Employe Voice and Employee Retention

^Academy of Management Journal1986, Vol, 29, No, 3, 488-502,

EMPLOYEE VOICE AND EMPLOYEE RETENTION

DANIEL G. SPENCERUniversity of Kansas

This study investigates the relationship between the extent to whichemployees have opportunities to voice dissatisfaction and voluntaryturnover in 111 short-term, general care hospitals. Results show that,whether or not a union is present, high numbers of mechanisms foremployee voice are associated with high retention rates. Implicationsfor theory and research as well as management practice are discussed.

The relationship between the job satisfaction and employee turnover hasbeen one of the most widely studied but least understood relationships inthe organizational behavior literature. The consistently low correlationsbetween job satisfaction and turnover found in empirical research on thetopic clearly indicate that the relationship is not direct. Many processes mayintercede either to cause satisfied employees to leave organizations or dissatis-fied employees to remain with organizations. Previous research has pro-posed and tested a number of moderators that might clarify the relationshipbetween job satisfaction and turnover (March & Simon, 1958; Mobley, Griffeth,Hand, & Meglino, 1979; Price, 1977). The two most prominent have been (1)the number of employment alternatives employees wbo are considering leav-ing an organization have, and (2) nonwork-related influences sucb as aspouse's immobility. Also, researcb bas recently centered around testingMobley's (1977) model of tbe process of employees' decisions on turnover inattempts to better understand tbose cognitive and bebavioral processes tbatmediate tbe relationship between affective reactions to jobs and turnover.But even tbougb scbolars bave expended much effort in theorizing aboutand researching tbe relationsbip between job satisfaction and turnover, con-sistently weak empirical relationsbips are still found.

Steers and Mowday (1981) suggested a beretofore neglected process tbatmay intervene in the relationsbip in question: employees' efforts to cbangedissatisfying work situations tbat would otberwise be major factors in theirdecisions to leave tbeir organizations. Employees wbo succeed in cbangingdissatisfying work situations will lower tbeir dissatisfaction; witb intent to

University of Kansas General Research Fund Grants 3616-0038 and 3219-0038 providedsupport for this research. In addition, it was supported by the University of Kansas School ofBusiness Research Fund provided by the Fourth National Bank & Trust Company, Wichita, Theideas and opinions expressed herein are solely those of the author. The comments of Richard T,Mowday, George F, Dreher, and Morris M, Kleiner on earlier drafts are greatly appreciated.Portions of this paper were presented at the 42nd Annual Meeting of the Academy ofManagement, New York, 1982,

488

1986 Spencer 489

leave changed to intent to remain, tbe likelihood increases tbat these employ-ees will ultimately remain with tbeir organizations. Up until recently, tben,tbe literature on turnover bas focused on only one of several alternativesopen to individuals faced witb dissatisfying work situations, leaving; it hasignored tbe alternative strategy of attempting to cbange sucb situations.

HIRSGHMAN'S GONGEPTS OF EXIT AND VOIGE

Tbe proposed relationsbip between employees' efforts to cbange dis-satisfying work situations and turnover is implicit in mucb of tbe work ofAlbert Hirscbman (1970, 1974), wbose perspective served as tbe theoreticalfoundation for tbe present researcb. He extensively studied responses todeclining firms in terms of exit and voice. Tbe two terms refer to two optionstbat members or clients of an organization bave wben responding to aneconomically dissatisfying relationsbip witb that organization. The sourceof dissatisfaction migbt be declining quality of tbe organization or decliningquality of its products. To voice dissatisfaction is to try to cbange a problem-atic situation, and to exit is simply to witbdraw from and leave tbe problem-atic situation. Most of Hirscbman's work is grounded in economics and tbusfocuses primarily on tbe clients of an organization using voice or exit ratbertban on its members using tbese options. Hirscbman's concepts are gainingincreasing attention in tbe organizational bebavior literature (Farrell, 1983).Tbis attention is not limited to studies of employee turnover; otber relevantwork includes researcb on absenteeism (Hammer, Landau, & Stern, 1981),political bebavior (Farrell & Peterson, 1982), organizational due process (Aram& Salipante, 1981), organizational dissent (Grabam, 1986), and tbe catbarticeffects of voice (Greenberg & Folger, 1983).

