attitudinal organizational commitment and job performance

11
Attitudinal Organizational Commitment and Job Performance: A Meta-Analysis Author(s): Michael Riketta Source: Journal of Organizational Behavior, Vol. 23, No. 3 (May, 2002), pp. 257-266 Published by: John Wiley & Sons Stable URL: http://www.jstor.org/stable/4093803  . Accessed: 24/01/2011 17:26 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at  . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at  . http://www.jstor.org/action/showPublisher?publisherCode=jwiley . . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].  John Wiley & Sons  is collaborating with JSTOR to digitize, preserve and extend access to  Journal of Organizational Behavior. http://www.jstor.org

Upload: da-niel

Post on 14-Apr-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 1/11

Attitudinal Organizational Commitment and Job Performance: A Meta-AnalysisAuthor(s): Michael RikettaSource: Journal of Organizational Behavior, Vol. 23, No. 3 (May, 2002), pp. 257-266Published by: John Wiley & SonsStable URL: http://www.jstor.org/stable/4093803 .

Accessed: 24/01/2011 17:26

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .http://www.jstor.org/action/showPublisher?publisherCode=jwiley. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of 

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

 John Wiley & Sons is collaborating with JSTOR to digitize, preserve and extend access to Journal of 

Organizational Behavior.

Page 2: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 2/11

Journal of Organizational Behavior

J. Organiz. Behav. 23, 257-266 (2002)

Published online 1 March 2002 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/job.141

Attitudinal organizational commitmentand job performance: a meta-analysis

MICHAEL RIKETTA*

UniversityfMannheim, ermany

Summary A meta-analysiswas conducted to estimatethe truecorrelationbetween attitudinalorganiza-tional commitmentandjob performanceand to identify moderatorsof this correlation.One-

hundredand eleven samples from 93 publishedstudies were included. The correctedmean

correlationwas 0.20. The correlationwas at least marginallysignificantly strongerfor: (a)extra-role performanceas opposed to in-role performance; (b) white-collar workers as

opposedto blue-collarworkers;and (c) performanceassessed by self ratingsas opposedto

supervisor atingsor objectiveindicators.Fourother assumedmoderators commitmentmea-

sure:Affective Commitment Scale versus OrganizationalCommitmentQuestionnaire, oblevel, age, and tenure) did not have at least marginally significant effects. Copyright ?2002 JohnWiley & Sons, Ltd.

Introduction

According to its most often cited definition, attitudinal (or affective) organizational commitment(AOC) is 'the relative strength of an individual's identification with and involvement in a particular

organization' (Mowday, Steers, & Porter, 1979, p. 226). This variable is one of the most often studied

variables in organizational behavior research (for recent reviews see Mathieu & Zajac, 1990; Meyer &

Allen, 1997). Probably the main reason for the extensive and long-lasting research interest in AOC is

that it is assumed to influence almost any behavior that is beneficial to the organization such as per-

formance, attendance, and staying with the organization (see Mathieu & Zajac, 1990; Meyer & Allen,

1997; Mowday, Porter, & Steers, 1982; Randall, 1990). The present study focuses on the relationship

between AOC and performance in particular.The assumption that employees who feel attached to and identify with their organization work

harder, is a popular one and may provide the rationale for many organizational attempts to foster

employees' organizational commitment or identification. Given its popularity, an empirical test of this

assumption is urgent. A prerequisite for a causal influence of AOC on performance is that both

variables are correlated. However, previous quantitative reviews suggest that the AOC-performance

correlation is moderate at best (Allen & Meyer, 1996; Cohen, 1991; Mathieu & Zajac, 1990; Mowday

et al., 1982; Organ & Ryan, 1995; Randall, 1990). For example, Mathieu and Zajac (1990) reported an

*Correspondenceo:MichaelRiketta, epartmentf SocialPsychology, niversityfMannheim,8131Mannheim,ermany.E-mail:[email protected]

Received 6 March2001Revised 12 September2001

Copyright? 2002 JohnWiley & Sons, Ltd. Accepted 22 January2002

Page 3: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 3/11

258 M. RIKETTA

estimatedtrueAOC-performance orrelationof r= 0.13 (k= 8) and Randall 1990) andCohen(1991)reportedestimated truecorrelationsbetween organizationalcommitment(affective as well as calcu-

lative) andperformanceof r= 0.21 (k= 7) and0.13 (k= 14), respectively.Although these correlations may appear disappointingly low, their relevance to the AOC-

performance elationship s limiteddue to a numberof shortcomingsof the mentionedreviews.First,

all of these reviews used only few samplesreportinganAOC-performance orrelation ks< 14). Thisis only a small partof the relevantempiricalresearchthat is availabletoday.Second,all of thementionedreviewsused eithertoo restrictedor too comprehensive amplesof stu-

dies. In particular,Randall 1990) andCohen(1991) didnotdistinguishbetweenAOC andotherformsof organizational ommitment normativeandcalculative) n theiranalysespertaining o performance.Given the conceptualand empiricaldifferencesbetweenthese three commitment ypes (see Allen &

