Cornell University ILR SchoolDigitalCommons@ILR

Faculty Publications - Human Resource Studies Human Resource Studies

7-1-2002

Employee Voice, Human Resource Practices, andQuit Rates: Evidence from theTelecommunications IndustryRosemary BattCornell University, rb41@cornell.edu

Alexander J.S. ColvinPennsylvania State University

Jeffrey KeefeRutgers University

This Article is brought to you for free and open access by the Human Resource Studies at DigitalCommons@ILR. It has been accepted for inclusion inFaculty Publications - Human Resource Studies by an authorized administrator of DigitalCommons@ILR. For more information, please contactjdd10@cornell.edu.

Batt, Rosemary; Colvin, Alexander J.S.; and Keefe, Jeffrey, " Employee Voice, Human Resource Practices, and Quit Rates: Evidencefrom the Telecommunications Industry" (2002). Faculty Publications - Human Resource Studies. Paper 4.http://digitalcommons.ilr.cornell.edu/hrpubs/4

573

Industrial and Labor Relations Review, Vol. 55, No. 4 (July 2002). © by Cornell University.0019-7939/00/5504 $01.00

H

EMPLOYEE VOICE, HUMAN RESOURCE

PRACTICES, AND QUIT RATES: EVIDENCE

FROM THE TELECOMMUNICATIONS INDUSTRY

ROSEMARY BATT, ALEXANDER J.S. COLVIN, and JEFFREY KEEFE*

The authors draw on strategic human resource and industrial relationstheories to identify the sets of employee voice mechanisms and human resourcepractices that are likely to predict firm-level quit rates, then empirically evaluatethe predictive power of these variables using data from a 1998 establishment-level survey in the telecommunications industry. With respect to alternativevoice mechanisms, they find that union representation predicts lower quit rates,even after they control for compensation and a wide range of other humanresource practices that may be affected by collective bargaining. Also predictinglower quit rates is employee participation in offline problem-solving groups andin self-directed teams. No apparent association is found between quit rates andthe availability of nonunion dispute resolution procedures. Regarding humanresource practices, higher relative wages and internal promotion policies pre-dict lower quit rates, and contingent staffing, electronic monitoring, and vari-able pay predict higher rates.

*Rosemary Batt is Associate Professor of HumanResource Studies, Industrial and Labor RelationsSchool, Cornell University; Alexander J.S. Colvin isAssistant Professor of Labor Studies and IndustrialRelations, Pennsylvania State University; and JeffreyKeefe is Associate Professor, School of Managementand Labor Relations, Rutgers University. This re-search is based on a multi-year study of the telecom-munications services industry conducted in collabo-ration with Harry Katz and generously funded by theAlfred P. Sloan Foundation. The authors thank HarryKatz for comments on prior versions of the draft andDanielle van Jaarsveld for her careful research assis-tance.

A data appendix with additional results, and cop-ies of the computer programs used to generate theresults presented in the paper, are available from thefirst author at the New York State School of Industrialand Labor Relations, 387B Ives Hall, Cornell Univer-sity, Ithaca, NY 14853.

igh quit rates are a significant cost toorganizations, both because they raise

labor costs (Oi 1962) and because theylower organizational performance (for ex-ample, Norsworthy and Zabala 1986). Re-search in industrial relations has shownthat unionized establishments have signifi-

cantly lower quit rates than nonunion es-tablishments because they provide a voicemechanism through which employees cannegotiate higher relative compensation andredress problems as an alternative to exit(Freeman and Medoff 1984). More recentresearch on “high commitment” or “highperformance” work systems has shown thatcoherent sets of human resource practiceslead to lower quit rates (for example, Arthur1994) and better organizational perfor-mance (for example, Ichniowski et al. 1996),and that high quit rates undermine perfor-

574 INDUSTRIAL AND LABOR RELATIONS REVIEW

mance (Alexander, Bloom, and Nuchols1994; Huselid 1995). High performancepractices include those that invest in theskills of the work force and provide theopportunity and incentives for employeesto use those skills effectively (for example,Appelbaum et al. 2000).

Over the past decade or so, however,firms have adopted a wide range of non-union voice mechanisms and cost-cuttinghuman resource practices that have notbeen considered in the industrial relationsor high performance literature. Alterna-tive voice mechanisms include differenttypes of nonunion dispute resolution pro-cedures as well as team-based work organi-zation. Cost-cutting human resource (HR)practices include downsizing and contin-gent staffing and pay, which are likely todecrease employee commitment to the firm.

In this study, we examine a comprehen-sive set of voice and human resource prac-tices that are likely to influence employeequit rates. We consider quit rates at theestablishment level rather than at the indi-vidual level of analysis so that we can iden-tify the specific management practices thatinfluence quit rates—a subject of particu-lar interest in periods of tight labor mar-kets. In contrast to the high performanceliterature, which generally examines clus-ters of HR practices taken together, weexamine individual practices in order todisaggregate the unique and potentiallycontradictory relationships between spe-cific management practices and quit rates.We consider these issues among a relativelyunder-studied group of workers: moder-ately and highly skilled service, sales, andtechnical employees. To do so, we examinea unique 1998 data set on a nationally rep-resentative sample of union and nonunionestablishments in the telecommunicationsindustry.

Voice Mechanisms asDeterrents to Turnover

The exit-voice framework (Hirschman1970) identifies a range of circumstancesin which markets may fail to provide orga-nizations with effective feedback and in

which customer or employee voice mightlead to improved organizational perfor-mance. Voice, as defined by Hirschman, isany attempt to change rather than escapefrom an unsatisfactory situation. In theemployment context, voice involves theexpression of dissatisfaction by employees;exit occurs when employees quit. The de-sign of work and human resource practicesis likely to influence the balance betweenemployee voice and exit. The exit-voiceframework provides a model of the politi-cal economy of organizations, where exitbelongs to the realm of economics andvoice to the realm of politics.

In recent years, employers have adoptedalternative forms of voice, both as part ofhigh performance work systems and as sub-stitutes for unions. Employers have viewedteam-based work systems as providing op-portunities for collective voice or directparticipation in shop-floor decisions, andnonunion dispute resolution proceduresas providing opportunities for individualvoice to redress employee grievances. Inthe following sections, we consider the theo-retical importance of these alternativemechanisms as predictors of quit rates.

Unions as Collective Voice

In the most important application of theexit-voice model to employment, Freemanand Medoff (1984) provided evidence link-ing union voice to higher productivity andlower quit rates. Unions reduce the prob-ability that workers will quit their jobs fortwo fundamental reasons. First, unionsprovide a voice mechanism through whichworkers gain higher compensation thanthey could earn in a similar nonunion job.Second, unions provide workers with a voicein determining an array of other rules andconditions of work, including policies thatreduce pay inequality, grievance and arbi-tration procedures for appealing manage-rial decisions, “just cause” for disciplineand discharge, and seniority clauses de-sired by workers.

Empirical research supports the view thatunion mechanisms affecting both wagesand voice lead to lower quit rates in union

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 575

establishments. Freeman (1980a), for ex-ample, found that union establishmentshad lower quit rates even when his analysiscontrolled for wage rates. Other research-ers have replicated this finding in the UnitedStates (for example, see Cotton and Tuttle1986; Wilson and Peel 1991) as well as inother countries (Miller and Mulvey 1991;Lincoln and Kalleberg 1996).

Recent research, however, has chal-lenged this general view. In a study oftruckers in the 1990s, for example, Deleryet al. (2000) found that the statistically sig-nificant relationship between quit rates andunionization disappeared when wages andbenefits were taken into account. They con-cluded that the union effect operated en-tirely through compensation. They sug-gested that union decline may be affectingunions’ ability to influence employee voiceand quit behavior, but they admitted thattheir findings are industry-specific andcould be accounted for by the independentnature of truck drivers’ jobs and their greaterattachment to their occupation than to theiremployers. Thus, Delery et al. (2000) re-open the union voice debate by question-ing whether unions in the 1990s were strongor effective enough as voice mechanisms toinfluence employee quit behavior.

Another stream of recent research hasfound that at the individual level, employeeuse of voice mechanisms is positively, notnegatively, correlated with employee in-tent to quit and subsequent turnover (Lewinand Boroff 1996; Boroff and Lewin 1997;Lewin and Peterson 1999). This latter bodyof research suggests a distinction not rec-ognized in much of the standard exit-voicetheory: whereas quit rates may be lower inenterprises where stronger voice mecha-nisms are available, among enterprises thathave similar voice mechanisms available toemployees, greater usage of these voicemechanisms may be associated with higherquit rates.

Teams as Collective Voice

In the high performance literature, team-based work systems generally are viewed asimproving performance by providing work-

ers (those closest to the point of produc-tion) with the opportunity to collaborate insolving problems and improving the pro-duction process. Much less attention hasbeen paid to teams as voice mechanismsthat allow dissatisfied workers to voice theirconcerns, and thereby reduce the likeli-hood that they will quit. Those studies thathave found an inverse relationship betweenhigh performance practices and quit rateshave included team participation only aspart of a cluster of practices (for example,Arthur 1994; Huselid 1995). Other studiesthat have examined individual HR prac-tices and quit rates have found mixed re-sults. For example, Delery et al. (2000)found no relationship between employeeparticipation and quit rates in the truckingindustry (see also Shaw et al. 1998).

A large body of research in organiza-tional behavior, however, has found thatemployees with “enhanced or enrichedjobs” (for example, those with greater au-tonomy, variety, or ability to complete awhole task) have higher job satisfactionand lower turnover (Hackman and Oldham1980; McEvoy and Cascio 1985; Cotton andTuttle 1986; Hom and Griffeth 1995). Thequestion, therefore, is whether the oppor-tunity for voice provided by different typesof team formations is sufficient to reducethe propensity to quit. If teams are merelysupervised groups with little discretion toaffect conditions at work, then their influ-ence on quits will be minimal.

In general, researchers have distin-guished between two types of team partici-pation: consultative and substantive (Levineand Tyson 1990). Consultative forms in-clude participation in problem-solvinggroups that meet periodically and allowemployees to voice their opinions to man-agement. Substantive forms include theorganization of work into self-managed orsemi-autonomous teams that have consid-erable responsibility for making day-to-daydecisions without consulting management.

