (Noe,Hollenbeck, Gerhart, & Wright, 2003). Such activi-ties are likely to leave employees with greater sig-naling capabilities. Although succession planningis often confined to the management and profes-sional levels, over 30 percent of our companiesreported using it for nonexempt populations. Fi-nally, the presence of a formal policy enabling ei-ther voluntary or required transfers “to supportlearning (e.g., international assignments)” shouldbroaden employee experience and improve signal-ing on the job market (we used an affirmative re-sponse to either the voluntary or required formaltransfer items as a single indicator). All four prac-tices were asked about for both exempt and nonex-empt populations, enabling us to weight the re-sponses by their respective proportions beforecombining. We then standardized and added thefour scores to create the index.3

Because the substitutable elements in theoreti-cally driven additive indexes are not equivalentmeasures of an underlying construct, it is inappro-priate to use an internal consistency reliabilitymeasure (e.g., coefficient alpha) to test the level ofagreement among the elements (Batt & Valcour,2003; Delery, 1998; Shaw et al., 2005). Test-retestcorrelations, however, yielded reasonable evidencefor the stability of the three indexes. For the 67companies in our analyses that provided surveys inboth 1998 and 1999, we found year-to-year corre-lations of .78 for the procedural justice index, .73for the job embeddedness index, and .75 for thecareer development index.

Organizational commitment. The 1998 and1999 employee attitude data from each companyincluded two slightly modifed items from Mowday,Steers, and Porter’s (1979) Organizational Commit-ment Questionnaire: “It would take a lot to get me

to leave this organization” and “I am glad I chosethis organization over others I was consideringwhen I joined.” These two items were scored on ascale anchored by “almost always untrue” (1) and“almost always true” (5). Coefficient alpha for allemployee responses to these two questions was .84.We created the individual-level commitment scoreby summing the two item scores and dividingby two.

To determine whether aggregating these commit-ment scores to the organization level was appropri-ate, we followed the approach of recent studiesusing aggregated employee attitudes (e.g., Currall,Towler, Judge, & Kohn, 2005; Fulmer et al., 2003;Whitener, 2001), including organizational commit-ment (e.g., Simons & Roberson, 2003). We first cal-culated the intraclass correlations (Shrout & Fleiss,1979) of our commitment variable to estimate thereliability of a single employee response, obtainingan ICC(1, 1) of .056 in 1998 and .064 in 1999.Because our focus was on the company averagescore as a reliable differentiator among firms, how-ever, the ICC(1, k) was more relevant (Fulmer et al.,2003; Glick, 1985). Bliese (1998) demonstrated thateven low ICC(1, 1) levels may lead to reliable ag-gregate means with sufficiently large groups. With anaverage of 137 responses per company in 1998 (min.� 41; max. � 754) and 147 responses per companyin 1999 (min. � 62; max. � 801), our ICC(1, k) was.89 in 1998 and .91 in 1999. These values indicatethat the mean level of commitment was indeed areliable differentiator of our companies, thus pro-viding a foundation for averaging individual com-mitment scores within companies to create organi-zation-level commitment.

Additional control variables. By controlling forthe use of early retirement in the downsizing pro-cess, we increased the probability that we wereseeing layoff effects in our downsizing measure.We included a dummy variable coded 1 if a com-pany indicated that an early retirement packagewas made available in association with its down-sizing event. Substantial organizational change, in-dependent of downsizing, may alter the relation-ship between employer and employees and may berelated to downsizing. We controlled for this po-tential bias by including restructuring (companyrespondents were asked, “Within the last 12months, has your organization experienced anymergers, acquisitions, divestitures, or spin-offs thataffected at least 5% of your employee popula-tion?”). Because labor costs often drive downsizingdecisions (Cascio, 1993), and because pay is oftenrelated to turnover, we accounted for pay level,measured as the response on a four-point scaleindicating whether the company targeted compen-

3 Although the company surveys included an item onhours devoted to training, there was no indication as tothe extent to which training was firm-specific (and likelyenhanced current employment utility) or general (andlikely enhanced alternative employment utility via sig-naling). Thus, we did not include the item, as its rele-vance to our theoretically driven index was unknown.


sation below market level, at market level, betweenthe 51st and 74th percentiles, or between the 75thand 100th percentiles. Also, given possible associ-ations with downsizing and turnover, we con-trolled for organization age, proportion male (em-ployees), proportion exempt, proportion union,proportion aged 25 years or less, proportion withless than 2 years service, number of employees(logged), and the proportion of change in employ-ees from year t to year t – 1. Industry health mayaffect both downsizing likelihood and turnoverrates. Thus, we also included dummy variables toaccount for the effects of the 48 industries in ourdata at the two-digit SIC code level. Finally, tomore precisely account for industry-specific labordemand, we controlled for two-digit industry un-employment rate data obtained from the Bureau ofLabor Statistics (http://www.bls.gov/).

Data Analysis

Researchers modeling turnover rates have usedboth ordinary least squares (OLS) regression anal-ysis (e.g., Shaw et al., 1998) and Tobit regressionanalysis (e.g., Batt et al., 2002). Tobit models wereoriginally derived because a concentration of obser-vations at some limiting lower or upper value vio-lates OLS assumptions. Absent such observations,however, OLS is appropriate (Tobin, 1958). Be-cause we had no turnover rates equal to zero, we

present OLS analyses. Unreported Tobit models,with censoring stipulated for very low rates,yielded highly similar results. Additionally, be-cause 67 companies provided company-year obser-vations for both 1998 and 1999, we used robustvariance estimators with the standard regressionmodel. As an extension to variance estimators cre-ated to account for heteroskedasticity (e.g., White,1980), these also provide standard error estimatesthat require only that observations be independentacross, but not necessarily within, the source of thedependence (Rogers, 1993).