Meyer, 1996; Mathieu & Zajac, 1991; Randall, 1990), it may be that Randall'sand Cohen'sresultswould not replicatefor AOC in particular.Allen and Meyer (1996), Mathieuand Zajac (1990), and

OrganandRyan (1995) did focus on AOC but consideredonly studiesthat,respectively,employeda

specificAOCmeasure theAffectiveCommitmentQuestionnaire ACS]by Allen & Meyer,1990),useda specific operationalization f performance objectiveindicators),and focused on a specific perfor-mance type (extra-rolebehavior).It is not clear whether the conclusions of these three studies are

generalizable o otherAOCmeasures,performancemeasuresandperformanceypes, respectively.The aim of thepresent tudy s to overcome heseshortcomings.treportshe resultsof ameta-analysis

that s based on a comprehensiveampleof studiesdealingspecificallywith theAOC-performanceela-

tionship.This meta-analysiss not only to providean updatedand specificestimate of the trueAOC-

performanceorrelation ut alsoto identifymoderators f thiscorrelation.nthisrespect, hisstudyrepli-cates andextendsthe meta-analyticmoderator nalyses by Cohen(1991) andRandall(1990).

Twoclasses of moderators reconsideredherein.The firstclass comprisestwo methodologicalvari-ables: the operationalizationof performanceand of AOC. Randall(1990) alreadyinvestigatedthesemoderators ndfoundstrongerorganizational ommitment-workbehaviorcorrelations orself-reportsand objective indicatorsthan for supervisorreportsof performanceand for Mowday et al.'s (1979)pervasiveOrganizationalCommitmentQuestionnaire(OCQ) than for other commitmentmeasures.

However,Randall'smoderatoranalyseshave the drawback hatthey

do notpertain

to AOC andper-formanceper se. Rather,she included studies pertaining o both affective and calculativeorganiza-

tional commitment andused a composite index of workbehavior,which encompassedperformance,tardiness,absentism, turnover,andeffort.

The second class of moderators omprisessubstantiveones. Theyfollow fromthepervasivehypoth-esis that the impact of AOC on performance s positively correlated with autonomyat work (e.g.,Kalleberg& Marsden,1995; Meyer & Allen, 1997, p. 39; van Knippenberg,2000; see also Judge,Thoresen,Bono, & Patton,2001). Cohen(1991) andRandall(1990) testedthis hypothesiswith threeindicatorsof autonomy.Randall assumed thatwhite-collarworkershavemoreautonomyat work thanblue-collarworkers. In line with this, she found thatthey displayeda strongercommitment-perfor-mance correlation thanblue-collar workers.However, the already mentioned shortcomingsof hermoderatoranalysisapplyalso to this finding.Cohenassumedthatemployees cumulaterelevantwork

experiencein the course of time andthusincrease theirautonomy.Providingsupport or this assump-tion, he found strongercommitment-performance orrelations or sampleswith older meanage and

longermean tenure.As alreadymentioned,however,he used only a small sampleof studies(k= 14)and,like Randall,didnotdistinguishbetweenaffective andother formsof commitment.Inthepresentstudy,the impactof the same three moderators s investigated.

In addition,this studydeals with the moderating mpactof two furtherpossible indicatorsof auton-

omy. The firstone is performance ype: in-role versusextra-role. In-roleperformance s defined asbehaviorrequiredby formaljob descriptions.Extra-roleperformance s defined as behaviorthat is

Copyright 2002JohnWiley& Sons,Ltd. J. Organiz. ehav. 3, 257-266(2002)

Page 4: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 4/11

ORGANIZATIONALCOMMITMENTAND PERFORMANCE 259

beneficial to theorganizationand also goes beyondformal ob requirements e.g., extrahours,altruis-

tic behavior,anddonating).Because extra-rolebehavioroften is voluntary,t shoulddependon intrin-

sic motivational actorsto a greaterextent than does in-rolebehavior.Thus,AOCshould relate more

stronglyto extra-rolebehaviorthan to in-role behavior.A comparisonof previousmeta-analysespro-vides preliminarysupportfor this hypothesis: Mathieu and Zajac's (1990) correlation (r= 0.13)

between AOC and performance obviously in-role behaviorfor the most part)is lower thanOrganand Ryan's (1995) correlations(r= 0.23, k=5, and r= 0.30, k = 4) between AOC and two facets

of extra-rolebehavior.

The second additionalassumedmoderators job level (supervisorversussubordinate). t is assumed

thatsupervisorshave moreautonomythan subordinates.So the AOC-performance orrelationshould

be strongeramong supervisors.

Organizational Context

Features of the Samples Included in the Meta-AnalysisThe resultsfor most (75 per cent) of the 111 samples analysedherein were publishedin the years

1990-2001, 20 percent in theyears 1980-1989, and 5 percent in theyears 1975-1980. The average

sampleconsisted of 59 percent men and41 percent women (gender proportionswere reported or

38 percent of the samples).Meanage and tenureacrosssampleswere 35.93 and 6.90 years,respec-

tively (ageandtenure nformationwas available or 56percentand 43 percentof thesamplesrespec-

tively). The huge bulk of the samples (86 per cent) was drawn from Anglo-Americancountries

(above all, the USA, 81 per cent), 4 per cent from from the Europeancontinent (in particular,

Germany,Belgium, andNetherlands),4 per cent fromeasternAsian countries(in particular, apan,

Korea,andSingapore),and 3 percent fromIsrael; he nationalitiesof the othersamples(4 percent)were mixed or not evident from the respectivestudies. Most sampleswere drawnfrom the service

sector, n particular: 8percent from financialserviceorganizationsbanks, nsurances,and account-ing firms), 16 per cent from health or social service organizations(above all, hospitals), and

14 per cent from othernon-publicservices (e.g., food, retailing,and research and development).In addition, 14 per cent of the samples were from the public sector, except health and social

services (e.g., education,police, and armedforces), and 8 per cent from manufacturingirms;the

other samples (27 per cent) were from unspecified or diverse industries.The most prominent

occupational groupsamong the analysed samples were salespeople (18 per cent of the samples)and nurses (5 per cent); the remainderof the samples (77 per cent) comprisedother,unspecified,or diverseoccupationalgroups.

Method

Search or relevant studies

Only publishedstudies were includedin the meta-analysis.It was decidednot to conducta search for

unpublished indingsbecause therewas reason to assume that such a searchwould resultonly in a very

Copyright 2002JohnWiley& Sons,Ltd. J. Organiz. ehav. 3, 257-266(2002)

Page 5: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 5/11

260 M.RIKETTA

selective sample of unpublishedstudies. For example, in their meta-analysisof the commitment-

loyalty literature,Tett andMeyer (1993, p. 266) mentionthat from 33 dissertationauthorssolicitedfor information,only four answered and eventually two providedusable information.Consideringsuch a small partof the unpublishedresearch,therefore,would not substantiallyreduce file-drawerbias (i.e., the bias resultingfrom consideringonly publishedresearch;Rosenthal, 1991); rather, t

may introduce otherbiases, dependingon the reasons for the solicited authors'cooperationor non-cooperation.The problemof file-drawerbias is addressedagainin the Discussion section.

To identifyrelevantpublishedstudies,a searchwas conducted n the electronicdatabasesPsycLIT(covering the years 1887-2001, April), ABI/INFORM (covering 1971-2001, June), and SocialSciences CitationIndex(covering1998-2001, June)for one of thekeywords 'organizational ommit-ment' and 'organizational dentification'alongside one of the keywords 'performance', 'in-role','extra-role', and 'organizationalcitizenship'. Moreover,the reference lists of previous reviews ofAOC researchwere inspected.

Only those studiesshould be consideredfurther hatdealt withAOC rather hancalculative or otherforms of organizational ommitment.To accomplishthis,all studiesusinga scale explicitlydevised tomeasure either AOC(e.g., ACS or OCQ)orthe relatedconstructsof organizationaldentificationandinternalizationof organizationalvalues (e.g., the scales by O'Reilly & Chatman,1986)wereretained,

whereas all studies using a scale explicitly devised to measurenon-affectiveforms of commitment(e.g., Hrebiniak& Alutto's, 1972,calculativecommitmentscale;Allen & Meyer's,1990,ContinuanceCommitmentScale andNormativeCommitmentScale) were discarded.When a studyuseda commit-mentscale the type of whichwas not specified,the studywas includedonly if the scale had some face

validityas an AOC measure(i.e., if at least one item seemed to tap affective attachment o the orga-nization).Face validitywasjudged by the author; ive of the samplesincludedin the final meta-ana-

lysis werejudgedthatway.Moreover,studiesusingmeasuresof effort as indicatorsof performance r

measuringAOCotherwise thanby self-reportswerediscarded.Fromthe remainingstudies,only thosestudies were retained hatreportedzero-ordercorrelationcoefficients or dataallowingcomputationofsuch correlationcoefficients (e.g., t and F values;cf. Hunter& Schmidt, 1990).

