WHENDOESVOICELEADTOEXIT.pdf

® Academy of Management Journal2013. Vol. 56. No. 2. 525-548.http://dx.doi.org/10.5465/amj.2011.0041

WHEN DOES VOICE LEAD TO EXIT? IT DEPENDS ONLEADERSHIP

ELIZABETH J. McCLEANCornell University

ETHAN R. BURRISUniversity of Texas-Austin

JAMES R. DETERTCornell University

We examine the unit-level relationship between employee voice and exit with multi-source data collected over two time periods in 136 restaurants. We find that threemanagerial characteristics that signal the ability and willingness to engage in change—management team change orientation, manager participation in decision making, andmanager access to organizational resources—moderate the unit-level relationship be-tween voice and exit: Employee voice is positively related to turnover when each ofthese factors is low and negatively related to turnover when each is high. Implicationsfor research on voice, leadership, and turnover are discussed.

When employees experience problems or ob-serve opportunities for improvement at work, dothey engage in improvement-oriented voice—thatis, speak up in ways that challenge the status quo tosomeone with the perceived power to act (Detert &Burris, 2007)? Or do they exit their organization inresponse to these suboptimal situations? Questionssuch as these about employee voice and exit havebeen of interest to organizational scholars at leastsince Hirschman (1970), in his exit-loyalty-voice(ELV) framework, laid the groundwork for exami-nation of these responses to dissatisfaction withsome aspect of an organization's functioning orproduct line. In the subsequent decades, scholarshave attempted to understand why employeesspeak up or exit by treating these behaviors asdiscrete, mutually exclusive choices that each in-dividual employee makes because of his or herdissatisfaction (Rusbult, Farrell, Rogers, & Main-ous, 1988; Rusbuh, Zembrodt, & Gunn, 1982). Be-cause voice can lead to examination of underlyingcauses and cures of employee dissatisfaction, incontrast to exit or silent loyalty, it is seemingly theresponse most likely to contribute directly to organ-izational learning (Withey & Cooper, 1989). It istherefore not surprising that scholars and practitio-ners alike have become increasingly focused onunderstanding the antecedents and outcomes ofdiscretionary, improvement-oriented input by em-ployees (Detert & Biu-ris, 2007; Morrison, 2011; Van

Dyne & LePine, 1998). Simultaneously, under-standing of the dynamics underlying employeettirnover (i.e., exit) continues to develop on thebasis of several decades of focused theory and re-search (Griffeth, Hom, & Gaertner, 2000; Lee &Mitchell, 1994; Maertz & Griffeth, 2004).

As noted first by Barry (1974), by suggesting thatvoice and exit are directly inversely related,Hirschman may have misspecified the model bycombining into one what is actually two distinctchoices for employees: (1) choosing voice or silenceand, (2) choosing to stay or exit their organization.Taking Barry's arguments seriously would implythat, in line with Hirschman's arguments, employ-ees may sometimes speak up and remain in theirorganization irrespective of how much or howquickly things change. Or, contrary to Hirschman'smodel, employees may speak up and subsequentlyexit the organization because of what happens (orfails to) in response to voice. Yet these possibilitiesremain largely unexplored in organizational re-search (Btirris, Detert, & Chiabiu'u, 2008; Morrison,2011), despite the costliness of turnover for organ-izations. Thus, we examine the relationship be,-tween voice and exit to begin to understand theconditions under which employee voice leads tohigher or lower rates of exit.

We further depart from prior research in theHirschman tradition, and some of the logic onwhich it rests, by examining the relationship be-

525

Copyright of the Academy of Management, all rights reserved. Contents may not he 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.

526 Academy of Management Journal April

tween voice and exit at the unit, rather than indi-vidual, level. We do so for two reasons. First, thevery nature of voice as a prosocial behavior in theorganizational literature (Organ, Podsakoff, &MacKenzie, 2006) suggests that the improvementsstemming from the input of any one employeeshould have spillovers that affect a broader workenvironment. That is, voice is a "discretionary ex-pression of change-oriented comments" intendedto "to benefit others, such as the organization" (VanDyne, Ang, & Botero, 2003: 1370-1371; see Detertand Burris [2007] and Morrison and Milliken[2000] for similar definitions). To be prosocial, andthus meet organizational scholars' definition ofvoice, an employee's input about needed improve-ments does not primarily benefit just the one indi-vidual who spe^cs up, but instead has the possibil-ity of bettering the situation for a broader set ofemployees around the speaker (Grant & Mayer,2009). Despite the fact that managers form in-groups and out-groups among their subordinates(Dansereau, Graen, & Haga, 1975; Liden, Sparrowe,& Wayne, 1997) and primarily attend to suggestionsmade bj' only those in their in-group (Biu-ris, Rod-gers, Mannix, Hendron, & Oldroyd, 2009), thechanges that do or do not occur in a unit as a resultof voice are not readily allocated only to favoredmembers. Organizational scholars have long recog-nized this collective aspect of voice in noting itspotential positive outcomes for a unit (versus onlyindividuals who spoke up); these outcomes includelearning (Edmondson, 2003), better error detection(Argyris & Scbön, 1978), innovation (Nemeth,1997), and change effectiveness (Morrison & Mil-liken, 2000). Likewise, many or all of a unit's em-ployees—not just those who spoke up—shouldcontinue to experience dissatisfaction and associ-ated negative outcomes when voice is ignored ordisregarded. Thus, both the nature of employeevoice behavior through its focus on collective im-provement, and decades of organizational theoryand research indicating that employee attitudes,behaviors, and outcomes in a defined unit are so-cially influenced and interdependent (e.g., Ibarra &Andrews, 1993; Salancik & Pfeffer, 1978), suggestthe importance of moving beyond the implied as-sumption in Hirschman's model that each employ-ee's satisfaction, voice, and relationship status withan organization can be considered independentlyand devoid of the larger social context.

Second, voice in itself does not automaticallymake things better or worse for those who speak upor anyone else. Instead, the potential value of voice

rests on what someone with the power to takeaction—which usually means "management"—doeswith the suggestions made. The next steps taken bya manager can result in improvement for many in aunit, in no meaningful change at all, or in a situa-tion actually getting worse in employees' minds.When managers have the ability and motivation totake action based on the prosocial suggestions oftheir employees, voice should improve the moraleand decrease the rate of exit of employees in a unitor an organization, including both those who spokeup and the many others who did not. Likewise,employees of all types should collectively feel thatemployees, as a group, have some control over out-comes (Barry & Shapiro, 2000; Greenberg, 2000), bemore satisfied with outcomes (Shapiro, 1993), andthus be more likely to remain with their organiza-tion because managers consider employees' opin-ions and make subsequent changes. Even if changesare not always made as a result of their own orothers' input, employees should be more likely tochoose to stay and less likely to be fired in envi-ronments where management demonstrates a gen-eral responsiveness to employee input (Tyler, 1987).

But not all managers are able and motivated totake action on the suggestions made by their em-ployees. The level of responsiveness and subse-quent action taken by management to address is-sues raised by employees varies (Detert & Burris,2007), maldng the prospect of voice directly influ-encing employee turnover contingent rather thancertain. When managers are not able or motivatedto address prosocial, improvement-oriented issuesraised, the broader set of employees, in addition tothe speaker(s), continues to experience the condi-tions that prompted voice by some. This, in turn,should lead more members of tbat unit to concludethat it is futile to speak up (Detert & Trevino, 2010;Morrison & Milliken, 2000), to lose faith that man-agers will make necessary improvements (Milliken,Morrison, & Hewlin, 2003), and, thus, to leave thefirm. To regain a sense of perceived control andequity, employees who feel impotent in the face ofnonresponsive, nonimproving conditions may be-gin to give less of themselves to their organization,perform less well (Blader & Tyler, 2009), or evenengage in deviant behaviors (Skarlicki & Folger,1997). As a result, managers may take steps to re-move employees who passively or actively under-mine the attainment of the goals of the managers'units (Giàcalone & Greenberg, 1997; Litsky, Ed-dleston, & Kidder, 2006; PariUa, Hollinger, & Clark,1988). Thus, voice may increase the level of subse-

2013 McGlean, Burris, and Detert 527

quent turnover in a unit if managers are notresponsive.

In this article we therefore examine the unit-levelrelationship between voice and exit among groupsof employees who share a work environment andargue that this relationship is contingent on mana-gerial responsiveness to voice. More specifically,we argue that the relationship between unit-levelvoice and exit depends on three characteristics ofmanagerial responsiveness: whether the managerswho receive it (the targets of voice) have access toorganizational resources to implement change;whether they are able to participate actively inorganizational decision making; and the changeorientation of the management team. These pro-posed moderators reflect whether management isable and motivated to respond to employee sugges-tions for improvement in ways that affect subse-quent turnover.

In what follows, we begin by explaining in detailwhy the proposed unit-level relationship betweenemployee voice and employee turnover is contin-gent on the responsiveness of managers. Then, be-fore testing the three proposed moderating hypoth-eses, we directly examine the premise thatemployees in general, not just speakers or favoredemployees, share improvements stemming frommanagerial responsiveness to voice. We do sothrough an analysis of 3,388 open-ended sugges-tions for change from employees of a national res-taurant chain wherein coders considered whethermanagerial responsiveness to a suggestion would 'benefit primarily the individual who provided theinput or, as argued here, multiple employees in thesame unit. Finally, we test our hypotheses usingmultisource, longitudinal perceptual and objectivedata from 5,200 employees, 372 managers, and 136general managers in the same organization.

Our work makes three primary contributions.First, we demonstrate the clear value of attendingtheoretically and empirically to the inherentlyprosocial, collective nature of voice and its out-comes in work settings. We extend Hirschman's(1972) framework by focusing on the social contextin which employee voice takes place, arguing thatthe benefits of successfully addressing employeevoice extend beyond the focal individual speakingup; voice can affect the rate of turnover for others inthe individual's unit whose performance or well-being are likewise affected by the issues raised.Second, oiu: study extends scholars' understandingof the complex relationship between voice and exitby showing how voice can increase or decrease

employee turnover depending on the degree towhich managers display characteristics reflectingthe motivation and ability to respond. We alsodemonstrate the importance to theory and practiceof considering management behaviors not only asantecedents to voice behavior (Detert & Burris,2007; Edmondson, 1999) but also as moderators ofits effects. Third, our study contributes to the unit-level turnover literatvire by considering how andwhy improvement-oriented voice, and managerialresponsiveness to it, help explain this critical or-ganizational outcome.

VOICE, MANAGERIAL RESPONSIVENESS, ANDEXIT: THEORY AND HYPOTHESES

Many voice scholars have argued that voice isgood for work units and organizations (e.g., Morri-son, 2011). High levels of voice can be indicative ofa learning environment in which employees andmanagers actively engage in activities to continu-ously reduce errors, improve organizational rou-tines, and produce innovations (Argyris & Schön,1978; Edmondson, 2003; Moscovici & Nemeth,1974). Employees in strong learning environmentsshould be more likely to stay (i.e., turnover shouldbe lower) because their situation likely improves asa result of the implementation of improvement-oriented ideas and because they likely feel invigo-rated by membership in an innovative environmentin which their input is regularly considered (Sha-piro, 1993). Voice scholars have also acknowl-edged, though, that not all such "high-voice" envi-ronments are likely to experience positiveoutcomes. Continued managerial inability or un-willingness to respond effectively to voice can re-sult in widely shared futility perceptions amongemployees (Detert & Trevino, 2010; Dutton & Ash-ford, 1993) and a broad climate in which employ-ees feel impotent, not valued, and no longer willingto put forth high levels of effort. In these environ-ments, employees should be more likely to volun-tarily remove themselves from the organization orbe forced to exit as a consequence of their deterio-rated attitude and work performance.