EMPLOYEE TURNOVER AND THE LABOR RELATIONS LITERATURE

Unions can be viewed as vebicles for tbe collective voicing of employees'dissatisfactions. Previous researcb bas demonstrated tbat unionism is clearlyassociated witb employee stability. Freeman and Medoff (1984), reviewingtbeir own and others' analyses of individual bebavior and industry aggregates,demonstrated tbat unionization is consistently and significantly associatedwitb retention of employees if wage rates and otber known predictors ofemployee exit are controlled for. Tbe control of wage rates is critical in orderto differentiate tbe effects of unions in terms of voice from any monopolyeffects due to union control of equivalent jobs.

Tbe researcb Freeman and Medoff (1984) reviewed is witbout exceptionbased on analyses of very large data sets like tbe National LongitudinalSurvey.^ Sucb data sets constrain tbe specific relationsbips tbat can be stud-ied and tbe levels of analysis at wbicb sucb relationsbips can be examined.In particular, very little researcb exists on tbe effects of unionization on

* This survey was conducted by the Genter for Human Resource Research, Ohio StateUniversity,

490 Academy of Management Journal September

quitting rates in organizations. Becker (1978) conducted sucb an organiza-tion level study, examining tbe effects of unionization on rates of quitting forindividuals in low-wage occupations in short-term, general care bospitals.Gontrary to results at otber levels of analysis, Becker found a weak, margin-ally significant relationsbip (p < .10) between unionization and rate of turn-over wbile controlling for monopoly effects and a wide variety of otberknown predictors of employee exit.

Tbe generalizability of researcb like Becker's (1978) is limited; sucbwork is usually specific to tbe occupation, organization, and region studied.^However, wben an investigator controls data collection, a study is morelikely to include predictors of exit tbat are not captured in studies usingexisting data bases. For example, Becker examined the effects of fringe bene-fits on rates of quitting in addition to tbe effects of v/ages. Freeman andMedoff (1984) noted tbat a major criticism of researcb on unions' effects interms of voice is tbat most existing data sets do not include information on aprincipal monetary reward sucb as fringe benefits and tbus do not fullycontrol for monopoly effects. Anotber significant criticism of researcb onvoice effects of unions is Ulman and Sorensen's (1984) demonstration tbat,for unionized employees, tbe effects of strikes and tbe tbreat of strikes pro-vide otber alternatives to quitting besides voice realized tbrougb tbe uniongrievance mecbanism.

Tbus, accurately assessing tbe effects of employee voice on numbers ofexits requires more tban assessing tbe effect of unionization. It requiresexploring tbe effects of mecbanisms, wbetber union mandated or not, tbatorganizations put in place to enbance tbeir employees' opportunities forvoicing dissatisfaction. Tbis approacb necessitates collecting data specific-ally for tbis purpose instead of using an existing data base. Tbis studyemployed a data collection strategy similar to tbat used by Becker (1978).

HYPOTHESES AND STUDY DESIGN

Study One

Two studies provided data for tbis researcb. Tbe first sougbt to deter-mine if tbere is a direct relationsbip between (1) tbe amount of opportunityan organization gives its employees to voice dissatisfaction and cbange dis-satisfying work situations, and (2) tbe organization's rate of employeeretention. Tbus,

Hypothesis 1; There will be a significant and negativerelationship between the total number of voice mecha-nisms for employees that an organization has and thevoluntary turnover rate among the organization's em-ployees.

^ Becker (1978) surveyed hospitals in three contiguous states in the north central UnitedStates,

1986 Spencer 491

Tbe focus bere was on tbe amount of opportunity available for voice and noton tbe quality of tbat opportunity; tbe second may, in fact, be a more viabledeterminant of retention. Employee voice mecbanisms examined in tbisresearcb included grievance procedures, suggestion systems, employee-management meetings, counseling services, ombudsman services, non-management task forces, question and answer programs, and survey feed-back.

Tbe following known predictors of rate of employee turnover were con-trolled for in tbis study: wage rate, tbe ratio of fringe benefits to totalcompensation, unemployment rate, number of grievances filed (a surrogatefor job dissatisfaction), percentage of minority employment, organizationalsize, and occupation-specific employment opportunities in tbe region (seeBecker, 1978). Tbe effect of unionization of tbe workforce on tbe rate ofemployee retention was examined separately, but for tbe purposes of tbisstudy, it is included as a control variable. Tbis study examined only tberetention rates of an occupation dominated by women tbat bas a docu-mented sbortage of workers: registered nursing. Sbort-term, general carebospitals tbat universally employ continuing education for registered nurseswere studied. Tbis population was cbosen to control for unwanted variancedue to differences in total employment accounted for by women, sbortage ofworkers, availability of continuing education, occupational type, and organi-zational type.