The finalsampleforthe meta-analysiscomprised111 individualsamples(n = 26 344) from 93 pub-lished studies. Sixty-ninestudies (74 per cent) have notbeen included in

anyof the

previous quanti-tativereviews of the AOC-performance elationship Allen & Meyer, 1996; Cohen, 1991;Mathieu&

Zajac, 1990;Mowdayet al., 1982;Organ&Ryan, 1995;Randall,1990).A list of the individualstudiesand theircharacteristics s available from the authoruponrequest.

Coding of sample characteristics

All sample characteristics(including correlationcoefficients and reliabilities) were coded by theauthorand anindependent ater. nterrater greementwas at least 87 percent foreveryvariable.Incon-sistencies thatwere not due to errorswere resolvedby discussion.

Performanceypewas codedin-role,extra-roleor mixed. If thestudyauthorsexplicitlystatedwhich

typeof performancehey soughtto measure, heirsamplewascodedcorrespondingly,.e. in each othercase, the codersinspectedtherespectiveperformancemeasuresto find outwhetherthey tapped n-role

or extra-roleperformanceor both. The above definitionsserved as guidelinesfor thisjudgment. Typeof workerwas coded blue-collar (i.e., all study participantswere blue-collarworkers),white-collar

(i.e., all studyparticipantswerewhite-collarworkers)or mixed/notstated. Job level was codedsuper-visor (i.e., all studyparticipantsheld supervisoryor managerialpositions),subordinate(i.e., none ofthe studyparticipantsheld supervisoryor managerialpositions),ormixed/notstated.Age andorgani-zational tenurewerecodedby theirsamplemeans into Cohen's(1991) categories:upto 29, 30-39, and

Copyright? 2002 JohnWiley & Sons, Ltd. J. Organiz.Behav.23, 257-266 (2002)

Page 6: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 6/11

ORGANIZATIONALCOMMITMENTAND PERFORMANCE 261

40 years and over for age and up to 2, 3-8, and 9 and over for tenure. Commitment measurewas

coded into two categories denoting the two most pervasive measures, ACS and OCQ, and into

others/mixed/not tated. The labels AOC and OCQ were used for both the full-length version and

shortenedversionsof eithermeasure.Withinthe OCQcategory, specificlabels were assignedto each

of the two mostpervasiveversions of theOCQ:the 9-item version and the 15-item version.Sourceof

performancedata was coded self-ratings, supervisor ratings,peer ratings, objective indicators, orothers/mixed/notstated. Self-reported supervisor ratings and self-reported objective indicators

were codedsupervisorratingsandobjectiveindicators,respectively.Because thepeer ratingcategory

comprisedonly five samples,it was not includedin the moderatoranalyses.However,as described n

the next section, peer ratings were corrected for unreliability in a special manner.Finally, the

percentageof women in the sample, type of organization,occupationsof participants,and countrywhere the investigationwas conductedwere coded for descriptivepurposes(see ContextualSidebar

for the categoriesandresults).When a study reported separatecorrelationspertainingto differentlevels of the same moderator

(e.g., for bothin-role andextra-rolebehavior),the correlationswere averagedacross moderatorevels

for all analyses except the analysisof the effect of that moderator. n the lattercase, the separatecor-

relationswere used. This was done with the moderators'performance ype' and 'source of perfor-

mance data'. In averaging across moderatorlevels, the same formula was used as in averagingacross samples (see next section).

Meta-analyticprocedure

The present study employed the meta-analyticmethods of Hunter and Schmidt (1990). Hunterand

Schmidtsuggestedthat a meta-analysisnotonly aggregatedataacross studiesbut also correctthe data

for artefactsas far as possible. The currentmeta-analysiscontrolledfor the artefactsof samplingand

measurementerror.

In the firststep, every individual correlationcoefficient was dividedby the square-rootof the reli-

abilities of the involved variables. With some exceptions, which are describedin turn,the sample-

specific reliability coefficients (usually internalconsistency coefficients) reportedin the respective

studywereused. When the authorsdid notreportreliabilitycoefficients,the averagereliabilitycoeffi-

cient for each variable acrossall samples includedin the meta-analysiswas used. Objective perfor-mance indicators for which no reliabilitycoefficient was reportedand factor scores were assigneda

reliabilitycoefficient of 1.00. Moreover,following the recommendationsby Viswesvaran,Ones, and