Empirical research in the teams or groupeffectiveness literature provides mixed re-sults on whether these different forms ofteam voice lower turnover. Consultativeforms have a significant but modest posi-

576 INDUSTRIAL AND LABOR RELATIONS REVIEW

tive relationship to employee behavior andperformance (Cotton 1993; Cohen andBailey 1997). Substantive participation inself-managed teams has a significant, andappreciably greater, positive relationshipto a range of employee behaviors and per-formance, but specific studies of turnoverhave yielded mixed results. For example,Cordery et al. (1991) found that autono-mous teams have higher turnover rates thannon-autonomous teams. Weisman et al.(1993) found that nurses in self-managedteams worked longer hours (increasingturnover), but had higher pay (decreasingturnover). Other studies have found thatalternative forms of voice can reduce em-ployee turnover (Bemmels 1997). For ex-ample, Spencer (1986) found a negativerelationship between turnover rates andnumber of voice mechanisms (such as griev-ance procedures, survey feedback, and sug-gestion systems). In sum, theory suggeststhat team participation programs may pro-vide employees with enough voice to re-duce turnover, but the research record ismixed.

Nonunion DisputeResolution Procedures: EmployeeVoice or Voice Suppression?

Employers also have designed a varietyof dispute resolution procedures in non-union settings, including procedures formanagement review of grievances, peerreview, and nonunion arbitration, that mayprovide an individualized form of voice inthe workplace and thereby reduce quit rates.There is no empirical research on the rela-tive effectiveness of these alternative proce-dures for reducing quit rates, but studies ofvariation in union grievance procedureshave found that procedures providing rela-tively stronger protections for workers areassociated with lower quit rates. Rees(1991), for example, examined collectivelybargained grievance procedures and foundthat relatively stronger procedures moreeffectively reduced quit rates than didweaker procedures. In Delery et al.’s truck-ing study (2000), formal grievance proce-dures were negatively related to quit rates,

but the relationship became statisticallyinsignificant in regression equations thatcontrolled for unionization. Their mea-sure of grievance procedures, however, didnot distinguish between weaker and stron-ger forms of procedures. We seek to under-stand which types of nonunion dispute pro-cedures provide workers with enough voiceto reduce quit rates.

Many nonunion dispute resolution pro-cedures provide a formal structure throughwhich employees can appeal disputes withsupervisors to higher levels of management.These may range from simple proceduresin which an individual manager reviewswritten complaints to more elaborate pro-cedures involving hearings of complaintsbefore appeal boards composed of senior-level managers. Under these types of non-union procedures, managers hold decision-making authority (Ewing 1989; Feuille andDelaney 1992). Under peer review proce-dures, by contrast, employee complaintsare evaluated by a panel composed pre-dominantly of the complainant’s peers.Presumably, employees are more willing totrust the effectiveness of a procedure inwhich their fellow employees decide griev-ances (Ewing 1989; Colvin 1999). How-ever, as employees of the company, peerreview panelists are not comparable to thirdparty neutrals used in arbitration or media-tion procedures. Also, peer review deci-sions lack the legal enforceability of uniongrievance-arbitration procedures, and thusdepend on the good faith of management.

Nonunion arbitration emerged duringthe 1990s as the most prominent and rap-idly expanding variety of nonunion disputeresolution procedure using non-manage-rial decision-makers. An important dangerposed by these individual employee arbi-tration agreements is their potential to sup-press employee voice. In Gilmer v. Inter-state/Johnson Lane, 500 U.S. 20 (1991), theSupreme Court deferred to employer-spon-sored arbitration procedures to resolve dis-putes over statutory rights. Individual em-ployee arbitration contracts often are com-pulsory because employees must agree toarbitration as a condition of employment atthe time of hire or, in some instances, as a

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 577

condition of continuing employment orfuture promotions or benefits (Stone 1996,1999). They require that an employee sub-mit any alleged violation of state or federallaw to a company-designed arbitration pro-cedure. By agreeing to arbitration, theemployee forgoes any opportunity to pur-sue his or her claim in court. Both thedesign of nonunion arbitration proceduresand the decisions of arbitrators are heavilyinsulated from judicial review (Stone 1996,1999). Whether this innovation enhancesor suppresses employee voice is at the cen-ter of a major controversy. Some observerssuggest that compulsory arbitration createsopportunities for due process in nonunionworkplaces (Zack 1999), while others sug-gest it represents a significant curtailmentof employee rights (Stone 1996, 1999).Current research suggests that the moreeffective employees perceive the grievanceprocedure to be, the more likely they willexercise the voice option rather than quit(Boroff and Lewin 1997). If employeesperceive nonunion arbitration to be di-rected primarily at limiting their statutoryemployment rights, they may be less likelyto use these procedures as mechanisms forvoice and more likely to quit.

Predictions about the relationship be-tween dispute resolution procedure usagerates and employee quit rates are uncer-tain. Higher usage rates may indicate thatemployees view nonunion procedures aseffective voice mechanisms. If so, then wewould expect higher usage rates to be asso-ciated with lower quit rates. However, ex-tensive research on unionized workplacesindicates that higher grievance rates areassociated with higher levels of workplaceconflict and reduced organizational per-formance (Katz, Kochan, and Gobeille 1983;Katz, Kochan, and Weber 1985; Norsworthyand Zabala 1986; Ichniowski 1986; Cutcher-Gershenfeld 1991). These higher levels ofworkplace conflict associated with highergrievance rates also may raise quit rates. Apositive relationship between grievancerates and quit rates is also suggested byresearch indicating that individuals whofile grievances are more likely to quit thanare other employees (Lewin 1987; Lewin

1990; Lewin and Peterson 1999). Theseindividual-level studies have focused pri-marily on variation in usage of voice mecha-nisms, rather than on variation in the pres-ence of voice mechanisms. This is in partbecause previous research has concentratedon unionized workplaces, which have lessvariation in grievance procedures thannonunion workplaces do, and hence hasnot examined the effect of simultaneousvariation in both the type of procedure andthe usage or dispute rate (Lewin 1999).The greater variation in the incidence andstructure of dispute resolution proceduresin the nonunion sector (Feuille and Delaney1992; Colvin 1999) permits examination ofthe question of how variation both in thetype of dispute resolution procedure andin dispute rates affects aggregate quit rates.Our prediction is that greater availability ofdispute resolution procedures to employ-ees should be associated with lower quitrates, but that, holding the availability ofprocedures constant, higher grievance ratesshould be associated with higher quit rates.

Human ResourcePractices and Quit Rates

Historically, the design of internal labormarkets encouraged long-term employ-ment, fostered by a set of interrelated hu-man resource policies designed to reducequit rates (Doeringer and Piore 1971;Schacht 1985; Jacoby 1985). Internal lad-ders provided a series of promotional stepsthat continually expanded the employees’firm-specific skill sets and rewarded themwith higher pay at each step. In exchange,employers gained a loyal and stable workforce with high levels of firm-specific skills,amortizing the quasi-fixed cost of employ-ment over many years of employee tenure.The high performance literature drawsheavily on internal labor market theory(emphasizing investment in skills and train-ing and long-term employment relations);but it also includes newer practices such asteams and performance-based pay (for ex-ample, Osterman 1994b).

Empirical research does show that com-mitment-enhancing internal labor market

578 INDUSTRIAL AND LABOR RELATIONS REVIEW

practices reduce quit rates. Firm-specifictraining (Miller and Mulvey 1991) andhigher pay (Leonard 1987; Powell, Mont-gomery, and Cosgrove 1994; Shaw et al.1998; Delery et al. 2000) are associated withlower quit rates. Cappelli and Chauvin(1991) found that wage premiums or highunemployment in the outside market re-sulted in greater reliance on grievance pro-cedures—presumably because the exit al-ternative was more costly. Meta-analyses ofresearch in organizational behavior alsofind support for internal labor markettheory (for example, Cotton and Tuttle1986; Hom and Griffeth 1995:40), withlower turnover among employees who aresatisfied with promotion opportunities,perceive their employment to be secure, orhave higher relative pay or pay satisfaction.Finally, studies of high performance sys-tems have shown that turnover is lower inworkplaces that adopt a cluster of highcommitment practices (Arthur 1994;Huselid 1995).

Over the last decade or so, however,firms have been dismantling internal labormarkets (for example, Cappelli 1999;Osterman 1999) in an attempt to enhancelabor flexibility or their ability to quicklyadjust labor costs to variation in productmarket conditions. The use of cost-cuttingHR practices (including downsizing, sub-contracting and outsourcing, contingentand part-time staffing, and contingent pay)has introduced labor market competitioninside organizations. The result is thatworkplaces often have a mix of HR prac-tices that provide contradictory incentivesto workers—commitment-enhancing invest-ments in training on the one hand, butgreater reliance on contingent staffing andpay, on the other. This use of contradictoryHR practices is at odds with the concept ofcoherence or “bundles” of consistent prac-tices found in the high performance litera-ture (MacDuffie 1995), and suggests thatHR practices must be examined individu-ally rather than as clusters or bundles.

Some research is emerging that showsthe effects of cost-cutting HR practices onthe behavior of “core” employees—thosewho are regular employees and critical to

the business enterprise. Firms thatdownsize, for example, often do so again(for example, Wyatt 1993), and therebysend a signal to their core employees thatfuture employment is insecure. In response,the more qualified or skilled employeeswho have alternative employment opportu-nities are likely to quit. Use of contingentand part-time employees may be used tobuffer core employees from job displace-ment; but recent studies show that firmsoften use contingent staffing as a cost-cut-ting strategy (Houseman 2001). In thesecases, firms send a signal to core employeesthat their jobs are insecure because futurecost-cutting may lead to converting morecore jobs to contingent positions. Again,the more qualified or skilled employees arelikely to leave. In the 1990s, firms oftendownsized and withdrew employment secu-rity commitments while simultaneously us-ing more contingent staffing.