Study Two

Tbe first study did not assess tbe quality of tbe bospitals' voicemecbanisms. A second analysis was conducted to examine tbe relationshipbetween tbe number of mechanisms offering employees tbe option of voicein an organization and employees' perceptions of tbe effectiveness of tbesemecbanisms. From four bospitals participating in tbe first study, data werecollected on registered nurses' expectancies for resolving work related prob-lems and tbeir perceptions of tbe general effectiveness of tbeir organizations'voice mecbanisms. It was proposed that.

Hypothesis 2: A high number of employee voice mecha-nisms will be positively related to high expectancies ofproblem resolution among employees and high levels ofeffectiveness for the organization's problem resolutionprocedures.

STUDY ONE

Methods

Sample. Organizations were tbe units of analysis in tbis study. Participat-ing organizations were sbort-term, general care bospitals ranging from 50 toover 1,000 beds in size. A systematic sample of presidents and executivedirectors of bospitals in tbe nortb central United States was drawn from tbeAmerican Hospital Association membersbip directory. Tbese administrators

492 Academy of Management Journal September

received questionnaire packages witb a cover letter requesting tbem to for-ward tbe questionnaire and tbe return envelope provided to tbeir hospitals'personnel directors. In most cases, personnel directors or members of person-nel staffs completed tbe questionnaires. Out of a sample of 278 bospitals,individuals in 129 of tbem completed and returned questionnaires for aresponse rate of 46.4 percent. Previous studies using similar strategies fordata collection bave yielded response rates below 30 percent (Becker, 1978).Of tbe bospitals participating in tbe study, 13.9 percent tbat did not supplydata on employee turnover were removed from tbe final sample. Data from111 sbort-term, general care bospitals, 15 of wbicb bad unionized registerednurses, were analyzed.

Employee voice mechanisms. Tbe questionnaire asked wbetber or not ahospital's registered nurses were subject to tbe following employee-relationspractices: (1) formal grievance procedure, (2) suggestion system, (3) employee-management meetings, (4) counseling service, (5) ombudsman, (6) non-management task forces, (7) question and answer program, and (8) surveyfeedback. A composite index of employee voice mecbanisms was created bysimply summing tbe number of tbese mecbanisms tbat an organizationemployed (Kuder-Ricbardson Formula 20 = .55).

Turnover among registered nurses. Eacb bospital administrator was askedto supply tbe rate of voluntary turnover for its registered nurses for tbeprevious 12-montb period.

Control variables. Tbe questionnaire sougbt tbe following numerical infor-mation on a bospital's registered nurses: (1) wage rate per bour, (2) ratio offringe benefits to total compensation, (3) number of grievances tbey filed,and (4) percentage of minority employment. It also asked tbe bospital's size(number of beds) and tbe number of hospital beds in tbe county to appraisealternative employment opportunities in tbe region. In addition, it asked ifregistered nurses were unionized. Finally, unemployment rates for eacbbospital's county and standard metropolitan statistical area (SMSA) for tbe12-montb period under study were obtained from tbe researcb and statisticsdepartments of state employment security divisions.

Results

Table 1 gives means and standard deviations of tbe study's variables.Tbe mean rate of turnover among registered nurses was 21.1 percent (s.d. =11.5). Table I also provides Pearson correlations between variables. All correla-tions between variables and turnover were in tbe predicted direction witbtbe exception of tbe positive correlation^ between wage rate and turnover.

^ There is a potential explanation for this finding that Becker (1978), who found similarresults, did not consider. For hospitals with production functions that require high levels ofstaffing at all hours, average hourly wages are higher because they pay premium differentials forstaffing undesirable shifts. In such circumstances, particularly with a shortage of nurses, itwould not be surprising that turnover would be high as employees would leave for moredesirable shifts with other organizations.