We argue that the responsiveness of unit manag-ers will play an important role in determiningwhen voice will lead to less or more exit in a unit.Añer all, employees speak up because they cannotfix problems or pursue opportunities by them-selves, needing instead someone with more formalpower to address the issues they have identified(Detert & Burris, 2007). Managerial responsiveness

528 Academy of Management Journal April

can he captured hy two dimensions. First, the ahil-ity to respond to voice may enable or impede man-agers (Armenakis & Bedeian, 1999; Kotter, 1996).Such ability may stem from access to organiza-tional resources necessary to investigate, imple-ment, and institutionalize organizational changes(Damanpour, 1991; Kraatz & Zajac, 2001; Singh,1986). Without access to resources, managers willbe less able to experiment with changes and thus belimited to a focus on efficient execution of currentpolicies and practices (March, 1991). And, hecausemost managers are not in a position to unilaterallydecide on changes, responsiveness to voice frombelow also likely hinges on a manager's influenceon decisions made hy more senior leaders (Detert &Trevino, 2010; Glauser, 1984). Second, managers'overall motivation or orientation toward changemay affect how they respond to voice (Beer, 2009).This managerial orientation involves a proactiveidentification of problems and opportunities and aproclivity to take action to address what has beensurfaced (Detert & Burris, 2007; Grant & Ashford,2008; Saunders, Sheppard, Knight, & Rofli, 1992).

We examine below three specific factors that sig-nal managers' ability and motivation to respond tovoice and therefore affect the likelihood of subse-quent turnover in a unit. We begin hy examiningmanagers' perceived access to the organizationalresources needed to respond to employee input as afactor influencing the unit-level voice-exit relation-ship. Next, we examine managers' level of partici-pation in higher-level decision making as a reflec-tion of their ability to enact change in their ownunits. Finally, we examine the change orientationof the management team that must address em-ployee voice. Figure 1 summarizes our theoreti-cal model.

Unit Manager Access toOrganizational Resources

Organizational resources, such as those used fortraining and development programs, marketing, ornew process implementation, are necessary for im-provement and adaptation in organizations (Cohen& Levinthal, 1990; Kraatz & Zajac, 2001; Nohria &Gulati, 1996). These resources, however, are oftenunequally distributed throughout an organization(Pfeffer, 1981), and accessing them is at least par-tially a function of manager proactivity and skill(Graen, Cashman, Ginsburgh, & Schiemann, 1977;Molm, 1990). Managers who are able to acquireresources will be more capable of making changes(Levinthal & March, 1993; March, 1991) becausethey will have the necessary capital, whether hu-man or financial, to devote to improvement activi-ties (Kraatz & Zajac, 2001). Without the slack cre-ated hy additional resources, managers axe less ableto experiment with changes and are confined to afocus on efficient execution of current policies andpractices (March, 1991).

The extent to which managers garner access toresources should affect the voice-exit relationshiphecause resources give managers the ability to acton or at least seriously consider employees' input.When a unit manager can access resources fromelsewhere in her or his organization, the supportthat the manager gets in the form of these resourcestrickles down to lower-level employees (Erdogan &Enders, 2007), who are thus likely to view hoththeir manager and the organization she/he embod-ies as supportive and responsive (Eisenberger et al.,2010). For example, if employees speak up about anissue that is detrimental to their work unit (suchas insufficiently trained new hires performing

FIGURE 1Theoretical Model

Manager Accessto

OrganizationalResources

ManagerParticipation in

Decision Making

Employee Voice(unit level)

ManagementTeam ChangeOrientation

Employee Turnover(unit level)

2013 McClean, Burris, and Detert 529

poorly], managers need resources to invest in moreor better training programs and to keep their unitsappropriately staffed diu-ing the training. When aunit's manager can obtain the necessary resourcesand properly train new hires, others in the unitwho were negatively affected by the performance ofnew hires—not just the one or few who spoke up—should feel better about the unit. Specifically, be-cause the manager was able to obtain resources torespond to voice, employees should feel that theorganization's agents care about tbeir well-beingand experience a heightened sense of organiza-tional support (Shanock & Eisenberger, 2006]. Thedecision to stay in the organization is one wayemployees who benefit from managerial respon-siveness can demonstrate reciprocity for organiza-tional support (Erdogan & Enders, 2007; Rhoades &Eisenberger, 2002].

Conversely, where managers lack access to re-sources needed to address upward input, employ-ees will grow increasingly frustrated by the lack ofresponsiveness and feel that an organization is lesssupportive (Shanock & Eisenberger, 2006]. Becausemanagers are seen as embodiments of an organiza-tion's values (Eisenberger et al., 2010], managerialinability to respond to employees' voice because oflack of resources is likely to be experienced byemployees as organizational disinterest in themand their ideas for improvement (Rhoades & Eisen-berger, 2002]. When employees reach the conclu-sion that they are not being supported by theirorganization, they are more likely to quit (Allen,Shore, & Griffeth, 2003]. Or they may withdrawfrom the organization in more indirect ways(Rhoades & Eisenberger, 2002]. As is well knownfrom equity theory, when employees perceive animbalance between what they give and what theyreceive, they often attempt to restore equity by en-gaging in counterproductive work behaviors, beingabsent, or underperforming (Adams, 1965]. In thisway, employees' legitimate frustration about workconditions cind lack of satisfying managerial re-sponse can lead to inappropriate responses thatultimately get employees fired (Giacalone & Green-berg, 1997; Litsky et al., 2006; Pariila et al., 1988].For example, employees who do not get promisedraises or necessary support through additional hir-ing or training because a manager lacks access toresources may end up being terminated for theshirking, stealing, or other unacceptable behaviorthat they resort to.

Taken together, these arguments suggest thatwhen employees speak up and managers do not

have access to resources to address their ideas andconcerns, employees will develop attitudes and en-gage in behaviors that lead to increased voluntaryand involuntary exit.

Hypothesis 1. A unit manager's access to or-ganizational resources moderates the unit-level relationship between employee voice andexit: The relationship between voice and exit isnegative when the manager's access to organ-izational resources is high and positive whenthe manager's access to organizational re-sources is low.

Unit Manager Participation in Decision Making

The structure of most organizations of even mod-est size is such that most managers are subordinateto other managers—that is, most bosses also havebosses. Thus, most middle managers serve as "link-ing pins" between organizational levels, working tosynchronize operating activities at one level withbroader strategic imperatives at higher levels (Lik-ert, 1967]. From a top-down perspective, this link-ing role may be seen as managers at one level im-plementing the decisions made at higher levels.However, middle- and lower-level managers oftenhave their fingers on the pulse of their organizationin such a way that effective linking also means theyparticipate in decision making by helping higher-level leaders understand what is needed to success-fully implement strategic directives and which di-rectives do not fit the reality faced by lower-levelemployees (Dutton & Jackson, 1987; Nonaka, 1988].

Managers who influence the decision-makingprocess of senior-level managers have the ability toadvocate for issues that employees raise. Havinginput from their ranks considered in higher-leveldecision making should increase employees' senseof procedural justice (Avery & Quiñones, 2002; Thi-baut & Walker, 1975] because it reaffirms that theirproximal manager and those higher in their organ-ization value employees and their ideas (Lind &Tyler, 1988]. For example, if some employees speakup with ideas for improving customer service rou-tines that repeatedly lead to customer complaints,participation in higher-level decision making in-creases the likelihood that a local manager canmake the changes needed to address employees'input. Because employees discuss among them-selves how their managers respond to their inputand recognize that they are collectively affected byany changes that result (Leung & Li, 1990; Nau-

530 Academy of Management Journal April

maim & Bennett, 2000), they often develop a sharedsense of procedural justice (Colquitt, Noe, & Jack-son, 2002). Thus, when a manager is able to suc-cessfully advocate for rank-and-file ideas for im-provement to higher-level leaders, a greater numberof employees within the manager's unit should feelthat decisions made in their work environment arejust and should thus be more likely to stay in theorganization (Masterson, Lewis, Goldman, & I'ay-lor, 2000; Simons & Roberson, 2003).

In contrast, if managers have limited ability tomake changes in their own units because they areunable to influence higher-level decision making,employees will be less likely to feel that the organ-ization considers problems and ideas of people attheir level (Thibaut & Walker, 1975). In this case,employees will .view the organization as less fairand will be more likely to exit (Masterson et al.,2000). For example, if their manager is unable toinfluence higher-level leaders to make improve-ments to a particular food preparation process thatis cumbersome, wasteful, and potentially danger-ous, a restaurant's employees are more likely to feeldisgruntled and disappointed and to subsequentlyleave the organization. When managers cannot par-ticipate in and influence decision making, someemployees (whether they spoke up about an issueor not) may also respond to the perceived unfair-ness of their situation, and the feelings this evokes,by taking action directly against the organization(Folger & Cropanzano, 1998). Counterproductivework behaviors (Fox, Spector, & Miles, 2001), un-derperformance (Bagozzi, 2003), and absenteeism(Hausknecht, Hiller, & Vance, 2008) are all poten-tial responses to the frustration and negative emo-tions employees experience when they feel they arebeing treated unfairly or they are unable to controltheir own environment. When employees engage inthese types of behaviors, they are more likely to getfired (Giacalone & Greenberg, 1997; Litsky et al.,2006; Parilla et al., 1988). Therefore, when employ-ees speak up to the manager of a unit who does notparticipate in higher-level decision making, em-ployees in that unit will collectively be more likelyto leave the organization either voluntarily orinvoluntarily.

Hypothesis 2. A unit manager's participationin decision making moderates the unit-levelrelationship between employee voice and exit:The relationship between voice and exit is neg-ative when the manager's participation in de-cision making is high and positive when the

manager's participation in decision makingis low.

Management Team Change Orientation

Despite popular deification of single leaders,strategic interpretation and decisions to make im-provements are usually not made exclusively byjust one person (Burgelman & Sayles, 1986; Chan-dler, 1962; Drucker, 1974; Mintzberg, 1973). In-stead, management teams, often comprised of lead-ers and their direct reports, share collectiveresponsibility for interpreting strategic issues andcreating change. Specifically, management teamswill be effective at creating change and signaling toemployees that they are willing to make changes tothe extent that members work well together to spotissues, make improvement plans, and commit col-lectively to implementing decisions (Chandler,1962; Hackman, 1987; March & Simon, 1958;Yukl, 2002).

When employees speak up and a managementteam is more change oriented, employees will viewtheir managers as willing to experiment and as notoverly committed to the status quo (Hambrick,Geletkanycz, & Fredrickson, 1993; Kanter, 1983).Given that one of the roles of those in charge is todirect and control change in aii organization (Mint-zberg, 1973) and that such an orientation fits mostpeople's "implicit leader theory" (Epitropaki &Martin, 2004; Sch)ais & Meindl, 2005), employeeslikely expect "good" managers to be motivated toidentify and work together to address opportiuiitiesto improve their organization (Porter & Steers,1973). When this occurs, employees will feel thattheir organization (through its management team)has met its obligations to respond to their ideas forimprovement (Morrison & Robinson, 1997). For ex-ample, if a group of workers is having trouble serv-ing customers quickly enough, the managementteam comprised of shift supervisors and the generalmanager will need to work together in a coordi-nated fashion to identify the underlying issues,make an improvement plan, and successfully im-plement it. If management teams can routinely dothese things, employees should, as a group, be morelikely to feel that managers have fulfilled their ob-ligations to address employees' concerns (Ho,2005), and thus be more likely to stay.