Schmidt(1996), interrater eliabilityrather han internalconsistencywas used to disattenuate orrela-

tions involving supervisorandpeerratingsof performance.Because no studyincludedin the present

meta-analysisreportedinterrater eliabilities, Viswesvaranet al.'s meta-analyticalestimates of the

interrater eliabilityof supervisorratings(0.52) andpeer ratings(0.42) were used to correct correla-

tions computedfrom suchratings.In the next step, the correlationcoefficientswere averagedacrosssamplesaccordingto the recom-

mendationsby Hunter and Schmidt (1990, pp. 148-150). Specifically, every correctedcorrelationcoefficient was weighted with the productof sample size and the reliabilitycoefficients for the two

correlatedvariables. Then the weighted coefficients were summed and divided by the sum of the

weights.The result is anestimate of the truepopulationcorrelation p). Note that this estimateis neces-

sarilyflawedby all artefactsnot correctedfor here:all artefactsbesides measurementerrorand sam-

pling error.Reliabilitycoefficients were averaged analogously,with sample sizes as weights.Anotherpopulationparameterof interest was the variance of the truepopulationcorrelations.The

estimate recommendedby Hunter and Schmidt (1990, p. 150) was employed here, that is, the

Copyright 2002 JohnWiley& Sons,Ltd. J. Organiz. ehav. 3, 257-266(2002)

Page 7: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 7/11

262 M. RIKETTA

difference between the variance of the corrected correlation coefficients and their averagesquaredstandarderror. The latterterm is an estimate of the varianceattributable o the correctedartefacts.

The statisticalsignificanceof the estimatedvariance of the populationcorrelationwas computedwith HunterandSchmidt's(1990, p. 151) Qtest. A significantresultpoints to the existence of mod-

erators.The statisticalsignificanceof specificmoderator ffects was testedwithHunterandSchmidt's(1990, pp. 437-438) z test. This test reveals the significanceof the difference in observedmean cor-relationcoefficients (correctedonly for samplingerror)between two subsamples n a meta-analysis(here:between two subsamplesrepresentingdifferent evels of the respectivemoderator).A prerequi-site for the z test is thatthe comparedsamplesareindependent.Therefore,fromeach samplecontri-

buting correlationsto more than one level of the moderator e.g., correlationsfor both in-role andextra-rolebehavior),only one correlationwas included in the moderatoranalysis.This was alwaysthe correlationat the moderator evel for which fewer sampleswere available.

All meancorrelations r) reportedn the followingarecorrected orsamplingerrorandattenuation.All ps reportedin the following are two-tailed, with a significance level of p < 0.05. Effects with

p < 0.10 are consideredmarginallysignificant.

Results

Table 1 shows the results of the meta-analysis.The mean corrected correlationbetween AOC and

performancewas 0.20 (k= 111). The 95 per cent confidence interval did not include zero; so thecorrelationwas statistically significant.

Moreover,62 percent of the varianceof theobservedAOC-performance orrelationswere not attri-butable to the controlledartefacts.HunterandSchmidt(1990) assume that if this proportionexceeds25 percent, the existence of moderatorss likely.The Qtest for

significanceof

unexplainedobserved

variancepointsto the samedirection-Q = 2(112)= 300.17, p < 0.001. Theseresultswereprerequi-site for the moderatoranalyses.

Fromthe methodologicalvariables,only source of performancedata had a marginallysignificanteffect: the correlationwas stronger or self-ratingsof performance r = 0.24) than for supervisorrat-

ings (r= 0.19) and objective indicators(r = 0.13) (ps = 0.09 and0.10, respectively).However,the

AOC-performancecorrelationdid not dependon the commitment measureused. Althoughthe ACS

yielded a slightly strongercorrelation r= 0.23) than the OCQin general (r= 0.18) and its two most

pervasiveversions (the 9-item and 15-item version, rs= 0.19 and 0.18), none of these differencesreachedsignificance(ps > 0.18).

From the substantiveassumedmoderators,only job type and workertype had at least marginallysignificanteffects. Both effects were in line with the predictions.First, AOC related signifcantly

more stronglyto extra-roleperformance r= 0.25) thanto in-role performance r= 0.18), p = 0.03.Second, the AOC-performancecorrelationwas significantlystrongeramong white-collar workers

(r= 0.20) thanamongblue-collar workers(r= 0.10) (p = 0.01). A problemwiththe worker ype ana-

lysis is that therewere only four samples in the blue-collarcategory.Nonetheless, the categorywas

analysedhere to allow for a tentative test of the relevanthypothesis.The differences for job level, age, and tenure were non-significant (p = 0.43, ps >0.13, and

ps > 0.16, respectively).Contraryo thepredictions, heAOC-performance orrelation vendecreasedas age and tenureincreased.