The availability of temporary workers assubstitutes for core employees also putsdownward pressure on the wages of incum-bent employees. Katz and Krueger (1999),for example, in an analysis controlling forunemployment, found that states with agreater share of temporary help employ-ment experienced lower wage growth. Theyargued that the growth of the temporaryhelp industry and other contingent arrange-ments facilitated wage restraint by increas-ing the ability of firms to locate substituteworkers. Similarly, establishments that userelatively large numbers of temporary work-ers as substitutes for core employees,thereby putting pressure on the core em-ployees to accept wage restraint, may findthat the wages of core employees fall be-hind those of competing employers in theexternal labor market. Thus, by using con-tingent labor contracts to hold down laborcosts, employers unintentionally may befacilitating the exit of their core employ-ees.

In conjunction with downsizing and con-tingent staffing, some employers havesought to reduce labor costs by standardiz-ing and simplifying job tasks. This re-engi-neering of tasks reduces training costs andpermits electronic monitoring that saves

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 579

on supervision and surfaces performanceproblems using objective measures. In thecontext of the telecommunications indus-try in this study, advanced information sys-tems have provided employers with a majortool for labor cost reduction through taskstandardization and continuous electronicmonitoring. Prior research has demon-strated that such standardization and sim-plification of jobs and electronic monitor-ing is associated with higher turnover (Wil-son and Peel 1991; Carayon 1993; Shaw etal. 1998).

The greater use of contingent pay mayaccelerate quit rates because performance-based pay plans increase the variability ofworkers’ pay, and thus their financial un-certainty. Performance-based pay linksworkers’ pay more closely to variability inmarket conditions, over which workers havelittle control. Given the relatively risk-aversepreferences reflected in the pay plans bar-gained by unions representing workers(Freeman 1980b and 1982), it is unlikelythat variable pay plans have been intro-duced at workers’ request. Most workersrely financially on their paycheck alone,and lack other sources of income or a diver-sified portfolio to hedge risk. Moreover,most variable pay plans are not accompa-nied by a compensating wage increase forthe greater risk borne by workers.

Expected Findings

In summary, in this paper we present amodel of the predictors of quit rates thatincludes alternative voice mechanisms andalternative human resource practices (seeFigure 1). We hypothesize that union pres-ence will have a negative relationship withquit rates, over and above the effects of theunion wage and other voice and humanresource practices. We hypothesize thatthe team-based voice mechanisms will bemodestly negatively related to quit rates.The prediction for nonunion dispute reso-lution procedures is uncertain, as they maybe designed to encourage or to suppressemployee voice. More effective procedures(for example, peer review panels) shouldbe able to reduce quit rates. Holding the

procedure constant, high grievance activityrates should be positively associated withquit rates because they signal higher levelsof conflict and greater numbers of prob-lems faced by employees in the workplace.

Finally, quit rates should be lower inworkplaces that increase the cost of exitthrough commitment-enhancing humanresource practices such as investment intraining, internal mobility opportunities,and high relative pay. Quit rates should behigher, by contrast, in workplaces that haveimplemented cost-cutting human resourcepractices—including downsizing, the useof contingent staffing, job standardizationand electronic monitoring, and variablepay—each of which signals a lack of em-ployer commitment to employee welfare orlong-term employment relations. We ex-pect these relationships to hold after wecontrol for market, organizational, anddemographic characteristics, which we dis-cuss in the methods section below.

Methods

Sample

The sample is a stratified random sampledrawn from the Dun and Bradstreet listingof establishments. Establishments werestratified by size (10–99 employees, 100-plus employees), by SIC code (4812, cellu-lar; 4813, wireline; 4841, cable), and bystate location. Sixty percent of the sampleis in the wireline segment, 15% in cabletelevision, 14.6% in cellular, and 6.9% inInternet services. All establishments withmore than 100 employees were sampled sothat the survey would cover a large percent-age of the industry’s work force. Samplingof the remaining smaller establishmentswas done so that the total sample reflectsthe relative proportion of establishmentsin each industry segment of the Dun andBradstreet listing. Because Internet serviceproviders (ISPs) are new and not systemati-cally captured by SIC code, additional ISPswere identified through the Directory ofNational Dial-up Providers and Area Codesof Operation.

In the fall 1998, a university-based survey

580 INDUSTRIAL AND LABOR RELATIONS REVIEW

team administered an initial telephone sur-vey covering questions related to basic in-dustry characteristics, management strate-gies, and work and human resource prac-tices. The general manager at each estab-lishment was the respondent. The tele-phone interview averaged 52 minutes, andyielded a 54% response rate with 636 us-able surveys. Respondents were asked ifthey were willing to take part in an addi-tional survey on dispute resolution proce-dures. This yielded a sample of 302 respon-dents, divided about equally between thetwo core sides of the business: networkoperations and service and sales.

Measures and Means

We used several techniques to reducemeasurement problems found in workplacestudies. Because human resource practicesvary by organizational unit and occupa-tional subgroup (for example, Osterman1987; Jackson et al. 1989), survey questionsasked managers about their “core” non-management group of workers: either net-

work technicians or service and sales work-ers. This strategy follows that of Osterman(1994a), among others. To reduce prob-lems impairing validity, we used only gen-eral managers as respondents, as some re-search shows that general managers are lessoptimistic than HR managers in their as-sessments of HR practices (Gerhart 1999).Moreover, general managers oversawworksites that were relatively small, with anaverage of 117 core employees per site inthe database.

Another way to improve validity and reli-ability is to base survey questions on fieldresearch and to frame questions in lan-guage that is context-specific. We con-ducted field research in numerous networkoperations and service and sales centers ina range of markets (local, long distance,cellular, cable) and customer segments(residential, small business, middle mar-ket). Our survey questions were based onquestions from our field interviews thatmanagers were readily able to answer with-out consulting other documentation. Insite visits, we checked these general manag-

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 581

ers’ responses against those of subject mat-ter experts in the same organization. Also,we checked the reliability of survey re-sponses against outside data, including theDun and Bradstreet listing, union contractsin specific companies, and the CurrentPopulation Survey (CPS).1

The primary dependent variable in thisstudy is the current annual quit rate, basedon the question, “As a percentage, what isyour annual voluntary quit rate among [net-work technicians or customer service andsales employees] (excluding discharges,retirements, transfers, and promotions)?”This definition differs from definitions usedin most prior studies, which have collapsedvoluntary and involuntary turnover (forexample, Bennett et al. 1993; Alexander etal. 1994; Arthur 1994; Huselid 1995). Itfollows the measure of quit rates used byWilson and Peel (1991), Powell et al. (1994),and Shaw et al. (1998), the last of whomshowed that the determinants of dischargesand quits are distinct. The independentand control variables fall into four catego-ries: employee voice mechanisms; cost-cutting human resource practices; commit-ment-enhancing HR practices; and con-trols for organizational and work force char-acteristics.

Table 1 reports the means, standard de-viations, number of observations, andranges for all variables used in the study.The average annual quit rate in this sampleis 11%. For collective voice, union pres-ence is captured by a dummy variable(where 1 = union, 0 = no union) becausethe core work force is defined narrowlyenough to fall completely within one bar-gaining unit. Twenty-four percent of theestablishments are union-represented. Forteam participation, we measured consulta-tive participation (whether the establish-ment makes use of off-line problem-solvinggroups) as the percentage of employees inthe establishment participating in prob-lem-solving teams. We measured substan-tive participation (whether the establish-ment makes use of self-directed teams) asthe proportion of the establishment’s em-ployees organized in self-directed teams.In the average establishment in this survey,46% of employees were in problem-solvinggroups and 19% were in self-directed workteams.

Dispute resolution procedures coveringnonunion core work forces are captured bythree dummy variables representing, re-spectively, any type of formal nonuniongrievance procedure, nonunion arbitration,and peer review panels. Usage of theseprocedures is measured by a single variableconsisting of the annual number of griev-ances or complaints per employee broughtunder the applicable union or nonunionprocedure. The average annual number ofdisputes was six per 100 employees. Due tothe length of the survey and its cost, only arandom subsample of 302 of the respon-dents answered the more detailed set ofquestions about their dispute resolutionprocedures.

To capture cost-cutting human resourcepractices, we included five items:downsizing, electronic monitoring, part-time workers, temporary employees, andvariable pay. Downsizing is measured bytaking the number of core employees dis-placed in the past five years as a percentageof the current core work force. Marketderegulation, technology advances, andmergers have led to ongoing downsizing

1First, we compared our survey items to data con-tained in the Dun and Bradstreet listing. There werethree variables (date establishment founded, primarymarket, and size) common to both data sets, and themeans for each variable in the two data sets were notstatistically significantly different. Second, we exam-ined union contracts and found that the wage ratesand job titles reported by managers at specific Bellcompanies were comparable to those in the corre-sponding union contracts. Third, we compared oursurvey data to data in the Current Population Surveyfor the telecommunications industry (CPS 1998). Forexample, for customer service and sales workers, themost comparable group in the CPS is clerical andsales workers. The 1998 median annual pay in theCPS is $31,200 for union clerical and sales workersand $26,000 for nonunion workers, compared to$33,000 for union workers and $28,000 in our sample.The 1998 CPS unionization rate is 32%, while that ofour sample is 38%. Given the fact that we over-sampled large establishments that tend to pay higherwages and to be more heavily unionized, these com-parisons are reasonably consistent.

582 INDUSTRIAL AND LABOR RELATIONS REVIEW

and restructuring in this industry. Elec-tronic monitoring is the percentage of atypical employee’s daily work time that iselectronically monitored; on average, thisvalue is 33%. Contingent staffing includesthe percentage of the work force that istemporary (on average, 4%) and part-time(on average, 5%), as opposed to perma-nent and full-time. Variable pay is thepercentage of pay of the typical core em-ployee that is variable—on average, 16%.

For commitment-enhancing HR prac-tices, we measured three dimensions: train-ing, internal mobility opportunities, andpay. Training is the number of weeks oftraining received by the typical core em-ployee, which is two weeks on average inthis sample. Our measure of internal mo-bility opportunities is the percentage ofcore employees who were promoted fromwithin the company or transferred fromother departments or business units within

Table 1. Telecommunications Survey Data:Means, Observations, Standard Deviations, and Ranges.

Variable Name Obs. Mean Std. Dev. Min. Max.