But not all managers are equally open-mindedabout or committed to change (Hambrick et al.,1993). When employees speak up and their localmanagement team is less change oriented, em-

2013 McClean, Burris, and Detert 531

ployee expectations about good management willbe unmet (Mintzberg, 1973; Porter & Steers, 1973).Because local managers are seen as embodiments oftheir larger organization (Eisenberger et al., 2010;Levinson, 1965), employees will likely also con-clude to some extent that that the organization as awhole is not responsive to employees' needs orideas for improvement. Such beliefs can lead toemployees' conclusion that their implicit or psy-chological contract with the organization is beingbreached (Robinson, 1996; Rousseau & McLeanParks, 1993). Employees may respond to this un-satisfactory state by quitting (Rousseau, 1995) or byexpressing their frustration and disappointmentthrough organizational deviance (Bordia, Restubog,& Tang, 2008; Kidwell & Bennett, 1993). Counter-productive work behaviors, such as theft or sabo-tage, often stem from unaddressed dissatisfyingwork conditions (Robinson & Bennett, 1997) andfrom employees' perception that the organizationhas failed to live up to obligations and expectations(Robinson & Rousseau, 1994). Alternatively, em-ployees may engage in less severe, yet still disrup-tive, behaviors such as slapping work (Hausknechtet al., 2008) or withholding effort (Kidwell & Ben-nett, 1993; Turnley, Bolino, Lester, & Bloodgood,2003) in response to a lethargic management re-sponse. Regardless of the specific means, most em-ployee responses to their frustration and disap-pointment are likely to be undesirable frommanagement's perspective and thus increase em-ployees' likelihood of being fired (Giacalone &Greenberg, 1997; Litsky et al., 2006; PariUa et al.,1988). Taken together, the above arguments suggestthat employee voice targeted to less change-oriented management teams will lead to a higherincidence of overall exit by those managementteams' employees.

Hypothesis 3. A unit management team'schange orientation moderates the unit-level re-lationship between employee voice and exit:The relationship between voice and exit is neg-ative when management team change orienta-tion is high and positive when managementteam change orientation is low.

METHODS

We collected both qualitative and quantitativedata from employees of 136 restaurants in fourdivisions of a corporation-owned chain located in21 states throughout the United States. Each restau-

rant employs three types of workers: one generalmanager (GM) per restaurant, restaurant shift su-pervisors (approximately 2-7 per restaurant), andhourly employees (e.g., cooks and servers; approx-imately 20-80 per restaurant).

We obtained 3,388 qualitative responses from(58 percent of) the hourly employees who com-pleted our entire survey in the 136 restaurants. Thecomments reflect written responses to the open-ended question. What is the one thing that wouldimprove this restaurant as a place to work? Wecoded these employee suggestion data to assesswhether, in fact, most employee voice can right-fully be considered prosocial—that is, whether sug-gestions have the potential for broad benefit ratherthan just improvement in the speaker's well-being.If employee suggestions have the potential to ben-efit many employees (and an organization itself)through actions taken by managers, this findingwould lend support to the contention that it isappropriate and important to examine the unit-level outcomes of voice.

We followed a rigorous process for coding theopen-ended comments to determine whether voiceis inherently prosocial—that is, likely to affect mul-tiple employees if leaders take action on the sug-gestions made. First, we used approximately 10percent of the open-ended comments to hone acoding scheme that categorized the comments intothree categories: (1) "speaker only": comments that,if responded to, would only affect the one em-ployee speaking up; (2) "multiple people": com-ments that, if responded to, would affect multipleemployees in the speaker's work unit; (3) "un-clear": comments for which the scope of beneficia-ries could not be clearly determined. To be conser-vative, we also included "split cases," (when anemployee provided multiple ideas for improve-ment, at least one of which could affect multipleemployees and at least one other of which wouldonly affect the speaker) in the third ("unclear")category. Next, the first author and a research as-sistant (not otherwise involved in this research)worked independently to code the remaining 3,028employee comments. Interrater agreement on theclassification of these 3,028 employee improve-ment suggestions was high, as evidenced by aCohen's kappa of .92 (Landis & Koch, 1977). In asubsequent discussion, the coders reached agree-ment on the best classification for each of the smallnumber of comments they had initially codeddifferently.

532 Academy of Management Journal April

The results of this coding overwhelmingly sup-port organizational scholars' definition of voice as aprosocial behavior—that is, as one with the poten-tial to benefit more than just the speaker if actedupon: Approximately 94 percent of employees'comments were coded as affecting the well-being ofmore than just the speaker, with just 3 percentbeing coded as affecting only the speaker and theremaining 3 percent being coded as unclear in re-gard to potential breadth of benefit. Sample com-ments from each of these categories are shown inTable 1. This result lends support to the notion thatan employee's exit can be affected by managerialresponsiveness, or lack thereof, to the input pro-vided by coworkers in the same unit as theemployee.

To test our three unit-level moderator hypothe-ses, we used quantitative survey and objective datafrom the 136 restaurants in our sample. We usedfour different sources to avoid common methodand single soiu:ce problems. We collected our inde-pendent variable (voice) from questionnaires ask-ing the shift supervisors to rate each employee'slevel of voice. We assessed the moderator variables(unit manager access to organizational resources,unit manager participation in decision making, andlocal management team change orientation) usingsurvey responses from GMs. Gontrol variables were

collected from a survey given to all employees andfrom the GM surveys. The sin-vey data come from5,200 employees, 372 managers, and the 136 res-taurant general managers. The average survey re-sponse rate per store was 74 percent, with a rangeof from 24 to 100 percent. Objective turnover datawere provided by the corporate HR department.

Measures

Unit-level voice. We measured employees' voiceexpressed to their supervisors using three items(a = .95) rated by their shift supervisors on a fre-quency scale ranging from 1 ("almost never") to 5("almost always"). In most cases, multiple supervi-sors rated an employee's level of voice (on average,1.85 shift supervisors rated each employee, with amaximum of 7 raters). We adapted items from VanDyne and LePine's (1998) prosocial voice scale.The items (prefaced by the words "This employee")were, "develops and makes recommendations con-cerning the restaurant," "speaks up and encouragesothers to get involved in issues that affect the res-taurant," and "speaks up with ideas for new proj-ects or changes in procedures." We then aggregatedthe ratings for employee voice to the restaurantlevel and calculated aggregation statistics to test forsignificant differences between restaurants. Specif-

TABLElQualitative Data Analysis Coding Examples

Code

Impact broadertJaan inputprovider

Impact onprovider only

Unclear ormixed

Example

"I have been trying for the last 6 months to be cross-trained and the last 3 months to be a trainer andneither has happened. I feel like I am being held back. I would like to be paid more for the work I do.'

"I have worked for the company for 3 years. I make $7.25 an hour and I do EOD, inventory, anddeposits. Why is this? I feel that I am not important to this company and what I do here means squat."

"I would like to have flexible hours so I could have the days off that I want.""This place would be most improved as a place to work if I got off on time."

"There should be two people in the dish room. One to wash and one to put away. I want more hours.""I think the guest count should be determined on those using the daily ACE.""I hate server side work! We should have a kickball team. I hate throwing away good food when

someone could eat it but we have to throw it away.""Seating charts."

Percent

"Not paying your workers that have been here for 3+ years $7-58. Yet new hires sometimes walkthrough the door making $8.50 and $9 and up—so this isn't right. We as a company need to work onpaying our old hires more money for their jobs. Let's make a change."

"Better training for new associates. Our store has recently hired a bunch of new people but without theadequate number of trainers available to train them. The results have been poorly trained anduninformed new employees and less than adequate service."

"The place needs to be better organized. There are too many mistakes, tbe rules aren't followed strictly.We run out of things (such as straws) and people leave without doing their sidework."

"We need better service and faster food preparation. There are days where it takes 20 minutes for all ourfood and drinks to come up. Tbis makes our servers look bad."

94%

3%

3%

2013 McClean, Burris, and Detert 533

ically, we computed ICCl, which reveals the extentto which group membership predicts variability inindividual level responses, and ICC2, which esti-mates the reliability of between-group differencesin a measure. Aggregation statistics indicated thatsignificant differences existed between restaurantsand that the unit-level voice construct was mea-sured reliably (ICCl = 0.11, ICC2 = 0.76, p < .01)(Bliese, 2000).

Unit managerial access to organizational re-sources. We measured managerial access to organ-izational resovirces using six items (a = .93). Be-cause the measurement of resources variesconsiderably across studies (Crook, Ketchen,Comhs, & Todd, 2008), which highlights the con-text specificity of this construct, we developed theitems through extensive interviews with seniorleaders from multiple levels of the studied organi-zation. These leaders identified the specific re-sources that they deemed necessary to make im-provements and changes affecting the restaurants.For instance, restaurant GMs often need access toresources from higher in the organization for stafftraining or to successfully make improvements tooperating policies or procedures in a specific res-taurant. Thus, in conjunction with these seniorleaders, we developed items tapping the specificresources that GMs could use to make materialchanges internal to their restaurants: "The DM pro-vides the necessary support for . . . "training,""marketing," "new process implementation,""growth/development programs," "evaluation andplanning," and "communication to crew mem-bers." The items were rated by GMs on a scaleranging from 1 ("never") to 5 ("always").

Unit manager participation in decision mak-ing. We measured managerial participation in de-cision making using three items (a =.91) based onSiegel and Ruh's (1973) five-item measure. The GMof each restaurant rated his/her participation indecision making at the division level (i.e., the man-agement level above him/her). The items were, "Ihave influence on division managers' (DM) deci-sions affecting the restaurant"; "I participate in DMdecisions ahout my job"; and "I have influence onDM decisions affecting me and my job." GMs ratedthese items on a scale ranging from 1 ("stronglydisagree") to 5 ("strongly agree").

Management team change orientation. We de-veloped the items for management team changeorientation on the basis of the specific behaviorsthat the organization's leaders identified as criticalto creating change at the restaurant level. Before we

created these items, we scoured the literature forappropriate items tapping improvement-orientedmanagement team behaviors. However, we did notfind a suitable measure. Most leadership measurestap specific styles (behaviors) (e.g., relational ortask-oriented leadership [Yukl & Nemeroff, 1979]),personality-driven behaviors (e.g., charismaticleadership theories [Conger & Kanungo, 1998]), orbroad sets of hehavior that encompass but do notdirectly measure specific behaviors that promotechanges (e.g., transformational leadership theories[Bass & Avolio, 1994]). We thus developed a mea-sure to assess specific behaviors that tap a manage-ment team's motivation to act upon and followthrough with improvement opportunities. We mea-sured local management team's change orientationusing two items [a = .68) rated hy the GM associ-ated with a focal team. The GM is the most appro-priate respondent to measiire a restaurant manage-ment team's change orientation hecause he or sheleads the team and is familiar with the functioningof the entire team. The items were, "The manage-ment team (including yourself). . ." "spots issuesquickly and does something ahout them" and"makes improvement plans and sticks to them."GMs rated these items on a scale ranging from 1("never") to 5 ("always").

Unit-level turnover. We measured employeetiunover as the rate at the restaurant level for thesix months after employee voice ratings were col-lected. We used a six-month period hecause welearned from discussions with the restaurant's ex-ecutive team (e.g., CEO, CFO, VP of operations) thatthe restaurants operate in quarters and that theybelieved that if employees spoke up with an ideafor improvement, most likely it would take one totwo quarters for the issue to he heard, vetted, andaddressed to the extent that improvement (or lackof) could be assessed. Second, the average annualturnover rate in this restaurant chain is 180 percentper year, which means that a high percentage of theemployees who answered our surveys would nolonger he employed after six months (two quarters)had passed. Thus, using a period shorter thansix months risked having insufficient time pass toassess the potential impact of managerial respon-siveness, hut using a longer one would lead tousing survey responses (i.e., independent and con-trol variables) from employees who were predomi-nantly no longer employed in the restaurant. Turn-over data for each restaurant were captured fromthe human resotu-ce records cleaned (i.e., strippedof identifying personal information) and trans-

534 Academy of Management Journal April

ferred to us by the sponsor organization. Turnoverwas measured as the number of employee exits as apercent of the average number of employees be-tween time 1 (beginning of month 1] and time 2(end of six months after voice ratings were col-lected]. We used the average number of employeesacross the six months to account for fluctuations inemployee number; this measure is consistent withprevious computations of turnover (Price, 1977].