RESULTS

Main and Mediated Downsizing Effects

Table 1 presents means, standard deviations, andcorrelations. The essentially zero correlation betweendownsizing rate and voluntary turnover rate is nei-ther surprising nor indicative of lack of support forour focal relationship. The covariation among the twovariables and several of our controls illustrates thepotential for a zero-order correlation to inflate or de-flate the more valid relationship estimate from a cor-rectly specified multivariate model.

Table 2 reports the results of the multivariate tests.In support of Hypothesis 1, model 1 reveals thatdownsizing rate has a statistically significant effect onvoluntary turnover rates. Specifically, a one-unit in-

TABLE 1Correlations and Descriptive Statisticsa

Variable Mean s.d. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

1. Voluntary turnover rate �2.15 0.832. Downsizing rate �7.24 2.66 �.013. Procedural justice index 1.30 0.94 �.07 �.034. Job embeddedness index �0.01 0.50 �.13 .08 .155. Career development

index0.03 0.65 �.03 .07 .20 .16

6. Commitment 1.69 0.10 �.21 �.27 �.16 .01 .137. Early retirement 0.12 0.32 �.13 .49 .00 .15 .01 �.128. Restructuring 0.26 0.44 .06 .17 .01 .03 .01 �.03 .059. Organization age 56.52 41.22 �.27 .00 .20 .19 .09 .02 .24 �.10

10. Pay level 2.65 0.72 .02 �.07 .00 .02 .11 .22 �.10 �.01 �.0811. Unemployment rate 3.64 1.26 .34 �.15 �.11 �.18 �.18 �.14 �.13 �.03 �.18 .0712. Employees 8.34 1.32 �.17 .14 .32 �.04 .23 �.20 .13 .05 .24 �.08 �.0413. Change in employees 0.11 0.22 .29 �.13 �.11 �.03 .03 .15 �.17 .11 �.32 .11 �.01 �.2314. Proportion with less than

2 years service0.35 0.17 .72 �.18 �.09 �.14 �.03 �.01 �.25 .05 �.41 .04 .31 �.26 .53

15. Proportion age 25 yearsor less

0.13 0.12 .55 �.17 .01 �.20 �.11 �.22 �.17 �.07 �.16 �.09 .45 .03 .18 .62

16. Proportion male 0.54 0.17 �.30 .13 �.04 �.05 .05 .01 .08 .06 �.10 .04 .03 .14 �.08 �.18 �.1717. Proportion exempt 0.49 0.23 .01 .05 �.21 .03 .15 .22 �.04 .08 �.19 .16 �.18 �.20 .08 .15 �.20 .1918. Proportion union 0.06 0.16 �.38 .09 .07 .08 �.05 .07 .15 �.04 .28 .03 �.01 .16 �.17 �.34 �.24 .20 �.28

an � 267. Voluntary turnover rate, downsizing rate, and employees are natural logarithms. Correlations with absolute values above .11are statistically significant at p � .05.


crease in logged downsizing rate predicts a .058 in-crease in logged turnover rate (coefficients in such“log-log” modeling are often referred to in economicsas “elasticities” and are interpreted as the percentchange in the raw dependent variable predicted by a1 percent change in the raw independent variable).Marginal effects translated back into the original met-rics of both variables yield more intuitive interpreta-tion: the predicted raw turnover rate increases 25percent (from .104 to .130) when going from non-downsizing status to a downsizing rate of .005 (the21st percentile of downsizing rate for companies thatdid downsize). In comparison, companies that down-size at rates of .02 (50th percentile of downsizers), .05(75th percentile of downsizers), and .10 (93rd percen-tile of downsizers) are predicted to have raw turnover

rate levels and increases (relative to nondownsizingstatus) of .141 and 36 percent, .149 and 43 percent,and .155 and 49 percent, respectively. Put anotherway, over half of the raw turnover rate increase pro-jected from reducing workforce by 10 percent occursas a result of cutting the first one-half of 1 percent ofemployees. Figure 2 shows this relationship in thegraph on the top left.

As indicated in the figure, and in keeping withCohen et al.’s (2003) discussion of log transforma-tions, our logging of downsizing rate allowed us totransform into linearity an underlying nonlinearrelationship in which the effects on raw (or logged)voluntary turnover rates are considerably larger atlow levels of raw downsizing rates. Because thisnonlinearity is an important empirical and concep-

TABLE 2Main and Mediated Effects of Downsizinga

VariablesTurnover,Model 1

Commitment,Model 2

Turnover,Model 3

Turnover,Model 4

Turnover,Model 5

Intercept �2.07*** (.55) 1.56*** (.10) �0.31 (.79) �1.59** (.58) 0.06 (.72)Two-digit industry Yes Yes Yes Yes YesOrganization age 0.00 (.00) 0.00 (.00) 0.00 (.00) 0.00 (.00) 0.00 (.00)Unemployment rate �0.08 (.10) 0.02 (.02) �0.05 (.11) �0.08 (.10) �0.06 (.11)Employees �0.02 (.03) �0.01 (.01) �0.03 (.03) �0.03 (.03) �0.04 (.03)Employees change �0.01 (.21) 0.05 (.04) 0.04 (.19) 0.01 (.20) 0.06 (.19)Restructuring 0.11 (.08) 0.00 (.01) 0.11 (.08) 0.07 (.07) 0.08 (.07)Proportion with � 2 years service 2.20*** (.49) 0.02 (.08) 2.22*** (.47) 2.21*** (.48) 2.23*** (.47)Proportion age 25 years or less 1.00 (.71) �0.18 (.10) 0.80 (.72) 0.77 (.66) 0.58 (.67)Proportion male �0.50 (.53) �0.01 (.07) �0.52 (.53) �0.47 (.49) �0.49 (.49)Proportion exempt �0.14 (.28) 0.06 (.04) �0.08 (.27) �0.19 (.26) �0.13 (.24)Proportion union �0.75* (.34) 0.13** (.05) �0.61 (.34) �0.71* (.33) �0.57 (.33)Pay level 0.04 (.05) 0.02 (.01) 0.06 (.05) 0.02 (.05) 0.04 (.05)Early retirement 0.03 (.13) 0.03 (.02) 0.06 (.13) 0.05 (.11) 0.09 (.11)