Annual Quit Rate %a 598 0.108 0.164 0 1

Employee VoiceUnion % 635 0.241 0.428 0 1

Teams:Problem-Solving Groups % 632 0.462 0.385 0 1Self-Directed Teams % 632 0.193 0.341 0 1

Dispute Resolution Procedures:Any Nonunion Procedure % 300 0.357 0.480 0 1Nonunion Arbitration % 291 0.162 0.369 0 1Peer Review % 294 0.160 0.367 0 1Dispute Rate per Employee 285 0.061 0.184 0 2

Cost-Cutting HR PracticesDownsizing Rate in Last Five Years % 633 0.092 0.011 0 0.261Part-Time Workers % 630 0.053 0.135 0 1Temporary Workers % 611 0.041 0.137 0 1Electronic Monitoring % 622 0.326 0.405 0 1Variable Pay % 616 0.158 0.530 0 1

Commitment-Enhancing HR PracticesMobility % 611 0.438 0.367 0 1Training (Days per Year) 636 2.025 1.652 0 10Ratio: Pay to Local Cost of Living 568 0.732 0.356 0.233 3.355

Control VariablesFemale % 625 0.401 0.360 0 1Ln of Establishment Employmentb 634 4.124 1.600 0 10.959Branch % 636 0.794 0.405 0 1Former Bell Company % 636 0.264 0.441 0 1Human Resource Department % 633 0.428 0.495 0 1College Graduates % 622 0.220 0.415 0 1Field Technicians % 636 0.275 0.447 0 1Office Technicians % 636 0.168 0.374 0 1Service and Sales Workers % 636 0.557 0.497 0 1

aAll variables listed as % are scaled from 0 to 1 for consistency. The mean for these variables should bemultiplied by 100 to get the actual percentage.

bEstablishment employment is defined as number of full-time, permanent employees; minimum log ofemployment 0 because some establishments have only one full-time employee.

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 583

the company; the average is 44%. We in-clude lateral as well as vertical opportuni-ties because lateral opportunities often arevalued by employees—for example, formoving to preferred jobs at the same pay,for gaining broader experience for subse-quent promotions, or for more convenientwork locations closer to home. Pay is theratio of median pay to the local cost ofliving. Median pay (the pay received by thetypical core worker) was provided by surveyrespondents. We divided this pay level bythe local cost of living, drawn from theEconomic Research Institute’s (ERI’s) Geo-graphic Reference Report, which provides costof living data for 3,568 North Americancities using the “Urban Family of Four”model of analysis. The average ratio ofworker pay to local cost of living ($35,040/$48,000) is 0.73.

Four variables control for variation inorganizational characteristics: size (thenatural log of the total number of employ-ees—4.12, on average, across these estab-lishments); branch location (whether theestablishment is part of a larger organiza-tion—79%); Bell company (whether theestablishment is owned by a former Bellcompany—26%); and presence of an HRdepartment at the establishment (43%).Prior research has shown that these orga-nizational characteristics are associatedwith long-term employment contracts(Rebitzer 1986; Osterman 1994a; Jacoby1985).

We also controlled for human capitalcharacteristics that are known to influenceemployees’ labor market power and exter-nal opportunities. Researchers, for ex-ample, have found statistically significantdifferences in employment tenure by gen-der, although not when other factors suchas occupation or family status are takeninto consideration (Osterman 1987; Homand Griffeth 1995). We measure femalecomposition as the percentage of the corework force that is female, which is 40% onaverage. Education also has been found tobe associated with higher turnover (Homand Griffeth 1995:38). In this study, weincluded a dummy variable for college(where 1 = college degree) because the

major educational dividing line is whetheremployees have a college degree. Twenty-two percent of respondents reported thatthe typical core employee has a collegedegree. Occupations are captured by aseries of three dummy variables: whetherthe employee is a service and sales repre-sentative (the omitted category), an officetechnician (17%), or a field technician(28%).

Results

In this section, we first report the resultsof our analysis of voice mechanisms andhuman resource practices using the fulldata set (Tables 2 and 3). Then, becausethe alternative dispute resolution variablesrely on a subset of the data, we report thoseresults separately (Table 4).

Table 2 reports the correlation matrix.The union and team voice variables aresignificantly correlated with lower quit rates,but the nonunion procedures and disputerates are not. Each of the cost-cuttinghuman resource practices, except electronicmonitoring, is associated with higher quitrates. Internal labor market practices thatare associated with lower quit rates at statis-tically significant levels are internal mobil-ity opportunities and high relative pay; train-ing is not correlated with quits. Some con-trol variables have a statistically significantpositive association with higher quit rates,including the percentage of the work forcethat is female, having a human resourcedepartment on site, and the percentage ofthe work force with college degrees. Othercontrols, including the percentage of thework force that is field technicians andwhether the establishment is part of aformer Bell company, are correlated withlower quit rates.

Table 3 presents the results from Tobitanalyses in which the dependent variable isregressed on four sets of independent vari-ables: employee (union and team) voice,cost-cutting human resource practices, com-mitment-enhancing HR practices, and thefull model, each with the appropriate con-trols. We used a Tobit procedure becausethe dependent variable Quit Rate is trun-

584 INDUSTRIAL AND LABOR RELATIONS REVIEW

cated on the left end of the distribution atzero (Maddala 1992)—no establishments,obviously, have fewer than zero annual quits.Tobit models correct for biasing in thecoefficient estimates due to the truncationof the dependent variable; however, inter-

pretation of the coefficients also is differ-ent under a Tobit model. McDonald andMoffit (1980) suggested a decompositionof the Tobit coefficients into changes in theprobability of observing an outcome abovethe left limit and changes in outcomes above

Table 2. Telecommunications Survey Data: Correlations.

1 2 3 4 5 6 7 8

1 Quit Rate 1.000 2 Union –0.244*** 1.000 3 Problem-Solving Groups % –0.105** –0.253*** 1.000 4 Self-Directed Teams % –0.104** –0.089* 0.261*** 1.000 5 Downsizing in Last 5 Yrs. 0.202*** –0.038 0.046 –0.029 1.000 6 Part-Time Workers % 0.266*** –0.150*** –0.026 0.009 0.101 1.000 7 Electronic Monitoring % 0.060 0.164*** –0.002 0.026 –0.037 0.122*** 1.000 8 Variable Pay % 0.131*** –0.251*** 0.166*** 0.062 0.015 0.039 –0.137*** 1.000 9 Temporary Workers % 0.171*** –0.035 –0.017 0.036 –0.013 0.057 –0.046 –0.05010 Mobility % –0.169*** 0.124*** –0.012 0.096** 0.048 –0.101** –0.053 –0.083*11 Training –0.039 –0.010 0.116* 0.081** –0.045 –0.078* 0.035 0.110***12 Pay to Cost of Living –0.140*** 0.035 0.072* 0.106** –0.069 –0.244*** –0.264*** 0.268***13 Any Nonunion Procedure 0.076 –0.480*** 0.051 0.026 –0.021 0.011 –0.053 0.00114 Nonunion Arbitration –0.018 –0.289*** 0.175*** 0.145** –0.050 –0.045 –0.066 0.07615 Peer Review –0.033 –0.285*** 0.089 0.196*** 0.044 0.030 0.003 0.10116 Dispute Rate per Employee 0.032 0.212*** –0.012 –0.066 0.004 –0.080 0.039 –0.10617 Female % 0.160*** –0.091** 0.084** –0.101** 0.076* 0.204*** 0.179*** 0.103*18 Ln Estab. Employment 0.003 0.353*** –0.205*** –0.140*** –0.078* –0.081** 0.238*** –0.147***19 Branch –0.007 0.242*** –0.196*** –0.090** –0.081** –0.116*** 0.130*** –0.00620 Former Bell Company –0.086** 0.567*** –0.175*** –0.069 0.051 –0.088** 0.109*** –0.099*21 HR Dept. on Premise 0.109*** 0.115*** –0.069* –0.046 –0.048 0.023 0.248*** –0.05622 College Graduate 0.084** –0.270*** 0.159*** 0.102** –0.028 –0.121*** –0.204*** 0.371***23 Field Technicians –0.216*** 0.295*** –0.187*** –0.013 –0.028 –0.197*** –0.046 –0.231***24 Office Technicians –0.054 –0.035 0.041 0.114*** –0.022 –0.073* –0.077 –0.122***

9 10 11 12 13 14 15 16

9 Temporary Workers % 1.00010 Mobility % –0.088* 1.00011 Training –0.019 0.008 1.00012 Pay to Cost of Living –0.047 0.068 0.064 1.00013 Any Nonunion Procedure 0.000 0.018 0.001 –0.022 1.00014 Nonunion Arbitration –0.053 0.073 0.171*** 0.028 0.269*** 1.00015 Peer Review –0.094 0.050 0.128* 0.021 0.419*** 0.361*** 1.00016 Dispute Rate 0.194*** –0.031 –0.025 –0.001 0.058 –0.077 –0.013 1.00017 Female % –0.086* –0.188*** –0.006 –0.294*** 0.070 0.078 0.003 –0.105*18 Ln Estab. Employment –0.067* 0.068* 0.017 0.028 –0.012 –0.097* –0.074 0.01019 Branch 0.034 0.002 0.077* 0.118*** 0.014 –0.091 –0.089 0.123*20 Former Bell Company –0.035 0.084* –0.038 0.088 –0.258*** –0.130* –0.168*** 0.106*21 HR Dept. on Premise 0.063 0.005 0.022 –0.050 0.044 –0.047 –0.042 0.013*22 College Graduates –0.020 –0.043 0.024 0.493*** 0.097* 0.046 0.093 –0.106*23 Field Technicians 0.027 0.215*** –0.032 0.003 –0.083 –0.095 –0.034 0.09724 Office Technicians 0.074* –0.010 0.053 0.109*** 0.006 0.023 0.025 0.085

17 18 19 20 21 22 23 24

17 Female % 1.00018 Ln Estab. Employment 0.085* 1.00019 Branch 0.048 0.333*** 1.00020 Former Bell Company 0.037 0.280*** 0.288*** 1.00021 HR Dept. on Premise 0.138*** 0.481*** 0.192*** 0.062 1.00022 College Graduates –0.042 –0.095* –0.063 –0.043 –0.026 1.00023 Field Technicians –0.576*** 0.120*** 0.105*** 0.078* –0.042 –0.305*** 1.00024 Office Technicians –0.252*** –0.166*** –0.124*** –0.031 –0.058 0.079* –0.277*** 1.000

*Statistically significant at the .10 level; **at the .05 level; ***at the .01 level.