Control variables. Recent meta-analyses suggestseveral antecedents that promote or inhibit turn-over (Griffeth et al , 2000; Hom & Griffeth, 1995]. Toaccount for these mechanisms, as well as those thatmight influence the relationship between voice andexit, we used a variety of control variables. We firstcontrolled for several characteristics of each restau-rant's employees. We controlled for average em-ployee tenure in each unit, which is one of thestrongest demographic predictors of turnover, be-cause longer-tenured employees are less likely toquit in subsequent periods (Griffetb et al., 2000].Given that minority groups in the U.S. are morelikely to quit (Hom & Griffeth, 1995], we also con-trolled for minority racial groups, measured as thepercentage of African American and Hispanic em-ployees in each iniit.

Second, we controlled for several different em-ployee attitudes—job satisfaction and justice per-ceptions—as these can influence turnover. Job sat-isfaction is the best job attitude predictor ofturnover; when employees are more satisfied, theyare less likely to leave or engage in actions that leadto involuntary termination (Griffeth et al., 2000;Hom & Griffeth, 1995]. We used one item ("OveraflI am satisfied with this restaurant as a place towork"] to measure job satisfaction, as global mea-sures are appropriate to measinre an employee'ssatisfaction with her/his overall position within acompany (Ironson, Smith, Brannick, Gibson, &Paul, 1989]. We controlled for interactional justiceperceptions, which refer to interpersonal treatmentreceived at a workplace, using items adapted fromBies and Moag (1986]. Those employees who feelthat they are not treated in a polite manner or withrespect are more likely to voluntarily or involun-tarily exit an organization (Aquino, Griffeth, Allen,& Hom, 1997].

Third, we controlled for two leader characteris-tics (Griffeth et al., 2000]. We controlled for abusivesupervision because previous research has shownthat employees who work under abusive leadersare more likely to exit an organization (Tepper,2000], We therefore used items adapted from Tep-

per's (2000] abusive leadership scale to account forthis alternative explanation of turnover. Note thateach of these three individually measured percep-tual control variables (overall satisfaction, justice,and abusive leadership] also met commonly recom-mended statistical thresholds for aggregation (e.g.,all F-test values for ICCl significantly differentfrom zero at p < .01]. Second, we also controlledfor each GM's teniure at the company to account forexperience managing a restaurant and its employ-ees. New GMs may also have not yet developed thepolitical capital in the organization needed to ob-tain resources for making improvements or to beinfluential in decisions affecting their restaiurant.

Fourth, we controlled for other restaurant char-acteristics that can affect work environment andsubsequent tvu-nover rates, including size, age, dateof most recent store remodel, and being a trainingrestaurant. We controlled for restaurant size (mea-sured as number of employees] because size hasbeen shown to be related to administrative com-plexity and ttirnover in previous studies (Shaw,Delery, Jenkins, & Gupta, 1998]. Restaurants wereconsidered new by senior leaders, and therefore notyet comparable to established restaurants, if theyhad been open for less than 18 months. We thusincluded a dummy variable (0 = "open for morethan 18 months," 1 = "open for less than18 months"] to account for restaurant newness. Wealso controlled for the number of months sinceeach restaurant had last been remodeled because(as learned through interviews in the company],this could affect work environment and subsequentexit behavior. This association is consistent withaspects of socio-technical systems theory (Trist,Higgins, Murray, & Pollock, 1963] noting the im-portance of the physical work environment to em-ployee attitudes and behavior. Finally, we alsofound that several restaurants were used for train-ing new employees. New employees placed inthese restaurants for training purposes do not countas full members of the training restaurants; they arecounted as members of their home restaurants.Nonetheless, these training restaurants are chosenbecause of tbeir exceptional performance and, thus,the turnover rate for these restaurants should belower than those for others. Given their uniquepurpose, we included a dummy variable (0 = "nota training restaurant," 1 = "training restaurant"] tocontrol for this condition.

Fifth, we accounted for external labor marketconditions, which have been recognized to affecttiurnover rates (Griffeth et al., 2000]. We measured

2013 McClean, Burris, and Detert 535

both the number of blue-collar laborers and theaverage household income within a three-mile ra-dius of the restatirant. We obtained these statisticsfrom the marketing department of the chain's cor-porate headquarters, which drew them from pub-licly available economic data (e.g., from the U.S.Census Bureau and the Bureau of Labor Statistics).Each of these variables controls for differences inthe external labor market that could affect oppor-tunities for alternative employment. For example,employees in areas with fewer blue-collar laborersand higher average income may have differentialopportunities for alternative forms of similar em-ployment, which could affect individuals' deci-sions to quit or to engage in risky behaviors thatmight get them fired.

Analysis Strategy

Our analyses are conducted at the restaurantlevel. Each restaurant is embedded in a district and,further, in a division containing multiple districts.Such groupings call into question whether the res-taurant-level data are independent. Therefore, priorto conducting our analyses, we employed multi-level analyses to explicitly model the nonindepen-dence resulting from restaurant division [n = 4)and district [n = 32) groupings (Raudenbush &Bryk, 2002). The between-division and between-district variance are not statistically different fromzero, meaning that the variance attributable to eachhigher level is insignificant in explaining variancein otir restaurant-level dependent variable. There-fore, we used ordinary least squares regression atthe restaurant level in the analyses reported here(Gelman & Hill, 2007). The pattern and significanceof our results are unchanged if we employ multi-level models.

RESULTS

We conducted confirmatory factor analysis(CFA) to examine the discriminant validity of ourthree moderator variables. Fit indexes from a CFAmodel (e.g., RMSEA = 0.06, CFI = 0.98, NFI =0.94) indicate that the hypothesized three-factorstructure fits the data well. We also estimated threetwo-factor models (one combining managerial par-ticipation in decision making and managerial ac-cess to organizational resources, one combiningmanagerial participation in decision making andmanagement team change orientation, and onecombining managerial access to organizational re-soiurces and management team change orientation)and a one-factor model (all three moderators load-ing on the same factor) to assess more parsimoniousmodels. All indexes from the more parsimoniousmodels indicate that each had a significantly worsefit to the data than did the hypothesized three-factor model. We conducted a chi-square differencetest for each model (relative to the hypothesizedmodel) and foimd that the more parsimoniousmodels were significantly different (worse-fitting)at p < .001. This analysis provides statistical sup-port for construct differentiation. See Table 2 for allresults.

Table 3 shows the correlations, means, and stan-dard deviations for the study variables. The pat-terns of the raw correlations are consistent withpast research. For instance, employee satisfaction (r= – .23, p < .01) and interactional justice (r = -.14,p < .05) are negatively related to turnover. Addi-tionally, restaurant remodeling date is positivelyrelated (r = .19, p < .01) to turnover, and trainingstatus is negatively related to turnover (r = -.20,p < .01). One might also expect that the managerialresponsiveness behaviors would be related to turn-over directly. However, although management

TABLE 2Confirmatory Factor Analysis Results"

Factor Structure x'/df C n RMSEA RMSEA CI NFI NfNFI AIC

1. Three-factor model2. Two-factor model: PDM + CO3. Two-factor model: PDM + AOR4. Two-factor model: CO + AOR5. One-factor model

" "AOR" is unit manager access to organizational resources; "CO" is local management team change orientation; "PDM" is unit managerparticipation in decision making.

**p < .01

62.76

130.99

322.19

99.25

358.52

4143434344

1.53

3.05

7.49

2.31

8.15

0.98

0.91

0.72

0.94

0.68

0.06

0.12

0.22

0.09

0.23

(0.02-0.09)

(0.09-0.15)

(0.19-0.24)

(0.07-0.12)

(0.20-0.25)

0.94

0.87

0.69

0.90

0.66

0.97

0.88

0.64

0.92

0.61

-19.24

44.99

236.19

13.25

270.62

68.23(2)**

259.43(2)**

36.49(2)**

295.76(3)**

• •g'sgu•9

I l I

M

1CO

25**

12

1

t.s at

1

CO CM

1

O

1

13

CO

1

o

1

CO CO

1

11 17*

CO

14*

1

rH

15*

CO

19**

CM

1

05

CO M

11 24**

CM

O

rH

CO

O

rH

1

gO

OO

rHCO

O

rH

1

O)COo

oo

COCMO

ôo

s1COCO

en

roCM

1

COrH

O

in

o

r-irH

CM

O

COCM

in

CO

o1

COrH

O

h»o1

CM

O

o

COCO

o

CO

or-1

.s

l g

no Si'S -ä

1

•a

«COCO

in COO o

1

* rr CM

) O rH

16*

1«CD

1

.04

14*

1

o1«rHCI

16*

.12

CO

o

r

28**

o

CO

o1

CO

o1

CMO

-.07

rH

rHO1

23**

«

CM

O

CO

o

CMO

r

-.10

CO

o

r. oo

CO

o1

CO

o1

in

o1

.11

tN.

o

rt^

o'

.05

CO

ino

oo

s

CMO

CO

o

o

inrH

r

.06

o

roo

1

CO

1

COO

«OO

o

rin

r

.08

tnq

CM

q

q

mol'

CO

o

r

.07

CMO

r

.19*

*

1

.09

,19*

*

1

16*

q

«rHCO

l"

CO

o

rinq

„ „ s s 'C , _

_, É I I 1§ 9̂ §°S H ë 3 < c o M 2 S Ë 2 2 l

in rH

" V V

2013 McGlean, Burris, and Detert 537

team change orientation is negatively related toturnover (r = -.18, p < .05), both managerial par-ticipation in decision making and access to organ-izational resources are not significantly associatedwith imit turnover at the bivariate level. Addition-ally, unit-level voice is not significantly directlyrelated to employee turnover (r = .004, n.s.).

We conducted four regression analyses to assessthe interaction between voice and each variable ofinterest (see Table 4); the first three include oneinteraction term only, and the final model includesthe three interaction terms simultaneously. To en-hance the interpretation of the main effects andreduce multicoUinearity concerns, we centered allvariables involved in the interaction terms (Aiken& West, 1991). First, we examined the relationshipbetween the control variables and unit-level turn-over. As model 1 shows, collectively these vari-

ables significantly contribute to the model's ex-planatory power (AF = 5.05, df = 136, p < .001)and explain 35 percent of the variance in laggedturnover. Several individual control variables re-main significantly related to turnover in the regres-sion, such as average employee tenure (ß = -0.34,p < .01), percent African American employees [ß =-0.22, p < .01), percent Hispanic employees (ß =-0.24, p < .01), number of employees (ß = -0.32,p < .01), months since the restaurant had beenremodeled (ß = 0.21, p < .05), and number ofblue-collar workers within a three-mile radius ofthe restainrant (ß = 0.20, p < .05).