HR practice indexesProcedural justice 0.06 (.05) �0.01 (.01) 0.05 (.05) �0.18* (.09) �0.17 (.09)Job embeddedness �0.05 (.07) 0.00 (.02) �0.05 (.06) �0.55*** (.15) �0.56*** (.14)Career development 0.00 (.06) 0.02* (.01) 0.03 (.06) 0.39* (.15) 0.43** (.16)

Commitment �1.127** (.425) �1.083** (.396)

Downsizing rate 0.058*** (.015) �0.013*** (.004) 0.044** (.016) 0.103*** (.022) 0.084*** (.023)

InteractionsDownsizing rate � procedural justice �0.035** (.013) �0.032** (.013)Downsizing rate � job

embeddedness�0.069*** (.021) �0.070*** (.020)

Downsizing rate � careerdevelopment

0.055** (.019) 0.057** (.020)

R2 .74 .47 .75 .77 .78Adjusted R2 .66 .30 .68 .69 .70Change in downsizing effect �25%Observations 267 267 267 267 267

a Robust standard errors are in parentheses. One-tailed tests for hypothesized effects; two-tailed tests for all other effects (including HRpractice indexes).

* p � .05** p � .01

*** p � .001


tual consideration that has yet to be explored in thedownsizing literature, we conducted an additionalanalysis (available from the authors) to shed furtherlight on why the effects appeared to be so muchstronger at the point of transition from nondown-sizer to downsizer. We reran models 1 and 3 afterreplacing the logged downsizing rate with both adownsizing dummy and the raw downsizing rate,thus capturing downsizing events and their sever-ity in separate variables in the same model. Bothvariables had positive, statistically significant ef-fects, further suggesting that both a downsizingevent and its severity uniquely contribute to turn-over via shock. We return to this issue in the Dis-cussion section.

In Hypothesis 2, we predict that aggregated or-ganizational commitment will mediate the effect ofdownsizing on turnover. Model 2 illustrates thesignificant relationship between logged downsiz-ing rate and commitment. In model 3, in which

organizational commitment is added to model 1 asa predictor, commitment is a statistically signifi-cant predictor of logged turnover rates, and thedownsizing coefficient is reduced by 25 percent,thereby exhibiting partial mediation. More for-mally, Sobel’s (1982) test revealed statistical sup-port for mediation (p � .05).

Downsizing by HR Practice Index Interactions

Models 4 and 5 in Table 2 provide results sup-porting the downsizing by procedural justice indexinteraction predicted in Hypothesis 3 (the twomodels are differentiated by the presence of com-mitment). In model 4, for example, evaluating thedownsizing effect at one standard deviation aboveand below the moderator mean indicates that therelationship between logged downsizing rates andlogged voluntary turnover rates is about 3.5 timesgreater when the procedural justice index is low

FIGURE 2Main and Moderating Downsizing Rate Effectsa

a Low and high x-axis levels of logged downsizing rates represent nondownsizers and the median downsizing rate among companiesthat did downsize, respectively.


than it is when the index high. The graph on thebottom left of Figure 2 illustrates this relationship.In terms of marginal effects, downsizing .02 of aworkforce, which was the median rate for compa-nies that downsized, predicts a raw turnover rateincrease of 63 percent when the procedural justiceindex is low, but an increase of only 15 percentwhen it is high.4

Models 4 and 5 also support Hypothesis 4. Inmodel 4, the relationship between logged downsiz-ing rates and logged voluntary turnover rates isalmost four times greater when the job embedded-ness index is low than when it is high, as the graphon the top right of Figure 2 shows. Marginal effectsreveal that downsizing .02 of a workforce predicts araw turnover rate increase of 65 percent when thejob embeddedness index is low, but only an in-crease of 14 percent when it is high.

Hypothesis 5 was also supported in both models.The relationship between logged downsizing ratesand logged voluntary turnover rates in model 4 isalmost four times greater when the career develop-ment index is high than when it is low; the bottomright graph in Figure 2 illustrates this relationship.Marginal effects reveal that downsizing .02 of aworkforce predicts a raw turnover rate increase of13 percent when the career development index islow, but an increase of 66 percent when it is high.

Alternative Analyses

We conducted four additional variants of ourprimary analyses to check the robustness of ourfindings. First, although our robust standard errorsapproach only requires independence between, butnot within, companies, we also conducted analysesto ensure that the 67 companies that were presentin both 1998 and 1999 were not driving our results.Dropping one of the observations for each of these67 companies (e.g., dropping the 1999 observation)yielded no meaningful changes, indicating that de-pendent observations were not a major concern.Second, we reran our analyses after reconfiguringthe downsizing variable so as to exclude year t,

thus leaving only years t � 1 and t � 2 downsizingto predict year t turnover. Support was weakened,indicating that excluding shared variation betweensame-year downsizing and turnover is problematicin that much of the turnover driven by downsizinglikely occurs reasonably close to the event (see thedownsizing measure description for our rationalefor including year t downsizing). Third, we reex-amined the interaction of downsizing and the ca-reer development index after using only the trans-fer policy and succession planning items tocomprise the index. The dropped items (career re-source center and employee assessment) were in-tended to reflect practices that would enhance em-ployee search capabilities, but we wanted to checkthat our career development index was not insteadcapturing downsizers’ attempts to help downsizingvictims and other potential leavers find externalemployment. The streamlined measure replicatedthe original interaction finding, alleviating con-cerns that we were misinterpreting downsizer en-couragement of turnover as search-and-signal-driven turnover. Finally, although our focus was onthe whole downsizing effect, regardless of whetherthe event was motivated by financial reasons, in-cluding controls for return on assets and totalshareholder return did not materially change theresults.