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 585

the left limit. The second part of thisdecomposition represents the effect ofchanges in the independent variables onthe dependent variable in the positiverange, which provides an interpretationequivalent to OLS estimates for this rangeof outcomes once the appropriate calcula-tion of the decomposition is made.2 Esti-mates of effect sizes reported here are basedon this second part of the decompositionand provide an estimate of the expectedchange in the observed values of the depen-dent variable for a given change in an inde-pendent variable. Also, because quit ratesare measured as a proportion of the cur-rent work force, the coefficients may beinterpreted, once multiplied by 100, as per-centage point changes in the quit rate at-tributable to the given independent vari-able.

In the first Tobit equation in Table 3, weanalyze the effects on quit rates of the threeemployee voice mechanisms. Union repre-sentation alone is associated with a reduc-tion of 6.5 percentage points in the quitrate (Tobit 1), which drops to a 4.9 percent-age point reduction in the quit rate in thefull model (Tobit 4). Quit rates are 3.9percentage points lower in establishmentswhere the entire work force is in problem-solving groups than in establishments withno employees in problem-solving groups(Tobit 1), and are 3.6 percentage pointslower in the full model (Tobit 4). Self-directed teams are only significant at the10% level in equation (1), but are signifi-cant at the 5% level in the full model, whichindicates quit rates are 2.5 percentage pointslower in establishments where the entirework force is organized into self-directed

teams than in establishments with no em-ployees in teams.

In the second equation (Tobit 2), weanalyze the relationship between cost-cut-ting human resource practices and quitrates. Each of the five practices indepen-dently is associated with increased quit rates.An increase in the size of work force reduc-tions equivalent to 1% of the current workforce is associated with an increase of 1.4percentage points in the quit rate, accord-ing to the results reported in equation (2);this association rises in the full model to a3.5 percentage point increase in the quitrate. Increasing the proportion of part-timers by 10 percentage points is associatedwith a 0.9 percentage point (Tobit 2) or 1.1percentage point (Tobit 4) rise in the quitrate for the core labor force. Increasingthe proportion of temporary employees by10 percentage points is associated with a1.2 percentage point (Tobit 2) or 1.0 per-centage point (Tobit 4) higher quit rate.Electronic monitoring of the work force atall times during the work day is associatedwith a 2.1 percentage point higher quit ratethan for workers who are not electronicallymonitored (Tobit 4). Variable pay also isassociated with higher quit rates at a barelysignificant α = 0.10 level in equation (2). Inequation (4), however, it becomes signifi-cant at the α = 0.01 level and indicates thatan increase in the portion of pay that isvariable by 10 percentage points would raisethe quit rate by 0.5 percentage points.

The third equation (Tobit 3) estimatesthe relationship between the three com-mitment-enhancing internal labor marketvariables and quit rates. Both internalmobility opportunities and a higher ratioof pay to the local cost of living significantlyreduce quit rates. Training, by contrast, isnot associated with any statistically signifi-cant difference in quits. If the portion ofemployees who have mobility opportuni-ties within the organization increased by 10percentage points, quit rates would be re-duced in the range of 0.3 percentage points(Tobit 3) to 0.2 percentage points (Tobit4). By increasing annual wages by $4,800(adjusted to reflect local cost of living), anestablishment can reduce quit rates in the

2The adjustment based on the second term in theMcDonald and Moffit (1980) decomposition is calcu-lated by multiplying the Tobit coefficients by[1 – z∗f(z)/F(z)–f(z)2/F(z)2], where F(z) is the cumu-lative normal distribution function associated withthe probability of cases being above the left limit; f(z),the first derivative of F(z), is the unit normal densityassociated with this probability; and z is the correspond-ing z-score for this probability (see Roncek 1992).

586 INDUSTRIAL AND LABOR RELATIONS REVIEW

range of 0.6 percentage points (Tobit 3) to0.4 percentage points (Tobit 4).

In each of the models, some controls arealso statistically significant. Establishmentsthat are larger are significantly more likelyto have higher quit rates. With regard tooccupational characteristics, field techni-cians and central office technicians areconsiderably less likely to quit than are callcenter workers in customer service and sales.Finally, once occupational and other char-acteristics are taken into account, the fe-male composition of the work force be-comes marginally significant and is nega-tively associated with quits.

In analyses not shown, we controlled forvariation in industry segments (for example,wireline, wireless, and internet services).We also controlled for local labor markets,because prior research has shown thathigher unemployment rates or alternativeopportunities in the external market arenegatively correlated with turnover (forexample, Hulin et al. 1985; Carsten andSpector 1987; Gerhart 1990). For eachestablishment surveyed, we included theaverage 1998 unemployment rate for thecounty or city where the establishment islocated, based on the Local Area Unem-ployment Statistics of the Bureau of LaborStatistics (http://stats.bls.gov.laumthd.htm). However, we found that includingthese market measures had no statisticallysignificant effect and lowered the varianceexplained, so we excluded them in ourfinal analyses.

Turning to the nonunion dispute resolu-tion mechanisms, we noted that in the fullcorrelation matrix reported in Table 2,none of the variables representing non-union dispute resolution procedures anddispute rates are significantly correlatedwith the quit rate. However, the full sampleincludes observations from both union andnonunion establishments. Therefore, wealso investigated the possibility that estab-lishments with nonunion arbitration andpeer review procedures have higher quitrates than union establishments, but at thesame time lower quit rates than nonunionestablishments that lack these procedures.Using a subsample consisting of only the

nonunion establishments, we found mar-ginally significant negative correlations (p< .10) between quit rates and the presenceof nonunion arbitration procedures, peerreview procedures, and the dispute rate.To further investigate the relationship be-tween nonunion dispute resolution proce-dures and quit rates, we ran Tobit regres-sions for the quit rate on our dispute reso-lution variables, using the variables fromthe full model in Table 3 as control vari-ables.

Table 4 presents the findings from Tobitregressions run on the subsample of re-spondents who answered the more detailedquestions on dispute resolution proceduresand activity. The focus of this analysis is onthe question of whether there is evidencefor lower quit rates for nonunion establish-ments that have dispute resolution proce-dures than for those lacking procedures.Equation (5) presents the results for thesame variables as in the full model withoutany dispute resolution variables includedto provide a basis for comparison. Thesmaller sample size used here reduces thesignificance levels for some variables, butthe coefficient estimates in equation (5)are generally similar to those for the fullsample, supporting the representativenessof the subsample when compared to thefull sample.

In each of the other models in Table 4,the quit rate is regressed on a series ofcombinations of dispute resolution vari-ables added to the set of independent vari-ables from the full model presented inTable 3. In each of these models, theprocedure usage rate is included along withthe dummy variables representing the typeof dispute resolution procedure, so that weare simultaneously accounting for the ef-fects of the presence and usage of proce-dures on quit rates. The only dispute reso-lution procedure type variable that reachesconventional levels of statistical significancein Table 4 is the variable representing thepresence of peer review procedures. Nei-ther the presence of any type of nonunionprocedure nor the inclusion of nonunionarbitration procedures has a statisticallysignificant association with the quit rate in

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 587

Table 3. Predictors of Annual Quit Rates.

Variable Tobit 1 Tobit 2 Tobit 3 Tobit 4

Employee Voice

Union –0.152**** –0.114****0.027 0.026

Problem-Solving Groups % –0.092*** –0.084***0.023 0.024

Self-Directed Teams % –0.049* –0.058**0.026 0.028

Cost-Cutting HR Practices

Downsizing in Last Five Years 3.267**** 8.088****0.687 2.309

Part-Time Workers % 0.212*** 0.251***0.074 0.079

Temporary Workers % 0.281**** 0.236****0.058 0.059

Electronic Monitoring % 0.047** 0.050**0.022 0.023

Variable Pay 0.066* 0.107***0.036 0.036

Commitment-Enhancing HR Practices

Mobility % –0.073*** –0.040*0.026 0.023

Training –0.006 –0.0050.006 0.005

Pay to Local Cost of Living –0.152*** –0.102***0.034 0.033

Control Variables

Female % –0.026 –0.030 –0.071* –0.067*0.036 0.036 0.040 0.037

Ln of Estab. Employment 0.0130** 0.016** 0.010 0.020***0.007 0.007 0.007 0.007

Branch 0.025 0.033 0.051 0.0220.023 0.023 0.025 0.024

Former Bell Company 0.014 –0.034* –0.027 0.0170.024 0.020 0.022 0.023

Human Resource Department 0.046** 0.017 0.037* 0.0250.019 0.020 0.021 0.020

College Graduates –0.011 0.009 0.059** 0.0350.023 0.024 0.028 0.027

Field Technicians –0.121**** –0.134**** –0.148**** –0.111****0.031 0.032 0.034 0.032

Office Technicians –0.063** –0.066** –0.078*** –0.0400.028 0.029 0.031 0.029

Constant 0.105*** –0.024 0.199**** 0.124***0.039 0.040 0.046 0.048

Observations 576 545 521 492Likelihood Ratio Chi Square 113.150 124.560 84.340 173.120Pseudo R Squared 0.674 0.767 0.509 1.303Likelihood Ratio –27.3/4149 –18.91/1291 –40.62/8212 20/1309

*Statistically significant at the .10 level; **at the .05 level; ***at the .01 level; ****at the .001 level.

588 INDUSTRIAL AND LABOR RELATIONS REVIEW

any of the specifications in which they areincluded. Even in the case of peer reviewprocedures, significance is achieved only atthe p < .10 level in equation (8).

Thus we do not find strong evidence thatnonunion dispute resolution proceduresprovide voice mechanisms that reduce quitrates in comparison to establishments lack-

ing such procedures. Higher dispute rates,however, are significantly associated withhigher quit rates at the p < .05 level in allspecifications. As discussed earlier, previ-ous research on unionized workplaceslinked high grievance rates to heightenedlevels of conflict and reduced organiza-tional performance (Katz, Kochan, and

Table 4. Dispute Resolution and Quit Rates.