Second, we entered the variables for employeevoice, managerial access to organizational re-sources, managerial participation in decision mak-ing, and local management team change orienta-tion. Results are presented in models 2-4 of Table

TABLE 4Hierarchical Regression Analysis Results for the Six-Month Unit-Level Turnover Rate"

Variables'"

Gontrol variablesAverage tenurePercent African AmericanPercent HispanicEmployee satisfactionJustice perceptionsAbusive leadershipGM tenureNumber of employeesNew restaurantMonths since last remodelingTraining restaurantBlue-collar workers within 3 milesAverage household income within 3 milesMain effectsEmployee voiceManager AORManager PDMManagement team COInteractionsVoice X manager AORVoice X manager PDMVoice X management team CO

Adjusted iî^F

Control Model

Model 1

-0.34*-0.22**-0.24**-0.17

0.050.030.08

-0.32**-0.05

0.21*0.010.20*

-0.09

0.35

0.285.05**

Model 2

-0.34**-0.22**-0.24**-0.17

0.050.040.08

-0.33**-0.05

0.21*0.020.20*

-0.09

-0.040.01

0.350.000.274.33**

Main Effects

Model 3

-0.36**-0.18*-0.23**-0.12

0.020.020.09

-0.34**-0.07

0.21*0.010.21*

-0.12

-0.02

-0.19*

0.380.030.304.94**

Model 4

-0.35**-0.19*-0.23**-0.17

0.040.010.07

-0.30**-0.05

0.24*0.010.20*

-0.11

-0.04

-0.16*

0.370.020.294.76**

Model 5

-0.36**-0.23**-0.21**-0.13

0.030.030.08

-0.31**-0.05

0.26*0.010.19*

-0.14

-0.040.03

-0.19*

0.380.030.294.56**

Moderation

Model 6

-0.35**-0.20*-0.20*-0.14

0.050.020.09

-0.35**-0.06

0.25*0.010.18*

-0.15

-0.03

-0.17*

-0.17*

0.400.050.325.05**

Model 7

-0.39**-0.24*-0.18*-0.19*

0.08-0.01

0.04-0.28**-0.07

0.30**0.010.24**

-0.14

-0.11

-0.15

-0.26**0.420.070.345.50**

Model 8

-0.43**-0.21**-0.16*-0.13

0.030.000.08

-0.28-0.28**

0.32**0.000.22**

-0.18*

-0.050.16

-0.17*-0.16*

-0.12-0.06-0.16*

0.460.110.374.97**

" n = 136. Standardized coefficients are reported.^ "AOR" is unit manager access to organizational resources; "CO" is local management team change orientation; "PDM" is unit manager

participation in decision making.* p < .05

**p < .01

538 Academy of Management Journal April

4. Employee voice is not significantly related toturnover at the unit level. Both unit manager par-ticipation in decision making and managementteam change orientation are significantly related toturnover (ß = -0.19, p < .05, and ß = -0.16, p <.05, respectively).

Finally, we examined the relationship betweenvoice and turnover as moderated by our three lead-ership variables in models 5-8. We proposed inHypothesis 1 that manager access to organizationalresources would moderate the relationship be-tween employee voice and exit at the restaurantlevel. As shown in model 5 of Table 4, the interac-tion term between voice and manager access toorganizational resources is significantly related toemployee turnover (ß = -0.19, p < .05) and ex-plains a significant amount of variance over andabove the control variables and main effects of themoderator variables (AJF^ = 5.60, df= 136, p < .05).To assess whether the form of this, and the other,interactions is consistent with our hypotheses, weplotted the interactions according to the guidelinesprovided by Aiken and West (1991), plotting re-sults at high and low levels of each variable (onestandard deviation above and below the mean). Asshown in Figure 2, the relationship between voiceand turnover when manager access to resources ishigh versus low is in the predicted direction. Wefollowed the approach of Preacher, Curran, andBauer (2006) to compute the region of statisticalsignificance of the simple slopes of the interactionin Z-score units. Specifically, we found that whenZ-scores are higher than 0.88, (i.e., high access toorganizational resources), the negative relationship

between voice and turnover is statistically signifi-cant (p < .05). When the Z-scores are lower than-2.00 (i.e., low access to organizational resources),the positive relationship between voice and turn-over is statistically significant (p < .05). These re-sults are consistent with Hypothesis 1.

We propose in Hypothesis 2 that managerial par-ticipation in decision making moderates the rela-tionship hetween employee voice and exit at theunit (here, restaurant) level. As shown in model 6of Table 4, the interaction term between voice andmanagerial participation in decision making is sig-nificantly related to employee turnover (ß = -0.17,p < .05) and explains a significant amount of vari-ance over and above the control variables and maineffects of the moderator terms (AF = 4.62, df= 136,p < .05). As shown in Figure 3, the relationshipbetween voice and turnover when manager partic-ipation in decision making is high versus low is inthe predicted direction. Specifically, usingPreacher et al.'s (2003) regions of significancemethod, we found that when Z-scores are higherthan 1.77, (i.e., high participation in decision mak-ing), the negative relationship between voice andturnover is statistically significant (p < .05). Whenthe Z-scores are lower than -2.68 (i.e., low accessto organizational resources), the positive relation-ship between voice and turnover is statistically sig-nificant (p < .05). These results are consistent withHypothesis 2.

We proposed in Hypothesis 3 that the manage-ment team change orientation moderates the rela-tionship between unit-level employee voice andexit. As shown in model 7 of Table 4, the interac-

FIGURE 2Interaction between Unit-Level Voice and Unit Manager Access to Organizational Resources

2 . 3 –

2 .2 –

2 . 1 –

2 –

1.9-Turnoverat

Six Months

1.7-

1.6-

1.5-

1.4

>— Low unit manageraccess toorganizationalresources

I— High unit manageraccess toorganizationalresources

Low Voice High Voice

2013 McClean, Burris, and Detert 539

FIGURE 3Interaction between Unit-Level Voice and Unit Manager Participation in Decision Making

•Turnoverat

Six Months

2.3-|

2.2-

2.1-

2 –

1.9-

1.7-

1.6-

1.5-

1.4

>— Low unit managerparticipation indecision making

I— High unit managerparticipation indecision making

Low Voice High Voice

tion • term between voice and management teamchange orientation is significantly related to em-ployee turnover (ß = -0.26, p < .01] and explainsa significant amount of variance over and above themain effects (AF^ = 10.77, df = 136, p < .01].Figure 4 displays the interaction pattern supportingour prediction. Specifically, we found that whenZ-scores are higher than 0.13, (i.e., high manage-ment team change orientation], the negative rela-tionship between voice and tvurnover is statisticallysignificant (p < .05]. When the Z-scores are lowerthan -0.84 (i.e., low management team change ori-entation], the positive relationship between voiceand tiurnover is statistically significant (p < .05].These results offer support for Hypothesis 3.

Last, to test their relative strength we entered allthree interactions into the regression model simul-taneously. As shovra in model 8 of Table 4, theinteraction between voice and management teamchange orientation remains significantly related toemployee turnover (ß = -0.16, p < .05], but theother two interactions do not. We consider theseresults further in the Discussion section.

Robustness Checks

We conducted additional analyses to ensure therobustness of tbe results presented above. First,despite our attempts to control for many differentalternative explanations for turnover and for factors

FIGURE 4Interaction between Unit-Level Voice and Management Team Change Orientation

Turnoverat

2.2-

2.1-

2 –

1.9-

Six Months 1-8-

1.7-

1.6-

1.5-

1.4

Lowmanagementteam changeorientationHighManagementteam changeorientation

Low Voice High Voice

540 Academy of Management Journal April

that might affect the relationship between unit-level employee voice and tinrnover, there is a rea-sonable possibility that other explanations mightremain unaccounted for. We therefore ran threeadditional models in which we also controlled foreach prior year's turnover rate in each restaurant toaccount for other sources of unobserved heteroge-neity. All interactions remained significant and thepattern of results stayed the same (manager accessto organizational resources X voice: ß = -0.14, p <.05; manager participation in decision making Xvoice: ß = -0.11, p < .05; management teamchange orientation X voice: ß = -0.25, p < .01).Second, because our turnover data are slightlyskewed, we reran all models using a log transfor-mation of turnover to account for nonnormality.Again, all interactions remained significant and thepattern of results stayed the same (manager accessto organizational resources X voice: ß = -0.17, p <.05; manager participation in decision making Xvoice: ß – 0.13, p < .05; management team changeorientation X voice: ß = -0.29, p < .01). Finally, toreduce concerns that spurious suppression couldaffect oiu: results, given the number of control vari-ables that we included in our analyses (Becker,2005), we reran the analyses taking out one controlvariable at a time to examine the effects on thesignificance levels of the interactions. Of the 39regressions run in this manner, 97 percent of theinteractions stay significant (p < .05) or marginallysignificant (p < .10) when one control variable isremoved. Although these results alone cannot com-pletely rule out the possibility of suppression,along with the theoretical rationale for each controlvariable's inclusion, they bolster confidence in therobustness of our results.

DISCUSSION

In this study, we used theory from the organiza-tional sciences and empirical evidence to recon-sider and build upon several aspects ofHirschman's (1970) seminal framework about voiceand exit in organizations. First, we challenged thenotion that the outcomes of voice, including em-ployee turnover, should be generally conceptual-ized at the individual level (i.e., the level of thespeaker). Instead, we argued and demonstratedthrough findings from analysis of 3,388 employeeinstances of improvement-oriented input that thevery nature of voice is prosocial and thus voice hasthe potential to affect outcomes for not just thespeaker but also for others whose performance or

well-being are likewise affected by the issuesraised. Second, instead of considering voice andexit as mutually exclusive alternatives, we theo-rized the importance of examining the relationshipbetween employee voice and exit and, in particular,argued and demonstrated empirically that the rela-tionship between the amount of speaking up byemployees and subsequent turnover in an appro-priately defined unit is contingent on the abilityand motivation of that unit's management to ad-dress the issues raised.

Our empirical results not only speak to long-standing theoretical frameworks, but are also ofsignificant practical import, given the costly andvexing problem that turnover represents for organ-izations. In the food service industry, the context ofthis study, turnover rates generally exceed 100 per-cent annually, with employees citing poor manage-ment as a key reason for leaving (Kacmar, Andrews,Van Rooy, Steilberg, & Cerrone, 2006; Zuber, 2001).Though turnover rates range significantly, and areobviously lower in other industries, turnover isdisruptive and expensive everywhere (Mueller &Price, 1989; Staw, 1980) because it inevitably raiseslabor costs related to recruiting, selecting, andtraining new hires (Oi, 1962; Sagie, Birati, & Tziner,2002) and lowers organizational performance (Kac-mar et al., 2006; Shaw, Gupta, & Delery, 2005). Thisstudy highlights yet another significant and ironiccost of turnover: places with high levels of em-ployee voice and thus many improvement-orientedideas can be the same places that lose the mostemployees when management is unwilling or un-able to respond. Even at a conservative estimate of$500 to replace an employee in this context (Kac-mar et al., 2006), the annual savings associatedwith, for example, a management team that is onestandard deviation higher than the mean on changeorientation compared to one that is one standarddeviation lower would exceed $1,600,000 for anorganization with the same number and size ofrestaurants as investigated here.^

^ To obtain the cost value, first we calculated the pre-dicted turnover from the regressions. We then multipliedthis coefficient by two to obtain the predicted annualturnover rate (recall that our dependent variable was theturnover for six months). Then, using the average numberof employees at each store (48), we calculated the num-ber of employees to be replaced when tbe moderator ishigh and low using the annual rate calculation. We thencalculated tbe cost per store using an average cost of $500per employee (Kacmar et al., 2006). To do tbis, we mul-

2013 McClean, Burris, and Detert 541

Contributions

Our theory and findings extend Hirschman'sframework by acknowledging the social context forvoice and the collective impact of suggestionsmade by employees. Our research seeks to alignsome of Hirschman's initial insights with the con-temporary understanding among voice scholars thatvoice in work contexts is a prosocial behavior— t̂hatis, to qualify as voice, the raising of an issue shouldhave the potenüal to benefit a larger collective such asa work group or organization (Van D)nie et al., 2003).Whereas most prior research has focused on individ-ual-level antecedents to voice (Burris, Detert, & Chia-buxu, 2008; Tangirala & Rumanujam, 2008), we ex-amined tbe unit-level outcomes of voice, explicitlytheorizing that the benefit or harm of different levelsof management responsiveness to employee input ex-tend to many employees in a unit. Upon systematicexaminafion of open-ended employee improvementsuggesfions, we found that the overwhelming major-ity of employee comments were indeed consistentwith the "would benefit more than just me if ad-dressed" criterion that defines voice as a prosocialbehavior (Grant & Mayer, 2009; Van Dyne et al.,2003). This finding highlights the importance of con-sidering the outcomes of managerial responsivenessto employee voice among groups of employees, ratherthan just those who speak up about certain issues orjust those who managers are more likely to be respon-sive to.