DISCUSSION

In sum, we found a positive relationship betweendownsizing rates and voluntary turnover rates, me-diation of that relationship by aggregated levels oforganizational commitment, and moderation of thedownsizing effect by procedural justice, job embed-dedness, and career development indexes. We nextaddress these findings in terms of modeling thedownsizing construct, implications for theory andpractice, limitations, and future research.

Modeling the Downsizing Construct

Although a larger downsizing event might bethought likely to generate proportionally greatershock than a small one, the unfolding model’s char-acterization of shock as a very distinguishableevent (Lee & Mitchell, 1994) suggests that thedownsizing event in and of itself may be shocking.That is, comparing the utilities of current and al-ternative employment may be partially initiatedsimply by the presence of downsizing, somewhatirrespective of its severity. Our use of logged down-sizing rate allowed us to model a linear relation-ship that, when converted back into raw originalmetrics, revealed a nonlinear relationship that sup-

4 A counterintuitive aspect of this interaction meritsmention. In the absence of downsizing, turnover rateswere higher when the procedural justice index was high(see Figure 2, bottom left). A potential explanation is thatthe practices in this index deal with the correction ofinjustices (e.g., via grievance or appeals). As such, moreof these practices may be present when there are morewrongs in need of righting. Such an environment maycontribute to voluntary turnover, especially in the ab-sence of the focus on fair procedures that a downsizingshock should bring.


ports this interpretation. Moreover, our supple-mentary analyses (available upon request) with adownsizing dummy and the raw downsizing rateprovide further insight. Because the downsizingevent and its severity each contribute unique vari-ance to voluntary turnover rates, shock appears tomanifest from the mere transition from nondown-sizer to downsizer, as well as from workforce re-duction magnitude. The effect size, which indi-cates that the transition alone predicts a 19 percentincrease in turnover rates, highlights the impor-tance of the transition to downsizer status. Thus,researchers should carefully consider potentialnonlinear, event, and severity considerations whenconducting downsizing studies. Similarly, unfold-ing model research would benefit from work thatrecognizes the potentially unique shocks of anevent and its severity.

Implications for Theory

Our findings are consistent with traditional ap-proaches to turnover that emphasize the compari-son between current and alternative employmentutility that employees make when deciding to quitor stay (e.g., March & Simon, 1958; Mobley, 1977).The results also support the general logic of theunfolding model of turnover (Lee et al., 1996), inwhich shocks such as downsizing are argued toprompt employees to engage in this comparison. Assuch, we provide evidence that downsizing, andperhaps shocks in general, addresses the same lackof explained turnover variance that motivated theunfolding model’s derivation (see Lee, Gerhart,Weller, & Trevor [in press] for a related perspectivethat focuses on unsolicited job offers as shocksprecipitating turnover).

We confirmed earlier speculation that downsiz-ing should precipitate voluntary turnover throughreductions in commitment (e.g., Batt et al., 2002;Mishra & Spreitzer, 1998), yet the mediation issueremains complex. The partial (25%) mediation as-pect is consistent with the unfolding model princi-ple that shocks lead to turnover via attitudes undercertain conditions but lead to turnover irrespectiveof attitudes under others. In addition to reflectingunfolding model paths that do not involve atti-tudes, the nonmediated downsizing associationwith voluntary turnover could indicate unmea-sured downsizer tendencies to actively encourageleaving behavior. Interestingly, we do not knowexactly where in our Figure 1 model, relative to theinteractions, the commitment mediation occurs.One possibility is that commitment mediates themoderated effects (i.e., the mediation occurs to theright of the points of moderation in the model).

We would interpret this as HR practices moder-ating the relationship between downsizing andthe commitment that ultimately predicts turnover.Replacing the turnover-dependent variable withcommitment and rerunning the basic moderatedanalysis, however, reveals little evidence of this, asthe interactions do not emerge. Moreover, commit-ment mediation could occur at more than one placein the model, suggesting that detecting its exactlocation(s) may require finer-grained individual-level data. The topic is a fruitful area for futurestudy.

Regardless of exactly where they fall relative tocommitment, our interactions provide nuancedsupport for the unfolding model’s general logic.That is, certain sets of HR practices reduced down-sizing shocks’ effects on turnover, and othersstrengthened it. Presumably, job embeddednessand procedural justice practices mitigate the extentto which downsizing reduces current employmentutility. Similarly, we assume that career develop-ment practices magnify the extent to which down-sizing enhances the utility of alternative employers(see Figure 1). These interactions support the or-ganization-level extensions of several individual-level models that we invoked to illuminate theunderlying processes through which the modera-tion might occur (e.g., Brockner & Wiesenfeld,1996; Carsten & Spector, 1987; Gerhart, 1990; Lee etal., 1996; Trevor, 2001; Trevor & Wazeter, 2006).

Although we used some of these individual-levelframeworks to provide distinct rationales for mod-eration by the procedural justice index and jobembeddedness index, at a very broad level it is alsoreasonable to infer a single explanation. The buff-ering of downsizing’s turnover effects appears tooccur at high levels of HR practices that are gener-ally considered to reflect well on current employ-ment, as these practices constrain downsizing-driven reductions in current employer utility(albeit in different ways). Research that character-izes HR on a single dimension could be viewed assuggesting, in light of this general explanation, thatthe greater the degree of, for example, high-involve-ment practices (e.g., Guthrie, 2001) in an HR sys-tem, the more protected turnover tendencies willbe from downsizing shocks (see Zatzick and Iver-son [2006] for a somewhat opposing take on pre-dicting productivity). This unidimensional HR ap-proach, however, would not explain the observedmoderation by the career development index,which suggests that downsizing triggers a greaterrole for the search and signaling benefits from ca-reer development practices than for any enhancedcurrent employment utility that these practicesbring. Despite this qualification, formal testing of


unidimensional HR factors in future downsizingand turnover research would be of value.