Variable Tobit 5 Tobit 6 Tobit 7 Tobit 8 Tobit 9

Employee VoiceAny Nonunion Procedure –0.028 –0.011

0.027 0.028Nonunion Arbitration –0.026 –0.012

0.033 0.035Peer Review –0.059* –0.047

0.033 0.037Annual Dispute Rate 0.139** 0.129** 0.138** 0.142**

0.057 0.056 0.056 0.057Union –0.110*** –0.133**** –0.125*** –0.128**** –0.140****

0.034 0.040 0.037 0.036 0.041Problem-Solving Groups % –0.028 –0.042 –0.042 –0.040 –0.043

0.031 0.032 0.032 0.032 0.033Self-Directed Teams % –0.031 –0.022 –0.015 –0.012 –0.008

0.035 0.035 0.036 0.036 0.037

Cost-Cutting HR PracticesDownsizing in Last Five Years 7.19*** 7.32*** 7.25*** 7.37*** 7.20***

2.24 2.38 2.42 2.38 2.42Part-Time Workers % 0.071 0.091 0.050 0.089 0.052

0.118 0.118 0.128 0.118 0.129Temporary Workers % 0.248*** 0.221*** 0.228*** 0.207** 0.214***

0.079 0.080 0.082 0.081 0.082Electronic Monitoring % 0.028 0.032 0.030 0.035 0.034

0.029 0.029 0.030 0.030 0.030Variable Pay 0.147*** 0.157*** 0.161*** 0.164*** 0.158***

0.054 0.055 0.055 0.056 0.057

Commitment-Enhancing HR PracticesMobility % –0.035 –0.025 –0.027 –0.027 –0.026

0.031 0.032 0.032 0.032 0.033Training –0.007 –0.007 –0.006 –0.006 –0.006

0.007 0.007 0.007 0.007 0.007Pay to Local Cost of Living –0.105** –0.103** –0.100** –0.102** –0.097**

0.044 0.045 0.046 0.045 0.046

Control VariablesIncluded, but not shown

Observations 253 243 238 241 237Likelihood Ratio Chi Square 110.99 112.75 107.45 113.45 108.82Pseudo R Squared 1.732 1.809 1.707 1.757 1.705–2 Log Likelihood –46.907 –50.425 –44.490 –48.875 –45.012

*Statistically significant at the .10 level; **at the .05 level; ***at the .01 level; ****at the .001 level.

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 589

Gobeille 1983; Katz, Kochan, and Weber1985; Norsworthy and Zabala 1986;Ichniowski 1986; Cutcher-Gershenfeld1991). This result supports that finding bylinking high dispute rates to higher quitrates and extends it to include both non-union and union workplaces.

Discussion

Our results indicate that union institu-tions and managerial policies that facilitatevoice can significantly reduce exits. Unionshave a strong influence on quit rates, de-spite significant declines in union densityin this industry—from roughly two-thirdsof all employees in 1984 to one-third in1998. The inverse relationship betweenunions and quit rates is statistically signifi-cant even after we control for team-basedvoice mechanisms, pay, and other humanresource practices that are affected by col-lective bargaining. These findings are incontrast to those of Delery et al. (2000),who found that the union effect on turn-over in the trucking industry disappearedwhen wages and benefits were taken intoaccount. Participation in problem-solvingteams is negatively related to quit rates; andparticipation in self-directed teams has asmaller effect. Union representation anddirect participation via shop-floor teamsmay be viewed as complementary vehiclesfor employee voice at work. By contrast, wefound that nonunion dispute resolutionprocedures, in some cases designed as analternative to union grievance procedures,did not significantly reduce quit rates. Ifnew workplace dispute resolution proce-dures are going to help close the represen-tation gap in nonunion workplaces, ourresults suggest they will need to be some-thing stronger than the present employer-designed procedures.

It is useful to discuss the findings fornonunion dispute procedures in the con-text of the telecommunications industry,which despite the decline in union densityremains more highly unionized than mostother industries in the country. Prior re-search suggests that firms adopt disputeresolution procedures as a union avoid-

ance strategy (Colvin 1999). To the degreethat this is true, we would expect nonuniondispute resolution procedures to be par-ticularly strong in this industry because thethreat of union organizing is high. In addi-tion, we examined stronger forms of non-union procedures, including nonunion ar-bitration and peer review panels. Even withthis relatively favorable setting for non-union procedures, we found no statisticallysignificant results for nonunion arbitrationprocedures and only marginally significantresults for peer review procedures, whichwere not robust with respect to alternativespecifications of the model.

One alternative explanation is that man-agement may introduce some nonunionprocedures in response to high levels ofemployee complaints in the workplace. Ifthese complaints reflect high levels of work-place conflict that are also associated withhigher quit rates, we might have an effect inthe opposite direction, reducing any nega-tive relationship between procedure pres-ence and quit rates. Although we cannotdefinitively exclude the possibility of thisreverse mechanism, our inclusion of thedispute rate in all of the dispute resolutionequations helps control for the possibilitythat some of the procedures were intro-duced in response to high levels of work-place conflict.

Finally, our results are strong and consis-tent on the question of the relationshipbetween grievance rates and quit rates:holding the type of procedure constant, wefind a positive relationship. This result isconsistent with the findings of Lewin andcolleagues in union settings and indicatesthat future research on dispute resolutionprocedures needs to disaggregate the ef-fects of the presence of dispute resolutionprocedures and the usage of those sameprocedures. One should not make themistake of assuming that because the avail-ability of voice mechanisms in the work-place is associated with lower quit rates,increased usage of these voice mechanismswill be associated with lower quit rates;indeed, our results suggest that the reverseis true.

With respect to human resource prac-

590 INDUSTRIAL AND LABOR RELATIONS REVIEW

tices, our results show that internal mobil-ity opportunities and higher wages are as-sociated with lower quit rates, but cost-cutting human resource practices that treatlabor as a variable cost and introduce labormarket competition into the workplace areassociated with higher employee quits.Downsizing, contingent staffing, electronicmonitoring, and variable pay are signifi-cantly associated with higher quit rates.Ongoing downsizing creates uncertaintyand demoralization that often lead the mostqualified to obtain jobs elsewhere. Higheruse of temporary workers may signal tocore employees that the employer has aready supply of substitute workers and isnot committed to long-term employmentcontracts. Greater reliance on electronicmonitoring also significantly drives em-ployee quits because it signals mistrust andcreates more onerous working conditions.Call center employees experience elec-tronic monitoring as unremitting surveil-lance of their work and personal activities,including their use of the restroom. Wherepossible, they are likely to seek alternativeworking conditions.

With respect to variable pay, some re-search on high performance work systemsfinds that variable pay clusters with teamsand investment in training in a bundle ofcoherent practices. By contrast, our resultsindicate that variable pay plans are exitdrivers. One explanation for these contra-dictory findings is that in most of the highperformance literature, performance-basedpay is in the form of group incentives suchas gain-sharing or profit-sharing that areadded to base pay and that rarely exceed 5–10% of total pay. In our case, more thanhalf of variable pay is individual commis-sion pay for service and sales workers, andthe mean percent of pay at risk is 27% (witha range of variation between 0% and 100%).

Commissions are a type of piece ratesystem whereby employees are directly com-pensated for each item they sell. Whilesome workers may thrive on risk and indi-vidual competition and benefit throughhigher earnings under these plans, priorresearch suggests that most workers arerisk-averse. Because most do not have di-

versified portfolios, they are likely to bedissatisfied with the greater risk and earn-ings volatility that are associated with vari-able pay. In addition, prior research showsthat piece rate systems are structured asprisoner dilemma games (see, for example,Levine 1992 or Gillespie 1991). That is, asemployees learn how to make more piecesto increase their earnings, they often adjustthe piece rate to capture the gains in pro-ductivity. This dynamic of employees im-proving performance leading to rate re-ductions is well documented. It causesemployees to game the system to “makeout.” Under piece rate systems, gaming thepay plan often became more important toworkers than getting the job done (Roy1952; Burawoy 1979). To defeat the gameof making out, management often makesthe incentive plan more complex, but thegame continues under the new more elabo-rate rules. Peer pressure restrains highperformers because high performance mayincrease an individual’s earnings, but onlytemporarily. In the longer run, high per-formance leads to rate reductions, whichdisadvantage all employees—hence theprisoner dilemma quality of piece rates andcommissions.

There are also reasons to questionwhether commission pay in service and salesenvironments improves organizational per-formance. On the one hand, commissionsprovide incentives to individuals to sellmore. On the other hand, shirking underthese plans takes the form of a tendency forthe quality of the service to fall (Pencavel1977). In field research for this study, forexample, sales reps reported that they triedto close a sale to get their commission evenwhen customers said they would probablydrop the feature after the promotion ended.Sales reps also reported knowingly sellingto customers who had been recently dis-connected for failure to pay their bills. Inaddition, individual commission plans un-dermine the social organization of workand discourage participation in direct voicemechanisms such as teams, decreasing jointproblem-solving and sharing of knowledgeand information. Recent research in callcenters, for example, has shown that both

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 591

service quality and sales are higher wherework organization and group incentivesemphasize team work (Batt 1999; Moynihanand Batt 2000). Thus, there is reason tobelieve that individual commission plansnot only raise quit rates but also are detri-mental to service and sales performancebecause they undermine group problem-solving and shared expertise. While vari-able pay plans are often adopted because ofthe intuitive appeal of paying-for-perfor-mance, we need more research to under-stand how different types of plans affectworker quit rates, the social organization ofwork, and organizational performance.

Gender differences also merit some dis-cussion. Once occupational and humanresource practices are taken into account,the female composition of the work force isassociated with lower quit rates (at a mar-ginally statistically significant level). This isconsistent with the prior literature as wellas our field research. Occupational segre-gation is widespread in this industry, withwomen concentrated in the jobs with lowdiscretion, continuous monitoring, and lowpay. In our sample, women comprise 56%of customer service and sales jobs but only12% of technical jobs. Within call centers,they are concentrated in operator services(80%) and the residential mass market(60%). By contrast, service and sales repsserving business clientele are 55% male.Thus, it is women’s concentration in loweroccupational titles, rather than gender perse, that is associated with higher quit rates.Once occupational differences are takeninto account, women have lower quit ratesthan men.