1986 Spencer 493

oo

03aO•V

Vi

e

.a

I °I

o o

r I

OI rH

O ) i n

r 'r H C n

' r« «T-t r H O t ^ C O C O OC S r H C N j q q r H r H

I

.i3

Q

1•B

rH O

•* o

CO a>m q

ro COrH r.J

r rO OO * • i ' CO CMrH O M CM O O

«CNl in O CO rH Oq q eg c»5 'n ^,

r r

« -K « *

I c v i r H c s j M p q O r H o q q o

\' ' \' ' \ \'

I O r - t r H r H q ^. ^. ^, ^. ^. ^. PP

r * ' r ' r« * M M *

^ TH C O C D i n ^ tN.CO M ( C O r H C O Ot o m r H r H r H CSI ^ r H T H q w q P ' ^

I r \' (ts.CO in l O N r H O lOr-t CDcntOCD COO)

i q q q T- jMq rH ' -^^ q i ; q q ^ p

' I r ' ' r r ' r« « « « «

I cs jcN i rH rH q r n q q ^ q ^ ' ^ . ^ ' ^ ^ P

r r r r r { { r * i' i rC C O r H O O 5 r H O 5 O>

in M i n r t in f ^ * ^ ^ t ° ° ^ 99"^^ " ^ ^r H • • • - . . . . Q ( o C O ^ C D O O t ^ OT^ tH T-I in CTi^ T T O r H CO r H C O O COTH q i n r H in cq t * . ^ c ^ ^ ' I ^ ^ ^ ^. ^.^ ^ ^ ^ ^ ^ ^ ^ ^ coin cdo>odin cvirnCM CM ts. O5

xi CO in

ll I I _1 I I "lilii.l I III M fill 1

cico^iri

2

XSap.

CO

•a

V

S CO

T3 S

II

s •" ;H

I ° "B 2 "eo s o

^ - ' Oi "H00 U 3S O . .

N §1(A 03 5 (U

"^J cn <0 {J

< a> > o

lit?ill Ia fl s a

I ^ 11 _DH _ 2 a "B ^ g S =!S i l - * VP U S .« D.

494 Academy of Management /ournaJ September

It was hypothesized that the more opportunities employees have to voicedissatisfaction and have input into changing dissatisfying work situations,the greater the numher of employees that will remain with an organization.The composite index of employee voice mechanisms correlated negativelywith turnover among registered nurses (r = -.24, p < .05); the greater thenumber of employee voice mechanisms that a hospital employed, the lowerthe turnover. This significant relationship hetween the composite index andrate of turnover holds when a numher of known predictors of employeeturnover are statistically partialled out (see Tahle 2).

Ordinary.-least-squares regression was considered appropriate for multi-variate analysis of the data (cf. Kochan & Helfman, 1981). As indicated inTahle 2, the multiple correlation of the linear combinations of all controlvariahles, except for unionization, with turnover was fl = .36. Adding union-ization to the predictive equation did not increase the multiple correlationwith turnover. Finally, adding the composite index of employee voicemechanisms, including the unionization variable, to the predictive equationsignificantly increased the multiple correlation with turnover to R = .45 andthe explained variance in the dependent variable to AR = .07. This signifi-cant (t = 2.6, p < .05) increase is reflected by the twofold increase in Rcorrected for shrinkage. Therefore, number of employee voice mechanismsis significantly associated with rate of turnover among registered nurses

TABLE 2Results of Hierarchical Regression Analyses of Unweighted

Composite Index of Employee Voice Mechanisms, Unionization,and Control Variables on Registered Nurse Turnover^

Independent Variables

Hourly wage rateFringe benefitsCounty unemployment rateGrievances filedMinority employmentBeds in hospitalHospital beds in countyUnionizationEmployee voice mechanismsConstantRR^Adjusted R

Step 1:Control

b

3.31-.10-.53

.04

.09

.00

.00

-2.67

1 VariablesP.23

-.06-.14

.15

.13

.03

.03

.36

.13

.05

t*>

1.8-0.5-1.3

1.41.10.20.2

Step 2:Assessmen

Unionizationb

3.37-.10-.52

.04

.09

.00

.00-.72

-3.02

P.23

-.06-.14

.15

.13

.02

.03-.02

.36

.13

.04

tofEffect

fb

1.8-0.5-1.2

1.31.10.20.2

-0.2

Step 3:Assessment ofVoice Effect

b

3.55-.03-.73

.04

.11

.00

.00-.66

-1.923.66

.24-.02-.19

.16

.16

.05

.00-.02-.28

.45*

.20

.10

t*>

1.9-0.1-1.7

1,41.30.40.0

-0.2-2,6*

" Deletion of missing data reduced the number of hospitals on which regression analyseswere conducted to 81.

b Partial t.*p < ,05.

1986 Spencer 495

when a large number of known predictors of employee turnover are con-trolled.

In the composite index employed, each employee voice mechanismreceived equal weight. When a weighted linear combination of the compo-nents of this index was created through regression analysis, the multiplecorrelation with rate of turnover among registered nurses was R = .41. Littleconfidence can be placed in this weighted combination of variables, however,because when the sample was split into two subsamples composed of evenand odd cases, the regression coefficients were not stable and the two sub-samples failed to double-cross-validate at an acceptable level (r = .16, n.s.;r = .31, p < .05).