Implications for Practice

Although downsizing is sometimes an integralpart of a broader restructuring strategy (Cascio,2002a), its allure is its potential to quickly yieldsubstantial savings in employment costs. Thispromise, however, can be compromised, at least inpart, by the considerable costs associated with un-anticipated increases in voluntary turnover. Thesecosts stem from the inefficiencies associated withunderstaffing and from the likely need for subse-quent hiring and training of replacements to returnan organization’s workforce to its target level. In aneffort to avoid the inefficiency and expense of post-downsizing turnover, company decision makersshould be mindful of turnover implications whendeciding how many people to cut. Our results sug-gest that companies can expect a postdownsizingturnover increase, with the size of this increasedependent on the levels of job embeddedness, pro-cedural justice, and career development HR prac-tices. Being cognizant of these relationships pro-vides a means for an improved projection ofpostdownsizing employment levels and, in turn,more accurate determination of how much of thework force needs to be downsized.

An issue potentially more fundamental than howmuch to downsize is whether to downsize. It is wiseto consider turnover implications when making thisdecision. For example, our marginal effects analysispredicts, for an average company, a 31 percent in-crease in postdownsizing voluntary turnover rateseven if just .01 of the workforce is downsized; if anorganization also has low levels of practices that pro-mote procedural justice and job embeddedness per-ceptions, the predicted increase in quitting becomes112 percent. This daunting prediction illustrates howdownsizing strategies in certain contexts may fail,given understaffing inefficiencies and replacementcosts. Hence, our results not only suggest how HRpractices may mitigate downsizing effects on turn-over, but also how HR practices may, via turnoverimplications, hinder downsizing success. If downsiz-ing is inevitable, adopting job embeddedness andprocedural justice practices prior to the event couldpay dividends via reducing downsizing’s effect onturnover. The appropriateness of implementing orterminating HR practices in a downsizing context, ofcourse, should be considered in relation to the esti-mated costs.

Limitations and Future Research

Although our results largely confirmed our hy-potheses, several limitations to the study are im-portant to consider. Certainly, external validity isan important concern. On the one hand, our samplespans multiple industries, both publicly and pri-vately held companies, and companies of a widevariety of sizes. Yet the fact that our entire sampleis made up of relatively admirable, employee-friendly companies (each seeking a place inFortune’s 100 best companies to work for in Amer-ica) raises the question of whether our findingstell us anything about downsizing effects in lessemployee-friendly places. One might speculatethat, relative to such companies, our main effectsare an overestimate; that is, downsizing in ouremployee-friendly organizations might constitute amore severe breach of an employee’s perceivedpsychological contract (Ho, Weingart, & Rousseau,2004), as job security expectations may be muchhigher than in an average company. Alternatively,the potential for our results to underestimate effectsin a broader population appears to be just as plau-sible. In our sample of the “best” places to work,there may have been a layer of employee good willand trust that buffered the downsizing fallout.Hence, the downsizings could have been seen assomewhat justified or unavoidable, with employ-ees reacting less severely than they might havewithout the good will and trust. Similarly, the na-ture of the overall sample may have restricted therange of downsizing, HR practices, commitment,and turnover rates, which would in turn likelyconstrain the effect sizes (and generalizability) ofour relationships. Clearly, research on downsizingeffects on turnover in other company populationswould be a welcome addition to our study.

Second, although our data allowed us to make in-ferences about downsizing and turnover rates, know-ing exactly who was being downsized and who quitwould add considerably to our findings. We wereunable to examine, for example, the reasonable ex-pectation that survivors will react more stronglywhen they are more similar to downsizing victims.Relatedly, knowing the job performance levels ofpostdownsizing leavers would be instructive, as turn-over implications change dramatically according toleaver job performance (e.g., Sturman et al., 2003;Trevor et al., 1997). Indeed, to the extent that leaversare easily replaceable by better-performing and/orless expensive workers, turnover may be desirable(Dalton, Todor, & Krackhardt, 1982; Hollenbeck &Williams, 1986; Martin & Bartol, 1985), suggesting amore positive interpretation of our downsizing ef-fects. In contrast, any postdownsizing increase in


strong performers’ leaving would be particularly det-rimental to an organization. Given that high-potentialand high-performing employees enjoy greater mobil-ity and are the most likely to quit in response toundesirable circumstances (Trevor, 2001; Trevor etal., 1997), such an increase is likely. Future researchon exactly who leaves following a downsizing, andon how context affects the distributional characteris-tics of this group, would be of great interest to re-searchers and practitioners.

Third, although companies were instructed torandomly select employees for the survey fromwhich we obtained the organizational commitmentmeasures, we have no actual proof that this wasdone. Hence, to the extent that company represen-tatives, influenced by the wish to be chosen for the“100 best” list, selectively chose employees to takethe survey, our mediation results may reflect bias.The safeguards employed by the administrators ofthe list likely discouraged self-serving tactics, yetwe cannot dismiss the possibility. Fourth, the tim-ing of the measures was not ideal. HR practiceswere reported in the summer of year t, but turnoverrate was calculated from the summer of year t – 1 tothe summer of year t. This raises simultaneity ques-tions. The complexity of moderated relationshipsmay reduce these concerns, as the alternative inter-pretation of our findings would be that turnoverrate effects on the HR practice indexes dependedon downsizing in a manner that produced resultsthat correspond to our conceptually grounded hy-potheses. Although this relationship seems un-likely, we cannot rule it out. Another factor thatmay partially alleviate simultaneity concerns isthat, for the 67 companies in our analyses thatprovided surveys in both 1998 and 1999, the re-ported values for the HR practices remained rela-tively constant across years; year-to-year correla-tions averaged .75 for our three HR practiceindexes. Fifth, common method bias is possible inthat the same employee might have provided thedata on number of voluntary leavers, total employ-ees, downsizing occurrence and timing, and HRpractices. The relatively objective nature of thesequestions reduces this concern; so also does thefact that independently collected organizationalcommitment, aggregated to the company level, me-diated the downsizing–turnover rate relationshipin a manner consistent with our rationale.