Limitations and Conclusions

There are several limitations to this study.Because the data were collected at the es-tablishment unit of analysis, it was not pos-sible to obtain the kind of individual-leveldata that capture variation in human capi-tal characteristics or in the use of humanresource and dispute resolution procedureswithin the establishment. Moreover, thecross-sectional nature of the data limitscausal inferences. It could be, for example,

that high unemployment in some localitiesreduces the likelihood of quits, therebyallowing the adoption of work innovationssuch as teams because managers know thatthe work force is experienced and stable.Using unemployment data from the Bu-reau of Labor Statistics for each countywhere an establishment was located, weexplored this explanation in analyses notshown, and found no statistically signifi-cant correlation between local unemploy-ment and adoption of HR practices. Theseanalyses suggest that at least in this indus-try, adoption is not driven by labor marketconditions.

We believe that our approach to thequestion of aggregate quit rates is promis-ing for a number of reasons. Our findingssuggest that it is important to disaggregatethe differential influences of alternativevoice mechanisms and human resourcepractices on employee quit rates becausethey do not appear to have a consistent orcoherent effect. Quit rates are substan-tially lower in unionized workplaces thanin nonunion workplaces even in the pres-ence of a broad set of controls for compen-sation and a range of human resource prac-tices. In addition, both self-directed teamsand off-line problem-solving groups havestrong and statistically significant negativerelationships with quit rates, lending sup-port to the idea that these mechanisms fordirect participation can provide employeeswith enhanced voice in the workplace. Incontrast, the absence of consistent statisti-cally significant results for nonunion dis-pute resolution procedures suggests theneed for further examination of the ques-tion of what features need to be incorpo-rated into procedures in order to ensurethat they can serve as effective voice mecha-nisms that provide employees with alterna-tives to quitting. In addition, differences inthe results for nonunion arbitration andpeer review panels provide support fordistinguishing between different types ofnonunion procedure in future research.Put simply, results based on research onone type of nonunion procedure may nothold for other types of procedures. Amongdifferent types of procedures, peer review

592 INDUSTRIAL AND LABOR RELATIONS REVIEW

REFERENCES

panels have received very little attention inacademic research, and our results suggestthe need for further investigation of theirrole in the nonunion workplace. Our re-sults also support recommendations (forexample, Lewin 1999) that have been madefor including measures of both the pres-ence and usage of procedures in futureresearch. Whereas the presence of peerreview procedures had a marginally signifi-cant negative association with quit rates,greater usage of procedures had a consis-tent and statistically significant positive as-sociation with higher quit rates. This sug-gests that it is important to treat availabilityand usage of procedures as conceptuallydistinct phenomena.

While this research is based on the tele-communications services industry, we be-lieve the findings have broader implica-tions. In the U.S. context, cost-cuttinghuman resource practices diffused duringa period from the late 1970s to the mid-1990s when labor supply exceeded demand,real wages of workers were declining, andworkers had few labor market alternatives.These HR practices disrupted the relatively

stable internal labor market practices inlarge firms that were developed during twoperiods of full employment—World War Ithrough the 1920s, and World War IIthrough the 1970s. In the tight labor mar-kets of the 1990s and after, however, highlytrained employees found it relatively easyto find comparable jobs in related indus-tries. Firms may be losing the more quali-fied workers and may find it necessary toreturn to the kinds of commitment-enhanc-ing or internal labor market practices thatfirms developed in prior periods of fullemployment. At times of near-full employ-ment, quit rates may be the critical factorraising operating costs and preventing or-ganizations from achieving their perfor-mance objectives, as is suggested by recentresearch that shows an inverse relationshipbetween quit rates and organizational per-formance. In a near-full-employmenteconomy, employers who have retained in-ternal labor market practices and employeevoice mechanisms may have a competitivecost and performance advantage over thosewho use cost-cutting human resource prac-tices.

Alexander, Jeffrey A., Joan R. Bloom, and Beverly A.Nuchols. 1994. “Nursing Turnover and HospitalEfficiency: An Organization Level Analysis.” Indus-trial Relations, Vol. 33, No. 4 (October), pp. 505–20.

Appelbaum, Eileen, Thomas Bailey, Peter Berg, andArne Kalleberg. 2000. Manufacturing Advantage:Why High Performance Work Systems Pay Off. Ithaca,N.Y.: ILR Press (an imprint of Cornell UniversityPress).

Arthur, Jeffrey. 1994. “Effects of Human ResourceSystems on Manufacturing Performance and Turn-over.” Academy of Management Journal, Vol. 37, No. 3(June), pp. 670–87.

Batt, Rosemary. 1999. “Work Organization, Technol-ogy, and Performance in Customer Service andSales.” Industrial and Labor Relations Review, Vol. 52,No. 4 (July), pp. 539–64.

Bemmels, Brian. 1997. “Exit, Voice, and Loyalty inEmployment Relationships.” In David Lewin, DanielJ.B. Mitchell, and Mahmood A. Zaidi, eds., TheHuman Resource Management Handbook, Part II. Green-wich, Conn.: JAI Press, pp. 245–59.

Bennett, Nathan, Terry C. Blum, Rebecca G. Long,and Paul M. Roman. 1993. “A Firm Level Analysis

of Employee Attrition.” Group and OrganizationManagement, Vol. 18, No. 4, pp. 482–99.

Boroff, Karen E., and David Lewin. 1997. “Loyalty,Voice, and Intent to Exit a Union Firm: A Concep-tual and Empirical Analysis.” Industrial and LaborRelations Review, Vol. 51, No. 1 (October), pp. 50–63.

Burawoy, Michael. 1979. Manufacturing Consent:Changes in the Labor Process under Monopoly Capital-ism. Chicago: University of Chicago Press.

Cappelli, Peter. 1999. The New Deal at Work. Cam-bridge, Mass.: Harvard Business School Press.

Cappelli, Peter, and Keith Chauvin. 1991. “A Test ofan Efficiency Model of Grievance Activity.” Indus-trial and Labor Relations Review, Vol. 45, No. 1 (Octo-ber), pp. 3–14.

Carayon, Pascale. 1993. “Effect of Electronic Perfor-mance Monitoring on Job Design and Worker Stress:Review of the Literature and Conceptual Model.”Human Factors, Vol. 35, No. 3, pp. 385–95.

Carsten, Jeanne M., and Paul E. Spector. 1987. “Un-employment, Job Satisfaction, and Employee Turn-over.” Journal of Applied Psychology, Vol. 72, No. 3(June), pp. 374–81.

EMPLOYEE VOICE, HR PRACTICES, AND QUIT RATES 593

Cohen, Susan, and Diane Bailey. 1997. “What MakesTeams Work: Group Effectiveness Research fromthe Shop Floor to the Executive Suite.” Journal ofManagement, Vol. 23, No. 3 (June), pp. 239–90.

Colvin, Alexander J.S. 1999. Citizens and Citadels:Dispute Resolution and the Governance of EmploymentRelations. Diss., Cornell University.

Cordery, John L., Walter S. Mueller, and Leigh M.Smith. 1991. “Attitudinal and Behavioral Effects ofAutonomous Group Working: A Longitudinal FieldStudy.” Academy of Management Journal, Vol. 34, No.2 (June), pp. 464–76.

Cotton, John L. 1993. Employee Involvement: Methodsfor Improving Performance and Work Attitudes. NewburyPark, Calif.: Sage.

Cotton, John L., and Jeffrey M. Tuttle. 1986. “Em-ployee Turnover: A Meta-Analysis and Review withImplications for Research.” Academy of ManagementReview, Vol. 11, No. 1 (January), pp. 55–70.

Cutcher-Gershenfeld, Joel. 1991. “The Impact onEconomic Performance of a Transformation inWorkplace Relations.” Industrial and Labor RelationsReview, Vol. 44, No. 2 (January), pp. 241–60.

Delery, John, Nina Gupta, Jason Shaw, G. DouglasJenkins, and Margot Ganster. 2000. “Unionization,Compensation, and Voice Effects on Quits and Re-tention.” Industrial Relations, Vol. 39, No. 4 (Octo-ber), pp. 625–46.

Doeringer, Peter B., and Michael J. Piore. 1971.Internal Labor Markets and Manpower Analysis. Lex-ington, Mass.: Heath.

Ewing, David W. 1989. Justice on the Job: ResolvingGrievances in the Nonunion Workplace. Boston: HarvardBusiness School Press.

Feuille, Peter, and John T. Delaney. 1992. “TheIndividual Pursuit of Organizational Justice: Griev-ance Procedures in Nonunion Workplaces.” Re-search in Personnel and Human Resources Management,Vol. 10, pp. 187–232.

Freeman, Richard B. 1980a. “The Exit-Voice Tradeoffin the Labor Market: Unionism, Job Tenure, Quits,and Separations.” Quarterly Journal of Economics, Vol.94, No. 3 (August), pp. 643–73.

____. 1980b. “Unionism and the Dispersion of Wages.”Industrial and Labor Relations Review, Vol. 34, No. 1(October), pp. 3–23.

____. 1982. “Union Wage Practices and Wage Disper-sion within Establishments.” Industrial and LaborRelations Review, Vol. 36, No. 1 (October), pp. 3–20.

Freeman, Richard B., and James L. Medoff. 1984.What Do Unions Do? New York: Basic Books.

Gerhart, Barry. 1990. “Voluntary Turnover andAlternative Job Opportunities.” Journal of AppliedPsychology, Vol. 75, pp. 467–73.

____. 1999. “Human Resource Management andFirm Performance: Challenges in Making CausalInferences.” In Patrick Wright, Lee Dyer, JohnBoudreau, and George Milkovich, eds., Research inPersonnel and Human Resources Management: StrategicHRM in the 21st Century (Supplement 4). Green-wich, Conn: JAI Press.

Gillespie, Richard. 1991. Manufacturing Knowledge: AHistory of the Hawthorne Experiments. Cambridge:

Cambridge University Press.Hackman, J. Richard, and Greg R. Oldham. 1980.

Work Redesign. Reading, Mass.: Addison-Wesley.Hirschman, Albert O. 1970. Exit, Voice, and Loyalty.