STUDY TWO

Methods

Sample. Four hospitals that participated in study one volunteered toallow a survey of nonsupervisory registered nurses' perceptions of theirorganizations' voice mechanisms. In two hospitals, one with 53 beds and theother with 250, questionnaires were administered on site during workinghours; response rates were 44.4 and 23.7 percent. In the second two hospi-tals (107 and 112 beds), personnel departments distributed questionnaireswith return envelopes addressed to the researcher to be filled out at theemployees' convenience; response rates were 42.8 and 27.3 percent. Allnonsupervisory registered nurses currently on staff at the hospitals receivedquestionnaires; participation was voluntary.

Nurses' perceptions. Nurses estimated how often they felt they would beeffective in resolving eight work-related problems. The response format rangedfrom 10 to 100 percent of the time on 10-percent intervals. Problems givenwere related to supervision, tasks, co-workers, pay, or promotions. Twoitems represented each of the first three types of problems and one itemrepresented each of the last two types. The reliability for the eight-itemcomposite measure of expectancy of problem resolution was satisfactory(a = .86).

The questionnaire also asked the extent to which nurses agreed or dis-agreed with a series of ten statements (see Appendix) designed to reflect theextent to which the management of their organizations effectively respondedto and used employee voice (a = .89). Finally, three one-item measures askedemployees the extent to which they agreed or disagreed with three state-ments (see Appendix) designed to assess their perceptions of the effective-ness of voice mechanisms, the utility of efforts to change dissatisfyingsituations, and their preference for leaving versus attempting to resolvedissatisfaction.

Results

Table 3 gives means and standard deviations of these variables as well asPearson correlations. All correlations are in the expected direction.

496 Academy of Management /ournal September

1

t1 CO

, to o>1 in in

^ t^ (0CO CO t^

1

1 '^1 in

^ 0)

^ o

in rHCM r-t

O) ^rH O

•* COCO O

1

r

roCM

rinrH

rino1

0>

CO

rH

rCOCO

rtoCM

rCOCM

1

torH

rCOCM

,05

Vato

A

o

II

2_T33CO

_g

a"nia'o

1

mCM

CM

1

O>

m

<DtM

CO

m

COin

a>CM

inCM

in

oCO

•91

CM

CO

CM

CMCO

0CO

to•V

toCM

00CO

0

ts

to

0

CO

CMrH

inCO

CM0

1

I^0

inT-t

COCM

inrH

1

COr-t

1

0CM

0

1986 Spencer 497

TABLE 4Registered Nurses' Perceptions and Relevant Organizational Characteristics

of Four Nonunion, Short Term, General Care Hospitals with VaryingNumbers of Employee Voice Mechanisms"

Variables

Relevant organizational characteristicsNumber of bedsGrievance procedureSuggestion systemEmployee-management meetingsCounseling serviceOmbudsmanNonmanagement task forcesQuestion-and-answer programSurvey feedbackTotal number of mechanismsTurnover among registered nursesTurnover among licensed practical nursesTotal organizational turnover

Perceived expectancy and effectivenessProblem resolution expectancies

Supervisory problemsTask problemsCo-worker problemsPay problemsPromotion problemsComposite expectancy

Management's responsiveness to voiceEffectiveness of voice mechanismsUtility of efforts to change dissatisfying situationsPreference for leaving versus attempting to resolve

dissatisfaction

High Opportunityfori

Hospital1

52YesYesYesNoYes 'YesNoYes

68,3%

16,6%15,0%

59,4%67,5%''75,6%''28,7%36,2%58,7%''4,20''2.87'>3.88>'

2,62''

/oice

Hospital2

250YesNoYesYesNoYesNoYes

520,7%19,4%23,1%

43,5%58,7%56,2%30,9%42,4%48,7%

3,883,483,68

2,20

Low Opportunityfor Voice

Hospital3

112YesNoNoNoNoNoYesYes

310.8%21,6%26,4%

34,5%41,3%47,7%30,9%31,5%38,7%2,811,912,16

3,34

Hospital4

107YesYesYesNoNoNoNoNo

333,3%28,6%40,0%

45,0%55,0%49.7%37,3%43,3%47,5%

3.452.602.47

2,80

a Ns = 8, 46, 33, 15, respectively, for Hospitals 1-4.'' t-test significant at the p < ,01 level (combined perceptions for Hospitals 1 and 2 compared

to combined perceptions for Hospitals 3 tuid 4),

Table 4 reports levels of expectancy and perceptions of the effective-ness of voice mechanisms for hospitals with varying numbers of voicemechanisms. Hospitals 1 and 2 had six and five mechanisms respectively,and hospitals 3 and 4 had three mechanisms apiece. When these hospitalsare treated as two groups, t-tests for differences between means indicatesignificantly higher levels of expectancy and perceptions of voice effective-ness in the pair of hospitals with the larger numbers of voice mechanisms.These differences emerged for all variables except expectancies for resolvingproblems related to supervision, pay, and promotion. These findings sup-port the hypothesis that employees of hospitals with large numbers of

498 Academy of Management Journal September

employee voice mechanisms will have high expectancies for problem reso-lution and will perceive a high level of effectiveness in their organizations'responses to employee voice.

These results should be interpreted with caution. First, the measures ofemployee perception are intercorrelated (see Table 3). Second, Table 4 reportsseparate means for each hospital in order to indicate the extent to whichthese means differ within each of the two designated groups. As these dataclearly indicate, hospital 1 tended to have the highest levels of expectancyand perceptions of voice effectiveness and hospital 3 consistently had thelowest levels. More important, although hospital 2 had more voice mecha-nisms than hospital 4, the levels of expectancy and perceived effectivenessof voice mechanisms were only slightly higher for hospital 2 in most cases.The rate of turnover among registered nurses was also higher for both hospi-tals 2 and 4 than for the other hospitals. Although extrapolation to otherturnover figures is somewhat problematic. Table 4 does indicate that volun-tary turnover for licensed practical nurses and the organization as a wholewas higher for the hospitals with the fewest employee voice mechanisms.

DISCUSSION

The results of study one suggest that the more an organization givesemployees the opportunity to voice dissatisfaction over aspects of their workin order to change dissatisfying work situations, the greater the likelihoodthat its employees will remain with the organization. Registered nurses'turnover rates in short-term, general care hospitals were significantly lowerin hospitals with many mechanisms for the voicing of employees' dissatis-faction. In addition, this relationship held with statistical control for a widevariety of known predictors of rate of employee turnover.

The results of study two, although only suggestive, indicate that highnumbers of voice mechanisms are associated with high levels of employees'expectancies for problem resolution and high perceived effectiveness of anorganization's procedures for resolving problems. These findings also lendcredibility to the construct, number of employee voice mechanisms. Thisimplies that when employees have many opportunities to voice dissatisfac-tion they (1) appear to be predisposed to take advantage of those opportunities,and (2) appear to consider procedures for problem resolution effective.

Although previous research has consistently found an effect for union-ization on retention rate, this research did not find such an effect. Unique-ness in the types of organization and occupation examined may account forthese negative results. Unionization's effects on voice may have been washedout. Data indicated that unionized hospitals averaged 4.42 mechanisms andnonunionized hospitals averaged 4.52 mechanisms; in addition, no signifi-cant differences emerged when each mechanism was considered individually.Perhaps this finding is a result of a union-threat effect; nonunionized hospi-tals may provide voice mechanisms that would otherwise be provided by

1986 Spencer 499

unions in order to reduce dissatisfaction and thus reduce the threat ofunionization.'*

The strengths and limitations of this research are worth considering.This research extended previous work on unions' effects on rates of employeeretention by looking at the potential effects of voice mechanisms themselves.Organizations might institute these mechanisms as a result of unionizationor in attempts to improve management process. Additional strengths are thatthis research controlled for a wide variety of potential predictors of rate ofturnover and included critical information on employees' perceptions of theorganizational characteristics under investigation.

A limitation is that this research does not represent a complete picture ofthe phenomena involved in the relationship between the options of voiceand exit. Two omissions are worthy of note. First, Hirschman's (1970) modelfocused on exiting in silence versus staying and voicing concerns. Addi-tional options open to employees, such as staying in silence and exiting withvoice, need to be incorporated into the model (Barry, 1974; Birch, 1975). Infact. Birch (1975) argued that in many cases people are more likely to exitwith voice than to remain and engage in voice, particularly when there is apossibility of retaliation. Second, this research did not address the issue ofloyalty. Hirschman (1970) hypothesized that those clients or employees whohave greater loyalty to a product or an organization will be more likely thanothers to respond with voice to correct a perceived decline in organizationalperformance. Barry has criticized the loyalty concept as an "ad hoc equationfiller" (1974: 95) that could be inversely as well as positively correlated withvoice (see also Laver, 1976). But it is difficult to deny that loyalty in the formof behavioral commitment to an organization—entrenchment due to length ofservice—is likely to have an inverse relationship with employees' exit rates,whether or not employees have successfully changed dissatisfying states ofaffairs (Mowday, Porter, & Steers, 1982).