Copyright 2002 JohnWiley& Sons,Ltd. J. Organiz. ehav. 3, 257-266(2002)

Page 8: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 8/11

ORGANIZATIONALCOMMITMENTAND PERFORMANCE 263

Table1. Results f themeta-analysis

Moderator k n r re SD p CI z

Total 111 26344 0.146 0.198 0.108 0.032,0.363Commitment easure

1. ACS 21 5072 0.174 0.233 0.086 0.071,0.395

2. OCQ all versions) 65 15511 0.132 0.181 0.093 0.014,0.348 1.363. OCQ-9 21 4322 0.126 0.178 0.047 -0.002, 0.358 1.21a4. OCQ-15 28 7099 0.142 0.191 0.116 0.032,0.351 0.73,b0.20c

Source f performanceata1.Objectivendicators 18 5801 0.111 0.125 0.123 -0.001, 0.2592. Self-ratings 32 8060 0.183 0.235 0.138 0.085,0.3853. Supervisoratings 59 14906 0.131 0.194 0.041 0.015,0.374 1.65,t0.64, 1.72td

Age1. Up to 29 years 10 1385 0.241 0.300 0.138 0.104, 0.4962. 30-39 years 34 8282 0.162 0.231 0.103 0.061, 0.4003. 40+ years 18 4879 0.145 0.198 0.121 0.042, 0.355 1.27, 1.46, 0.42d

Tenure1. Up to 2 years 8 1213 0.209 0.297 0.027 0.077, 0.5152. 3-8 years 24 5183 0.169 0.238 0.141 0.064, 0.412

3. 9+ years 16 4654 0.138 0.203 0.046 0.039, 0.367 0.68, 1.37, 0.70dJob level

1. Supervisor 9 1774 0.167 0.200 0.107 0.034, 0.3662. Non-supervisor 44 11 272 0.128 0.178 0.089 0.016, 0.339 0.78

Performance ype1. In-role 87 20973 0.130 0.178 0.096 0.011, 0.3442. Extra-role 42 10747 0.185 0.252 0.093 0.093, 0.412 2.20*

Worker ype1. Blue-collar 4 1024 0.067 0.098 0 -0.085, 0.2812. White-collar 84 17554 0.150 0.201 0.124 0.026, 0.375 2.55*

Notes: k-number of averagedcorrelations;n-number of individuals;r-mean correlationcorrected orsamplingerror; ,-mean correlationcorrected for sampling error and attenuation;SD p-estimated standarddeviation of the populationcorrelations;CI-95 per cent confidenceinterval for re;z-result of the significancetest on the difference in r between two

moderatorevels(levels 1

and2except

where statedotherwise).*p< 0.05;tp< 0.10.

aModeratorevels 1versus 3. bModeratorevels 1versus4. cModeratorevels 3 versus4. dModeratorevels 1versus2, 1versus3,and2 versus 3, respectively.

Discussion

When interpretinghe results,the readershouldkeep in mind three limitationsof this meta-analysis.

First,only publishedstudieswereconsidered.Exclusionof null findingsfrompublication(file-drawer

bias) may have inflatedthe estimatedtruecorrelation.However,file-drawerbias may be less of a pro-

blem herebecause in the analysedstudiesthe AOC-performancecorrelationwas often reportedonlyas an ancillaryresult.In this case, the non-significanceof this correlationmay not have affectedthe

publicationchancesof the respectivestudy.Resultsreportedby Allen andMeyer (1996, Table5) are

conclusive in this context. These authorsreportedsix AOC-in-role performancecorrelationsand

five AOC-extra-roleperformancecorrelations romfour andthreeunpublishedstudies,respectively,all of whichused the ACS (samplesizes and reliabilitieswere notreported).Theunweightedmeansof

these correlationswere 0.188 and 0.248, respectively,and thus even largerthan the corresponding

unweightedmean correlations or the publishedACS studies analysedherein (0.157 and0.183).

Copyright? 2002 JohnWiley & Sons, Ltd. J. Organiz.Behav.23, 257-266 (2002)

Page 9: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 9/11

264 M.RIKETTA

Moreover, heproblemof file drawerbiasdoes notchallengethepracticalsignificanceof thepresentmoderatoranalyses because file drawer bias is more likely to blur moderatoreffects ratherthanto

inflate them.If realmoderators xist, studiespertaining o the levels with the smaller true effect sizes

have less of a chance of yielding significantresults than studies pertainingto the other moderator

levels. Hence, providedthatsignificantresults have a better chance to be publishedthannon-signifi-

cantones, there aremoreunpublishednon-significantstudies

pertainingo the moderatorevels with

smallertrueeffect sizes thanto the othermoderatorevels. So the publishedeffect sizes for the former

moderatorevels shouldin totalbe morestrongly upwardlybiasedthanthe publishedeffects sizes for

theothermoderatorevels. Thus,thefact thatonly publisheddatawereconsideredherein ikely leadto

an underestimationof the true moderator effects. This renders the significant moderatoreffects

obtained hereineven moreremarkable.