Prior studies have often treated voice as the endresult, usually suggesting rather than testing anydownstream relationships between voice and desir-able or undesirable outcomes. Research building onHirschman's (1970) framework has characterizedvoice as replacing need to exit because it inevitablybrings about improvements that address thesources of employee dissatisfaction if given enoughtime (e.g., Withey & Cooper, 1989). Such logic ex-tends beyond turnover, as scholars have arguedthat voice should affect other group-level outcomessuch as change effectiveness (Morrison & Milliken,2000) and performance (Morrison, 2011). These ar-

tiplied the number of employees to be replaced by the$500 value for botb the low and high moderator situa-tions. To ohtain the total cost for the company, we mul-tiplied the cost per store value times 139, the number ofunits in this particular company. To obtain the cost sav-ings value, we then subtracted the value calculated fromthe high moderator situation from the low moderatorsituation.

guments rest largely on the assumption that actuallearning occurs and that substantive changes takeplace. We explore this view theoretically and em-pirically by acknowledging that managers oftenhave limited latitude and ability to initiate substan-tive cbange and that, as a result, they are oftenconstrained in addressing employee concerns. Ourresults show that a high level of voice alone is nota sign that employees are more likely to stay; in-stead, the relationship between voice and exitwithin work imits is dependent on whether man-agers are willing and able to do something withemployees' ideas for improvement. Thus, our re-search highlights the importance of consideringmanagement responsiveness as an important deter-minant of whether voice leads to positive or nega-tive outcomes for a work unit.

Our research also makes several important theo-retical contributions to the leadership literature. Ata broad level, our results add to contingency per-spectives on leadership. Much recent leadershipresearch has predominantly focused on direct rela-tionships between aspects of leadership and effec-tiveness outcomes (Judge, Piccolo, & Kosalka, 2009;Lowe, Kroeck, & Sivasubramaniam, 1996), includ-ing employee turnover (Graen, Liden, & Hoel,1982). Our research fits with a recent resurgence inthe leadership literature that examines the contin-gencies, rather than main effects, of leadership onvmit-level outcomes (Grant, Gino, & Hofmann,2011). For example. Grant and colleagues (2011)found that the relationship between leader extra-version and unit effectiveness is contingent on theproactivity of employees. Our study demonstrateshow the value of specific leadership behaviors forreducing exit depends, in part, on the level of voiceactivity by subordinates. More specific to the phe-nomenon of voice, our treatment of leadership rep-resents a significant departure from the focus onleadership characteristics as antecedents to voicebehavior. Past research shows, for example, thatspecific leader behaviors or styles like transforma-tional characteristics or leader openness (Detert &Burris, 2007) stimulate more voice. Complement-ing this research, our results show the implicationsof managerial actions when voice is already pres-ent. Namely, our findings suggest that leader re-sponsiveness to voice also affects the choices ofemployees after spealcing up has occurred.

Finally, our findings are also theoretically andpractically relevant for those interested in minimiz-ing turnover. Our study explicitly acknowledgesand demonstrates that turnover decisions are likely

542 Academy of Management Journal April

to be predictable at meaningful aggregate levelsbased on social factors—such as managerial re-sponsiveness to voice—that are similarly experi-enced by defined groups of employees (Krackhardt& Porter, 1986). Further, by examining when unit-level voice affects turnover rates, we uncover po-tential levers for leaders to use to minimize turn-over, such as providing managers with resources tomake changes or hiring managers who are moreoriented toward change. Also, beyond our focalvariables, the set of control variables we includedexplained fully 35 percent of the variance in laggedturnover. Of these alternative explanations, several,including the percentages of African American andHispanic employees and the number of monthssince a restaurant was last remodeled, are signifi-cantly related to exit. Gollectively, our findings ex-tend understanding of unit-level factors that influ-ence employee turnover and thus offer importantavenues for future research.

Strengths, Limitations, and Future Directions

Several features of this study bolster confidencein our results. First, we collected data from multi-ple sources, including observer ratings of employeevoice. Second, we used objective turnover data forthe six months after the assessment of voice, ratherthan concurrent turnover statistics or turnover in-tention reports (which may or may not result inactual exit). Third, we found support for our hy-potheses about the relationship between unit-levelvoice and exit after first controlling for an extensiveset of theoretically meaningful alternative explana-tions of turnover. Fourth, our results are robust tomultiple additional analyses, including modelsthat control for prior turnover, models that uselog-transformed turnover as the dependent vari-able, and models that check for suppression effects.

Despite these strengths, several limitations of ourstudy should be addressed in future research. First,we tested our hypotheses in a particular context—one where turnover is high and factors such asorganizational commitment are relatively low. Un-derstanding the generalizability of our results thusrequires additional research in other contexts. Forinstance, the relevant size of the group or unit ofpeople that managerial responsiveness to voice willaffect likely varies. We used single restaurants asthe unit of aggregation in this study because they fitthe nature of the types of issues raised; however, inother contexts, such as scientific research groups orproduct development teams, the types of issues

raised may make the appropriate level of analysismuch smaller. For example, in academia, we mightexpect the appropriate level of aggregation to be thedepartment rather than school level. Second, futurestudies could benefit from additional and more pre-cise measures. For example, the use of objectivemeasures of access to resources and other-sourceratings of manager participation in decision makingand management team change orientation wouldreduce the potential biases of self-report measures.Additionally, in arguing that our moderators areproxies for managerial responsiveness to employ-ees' concerns, we suggested that reduced turnoveris likely the result of actual improvements resultingfrom employees' voice and/or that employees feelsufficiently satisfied by managers taking their con-cerns seriously even when substantive changesdo not always occur. In this study, we were not ableto empirically assess each of these potential medi-ators between managerial responsiveness and sub-sequent employee exit. We were also unable todisentangle the voluntary from the involuntaryturnover rates for each unit and thus determinemore precisely to what extent when employeesspeak up, leader responsiveness is related to em-ployee decisions to quit, leader decisions to fire,or both.

Future research could also more robustly exam-ine both individual- and group-level effects by col-lecting individual-level responses for the modera-tors and individual-level turnover data. Such datawould allow for a multilevel analysis that moreprecisely disentangles the extent to which em-ployee exit is an individual decision made in reac-tion to leader behaviors directed to each employeewho speaks up or predominantly influenced by thelarger social context, including whether improve-ments were made (or not made), irrespective ofwho raises specific issues. Researchers might alsouse other designs and data analysis approaches toexamine whether employees exit after they or oth-ers have spoken up because they were bothered bya lack of response to a single, specific issue or by amore general pattern of unresponsiveness to an ar-ray of work issues. Ballinger and Rockmann (2010),for example, suggested that relationships canchange quickly, with the outcome of a particularevent dramatically and permanently altering thestate of a relationship. Similarly, the unfoldingmodel of turnover (Lee & Mitchell, 1994) identifiesshocks, or particularly jarring work-related events,as antecedent to voluntary exit. We suspectthat—in contrast to the picture painted by

2013 McClean, Burris, and Detert 543

Hirschman (1970)—turnover usually results froman ongoing and building sense of frustration, dis-satisfaction, and change in performance rather thanfrom a single issue or event, but the matter is anempirical question still open for adjudication. Fu-ture researchers should also consider measuringexit at different time periods. We chose a six-monthturnover window hecause interviews with seniormanagers suggested this was an appropriate lengthof time in this context for assessing whetherchanges were attempted and succeeding (Mitchell& James, 2001). However, by collecting turnoverrates at multiple time intervals, researchers wouldhe ahle to test more precisely how managerialresponsiveness affects subsequent employeebehavior.

Beyond addressing the limitations of this study,future research could build upon our study andexplore some of the unexpected findings. For ex-ample, we were surprised to find that when em-ployee voice was low and managerial access toorganizational resources was low, employees weremore likely to stay with the organization (as theywere under conditions of high voice and high man-agerial access to resources). One possihle explana-tion for these findings is that when employee be-havior matches manager ability to make changes,employees are more likely to stay with an organi-zation than they are when employees' behavior ismismatched to their manager's orientation andahility to make changes. Additionally, several ofour focal variables (voice, manager access to organ-izational resources, and manger participation indecision making) are not directly related to turn-over. We have already explained our perspectiveon these findings (i.e., that the relationships he-tween voice and outcomes depend on what manag-ers do next) but recognize the need for future re-search in this area given the practical importance ofsuch relationships. We were also surprised to findthat abusive leadership did not have a significantimpact on turnover in this study. This could stemfrom the low incidence of (and variance in) abusiveleadership captured here. Or, it could be that, aswith other aspects of leadership, the relationshipbetween abusive leadership and turnover is contin-gent on factors such as norms in a given environ-ment and employee mobility options (Tep-per, 2000).

Last, we note that all three interactions are sig-nificant when estimated independently (models5-7), but when all are included in one equation,only the interaction hetween voice and manage-

ment team change orientation remains significant(model 8). Statistically, this pattern of results maybe due to the lack of power that results from simul-taneously testing the significance of three interac-tions with a sample of the size of our sample. Theresults may also be due to shared variance amongour predictors, given that each interaction includesa common variable (voice). When predictors in-clude a common variable, it is harder to tease outthe individual effects of each (Aiken & West, 1991).A more substantive reason for these results may hethat modeling multiple interactions together is avery conservative test that speaks to the relativeimpact of each interaction when the others are con-trolled for. In our case, the test is more conservativehecause the three variables—managerial access toorganizational resources, participation in decisionmaking, and management team change orienta-tion—co-occur at a relatively high level in the ma-jority of our restaurants. For example, in 81 percentof the restaurants at which managers have ahovethe mean level of participation in decision making,they also have above the mean level of access toresources (even though the raw correlation hetweenparticipation in decision making and access to re-sources is .44). This fact—that our sample does notinclude many instances of all permutations of highversus low levels of each of our three moderators—makes it difficult to tease out unique effects in asingle model. So although model 8 provides anopportunity to examine the relative magnitude ofeach interaction's effect, models 5-7 provide aclearer picture of the independent effect of each.

Conclusion

Our findings suggest that the outcomes of em-ployee voice rest, to a nontrivial extent, in thehands of managers; their responsiveness impactswhether voice proves beneficial or detrimental forthe work unit and beyond. At the same time, ourresults emphasize a potential disadvantage of en-couraging employee voice: if leaders are not ahleand prepared to act on problems identified andopportunities suggested, employees may look tofind employment where their input has more im-pact or even engage in behaviors that provoke man-agers to remove them. Given the significant cost ofturnover, as well as the other losses associated withthe failure to respond to employees' knowledge andideas, managers would he wise to enhance their,and their management teams', ability and motiva-tion to take action when employees speak up.

544 Academy of Management Journal April

REFERENCES

Adams, J, S. 1965. Inequity in social exchange. Advancesin Experimental Social Psychology, 62: 335-343.

Aiken, L. S., & West, S. G. 1991. Multiple regression:Testing and interpreting interactions. NewhuryPark, CA: Sage.