An important limitation, addressed in the introduction, is that the actualquality of the voice mechanisms under study was unknown. The secondstudy indirectly addressed this issue by assessing employees' perceptions ofthe effectiveness of voice mechanisms. This assessment, however, cannotsubstitute for an approach that would ideally involve in-depth content analy-sis of the nature and quality of each employee voice mechanism at eachorganizational site under study. This research assessed employees' percep-tions of the effectiveness of voice mechanisms at only four of the organiza-tions represented in the larger study; in addition, sample sizes for theseanalyses were small and measures whose psychometric properties were rela-tively unknown were constructed and used.

* Ciirtin (1970) found that unsuccessful unionization attempts generally result in improvedcommunication practices; in over 60 percent of the companies these practices usually included:"establishment of regular meetings with the employees, institution of a formal grievanceprocedure, formation of an office committee, concerted effort to deal promptly with employeecomplaints" {1970: 67),

500 Academy of Management Journal September

Another limitation of this research is that although significant relation-ships emerged from the analysis, they are not very strong. Also, these rela-tionships are purely associational and thus causality cannot be assumed. Arelated problem concerns synchronicity of the data; assessment of number ofvoice mechanisms was made at the end of the one-year period for whichretention data was collected. Another limitation is that the research wasconducted in one particular type of organization and on one specific occupa-tional classification. A further limitation is the relatively low response rate;however, comparing the responses obtained against population demographicsprovided by each of the hospitals and bed sizes obtained from the NationalCenter for Health Statistics indicated that the responses were representativeof the population of individuals and of hospitals involved. Future researchshould incorporate tests of the hypotheses in different populations and useresearch methodologies where causality could be inferred as well as strate-gies of data collection that would improve rates of response.

The findings of this study reinforce Steers and Mowday's (1981) assump-tion that investigations of the process of employee turnover need to consideremployees' efforts to change unsatisfactory work situations. On the organiza-tional level of analysis, future research should consider not only formalvoice mechanisms and their quality, but also informal organizational cul-tures that create and sustain those mechanisms. It is not inconceivable thatthe rough assessment of total numbers of voice mechanisms that this studyused is tangible evidence of a potentially salient organizational componentsuch as managerial philosophy. The individual level of analysis needs amodel of the process of turnover decisions that incorporates the two majoralternatives available to employees: (1) searching for alternative employ-ment to exit from a dissatisfying work situation, and (2) attempting to changea dissatisfying work situation in an effort to remain. Mobley's (1977) modelof the process of turnover decisions focused only on the first course ofaction. But what causes employees to take one course of action before theother or to take both courses of action simultaneously? Perhaps employees'commitment and expectancies concerning the effectiveness of voice mecha-nisms and procedures for problem resolution would play a significant role inthis process.

If future research supports the hypotheses examined in this study, onepractical implication is that organizations may be able to effectively reduceemployee turnover by increasing the sophistication of their processes forresolving complaints. Employees who search for alternative employmentwithout attempting to change dissatisfying work situations give organiza-tions no indication that anything is wrong until after they find other workand hand in their resignations. By listening to, encouraging, and providingmechanisms for employees who want to change dissatisfying work situations,perhaps organizations could prevent employees' forming an intent to leave.In addition, organizations that engage in such activities should be able toobtain information from employees that could improve the effectiveness of

1986 Spencer 501

these organizations (Freeman, 1976; Hirschman, 1976), information that isnot available when employees silently leave.

REFERENCES

Aram, J. O., & Salipante, P. F., Jr. 1981. An evaluation of organizational due process in theresolution of employee/employer conflict. Academy of Management Beview, 6:197-204.

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

Becker, B. 1978. Hospital unionism and employment stability. Industrial Belations, 17: 96-101.

Birch. A. H. 1975. Economic models in political science: The case of "Exit, voice, and loyalty."British Journal of Political Science, 5: 69-82.

Curtin. E. R. 1970. VWiife collar unionism. Personnel policy study No. 220, National IndustrialConference Board, New York.