Finally, although we controlled for several as-pects of downsizing context (e.g., restructuring,early retirement), alternative explanations remainpossible. We were unable to account for whethercompanies were repeat downsizers, which mayhave shed additional light on our findings. We alsocould not control for voluntary buyouts, which

could conceivably have been figured into both thereported downsizing rate and the number of volun-tary leavers. This concern, however, is diminishedto the extent that voluntary buyouts covaried withour early retirement control. Additionally, becausethe total number of employees downsized tends tobe considerably smaller than the number of volun-tary leavers, any double-counting of employees tak-ing voluntary buyouts likely would have little ef-fect on the turnover rate measure.

These concerns notwithstanding, we believe thatour study’s evidence of variation in downsizing,HR practices, turnover rates, and aggregated atti-tudes is noteworthy. Despite their flaws, the dataprovide an unprecedented opportunity to betterunderstand the turnover ramifications of downsiz-ing and the HR context.

Conclusion

Evidence is amassing that turnover rates are neg-atively related to financial performance (e.g.,Glebbeek & Bax, 2004; Kacmar et al., 2006; McElroyet al., 2001; Shaw et al., 2005). Thus, businessstrategies that affect turnover rates should be exam-ined very carefully. In our study, downsizing,which is designed to reduce costs, predicted greaterturnover rates, which increase costs; moreover, cer-tain types of HR practices mitigated the downsiz-ing-turnover relationship, while other HR practicesstrengthened it. Hence, it appears that the HRcontext has considerable impact on whether andthe extent to which the shock of downsizing ulti-mately, via turnover rates, translates toorganizational performance.

Perhaps too easily lost in our analysis is theplight of the downsizing victims. The downsizing-turnover relationship suggests a sad irony in thatemployees are laid off by companies that may sub-sequently find themselves understaffed. Moreover,to the extent that turnover hinders organizationalperformance, the performance of downsizing com-panies may well suffer further through the leavingbehavior that the layoffs generate. The formidablecosts to downsizing victims (and perhaps to thesociety that absorbs them) may be seen as unavoid-able in today’s business environment, but this po-sition is much more palatable when the costs donot appear to be in vain. Hence, though we ac-knowledge the frequent necessity of downsizing,we also emphasize that downsizing wisely is im-perative. Careful attention to HR practices and thedownsizing-turnover relationship, given the impli-cations for costs, understaffing, and organizationalperformance, appears to be one avenue towarddownsizing wisely.


REFERENCES

Allen, T. D., Freeman, D. M., Russell, J. E. A., Reizen-stein, R. C., & Rentz, J. O. 2001. Survivor reactions toorganizational downsizing: Does time ease the pain?Journal of Occupational and Organizational Psy-chology, 74: 145–164.

Armstrong-Stassen, M. 2002. Designated redundant butescaping lay-off: A special group of lay-off survivors.Journal of Occupational and Organizational Psy-chology, 75: 1–13.

Arthur, J. B. 1994. Effects of human resource systems onmanufacturing performance and turnover. Academyof Management Journal, 37: 670–687.

Batt, R. 2002. Managing customer services: Human re-source practices, quit rates, and sales growth. Acad-emy of Management Journal, 45: 587–597.

Batt, R., Colvin, A. J. S., & Keefe, J. 2002. Employee voice,human resource practices, and quit rates: Evidencefrom the telecommunications industry. Industrialand Labor Relations Review, 55: 573–594.

Batt, R., & Valcour, M. 2003. Human resources practicesas predictors of work-family outcomes and employeeturnover. Industrial Relations, 42: 189–220.

Bliese, P. D. 1998. Group size, ICC values, and group-level correlations: A simulation. Organizational Re-search Methods, 1: 355–373.

Branch, S. 1999. The 100 best companies to work for inAmerica. Fortune, 139: 118–130.

Brockner, J., Grover, S., Reed, T., Dewitt, R., & O’Malley,M. 1987. Survivors’ reactions to layoffs: We get bywith a little help for our friends. AdministrativeScience Quarterly, 32: 526–541.

Brockner, J., Grover, S. L., & Blonder, M. D. 1988. Predic-tors of survivors’ job involvement following layoffs:A field study. Journal of Applied Psychology, 73:436–442.

Brockner, J., & Wiesenfeld, B. M. 1996. An integrativeframework for explaining reactions to decisions: In-teractive effects of outcomes and procedures. Psy-chological Bulletin, 120: 189–208.

Cappelli, P. 2000. Examining the incidence of downsiz-ing and its effect on establishment performance. InD. Neumark (Ed.), On the job: Is long-term employ-ment a thing of the past? 463–516. New York: Rus-sell Sage Foundation.

Carsten, J. M., & Spector, P. E. 1987. Unemployment, jobsatisfaction, and employee turnover: A test of theMuchinsky model. Journal of Applied Psychology,72: 374–381.

Cascio, W. F. 1993. Downsizing: What do we know?What have we learned? Academy of ManagementExecutive, 7(1): 95–104.

Cascio, W. F. 2000. Costing human resources: The fi-nancial impact of behavior in organizations (4thed.). Boston: Kent.

Cascio, W. F. 2002a. Responsible restructuring: Cre-ative and profitable alternatives to layoffs. SanFrancisco: Berret-Koehler.

Cascio, W. F. 2002b. Strategies for responsible restruc-turing. Academy of Management Executive, 16(3):80–91.

Chadwick, C., Hunter, L. W., & Walston, S. L. 2004.Effects of downsizing practices on the performanceof hospitals. Strategic Management Journal, 25:405–427.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. 2003.Applied multiple regression/correlation analysisfor the behavioral sciences. Mahwah, NJ: Erlbaum.