Allen, D. G., Shore, L. M., & Griffeth, R. W. 2003. The roleof perceived organizational support and supportivehuman resoinrce practices in the turnover process.Journal of Management, 29: 99-118.

Aquino, K., Griffeth, R., Allen, D., & Hom, P. 1997. Anintegration of justice constructs into the tiu-noverprocess: Test of a referent cognitions model. Acad-emy of Management Journal, 40: 1208-1227.

Argyris, G., & Schön, D. 1978. Organizational learning:A theory of action perspective. Reading, MA: Ad-dison Wesley.

Armenakis, A. A., & Bedeian, A. G. 1999. Organizationalchange: A review of theory and research in the1990s. Journal of Management, 3: 293-315;

Avery, D. R., & Quiñones, M. A. 2002. Disentangling theeffects of voice: The incremental roles of opportu-nity, hehavior, and instrumentality in predictingprocedxnral fairness. Journal of Applied Psychology,87: 81-86.

Bagozzi, R. P. 2003. Positive and negative emotions inorganizations. In K. S. Gameron, J. E. Dutton, & R. E.Quinn (Eds.), Positive organizational scholarship:176-193. San Francisco: Berrett-Koehler.

Ballinger, G. A., & Rockmann, K. W. 2010. Ghutes versusladders: A punctuated-equilibrium perspective onsocial exchange relationships. Academy of Manage-ment Review, 35: 373-391.

Barry, B. 1974. Review article: Exit, voice and loyalty.British Journal of Political Science, 4: 79-107.

Barry, B., & Shapiro, D. L. 2000. When will grievantschoose voice? A test of situational, motivational, andattributional explanations. International Journal ofConfiict Management, 11: 106-134.

Bass, B., & Avolio, B. 1994. Improving organizationaleffectiveness through transformational leader-ship. Thousand Oaks, GA: Sage.

Becker, T. E. 2005. Potential prohlems in the statisticalcontrol of variables in organizational research: Aqualitative analysis with recommendations. Organi-zational Research Methods, 8: 274-289.

Beer, M. 2009. High commitment, high performance:How to build a resilient organization for sustainedadvantage. San Francisco: Jossey-Bass.

Bies, R. ]., & Moag, J. S. 1986. Interactional justice: Com-munication criteria for fairness. In B. Sheppard (Ed.),Research on negotiation in organizations, vol. 1:43-55. Greenwich, GT: JAI.

Blader, S. L., & Tyler, T. R. 2009. Testing and extendingthe group engagenient model: Linkages hetweensocial identity, procedural justice, economic out-comes, and extrarole behavior. Journal of AppliedPsychology, 94: 445-464.

Bliese, P. D. 2000. Within-group agreement, non-inde-pendence, and reliahility: Implications for data ag-gregation and analysis. In K. J. Klein & S. W. Koz-lowski (Eds.), Multilevel theory, research, andmethods in organizations: 349-381. San Francisco:Jossey-Bass.

Bordia, P., Restuhog, S. L., & Tang, R. L. 2008. Whenemployees strike hack: Investigating mediatingmechanisms hetween psychological contract breachand workplace deviance. Journal of Applied Psy-chology, 93: 1104-1117.

Burgelman, R. A., & Sayles, L. R. 1986. Inside corporateinnovation: Strategy, structure, and managerialskills. New York: Free Press.

Bxnrris, E. R., Detert, J. R., & Ghiaburu, D. S. 2008. Quittingbefore leaving: The mediating effects of psychologi-cal attachment and detachment on voice. Journal ofApplied Psychology, 93: 912-922.

Burris, E. R., Rodgers, M. S., Mannix, E. A., Hendron,M. G., & Oldroyd, J. B. 2009. Playing favorites: Theinfluence of leaders' inner circle on group processesand performance. Personality and Social Psychol-ogy Bulletin, 35: 1224-1257.

Ghandler, A. D., Jr.. 1962. Strategy and structure: Chap-ters in the history of the American industrial en-terprise. Gambridge, MA: MIT Press.

Gohen, W., & Levinthal, D. 1990. Absorptive capacity: Anew perspective on learning and innovation. Ad-ministrative Science Quarterly, 35: 128-152.

Golquitt, J. A., Noe, R. A., & Jackson, G. L. 2002. Ante-cedents and consequences of procedin-al justice cli-mate. Personnel Psychology, 55: 83-109.

Gonger, J., & Kanungo, R. 1998. Charismatic leadershipin organizations. Thousand Oaks, GA: Sage.

Grook, T. K, Ketchen, D. J., Gomhs, J. G., & Todd, S. Y.2008. Strategic resources and performance: A meta-einalysis. Strategic Management Journal, 29:1141-1154.

Damanpour, F. 1991. Organizational inno.vation: A meta-analysis of effects of determinants and moderators.Academy of Management Journal, 34: 555-590.

2013 McGlean, Burris, and Detert 545

Dansereau, F., Jr, Graen, G., & Haga, W. J. 1975. A verticaldyad linkage approach to leadership within formalorganizations: A longitudinal investigation of therole making process. Organizational Behavior andHuman Performance, 13: 46-78.

Detert, J. R., & Burris, E. R. 2007. Leadership behaviorand employee voice: Is the door really open? Acad-emy of Management Journal, 50: 869-884.

Detert, J., & Trevino, L. 2010. Speaking up to higher-ups:How supervisors and skip-level leaders influenceemployee voice. Organization Science, 21: 249-270.

Drucker, P. F. 1974. Management: Tasks, responsibili-ties, practices. New York: Harper & Row.

Dutton, J. E., & Ashford, S. J. 1993. Selling issues to topmanagement. Academy of Management Review,18: 397-428.

Dutton, J. E., & Jackson, S. E. 1987. The categorization ofstrategic issues by decision makers and its links toorganizational actions. Academy of ManagementReview, 12: 76-90.

Edmondson, A. G. 1999. Psychological safety and learn-ing behavior in work teams. Administrative ScienceQuarterly, 44: 350-383.

Edmondson, A. C. 2003. Speaking up in the operatingroom: How team leaders promote learning in inter-disciplinary action teams. Journal of ManagementStudies, 40: 1419-1452.

Eisenberger, R., Karagonlar, G., Stinglhamber, F., Neves,P., Becker, T. E., Gonzalez-Morales, M. G., & Steiger-Mueller, M. 2010. Leader-member exchange and af-fective organizational commitment: The contribu-tion of supervisor's organizational embodiment.Journal of Applied Psychology, 95: 1085-1103.

Epitropaki, O., & Martin, R. 2004. Implicit leadershiptheories in applied settings: Factor structure, gener-alizability, and stability over time. Journal of Ap-plied Psychology, 89: 293-310.

Erdogan, B., & Enders, J. 2007. Support from the top:Supervisors' perceived organizational support as amoderator of leader-member exchange to satisfactionand performance relationships. Journal of AppliedPsychology, 92: 321-330.

Folger, R., & Gropanzano, R. 1998. Organizational jus-tice and human resource management. BeverlyHills, GA: Sage.

Fox, S., Spector, P. E., & Miles, D. 2001. Gounterproduc-tive work behavior in response to job Stressors andorganizational justice: Some mediator and modera-tor tests for autonomy and emotions. Journal of Vo-cational Behavior, 59: 291-309.

Gelman, A., & Hill, J. 2007. Data analysis using regres-sion and multilevel/hierarchical models. Gam-bridge, UK: Gambridge University Press.

Giacalone, R., & Greenberg, J. 1997. Antisocial behaviorin organizations. London: Sage.

Glauser, M. J. 1984. Upward information flow in organi-zations: Review and conceptual analysis. HumanRelations, 37: 613-643.

Graen, G., Gashman, D., Ginsburgh, S., & Schiemann, W.1977. Effects of linking-pin quality upon the qualityof working life of lower participants: A longitudinalinvestigation of the managerial understructure. Ad-ministrative Science Quarterly, 22: 491-504.

Graen, G. B., Liden, R., & Hoel, W. 1982. Role of leader-ship in the employee withdrawal process. Journal ofApplied Psychology, 67: 868-872.

Grant, A. M., & Ashford, S. J. 2008. The dynamics ofproactivity at work. In B. M. Staw & A. Brief (Eds.),Research in organizational behavior, 28: 3-34.Greenwich, GT: JAI.

Grant, A. M., Gino, F., & Hofmann, D. A. 2011. Reversingthe extraverted leadership advantage: The role ofemployee proactivity. Academy of ManagementJournal, 54: 528-550.

Grant, A. M., & Mayer, D. M. 2009. Good soldiers andgood actors: Prosocial and impression managementmotives as interactive predictors of afñliative citi-zenship behaviors. Journal of Applied Psychology,94: 900-912.

Greenberg, J. 2000. Promote procedural justice to en-hance acceptance of work outcomes. In E. A. Locke(Ed.), Handbook of principles of organizational be-havior: 181-195. Oxford, UK: Blackwell.

Griffeth, R. W., Hom, P. W., & Gaertner, S. 2000. A meta-analysis of antecedents and consequences of em-ployee txirnover: Update, moderator tests, cind re-search implications for the next millermium.Journal of Management, 26: 463-488.

Hackman, J. R. 1987. The design of work teams. In J.Lorsch (Ed.), Handbook of organizational behav-ior: 315-342. Englewood Gliffs, NJ: Prentice Hall.

Hambrick, D. G., Geletkanycz, M. A., & Fredrickson, J. W.1993. Top management commitment to the statusquo: Some tests of its determinants. Strategic Man-agement Journal, 14: 401-418.

Hausknecht, J. P., Hiller, N. J., & Vance, R. J. 2008. Work-unit absenteeism: Effects of satisfaction, commit-ment, labor market conditions, and time. Academyof Management Journal, 51: 1223-1245.

546 Academy of Management Journal April

Hirschman, A. O. 1970. Exit, voice, and loyalty: Re-sponses to decline in firms, organizations, andstates. Gambridge, MA: Harvard University Press.

Ho, V. 2005. Social influence on evaluations of psycho-logical contract fulflUment. Academy of Manage-ment Review, 30: 113-128.

Hom, P. W., & Griffeth, R. W. 1995. Theories of turnover:Employee turnover. Gincinnati: South-Western.

Ibarra, H., & Andrews, S. B. 1993. Power, social-influ-ence, and sense making: Effects of network centralityand proximity on employee perceptions. Adminis-trative Science Quarterly, 38: 277-303.

Ironson, G. H., Smith, P. G., Brannick, M. T., Gibson,W. M., & Paul, K. B. 1989. Gonstruction of a job ingeneral scale: A comparison of global, composite andspecific measures. Joumal of Applied Psychology,74: 193-200.

Judge, T. A., Piccolo, R. F., & Kosalka, T. 2009. The brightEind dark sides of leader traits: A review and theoret-ical extension of the leader trait paradigm. Leader-ship Quarterly, 20: 855-875.

Kacmar, K. M., Andrews, M. G., Van Rooy, D. L., Steil-berg, R. G., & Gerrone, S. 2006. Sure, everyone can bereplaced . . . but at what cost? Turnover as a predic-tor of unit-level performance. Academy of Manage-ment Joumal, 49: 133-144.

Kanter, R. M. 1983. The change masters: Innovation forproductivity in the American corporation. NewYork: Simon & Schuster.

Kidwell, R. E., & Bennett, N. 1993. Employee propensityto withhold effort: A conceptual model to intersectthree avenues of research. Academy of Manage-ment Joumal, 18: 429-456.

Kotter, J. P. 1996. Leading change. Boston: HarvardBusiness School Press.

Kraatz, M., & Zajac, E. 2001. How organizational re-sources affect strategic change and performance inturbulent environments: Theory and evidence. Or-ganization Science, 12: 632-657.