Fsirrell, D. 1983. Exit, voice, loyalty, and neglect as responses to job dissatisfaction: A multidi-mensional scaling study. Academy of Management Journal, 26: 596-607.

Farrell, D., & Peterson. J. C. 1982. Patterns of political behavior in organizations. Academy ofManagement Betriew, 7: 403—412.

Freeman. R. B. 1976. Individual mobility and union voice in the labor market. American Eco-nomic Beview, 66: 361-368.

Freeman, R. B., & Medoff. J. L. 1984. What do unions do? New York: Basic Books.

Graham, J. W. 1986. Principled organizational dissent: A theoretical essay. In B. M. Staw & L. L.Cummings (Eds.), Besearch in Organizational Behavior, vol. 8: In press. Greenwich,Conn.: JAI Press.

Greenberg, J., & Folger, R. 1983. Procedural justice, participation, and the fair process effect ingroups and organizations. In P. B. Paulus (Ed.), Basic group processes: 235-256. NewYork: Springer-Verlag.

Hammer, T. H.. Landau, J. C, & Stern, R. N. 1981. Absenteeism when workers have a voice: Thecase of employee ownership. Journal of Applied Psychology, 66: 561—573.

Hirschman, A. O. 1970. Exit, voice, and loyalty. Cambridge, Mass.: Harvard University Press.

Hirschman, A. O. 1974. Exit, voice, and loyalty: Further reflections and a survey of recentcontributions. Social Science Information, 13(1): 7-26.

Hirschman, A. 0.1976. Some uses of the exit-voice approach—Discussion. American EconomicBeview, 66: 386-391.

Kochan, T. A., & Helfman, D. E. 1981. The effects of collective bargaining on economic andbehavioral job outcomes. In R. G. Ehrenberg (Ed.), Besearch in labor economics: 321-365.Greenwich. Conn.: JAI Press.

Laver. M. 1976. "Exit, voice, and loyalty" revisited. British Journal of Political Science, 6:463-482.

March. J. G., & Simon. H. A. 1958. Organizations. New York: John Wiley & Sons.

Mobley. W. H. 1977. Intermediate linkages in the relationship between job satisfaction andemployee turnover. Journal of Applied Psychology, 62: 237-240.

Mobley, W. H., Griffeth, R. W., Hand. H. H., & Meglino, B. M. 1979. Review and conceptualanalysis of the employee turnover process. Psychological Bulletin, 86: 493-522.

Mowday, R. T., Porter, L. W., & Steers, R. M. 1982. Employee-organization linkages: Thepsychology of commitment, absenteeism, and turnover. New York: Academic Press.

502 Academy of Management /ournal September

Price. J. L. 1977. The study of turnover. Ames: Iowa State University Press.

Steers, R. M.. & Mowday. R. T. 1981. Employee turnover and post decision accommodationprocesses. In L. L. Cummings & B. M. Staw (Eds.). Besearch in Organizational Behavior:235-281. Greenwich. Conn.: JAI Press.

Ulman. L.. & Sorensen. E. 1984. Exit, voice, and muscle: A note. Industrial Relations, 23:424-428.

APPENDIX

All items used 7-point response formats; "R" indicates reverse scoring.Management's responsiveness to voice mechanisms:

The management of my organization is not interested in resolving individual employeeproblems. (R)

The management of my orgeinization encourages employees to voice their problems.The management of my organization has not provided enough mechanisms (for example,

suggestion systems, grievance procedures, etc.) to allow employees to effectively voicetheir dissatisfaction. (R)

I get the feeling that my superior does not want to hear about my complaints. (R)The personnel manager is open to receiving complaints.My boss comes around regularly to keep in touch with any complaints that I may have.The personnel manager makes an effort to keep in touch with any complaints that workers

have.I feel intimidated by my superiors when pursuing a grievance. (R)Management views grievances as a challenge to their authority. (R)The organization encourages suggestions to improve situations that are dissatisfying to

employees.Effectiveness of voice mechanisms:

The mechanisms to resolve employee problems of my organization are very effective.Utility of efforts to change dissatisfying situations:

Attempting to change something at work that dissatisfies me would be a waste of time. (R)Preference for leaving versus attempting to resolve dissatisfaction:

I would rather find a job elsewhere than attempt to change something at work that dissatis-fies me.

Daniel G. Spencer is an associate professor of business at the University of Kansas.Lawrence. Kansas. He received his Ph.D. degree in organization and management fromthe University of Oregon. His current research interests include turnover, absenteeism,and processes of resolution of employee-organization conflicts.