Currall, S. C., Towler, A. J., Judge, T. A., & Kohn, L. 2005.Pay satisfaction and organizational outcomes. Per-sonnel Psychology, 58: 613–640.

Dalton, D. R., Todor, W. D., & Krackhardt, D. M. 1982.Turnover overstated: The functional taxonomy.Academy of Management Review, 7: 117–123.

Delery, J. E. 1998. Issues of fit in strategic human re-source management: Implications for research. Hu-man Resource Management Review, 8: 289–309.

Folger, R., & Greenberg, J. 1985. Procedural justice: Aninterpretive analysis of personnel systems. InK. H. G. Ferris (Ed.), Research in personnel andhuman resources management, vol. 3: 141–183.Greenwich, CT: JAI Press.

Fulmer, I. S., Gerhart, B., & Scott, K. S. 2003. Are the 100best better? An empirical investigation of the rela-tionship between being a “great place to work” andfirm performance. Personnel Psychology, 56: 965–993.

Gerhart, B. 1990. Voluntary turnover and alternative jobopportunities. Journal of Applied Psychology, 75:467–476.

Gerhart, B., & Trevor, C. O. 1996. Employment variabilityunder different managerial compensation systems.Academy of Management Journal, 39: 1692–1712.

Glebbeek, A. C., & Bax, E. H. 2004. Is high employeeturnover really harmful? An empirical test usingcompany records. Academy of Management Jour-nal, 47: 277–286.

Glick, W. H. 1985. Conceptualizing and measuring organ-izational and psychological climate: Pitfalls in mul-tilevel research. Academy of Management Review10: 601–616.

Guthrie, J. P. 2001. High-involvement work practices,turnover, and productivity: Evidence from New Zea-land. Academy of Management Journal, 44: 180–190.

Ho, V. T., Weingart, L. R., & Rousseau, D. M. 2004.Responses to broken promises: Does personalitymatter? Journal of Vocational Behavior, 65: 276–293.

Hollenbeck, J. R., & Williams, C. R. 1986. Turnover func-


tionality versus turnover frequency: A note on workattitudes and organizational effectiveness. Journalof Applied Psychology, 71: 606–611.

Holtom, B. C., Mitchell, T. R., Lee, T. W., & Inderrieden,E. J. 2005. Shocks as causes of turnover: What theyare and how organizations can manage them. Hu-man Resource Management, 44: 337–352.

Hom, P. W., Caranikas-Walker, F., & Prussia, G. E. 1992.A Meta-analytical structural equations analysis of amodel of employee turnover. Journal of AppliedPsychology, 77: 890–909.

Huselid, M. 1995. The impact of human resource man-agement practices on turnover, productivity, andcorporate financial performance. Academy of Man-agement Journal, 38: 635–672.

Johnson, A. A. 1995. The business case for work-familyprograms. Journal of Accountancy, 180: 53–57.

Kacmar, K. M., Andrews, M. C., Rooy, D. L. V., Steilberg,R. C., & Cerrone, S. 2006. Sure everyone can bereplaced . . . but at what cost? Turnover as a predic-tor of unit-level performance. Academy of Manage-ment Journal, 49: 133–144.

Knudsen, H. K., Johnson, J. A., Martin, J. K., & Roman,P. M. 2003. Downsizing survival: The experience ofwork and organizational commitment. SociologicalInquiry, 73: 265–283.

Lavelle, L. 2003. After the jobless recovery, a war fortalent. BusinessWeek, 3851: 92–92.

Lee, T. H., Gerhart, B., Weller, I., & Trevor, C. O. In press.Understanding voluntary turnover: Path-specific jobsatisfaction effects and the importance of unsolicitedjob offers. Academy of Management Journal.

Lee, T. W., & Mitchell, T. R. 1994. An alternative ap-proach: The unfolding model of voluntary employeeturnover. Academy of Management Review, 19:51–89.

Lee, T. W., Mitchell, T. R., Sablynski, C. J., Burton, J. P.,& Holtom, B. C. 2004. The effects of job embedded-ness on organizational citizenship, job performance,volitional absences, and voluntary turnover. Acad-emy of Management Journal, 47: 711–722.

Lee, T. W., Mitchell, T. R., Wise, L., & Fireman, S. 1996.An unfolding model of voluntary employee turn-over. Academy of Management Journal, 39: 5–36.

Lepak, D. P., & Snell, S. A. 1999. The human resourcearchitecture: Toward a theory of human capital allo-cation and development. Academy of ManagementReview, 24: 31–48.

Leventhal, G. S. 1980. What should be done with equitytheory? New approaches to the study of fairness insocial relationships. In M. G. K. Gergen & R. Willis(Eds.), Social exchange: 27–55. New York: Plenum.

Levering, R., & Moskowitz, M. 2000. The 100 best com-panies to work for. Fortune, 141: 82–110.

Luthans, B. C., & Sommer, S. M. 1999. The impact of

downsizing on workplace attitudes. Group & Orga-nization Management, 24: 46–70.

Madrick, J. 1995. Corporate surveys can’t find a produc-tivity revolution, either. Challenge, 38(6): 31–34.

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

Martin, D. C., & Bartol, K. M. 1985. Managing turnoverstrategically. Personnel Administrator, 30(11): 63–73.

McElroy, J. C., Morrow, P. C., & Rude, S. N. 2001. Turn-over and organizational performance: A comparativeanalysis of the effects of voluntary, involuntary, andreduction-in-force turnover. Journal of Applied Psy-chology, 86: 1294–1299.

Mishra, A. K., & Spreitzer, G. M. 1998. Explaining howsurvivors respond to downsizing: The role of trust,empowerment, justice, and work redesign. Academyof Management Review, 23: 567–588.