Cambridge, Mass.: Harvard University Press.Hom, Peter W., and Rodger W. Griffeth. 1995. Em-

ployee Turnover. Cincinnati: South-Western CollegePublishing.

Houseman, Susan. 2001. “Why Employers Use Flex-ible Staffing Arrangements: Evidence from an Es-tablishment Survey.” Industrial and Labor RelationsReview, Vol. 55, No. 1 (October), pp. 149–70.

Hulin, Charles L., Mary A. Roznowski, and DonnaHachiya. 1985. “Alternative Opportunities andWithdrawal Decisions: Empirical and TheoreticalDiscrepancies and an Integration.” PsychologicalBulletin, Vol. 97, No. 2, pp. 233–50.

Huselid, Mark A. 1996. “The Impact of HumanResource Management Practices on Turnover, Pro-ductivity, and Corporate Financial Performance.”Academy of Management Journal, Vol. 38, No. 3 (June),pp. 635–72.

Ichniowski, Casey. 1986. “The Effects of GrievanceActivity on Productivity.” Industrial and Labor Rela-tions Review, Vol. 40, No. 1 (October), pp. 75–89.

Ichniowski, Casey, Thomas Kochan, David Levine,Craig Olson, and George Strauss. 1996. “WhatWorks at Work: Overview and Assessment.” Indus-trial Relations, Vol. 35, No. 3 (July), pp. 299–334.

Jackson, Susan E., Randall S. Schuler, and J. CarlsoRivero. 1989. “Organizational Characteristics asPredictors of Personnel Practices.” Personnel Psy-chology, Vol. 42, pp. 727–86.

Jacoby, Sanford. 1985. Employing Bureaucracy. NewYork: Columbia University Press.

Katz, Harry C., Thomas A. Kochan, and Kenneth R.Gobeille. 1983. “Industrial Relations Performance,Economic Performance, and QWL Programs: AnInterplant Analysis.” Industrial and Labor RelationsReview, Vol. 37, No. 1 (October), pp. 3–17.

Katz, Harry C., Thomas A. Kochan, and Mark R.Weber. 1985. “Assessing the Effects of IndustrialRelations and Quality of Work Life Efforts on Orga-nizational Effectiveness.” Academy of ManagementJournal, Vol. 28, No. 3 (September), pp. 509–27.

Katz, Lawrence F., and Alan B. Krueger. 1999. “TheHigh Pressure U.S. Labor Market of the 1990s.”Brookings Papers on Economic Activity, Vol. 1, No. 1,pp. 1–65.

Leonard, Jonathan S. 1987. “Carrots and Sticks: Pay,Supervision, and Turnover.” Journal of Labor Eco-nomics, Vol. 5, No. 4, Part 2 (October), pp. S136–52.

Levine, David. 1992. “Piece Rate, Output Restric-tion, and Conformism.” Journal of Economic Psychol-ogy, Vol. 13, pp. 473–89.

Levine, David, and Laura D’Andrea Tyson. 1990.“Participation, Productivity, and the Firm’s Envi-ronment.” In Alan Blinder, ed., Paying for Productiv-ity. Washington, D.C.: Brookings Institution.

Lewin, David. 1987. “Dispute Resolution in theNonunion Firm: A Theoretical and Empirical Analy-sis.” Journal of Conflict Resolution, Vol. 31, No. 3(September), pp. 465–502.

594 INDUSTRIAL AND LABOR RELATIONS REVIEW

____. 1990. “Grievance Procedures in NonunionWorkplaces: An Empirical Analysis of Usage, Dy-namics, and Outcomes.” Chicago-Kent Law Review,Vol. 66, No. 3, pp. 823–44.

____. 1999. “Theoretical and Empirical Research onthe Grievance Procedure and Arbitration: A Criti-cal Review.” Chapter 5 in Adrienne E. Eaton andJeffery H. Keefe, eds., Employment Dispute Resolutionand Worker Rights in the Changing Workplace. Madi-son, Wis.: Industrial Relations Research Associa-tion.

Lewin, David, and Karen Boroff. 1996. “The Role ofLoyalty in Exit and Voice: A Conceptual and Em-pirical Analysis.” In David Lewin, Bruce E. Kaufman,and Donna Sockell, eds., Advances in Industrial andLabor Relations, Vol. 7. Greenwich, Conn.: JAI Press,pp. 69–96.

Lewin, David, and Richard Peterson. 1999. “Behav-ioral Outcomes of Grievance Activity.” IndustrialRelations, Vol. 38, No. 4 (October), pp. 554–76.

Lincoln, James R., and Arne L. Kalleberg. 1996.“Commitment, Quits, and Work Organization inJapanese and U.S. Plants.” Industrial and LaborRelations Review, Vol. 50, No. 1 (October), pp. 39–59.

Maddala, G. S. 1992. Introduction to Econometrics, 2nded. New York: Macmillan.

McEnvoy, G. M., and Wayne F. Cascio. 1985. “Strat-egies for Reducing Employee Turnover: A Meta-Analysis.” Journal of Applied Psychology, Vol. 70, No.3 (June), pp. 342–53.

Miller, Paul, and Charles Mulvey. 1991. “AustralianEvidence on the Exit/Voice Model of the LaborMarket.” Industrial and Labor Relations Review, Vol.45, No. 1 (October), pp. 44–57.

Moynihan, Lisa, and Rosemary Batt. 2000. “Anteced-ents and Consequences of Transactive Memory inService Teams: Theory and Scale Development.”Paper presented at the Academy of ManagementMeetings, Toronto, Ontario, August 6–9.

Norsworthy, J. R., and Craig A. Zabala. 1985. “WorkerAttitudes, Worker Behavior, and Productivity in theU.S. Automobile Industry, 1959–76.” Industrial andLabor Relations Review, Vol. 38, No. 4 (July), pp. 544–57.

Oi, Walter. 1962. “Labor as a Quasi-Fixed Factor.”Journal of Political Economy, Vol. 70, No. 6 (Decem-ber), pp. 538–55.

Osterman, Paul. 1987. “Turnover, Employment Se-curity, and the Performance of the Firm.” In MorrisKleiner, Myron Roomkin, and Sidney W. Salsburg,eds., Human Resources and the Performance of the Firm.IRRA Series. Madison, Wis.: Industrial RelationsResearch Association.

____. 1994a. “How Common Is Workplace Transfor-mation and Who Adopts It?” Industrial and LaborRelations Review, Vol. 47, No. 2 (January), pp. 173–88.

____. 1994b. “Internal Labor Markets: Theory andChange.” In Clark Kerr and Paul Staudohar, eds.,Labor Economics and Industrial Relations: Markets andInstitutions. Cambridge, Mass.: Harvard UniversityPress, pp. 303–39.

____. 1999. Securing Prosperity. Princeton: Princeton

University Press.Park, Hong Y., Joseph Ofori-Dankwa, and Deborah

Ramirez Bishop. 1994. “Organizational and Envi-ronmental Determinants of Functional and Dys-functional Turnover: Practical and Research Impli-cations.” Human Relations, Vol. 47 (May), pp. 353–366.

Pencavel, John H. 1977. “Work Effort, On-the-JobScreening, and Alternative Methods of Remunera-tion.” In Ronald Ehrenberg, ed., Research in LaborEconomics, Vol. 1. Greenwich, Conn.: JAI Press, pp.225–58.

Powell, Irene, Mark Montgomery, and James Cosgrove.1994. “Compensation Structure and EstablishmentQuit and Fire Rates.” Industrial Relations, Vol. 33,No. 2, pp. 229–48.

Rebitzer, James B. 1986. “Establishment Size and JobTenure.” Industrial Relations, Vol. 25, No. 3 (July),pp. 292–302.

Rees, Daniel, I. 1991. “Grievance Procedure Strengthand Teacher Quits.” Industrial and Labor RelationsReview, Vol. 45, No. 1 (October), pp. 31–43.

Roy, Donald. 1952 “Quota Restriction andGoldbricking in a Machine Shop.” American Journalof Sociology, Vol. 57 (March), pp. 432–41.

Schacht, John 1985. The Making of Telephone Union-ism, 1920–1947. New Brunswick, N.J.: RutgersUniversity Press.

Shaw, Jason, John Delery, G. Douglas Jenkins, Jr., andNina Gupta. 1998. “An Organization-Level Analysisof Voluntary and Involuntary Turnover.” Academy ofManagement Journal, Vol. 39, No. 5, pp. 1–15.

Spencer, Daniel G. 1986. “Employee Voice andEmployee Retention.” Academy of Management Jour-nal, Vol. 29, No. 3, pp. 488–502.

Stone, Katherine V. W. 1996. “Mandatory Arbitra-tion of Individual Employment Rights: The YellowDog Contract of the 1990s.” Denver University LawReview, Vol. 73, pp. 1017–50.

____. 1999. “Employment Arbitration under theFederal Arbitration Act.” Chapter 2 in Adrienne E.Eaton and Jeffrey H. Keefe, eds., Employment DisputeResolution and Worker Rights in the Changing Work-place. Champaign, Ill.: Industrial Relations Re-search Association.

Weisman, Carole S., Dorothy L. Gordon, Sandra D.Cassard, Marilyn Bergner, and Rebeca Wong. 1993.“The Effects of Unit Self-Management on HospitalNurses’ Work Process, Work Satisfaction, and Re-tention.” Medical Care, Vol. 31, No. 5, pp. 381–93.

Wilson, Nicholas, and Michael J. Peel. 1991. “TheImpact on Absenteeism and Quits of Profit-Sharingand Other Forms of Employee Participation.” In-dustrial and Labor Relations Review, Vol. 44, No. 3(April), pp. 454–68.

Wyatt Data Services, Cole Surveys. 1991. Restructur-ing—Cure or Cosmetic Surgery? Results of CorporateChange in the ’80s with Rx’s for the ’90s. Boston: Wyatt.

Zack, Arnold M. 1999. “Agreements to Arbitrate andthe Waiver of Rights under Employment Law.”Chapter 3 in Adrienne E. Eaton and Jeffrey H.Keefe, eds., Employment Dispute Resolution and WorkerRights in the Changing Workplace. Champaign, Ill.:Industrial Relations Research Association.