Krackhardt, D., & Porter, L. W. 1986. The snowball effect:Turnover embedded in commimication networks.Journal of Applied Psychology, 71: 50-55.

Landis, J. R., & Koch, G. G. 1977. The measurement ofobserver agreement for categorical data. Biometrics,33: 159-174.

Lee, T. W., & Mitchell, T. R. 1994. An alternative approach:The unfolding model of volimtary employee turnover.Academy of Management Review, 19: 51-89.

Leung, K., & Li, W. 1990. Psychological mechanisms of

process-control effects. Joumal of Applied Psychol-ogy, 75: 613-620.

Levinthal, D., & March, J. 1993. The myopia of learning.Strategic Management Joumal, 14: 95-112.

Liden, R. G., Sparrowe, R. T., & Wayne, S. J. 1997. Leader-member exchange theory: The past and potential forthe future. Research in Personnel and Human Re-sources Management, 15: 47-119.

Likert, R. 1967. The human organization: Its manage-ment and value. New York: McGraw-Hill.

Lind, E. A., & Tyler, T. R. 1988. The social psychology ofprocedural justice. New York: Springer.

Litsky, B. E., Eddleston, K. A., & Kidder, D. L. 2006. Thegood, the bad, and the misguided: How managersinadvertently encourage deviant behaviors. Acad-emy of Management Perspectives, 20(1): 91-103.

Lowe, K. B., Kroeck, K. G., & Sivasubramaniam, N.1996. Effectiveness correlates of transformationaland transactional leadership: A meta-anal5^c reviewof the MLQ literature. Leadership Quarterly,7: 385-425.

Maertz, G. P., & Griffeth, R. W. 2004. Eight voluntaryforces and motivational behavior. Joumal of Man-agement 30: 667-683.

March, J. G. 1991. Exploration and exploitation in organ-izational learning. Organization Science, 2: 71-86.

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

Masterson, S. S., Lewis, K., Goldman, B. M., & Taylor,M. S. 2000. Integrating justice and social exchange:The differing effects of fair procedures and treatmenton work relationships. Academy of ManagementJoumal, 43: 738-748.

Milliken, F. J., Morrison, E. W., & Hewlin, P. F. 2003. Anexploratory study of employee silence: Issues thatemployees don't communicate upward and why.Joumal of Management Studies, 40: 1453-1476.

Mintzberg, H. 1973. The nature of managerial work.New York: Harper & Row.

Mitchell, T. R., & James, L. R. 2001. Building better the-ory: Time and the specification of when things hap-pen. Academy of Management Review, 26: 530-547.

Molm, L. 1990. Structure, action, and outcomes: Thedynamics of power in social exchange. AmericanSociological Review, 55: 427-447.

Morrison, E. W. 2011. Voice and silence within organi-zations: Literature review and directions for futureresearch. In J. P. Walsh & A. P. Brief (Eds.), Academy

2013 McGlean, Burris, and Detert 547

of Management annals, vol. 5: 373-412. Essex, UK:Routledge.

Morrison, E. W., & Milliken, F. 2000. Organizational si-lence: A barrier to change and development in apluralistic world. Academy of Management Re-view, 25: 706-725.

Morrison, E. W., & Robinson, S. L. 1997. When employ-ees feel betrayed: A model of how psychologicalcontract violation develops. Academy of Manage-ment Review, 22: 226-256.

Moscovici, S., & Nemeth, G. 1974. Social psychology:Classic and contemporary integrations. Oxford,UK: Rand McNally.

Mueller, G., & Price, J. 1989. Some consequences of turn-over: A work unit analysis. Human Relations, 42:389-402.

Naumann, S. E., & Bennett, N. 2000. A case for proce-din-al justice climate: Development and test of a mul-tilevel model. Academy of Management Journal,43: 881-889.

Nemeth, G. 1997. Managing innovation: When less ismore. California Management Review, 40(1): 59-74.

Nohria, Ñ., & Gulati, R. 1996. Is slack good or bad forinnovation? Academy of Management Journal, 39:1245-1264.

Nonaka, I. 1988. Toward middle-up-down management.Sloan Management Review, 29(spring): 9-18.

Oi, W. 1962. Labor as a quasi-fixed factor. Journal ofPolitical Economy, 70: 538-555.

Organ, D. W., Podsakoff, P. M., & MacKenzie, S. B. 2006.Organizational citizenship behavior: Its nature,antecedents, and consequences. London: Sage.

Parilla, P. E., Hollinger, R. G., & Glark, J. P. 1988. Orga-nizational control of deviant behavior: The case ofemployee theft. Social Science Quarterly, 69: 261-280.

Pfeffer; J. 1981. Power in organizations. Marshfield, MA:Pitman.

Porter, L. W., & Steers, R. M. 1973. Organizational, work,and personal factors in employee turnover and ab-senteeism. Psychological Bulletin, 80: 151-176.

Preacher, K. J., Gurran, P. J., & Bauer, D. J. 2006. Compu-tational tools for probing interactions in multiplelinear regression, multilevel modeling, and latentcurve analysis. Journal of Educational and Behav-ioral Statistics, 31: 437-448.

Price, J. L. 1977. The study of turnover. Ames: Iowa StateUniversity Press.

Raudenhush, S. W., & Bryk, A. S. 2002. Hierarchicallinear models: Applications and data analysismethods. Newbury Park, GA: Sage.

Rhoades, L., & Eisenberger, R. 2002. Perceived organiza-tional support: A review of the literatinre. Journal ofApplied Psychology, 87: 698-714.

Robinson, S. L. 1996. Trust and breach of the psycholog-ical contract. Administrative Science Quarterly,41: 574-599.

Robinson, S. L., & Bermett, R. J. 1997. A typology ofdeviant workplace behaviors: A multidimensionalscaling study. Academy of Management Journal,38: 555-572.

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

Rousseau, D. M. 1995. Psychological contract in organ-izations: Understanding written and unwrittenagreements. Newbury Park, GA: Sage.

Rousseau, D. M., & McLean Parks, J.. 1993. The contractsof individuals and organizations. In L. L. Gummings& B. M. Staw (Eds.), Research in organizationalbehavior, vol. 15: 1-43. Greenwich, GT: JAI.

Rusbult, G. E., Farrell, D., Rogers, G., & Mainous, A. G.1988. Impact of exchange variables on exit, voice,loyalty, and neglect: An integrative model of re-sponses to declining job satisfaction. Academy ofManagement Journal, 31: 599-627.

Rusbult, G. E., Zembrodt, I. M., & Gunn, L. K. 1982. Exit,voice, loyalty, and neglect: Responses to dissatisfac-tion in romantic involvements. Journal of Personal-ity and Social Psychology, 43: 1230-1242.

Sagie, A., Birati, A., & Tziner, A. 2002. Assessing thecosts of behavioral and psychological withdrawal: Anew model and an empirical illustration. AppliedPsychology: An International Review, 51: 67-89.

Salancik, G. R., & Pfeffer, J. 1978. A social informationprocessing approach to job attitudes and task design.Administrative Science Quarterly, 23: 224-253.

Saimders, D. M., Sheppard, B. H., Knight, V., & Roth, J.1992. Employee voice to supervisors. Employee Re-sponsibility and Rights Journal, 3: 241-259.

Schyns, B., & Meindl, J. R. 2005. Implicit leadershiptheories: Essays and explorations. Greenwich, GT:IAP.

Shanock, L. R., & Eisenherger, R. 2006. When supervisorsfeel supported: Relationships with subordinates'perceived supervisor support, perceived organiza-tional support and performance. Journal of AppliedPsychology, 91: 689-695.

548 Academy of Management Journal April

Shapiro, D. L. 1993. Reconciling theoretical differencesamong procedural justice researchers by re-evaluat-ing what it means to have one's views "considered":Implications for third-party managers. In R. Gropan-zano (Ed.), Justice in the workplace: Approachingfairness in human resource mimagement: 51-78.New York: Erlbaum.

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.

Siegel, A. L., & Ruh, R. A. 1973. Job involvement, partic-ipation in decision making, personal backgroundand job behavior. Organizational Rehavior and Hu-man Performance, 9: 318-327.

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

Singh, J. V. 1986. Performance, slack, and risk taking inorganizational decision making. Academy of Man-agement Journal, 29: 562-585.

Skarlicki, D. P., & Folger, R. 1997. Retaliation in theworkplace: The roles of distributive, procedural, andinteractional justice. Journal of Applied Psychol-ogy, 82: 434-443.

Staw, B. M. 1980. The consequences of turnover. Journalof Occupational Rehavior, 1: 253-273.

Tangirala, S., & Rumanujam, R. 2008. Employee silenceon critical work issues: The cross level effects ofprocedural justice climate. Personnel Psychology,61: 37-68.

Tepper, B. 2000. Gonsequences of abusive supervision.Academy of Management Journal, 43: 178-190.

Thibaut, J., & Walker, L. 1975. Procedural justice: Apsychological analysis. Hillsdale, NJ: Erlbaum.

Trist, E. L., Higgins, G. W., Murray, H., & Pollock, A. B.. 1963. Organizational choice. London: Tavistock.

Turnley, W. H., Bolino, M. G., Lester, S. W., & Bloodgood,J. M. 2003. The impact of psychological contractfulfillment on performance of in-role and organiza-tional citizenship behaviors. Journal of Manage-ment, 29: 187-206.

Tyler, T. R. 1987. Gonditions leading to value expressiveeffect in judgments of procedural justice: A test of

four models. Journal of Personality and Social Psy-chology, 52: 333-344.

Van Dyne, L., Ang, S., & Botero, I. G. 2003. Conceptual-izing employee silence and employee voice as mul-tidimensional constructs. Journal of ManagementStudies, 40: 1359-1392.

Van Dyne, L., & LePine, J. A. 1998. Helping and voiceextra-role behavior: Evidence of construct and pre-dictive validity. Academy of Management Journal,41: 108-119.

Withey, M. J., & Cooper, W. H. 1989. Predicting exit,voice, loyalty, and neglect. Administrative ScienceQuarterly, 34: 521-539.

Yukl, G. 2002. Leadership in organizations. SaddleRiver, NJ: Prentice Hall.

Yukl, G., & Nemeroff, W. 1979. Identification and mea-surement of specific categories of leadership behav-ior: A progress report. In J. G. Hunt & L. L. Larson(Eds.), Crosscurrents in leadership: 164-200. Gar-bondale: Southern Illinois University Press.

Zuber, A. 2001. A career in foodservice—Cons: Highturnover. Nation's Restaurant News, 35: 147-148.

Elizabeth J. McClean (ejm45@cornell.edu) is a Ph.D. can-didate in human resource management and organiza-tional behavior at the School of Industrial and LaborRelations at Cornell University. Her current research fo-cuses on the antecedents and outcomes of employeevoice and the role of HR and leadership on employee Eindfirm outcomes.

Ethan R. Burris (ethan.burris@mccombs.utexas.edu) isan associate professor of management at the McCombsSchool of Business at the University of Texas at Austin.He received his Ph.D. from Cornell University. His cur-rent research focuses on understanding the antecedentsand consequences of employees speaking up or stayingsilent in organizations; leadership behaviors, processes,and outcomes; and the effective management of conflictgenerated by multiple perspectives.

James R. Detert (jdetert@cornell!edu) is an associate pro-fessor of management at the Johnson Graduate School ofManagement at Cornell University. His current researchinterests include voice and silence in organizations, lead-ership influences on voice and ethical decision making,and cognitive moral disengagement as a predictor of un-ethical behavior. He received his Ph.D. in organizationalbehavior from HarvcU'd University.

Copyright of Academy of Management Journal is the property of Academy of Management and its content may

not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written

permission. However, users may print, download, or email articles for individual use.