A secondlimitationof thisstudy s thatthe correlationsoryoungemployees,low-tenureemployees,

supervisors, and blue-collar workers were based on only few (:510) samples. Hence, these

correlationsmay be alteredby few additionalstudiesor mayhave been substantivelybiasedby single

non-representativeindings.Therefore,the moderatoranalysesfor age, tenure, ob level, and worker

type are somewhatpreliminary.

Finally, in the populationof the analysed studies, employees from Anglo-Americancountries

(especially the USA) and white-collarworkers(especially salespeople)were clearly overrepresented(see ContextualSidebar).As a consequence,one shouldbe particularly autiouswith generalizing he

presentresults to other,especially collectivistic (e.g., Asian), culturesandto blue-collarworkers.

This having been said, the researchand practical implicationsof the results are outlined in the

following. The estimated true AOC-performancecorrelationobtained herein (0.20) was similarly

strongas the correspondingestimates reportedin the previous meta-analysesof the commitment-

performancerelationship(Cohen, 1991; Mathieu & Zajac, 1990; Randall, 1990; see introductory

section). Thus, after one decade of additionalresearch,one has still to conclude that the AOC-

performancecorrelation s weak. However,whereasthe correlationsreported n those meta-analysesare based on 14 samplesor less, the correlationreportedhereinis basedon 111 samples.Hence, it is

less likely than t was with thepreviousmeta-analyses hat additionalresearchwill altertheestimateof

the trueAOC-performancecorrelation.

Furthermore,he presentstudywas concernedwith moderatorsof the AOC-performance elation-

ship. One methodologicalvariable(source of performancedata) and two substantivevariables(job

type and workertype) turnedout to be at least marginallysignificantmoderators.One furthermeth-

odologicalvariable commitmentmeasure)and threefurther ubstantivevariables age, tenure,and ob

level) did nothavesignificantmoderator ffects, with the tendenciesfor age and tenurebeingcontraryto expectations.Thus, the autonomy-moderatorhypothesis,which was used to predictthe effects of

the substantivemoderators, ecieved only mixed support.A reasonfor thenon-significanceand thepartlyunexpecteddirectionsof theeffects by age, tenure,

andjob level may be that thosevariablesdo not constituteadequateoperationalizations f autonomy.

Rather, hey may be confoundedwith a numberof othervariablesthatmay moderatethe AOC-per-formancecorrelation e.g., economic dependencyon thejob [Brettet al., 1995],workload,andhealth

status).These variablesmay have effects thatrun counterto the effects

by autonomy.Hence,future

researchshould test the autonomy-moderatorhypothesis more directly,either by using self-report

measures of autonomy(but see Kalleberg& Marsden, 1995, for a null finding obtainedwith this

method)or by experimentalmanipulationsof this variable.

It shouldbe mentionedthatoriginallyit was intendedto explorethe moderating mpactof an addi-

tional methodological feature-study design (longitudinal versus cross-sectional). Also Randall

(1990) included this variable in her meta-analysisand found a non-significantlyweakercorrelation

for longitudinalstudies. However, a replicationof this analysis turnedout to be problematichere

Copyright 2002 JohnWiley& Sons,Ltd. J. Organiz. ehav. 3, 257-266(2002)

Page 10: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 10/11

ORGANIZATIONALOMMITMENTND PERFORMANCE 265

because the longitudinalstudies ncluded n thepresentmeta-analyseswereextremely heterogenous n

termsof the reported ime lag between measurementof commitmentandperformance rangingfrom

two weeks to fouryears).Hence, it would havebeen necessaryto divide the longitudinalstudiesinto

subgroupswith different time lags to allow for moremeaningful analyses.Yet, this was not possiblebecause the time lag was reported or only eight samples in the longitudinalcategory.This suggests

thatresearchers tudyingtheAOC-performanceinkprovidedetailed nformationaboutthetimewhentheirmeasureswere collected so thatthe moderating mpactof designcan be assessed in futuremeta-

analyses.Now that a reliable (thoughweak) correlation between AOC and performancehas been demon-

strated, he questionof causalityarises. Moderatoranalysesarebut one way to test causalhypotheses.Other suitable methods are experimentsand crucial tests of alternative structuralequationmodels

(see Farkas& Tetrick, 1989,for anexample).At a basic level, the researchagendaproposed by Judgeet al. (2001) could serve as a guideline and integrative ramework or such research.

Provided that AOCdoes cause performance, he resultsof this meta-analysishave practical mpli-cationsin two respects.First,the resultssuggest that AOCis a betterpredictorof performancewhen:

(a) performance s measuredby self-reportsrather hansupervisorreportsor objectiveindicators; b)extra-roleperformancerather than in-role performanceis predicted;and (c) white-collar workers

rather hanblue-collar workersare studied.Conclusion(c) is only tentative,given the small numberof analysedblue-collarsamples.Second,with the samecaveat,conditions(b) and(c) pointto circum-

stancesunderwhich attemptsto increaseproductivity hroughAOC may be particularlyeffective.