Mitchell, T. R., Holtom, B. C., & Lee, T. W. 2001. How tokeep your best employees: Developing an effectiveretention policy. Academy of Management Execu-tive, 15(4): 96–108.

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 ofManagement Journal, 44: 1102–1121.

Mobley, W. H. 1977. Intermediate linkages in the rela-tionship between job satisfaction and employeeturnover. Journal of Applied Psychology, 62: 237–240.

Moorman, R. H. 1991. Relationship between organiza-tional justice and organizational citizenship behav-iors—Do fairness perceptions influence employeecitizenship. Journal of Applied Psychology, 76:845–855.

Mowday, R. T., Steers, R. M., & Porter, L. W. 1979. Themeasurement of organizational commitment. Jour-nal of Vocational Behavior, 14: 224–247.

Noe, R. A., Hollenbeck, J. R., Gerhart, B., & Wright, P. M.2003. Human resource management: Gaining acompetitive advantage (4th ed.). New York:McGraw-Hill/Irwin.

Robinson, S. L., & Rousseau, D. M. 1994. Violating thepsychological contract: Not the exception but thenorm. Journal of Organizational Behavior, 15:245–259.

Rogers, W. 1993. Regression standard errors in clusteredsamples. Stata Technical Bulletin, 13: 19–23.

Schwab, D. P. 1991. Contextual variables in employeeperformance-turnover relationships. Academy ofManagement Journal, 34: 966–975.

Schwab, D. P., Rynes, S. L., & Aldag, R. J. 1987. Theoriesand research in personnel and human resource man-agement. In K. R. G. Ferris (Ed.), Research in per-


sonnel and human resource management, vol. 5:129–166. Greenwich, CT: JAI Press.

Shaw, J. D., Delery, J. E., Jenkins, G. D., & Gupta, N. 1998.An organization-level analysis of voluntary and in-voluntary turnover. Academy of Management Jour-nal, 41: 511–525.

Shaw, J. D., Gupta, N., & Delery, J. E. 2005. Alternativeconceptualizations of the relationship between vol-untary turnover and organizational performance.Academy of Management Journal, 48: 50–68.

Shrout, P. E., & Fleiss, J. L. 1979. Intraclass correlations:Uses in assessing rater reliability. PsychologicalBulletin, 86: 420–428.

Simons, T., & Roberson, Q. 2003. Why managers shouldcare about fairness: The effects of aggregate justiceperceptions on organizational outcomes. Journal ofApplied Psychology, 88: 432–443.

Sobel, M. E. 1982. Asymptotic confidence intervals forindirect effects in structural equations models. In S.Leinhart (Ed.), Sociological methodology: 290–312.San Francisco: Jossey-Bass.

Solomon, J. 1988. Companies try measuring cost savingsfrom new types of corporate benefits, Wall StreetJournal, December 29: B1–B3.

Spence, M. 1973. Job market signaling. Quarterly Jour-nal of Economics, 87: 355–374.

Sturman, M. C., Trevor, C. O., Boudreau, J. W., & Gerhart,B. 2003. Is it worth it to win the talent war? Evalu-ating the utility of performance pay. Personnel Psy-chology, 56: 997–1035.

Swaen, G. M. H., Bultmann, U., Kant, I., & Van Amels-voort, L. 2004. Effects of job insecurity from a work-place closure threat on fatigue and psychologicaldistress. Journal of Occupational and Environmen-tal Medicine, 46: 443–449.

Tepper, B. J., & Taylor, E. C. 2003. Relationships amongsupervisors’ and subordinates’ procedural justiceperceptions and organizational citizenship behav-iors. Academy of Management Journal, 46: 97–105.

Tett, R. P., & Meyer, J. P. 1993. Job satisfaction, organi-zational commitment, turnover intention, and turn-over: Path analyses based on meta-analytical find-ings. Personnel Psychology, 46: 259–293.

Thau, S., Aquino, K., & Wittek, R. 2007. An extension ofuncertainty management theory to the self: The re-lationship between justice, social comparison orien-

tation, and antisocial work behaviors. Journal ofApplied Psychology, 92: 250–258.

Tobin, J. 1958. Estimation of relationships for limiteddependent variables. Econometrica, 26: 24–36.

Trevor, C. O. 2001. Interactions among actual ease-of-movement determinants and job satisfaction in theprediction of voluntary turnover. Academy of Man-agement Journal, 44: 621–638.

Trevor, C. O., Gerhart, B., & Boudreau, J. W. 1997. Vol-untary turnover and job performance: Curvilinearityand the moderating influences of salary growth andpromotions. Journal of Applied Psychology, 82: 44–61.

Trevor, C. O., & Wazeter, D. 2006. A contingent view ofreactions to objective pay conditions: Interdepen-dence among pay structure characteristics and payrelative to internal and external referents. Journal ofApplied Psychology, 91: 1260–1275.

White, H. 1980. A heteroskedasticity-consistent matrixestimator and a direct test for heteroskedasticity.Econometrica, 48: 817–838.

Whitener, E. M. 2001. Do “high commitment” humanresource practices affect employee commitment? Across-level analysis using hierarchical linear model-ing. Journal of Management, 27: 515–535.

Zatzik, C. D., & Iverson, R. D. 2006. High-involvementmanagement and workforce reduction: Advantage ordisadvantage. Academy of Management Journal,49: 999–1015.

Charlie O. Trevor ([email protected]) is an associateprofessor and the Keller Fellow in the Management andHuman Resources Department at the University of Wis-consin–Madison School of Business. He received hisPh.D. in industrial and labor relations from Cornell Uni-versity. His research interests include employee turn-over, compensation, and staffing.

Anthony J. Nyberg ([email protected]) is a Ph.D. candi-date in the Management and Human Resources Depart-ment at the University of Wisconsin–Madison School ofBusiness. His research focuses on strategic human re-source issues with emphases on compensation and em-ployee turnover.


keeping your headcount when all about you are...

Documents