Author biography

Michael Riketta received diplomas (M.A. equivalent) in economics (University of Augsburg,

Germany,1997) and psychology (Catholic University of Eichstatt, Germany,1999). In 1999 and

2000, he was research assistantat the Departmentof Economic and Social Psychology, Catholic

University of Eichstitt. Since 2000, he has been research assistant at the Departmentof Social

Psychology, Universityof Mannheim,Germany.His areas of researchare context dependence of

the self-concept, organizationalcommitment and identification,and social psychological aspects of

European integration.

References

ABI/INFORM.ttp://www.il.proquest.com/l2000].Allen NJ, MeyerJP. 1990. The measurement nd antecedents f affective,continuance nd normative

commitmento theorganization.ournal f Occupational sychology3: 1-18.AllenNJ,MeyerJP.1996.Affective, ontinuance,ndnormativeommitmento theorganization.ournal f

Vocational ehavior 9: 252-276.BrettJF,CronWL,SlocumJW. 1995.Economicdependencyn work.Academy f Managementournal 5:

261-271.CohenA. 1991.Career tageas a moderator f therelationshipsetweenorganizationalommitment nd ts

outcomes: meta-analysis.ournal f Occupationalsychology 4: 253-268.FarkasAJ, TetrickLE. 1989. A three-waveongitudinal nalysisof the causalordering f satisfaction nd

commitmentnturnoverecisions. ournal f AppliedPsychology 4: 855-868.

Copyright 2002 JohnWiley& Sons,Ltd. J. Organiz. ehav. 3, 257-266(2002)

Page 11: Attitudinal Organizational Commitment and Job Performance

7/27/2019 Attitudinal Organizational Commitment and Job Performance

http://slidepdf.com/reader/full/attitudinal-organizational-commitment-and-job-performance 11/11

266 M. RIKETTA

Hrebiniak LG, Alutto JA. 1972. Personal and role-related factors in the development of organizationalcommitment.AdministrativeScience Quarterly17: 555-573.

HunterJE, Schmidt FL. 1990.Methodsof Meta-Analysis.Sage:NewburyPark,CA.

Judge TA, Thoresen CJ, Bono JE, Patton GK. 2001. The job satisfaction-job performancerelationship:a

qualitativeandquantitative eview.PsychologicalBulletin 127: 376-407.

KallebergAL, Marsden PV. 1995. Organizationalcommitment and job performance n the U.S. labor force.Research in the

Sociology ofWork5: 235-257.

MathieuJE, Zajac DM. 1990. A review and meta-analysisof the antecedents,correlates,and consequences of

organizational ommitment.PsychologicalBulletin 108: 171-194.

MeyerJP,Allen NJ. 1997. Commitmentn the Workplace.Sage: ThousandOaks,CA.

Mowday RT,PorterLW,Steers RM. 1982. Employee-organizationLinkages.Academic:New York.

Mowday RT, Steern RM, Porter LW. 1979. The measurement of organizationalcommitment. Journal ofVocationalBehavior 14: 224-247.

O'Reilly C III, ChatmanJ. 1986. Organizational ommitment andpsychologicalattachment. ournalof AppliedPsychology71: 492-499.

OrganDW, Ryan K. 1995. A meta-analyticreview of attidudinalanddispositionalpredictorsof organizationalcitizenshipbehavior.PersonnelPsychology48: 775-802.

PsycLIT.2001. http://www.psycinfo.com/psyclit.htmlApril2001].Randall DM. 1990. The consequences of organizationalcommitment:methodological investigation.Journalof

OrganizationalBehavior11: 361-378.Rosenthal R. 1991.

Meta-analyticProcedures

orSocial Research.

Sage: BeverlyHills,CA.

Social Sciences CitationsIndex. 2001. http://www.isinet.com/isi/products/citation/ssci/index.html2001].TettRP,MeyerJP.1993. Jobsatisfaction,organizational ommitment, urnover ntention,andturnover.Personnel

Psychology46: 259-293.van KnippenbergD. 2000. Workmotivationandperformance: social identity perspective.Applied Psychology:

An InternationalReview49: 357-371.ViswesvaranC, Ones DS, SchmidtFL. 1996. Comparativeanalysisof the reliabilityof job performance atings.

Journal of Applied Psychology 81: 586-597.

Copyright? 2002 JohnWiley & Sons, Ltd. J. Organiz.Behav.23, 257-266 (2002)