Journal of Organizational Behavior J - PDF document

JournalofOrganizationalBehaviorJ.Organiz.Behav.,293–315(2004)PublishedonlineinWileyInterScience(www.interscience.wiley.com).:10.1002/job.248Jobdemands,jobresources,andtheirrelationshipwithburnoutandengagement:amulti-samplestudyWILMARB.SCHAUFELI*ANDARNOLDB.BAKKERDepartmentofPsychologyandResearchInstitutePsychologyandHealth,UtrechtUniversity,Utrecht,TheNetherlandsThisstudyfocusesonburnoutanditspositiveantipode—engagement.Amodelistestedinwhichburnoutandengagementhavedifferentpredictorsanddifferentpossibleconsequences.Structuralequationmodelingwasusedtosimultaneouslyanalyzedatafromfourindependentoccupationalsamples(total1698).Resultsconrmthehypothesizedmodelindicatingthat:(1)burnoutandengagementarenegativelyrelated,sharingbetween10percentand25percentoftheirvariances;(2)burnoutismainlypredictedbyjobdemandsbutalsobylackofjobresources,whereasengagementisexclusivelypredictedbyavailablejobresources;(3)burnoutisrelatedtohealthproblemsaswellastoturnoverintention,whereasengagementisrelatedonlytothelatter;(4)burnoutmediatestherelationshipbetweenjobdemandsandhealthproblems,whereasengagementmediatestherelationshipbetweenjobresourcesandturnoverintention.Thefactthatburnoutandengagementexhibitdifferentpatternsofpossiblecausesandconsequencesimpliesthatdifferentinterventionstrategiesshouldbeusedwhenburnoutistobereducedorengagementistobeenhanced.Copyright2004JohnWiley&Sons,Ltd.IntroductionPositivestatesarenotpopularinpsychology.Basedonanelectronicsearchof,Myers(2000)calculatedthatnegativeemotionsoutnumberpositiveemotionsbyaratioof14to1.Thesameistrueforoccupationalhealthpsychology:asimplecountofarticlesthatappearedfrom1996onwardsintheJournalofOccupationalHealthPsychologyrevealsthatnegativework-relatedoutcomesoutnumberthepositiveoutcomesbyacomparableratioof15to1.So,itisnotsurprisingthattheemergingpositivepsychologyproposesashiftfromthistraditionalfocusonweaknessesandmalfunctioningtowardshumanstrengthsandoptimalfunctioning(Seligman&Csikszentmihalyi,2000).Asimilarswitchfromburnouttowardsitsopposite—engagement—hasrecentlybeenputforwardbyMaslach,Schaufeli,andLeiter(2001).InthecurrentarticleweproposeReceived10October2002Revised6March20032004JohnWiley&Sons,Ltd.Accepted27August2003 *Correspondenceto:WilmarB.Schaufeli,DepartmentofPsychology,UtrechtUniversity,POBox80140,3508TCUtrecht,TheNetherlands.E-mail:W.Schaufeli@fss.uu.nl acomprehensivemodeltopredictburnoutandengagementfromjobdemandsandjobresources,andtorelatebothtypesofemployeewell-beingtopotentialindividualandorganizationalcosts.BurnoutandengagementBurnoutisametaphorthatiscommonlyusedtodescribeastateofmentalweariness.Originally,burn-outwasconsideredtooccurexclusivelyinthehumanservicesamongthosewhodo‘peoplework’ofsomekind(Maslach&Schaufeli,1993).However,graduallyitbecameclearthatburnoutalsoexistsoutsidethehumanservices(Maslach&Leiter,1997).Consequently,theoriginalversionoftheMaslachBurnoutInventory(Maslach&Jackson,1986)wasadaptedforuseoutsidethehumanservices;thisnewversionwascalledMBI–GeneralSurvey(MBI-GS:Schaufeli,Leiter,Maslach,&Jackson,1996).ThethreedimensionsoftheMBI-GSparallelthoseoftheoriginalMBI,inthesensethattheyaremoregenericanddonotrefertootherpeopleoneisworkingwith.Therstdimension——measuresfatiguewithoutreferringtootherpeopleasthesourceofone’stiredness.Theseconddimension—cynicism—reectsindifferenceoradistantattitudetowardsworkingeneral,notnecessarilywithotherpeople.Finally,professionalefÞcacyencompassesbothsocialnon-socialaspectsofoccupationalaccomplishments.Inthepresentstudy,theMBI-GSisusedtomeasureburn-out;highscoresonexhaustionandcynicism,andlowscoresonprofessionalefcacy,areindicativeofJobengagementisassumedtobethepositiveantipodeofburnout.Or,asMaslachandLeiter(1997,p.34)putit:‘Energy,involvement,andefcacy—thesearethedirectoppositesofthethreedimen-sionsofburnout.’Intheirview,burnoutisanerosionofengagement,whereby‘Energyturnsintoexhaustion,involvementturnsintocynicism,andefcacyturnsintoineffectiveness’(p.24).Accord-ingtoMaslachandLeiter(1997),jobengagementisassessedbytheoppositepatternofscoresonthethreeMBIdimensions:thatis,lowscoresonexhaustionandcynicism,andhighscoresonefcacy,areindicativeofjobengagement.Hence,MaslachandLeiter(1997)considerburnoutandengagementtobetheoppositepolesofacontinuumthatisentirelycoveredbytheMBI.However,thiswayofusingtheMBIasabipolarinstrumentthatassessesburnoutaswellasengagementisratherquestionableinviewofthedebateonthepolarityofpositiveandnegativeaffect(Diener,1999).Forinstance,basedonsecondaryanalysesofasetofearlierstudies,RussellandCarroll(1999)showedconvincinglythatpositiveandnegativeaffectareindependentstates,ratherthantwooppositepolesofthesamebipolardimension.Inasimilarvein,itcouldbearguedthatinsteadofbeingtwooppositepolesburnoutandengagementareindependent,yetnegativelycorrelatedstatesofmind.Wetakethisperspectivebymeasuringburnoutandengagementindependentlyusingdifferentinstruments.ContrarytoMaslachandLeiter(1997)wedonotfeelthatengagementisadequatelymea-suredbytheoppositeproleofMBIscoressince—logicallyspeaking—thisimpliesthatbothcon-ceptsareeachother’scomplements.Apartfromtheanalogywiththeindependenceofpositiveandnegativeaffectthisalsodoesnotmakesense,giventhenatureandthescoringoftheMBIitems.Forinstance,feelingemotionallydrainedfromone’swork‘onceaweek’doesbynomeansexcludethatinthesameweekonemightfeelburstingwithenergy.Seenfromthisperspective,insteadofper-fectlycomplementaryandmutuallyexclusivestates,burnoutandengagementareindependentstatesthat—becauseoftheirantitheticalnature—aresupposedtobenegativelyrelated.Weexpecttherela-tionshipbetweenburnoutandengagementtobemoderatelytostronglynegative. AccordingtoCohenandHolliday’s(1982)ruleofthumb,correlationsbetween0.40and0.69areconsideredtobemoderate,whereascorrelationsexceeding0.69areconsideredhigh.W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) Wedeneengagementasapositive,fullling,work-relatedstateofmindthatischaracterizedbyvigor,dedication,andabsorption(seealsoSchaufeli,Salanova,Gonzalez-Roma,&Bakker,2002a).Engagementreferstoapersistentandpervasiveaffective–cognitivestatethatisnotfocusedonanyparticularobject,event,individual,orbehavior.Vigorischaracterizedbyhighlevelsofenergyandmentalresiliencewhileworking,thewillingnesstoinvesteffortinone’swork,andpersistencealsointhefaceofdifculties.Dedicationischaracterizedbyasenseofsignicance,enthusiasm,inspira-tion,pride,andchallenge.Vigoranddedicationarethedirectpositiveoppositesofexhaustionandcynicism,respectively.Thethirddimensionofengagementiscalledabsorption,whichwasfoundtobeaconstitutingelementofengagementin30in-depthinterviews(Schaufelietal.,2001).Absorp-ischaracterizedbybeingfullyconcentratedandhappilyengrossedinone’swork,wherebytimepassesquicklyandonehasdifcultieswithdetachingoneselffromwork.Beingfullyabsorbedinone’sworkcomesclosetowhathasbeencalled‘ow,’astateofoptimalexperiencethatischaracterizedbyfocusedattention,clearmind,mindandbodyunion,effortlessconcentration,completecontrol,lossofself-consciousness,distortionoftime,andintrinsicenjoyment(Csikszentmihalyi,1990).However,typically,owisamorecomplexconceptthatincludesmanyaspectsandreferstoshort-term‘peak’experiencesinsteadofamorepervasiveandpersistentstateofmind,asisthecasewithabsorption.Insum:burnoutandengagementareconsideredeachother’sopposites,particularlyasfarasexha-ustionandvigor,andcynicismanddedicationareconcerned.Theformertwoscalesspanadimensionthatmightbelabeledactivation,whereasbothlatterscalesconstitutetheoppositepolesofadimen-sionthatmightbelabeledidentiÞcation(foratheoreticaldiscussionseeSchaufeli&Bakker,2001).Inaddition,burnoutandengagementbothincludeathirdconstitutingcharacteristic:reducedprofessionalefcacyandabsorption,respectively.EngagementisoperationalizedwiththeUtrechtWorkEngagementScale(UWES),aself-reportinstrumentthatincludesthethreedimensionsthatwerementionedabove.TworecentstudiesusingconrmativefactoranalysisdemonstratedthefactorialvalidityoftheUWES(Schaufelietal.,2002a,2002b).Asexpected,theengagementandburnoutscalesweremoderatelynegativelycorre-lated.Schaufelietal.(2002a)observedthatacoreburnoutfactor—consistingofexhaustionandcyni-cism—andanextendedengagementfactor—includingprofessionalefcacyinadditiontothethreeoriginalengagementscales—ttedthedatabest.Inasimilarvein,usingdiscriminantanalysis,Demeroutietal.(2001a)foundthatlevelsofjobdemandsandjobcontrolwerepredictedbytwodis-criminantfunctions:acoreburnoutfunction(i.e.,exhaustionandcynicism—andhealthcomplaints)andanextendedengagementfunction(i.e.,includingallthreeengagementscalesplusprofessionalefcacy).ThepreviousresultsagreewithGreen,Walkey,andTaylor(1991),whocalledexhaustionanddepersonalization‘thecoreofburnout.’Furthermore,fromtheperspectivethatengagementistheoppositeofburnout,itmakessensethattheonlywordedMBIburnoutscale—professionalefcacy—loadsontheengagementfactor.Therstobjectiveofthecurrentstudyistotestwhetherornotamodelthatincludesthecoreofburnout(exhaustionandcynicism)aswellasanextendedengagementfactor(vigor,dedication,absorption,andefcacy)tsbettertothedatacomparedto:(1)amodelthatassumesthatandengagementscalesloadonasinglefactor,whichwouldagreewithMaslachandLeiter’s(1997)approach;(2)amodelthatincludestheoriginalburnout(exhaustion,cynicism,efcacy)andengage-mentscales(vigor,dedication,absorption).JobdemandsandjobresourcesGenerallyspeaking,twosetsofvariablescanbedistinguishedinanykindofjob:jobdemandsandjobresources.JonesandFletcher(1996,p.34)denedemandsas‘thedegreetowhichtheenvironmentBURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) containsstimulithatperemptorilyrequireattentionandresponse.Demandsarethe‘‘thingsthathavetobedone.’’’Clearly,ineveryjobsomethinghastobedone.Morespecically,werefertojobdemandsasthosephysical,psychological,social,ororganizationalaspectsofthejobthatrequiresustainedphy-sicaland/orpsychological(i.e.,cognitiveoremotional)effortandarethereforeassociatedwithcertainphysiologicaland/orpsychologicalcosts.Althoughjobdemandsarenotnecessarilynegative,theymayturnintojobstressorswhenmeetingthosedemandsrequireshigheffortandisthereforeasso-ciatedwithhighcoststhatelicitnegativeresponsessuchasdepression,anxiety,orburnout.Inthepre-sentstudy,weincludedquantitative(workoverload)aswellasqualitative(emotional)jobdemands.Jobresourcesrefertothosephysical,psychological,social,ororganizationalaspectsofthejobthateither/or(1)reducejobdemandsandtheassociatedphysiologicalandpsychologicalcosts;(2)arefunctionalinachievingworkgoals;(3)stimulatepersonalgrowth,learninganddevelopment.Hence,resourcesarenotonlynecessarytodealwithjobdemandsandto‘getthingsdone,’buttheyalsoareimportantintheirownright(Hobfoll,2002).Inthepresentstudy,weincludedjobresourcesatthetask-level(performancefeedback),theinterpersonal-level(supportfromcolleagues),andtheorganiza-tionallevel(supervisorycoaching).Intheirso-calledstructuralmodel,Maslach,Jackson,andLeiter(1986,pp.36–37)hypothesizethatthepresenceofspecicdemands(i.e.workoverloadandpersonalconicts)andtheabsenceofspecicresources(i.e.controlcoping,socialsupport,autonomy,anddecisioninvolvement)predictsburnout,whichinitsturnisexpectedtoleadtovariousnegativeoutcomessuchasphysicalillness,turnover,absenteeism,anddiminishedorganizationalcommitment.However,theirmodelispurelydescriptiveandexclusivelyservesasaheuristicframeworkforintegratingstudyresultsthatwereobtainedwiththeMBI.Recently,Demerouti,Bakker,Nachreiner,andSchaufeli(2001b)wentonestepbeyondbysuccessfullytestingtheso-calledJobDemand-Resources(JD-R)modelthatpositsthatjobdemands(i.e.,physicaldemands,timepressure,shiftwork)areassociatedwithexhaustion,whereaslackingjobresources(i.e.,performancefeedback,jobcontrol,participationindecisionmaking,socialsupport)areassociatedwithdisengagement.InsteadoftheMBI,Demeroutietal.(2001b)usedanewlydevelopedGermanquestionnaire—theOldenburgBurnoutInventory(OLBI;Demerouti,Bakker,Vardakou,&Kantas,2003)—thatincludesbothcoredimensionsofburnout(i.e.exhaustionanddisengagement)thatareconceptuallysimilartothoseoftheMBI-GS(i.e.exhaustionandcynicism).However,contrarytotheMBI-GS,bothOLBI-dimensionsaremeasuredbynegativelyphraseditemsaswellasbyposi-tivelyphraseditems.Bydoingso,Demeroutietal.(2001b,2003)offerathirdalternativetotheoper-ationalizationofburnoutandengagement.WhileMaslachandLeiter(1997)arguethatburnoutandengagementaretheendpointsofacontinuumthatisspannedbytheMBI,andSchaufelietal.(2002a,2002b)arguethatburnoutandengagementareindependentstatesthatshouldbemeasuredseparatelybydifferentinstruments,theOLBIisconstitutedbytwobipolarscalesthatincludepositiveengage-mentitemsaswellasnegativeburnoutitems.Theoreticallyspeaking,theJD-Rmodelassumestwoprocesses:(1)anenergeticprocessofover-taxingandwearingoutinwhichhighjobdemandsexhausttheemployee’senergybackup;(2)amoti-vationalprocessinwhichlackingrecoursesprecludedealingeffectivelywithhighjobdemandsandfostermentalwithdrawalordisengagement.Inthecurrentarticle,weextendtheJD-Rmodelaspro-posedbyDemeroutietal.(2001b)byincludingengagement—asmeasuredindependentlyfromburn-out—andbyaddingindicatorsforhealthimpairmentandorganizationalcommitment(i.e.,lowturnoverintention)aspossibleconsequencesofburnoutandengagement,respectively.TheresearchmodelBuildinguponpreviousworkourresearchmodel(seeFigure1)assumestwopsychologicalprocesses.W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) TheenergeticprocessThisprocesslinksjobdemandswithhealthproblemsviaburnoutandcanbeilluminatedbyHockey’s(1993,1997)stateregulationmodelofcompensatorycontrol.Thismodeloffersacognitive–emotionalframeworkforunderstandinghumanperformanceunderstress.Itisconcernedwiththemaintenanceofperformancestabilityunderdemandingconditions,whichrequiresthemobilizationandmanagementofmentaleffort.Whenconfrontedwithhighjobdemands,which,accordingtoourdenitionrequiresustainedeffort,employeeseitheradoptperformanceprotectionstrategieswhichareassociatedwithextracosts,ortheyacceptareductioninovertperformancewithnoincreaseincosts.Undernormalworkingconditions,performanceremainsstableandtheassociatedeffortremainswithinreservelim-its,thoughtheoveralllevelofenergyspentisincreased.Althoughanactivecopingresponseisbasi-callyadaptiveintheshortrun,itislikelytobemaladaptiveasahabitualpatternofresponsetoworkor,ifsustainedoveraprolongedperiod,becauseitmightdepletetheindividual’senergyresources.Whentheperceiveddemandsaretoohightobemetbytheusualworkingeffort,twooptionsareopen(Hockey,1997).Intheso-calledstraincopingmode,themaximumeffortbudgetisfurtherincreasedtoaccommodatethehighlevelofdemand.Thetargetperformanceismaintained,butonlyattheexpenseofanincreaseincompensatorycoststhataremanifestedpsychologically(e.g.,fatigueandirritability)aswellasphysiologically(e.g.,increasedexcretionofcortisol).Analternativeresponsetoexcessivedemandsistoadoptaso-calledpassivecopingmode,whichinvolvesdownwardadjustmentofperformancetargets,forinstancebyreducinglevelsofaccuracyandspeed.Byleavingthemaximumeffortbudgetatitsusuallevel,furtherpsychologicalandphysiologicalcostsarepre-vented,however,attheexpenseofperformancetargets.Inextremecases,completedisengagementfromthepursuitoftaskgoalsmaybeobserved.AlthoughHockey’smodelofcompensatorycontrolwasnotdesignedtoexplainburnoutitisnote-worthythatallthreeburnoutdimensionsareincluded:energydepletion(exhaustion),disengagement(cynicism),anddiminishedperformance(reducedefcacy).Moreover,theburnoutprocessmaybedescribedinawaythatiscompatiblewithHockey’smodel.AccordingtoMaslach(1993),burnoutiscausedbyhighjobdemandsthatdraintheemployee’senergyandinanattempttocopewiththeresultingexhaustiontheemployeewithdrawsmentally.However,thisisaninadequatestrategysincethispreventsproperperformance.Indeed,ithasbeenconsistentlyfoundthatemotionalexhaustionisrelatedtojobdemands,suchastimepressureandworkoverload,asisshowninameta-analysisbyLeeandAshforth(1996).Obviously,givingtoomuchtoolongisenergydepletingandleadstoexhaustion.Thisagreeswiththehabitualactivecopingmodeandperhapsevenmoresowiththestraincoping Figure1.TheresearchmodelBURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) mode,asdescribedbyHockey(1997).Thepassivecopingmodecomesclosetocynicism,thesecondcoredimensionofburnout.Indeed,somestudieshavefoundthatexhaustionleadstocynicism,ratherthantheotherwayaround(Bakkeretal.,2000;Lee&Ashforth,1993;Toppinen-Tanner,Kalimo,&Mutanen,2002).ThismightsuggestthatHockey’scopingmodesunfoldinadynamicwayacrosstime:afterhavinginvestedextraenergy,uptothepointofexhaustion,employees‘decide’that‘enoughisenough’andhenceswitchfromanactive/strainmodetoapassivecopingmodethatischaracterizedbyde-investmentanddisengagement.Psychophysiologicallyspeaking,thehabitualactivecopingmode,butparticularlythestraincopingmode,isrelatedtoincreasedsympatheticactivity.Andsoisapassivecopingmode,atleastwhenper-formanceonthejobisofexistentialimportanceforthepersoninvolved,whichtypicallyisthecaseinburnoutcandidates(Schaufeli&Enzmann,1998,pp.77–78).Theprocessofsympatheticoverloadresultingfromhighjobdemandsthatmightleadtohealthimpairmentisknownassustainedactivation(Knardahl&Ursin,1985).ThemotivationalprocessThisprocesslinksjobresourcesviaengagementwithorganizationaloutcomes(e.g.,turnoverinten-tion).Asfollowsfromourdenition,jobresourcesmayplayeitheranintrinsicmotivationalrolebecausetheyfosteremployees’growth,learninganddevelopment,ortheymayplayanextrinsicmoti-vationalrolebecausetheyareinstrumentalinachievingworkgoals.Intheformercase,jobresourcesfulllbasichumanneeds,suchastheneedsforautonomy(DeCharms,1968),competence(White,1959),andrelatedness(Baumeister&Leary,1995).Accordingtoself-determinationtheory(Deci&Ryan,1985)workcontextsthatsupportpsychologicalautonomy,competence,andrelatednessenhancewell-being(i.e.,vitality)andincreaseintrinsicmotivation(Ryan&Frederick,1997).Forinstance,properfeedbackfosterslearning,therebyincreasingjobcompetence,whereasdecisionlati-tudeandsocialsupportsatisfytheneedforautonomyandtheneedtobelong,respectively.Thisintrin-sicmotivationalpotentialofjobresourcesisalsorecognizedbyJobCharacteristicsTheory(JCT;Hackman&Oldham,1980).AccordingtoJCT,everyjobhasaspecicmotivationalpotentialthatdependsonthepresenceofvecorejobcharacteristics:skillvariety,taskidentity,tasksignicance,autonomy,andfeedback.Furthermore,JCThypothesizesthatthesejobcharacteristicsarelinked—throughso-calledcriticalpsychologicalstates—withpositiveoutcomessuchashigh-qualityworkper-formance,jobsatisfaction,andlowabsenteeismandturnover.Jobresourcesmayalsoplayanextrinsicmotivationalrole,because,accordingtotheso-calledEffort-Recoverymodel(Meijman&Mulder,1998),workenvironmentsthatoffermanyresourcesfos-terthewillingnesstodedicateone’seffortsandabilitiestotheworktask.Inthatcaseitislikelythatthetaskwillbecompletedsuccessfullyandthattheworkgoalwillbeattained.Forinstance,supportivecolleaguesandproperfeedbackfromone’ssuperiorincreasethelikelihoodofbeingsuccessfulinachievingone’sworkgoals.Ineithercase,beitthroughthesatisfactionofbasicneedsorthroughtheachievementofworkgoals,theoutcomeispositiveandengagement—afullling,positivework-relatedstateofmind—islikelytooccur.Initsturn,itisplausibletoassumethatengagedworkershavealowtendencytoleavetheorga-nization.Afterall,theorganizationprovidesthemwithvaluedjobresourcesthatenhancelearning,growth,anddevelopment(Houkes,Janssen,DeJonge,&Nijhuis,2001).Inotherwords,engagementisexpectedtomediatetherelationshipbetweenjobresourcesand(low)intentiontoleave.OurhypotheticalmodelisgraphicallydisplayedinFigure1.Itincludestheenergy-depletingprocessofgraduallywearingout(top)andthemotivationalprocessofpersonaldevelopmentandgoalW.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) attainment(bottom).Inaddition,Figure1includesvecross-linksbetweenbothprocesses.First,fol-lowingourdenitionofjobresourcesinwhichjobresourcespotentiallyreducejobdemands,weexpectanegativerelationshipbetweendemandsandresources.Second,asarguedbefore,weexpectanegativerelationshipbetweenburnoutandengagement.Third,basedonampleempiricalevidence,wehypothesizethatjobresourcesarenegativelyrelatedtoburnout,whereas—fourth—burnoutispositivelyrelatedtoturnoverintention(foranoverviewseeLee&Ashforth,1996;Schaufeli&Enzmann,1998,pp.89–91).Thetheoreticalrationalefortheformeristhatlackingresources—bydenition—increasejobdemandsandmayhencefosterburnout.Thepositiverelationshipbetweenburnoutandturnoverintentionwasdemonstratedinastudythatusedasocialexchangeperspectiveinwhichbothburnoutandturnoverintentionwereconsideredtooriginatefromaninequitableexchangerelationshipwiththeorganization(Geurts,Schaufeli,&DeJonge,1998).Finally,weexpectapositiverelationshipbetweenhealthproblemsandturnoverintentions,ashasbeenshowninpreviousresearch(e.g.,Jamal,1990).Insum:thisstudyseekstotestthefollowingfour(setsof)hypotheses:Hypothesis1:Atwo-factormodelthatincludesthecoreofburnout(exhaustionandcynicism)aswellasanextendedengagementfactor(vigor,dedication,absorption,andefcacy)tsbettertothedatathan:(a)aone-factormodeland(b)analternativetwo-factormodelthatincludestheoriginalburnout(exhaustion,cynicism,efcacy)andengagementscales(vigor,dedication,absorption).Hypothesis2:BurnoutmediatestherelationbetweenhighjobdemandsandexperiencedhealthHypothesis3:Engagementmediatestherelationshipbetweenjobresourcesand(low)turnoverintention.Hypothesis4:Variouscross-linksexistbetweentheenergeticalandmotivationaldrivenprocesses:(a)Jobdemandsandjobresourcesarenegativelyrelated.(b)Engagementandburnoutarenegativelyrelated.(c)Jobresourcesarenegativelyrelatedtoburnout.(d)Burnoutispositivelyrelatedtoturnoverintention.(e)Healthproblemsandturnoverintentionarepositivelyrelated.Hypotheses2–4aretestedsimultaneouslyacrossfourindependentsamplesbyttingtheresearchmodel(seeFigure1)tothedatausingstructuralequationmodelingmethods.OrganizationalContext EconomyTheNetherlandsisaprosperousandopenservice-orientedeconomydependingheavilyonforeigntrade.Themajorityoftheworkforce(73percent)worksinserviceoccupations,whichisthehigh-estrateamongallEUmemberstates.Atthetimeofthestudies(2001),theeconomyisnotedforstableindustrialrelations,moderateination,asizablecurrentaccountsurplus,andanimportantroleasaEuropeantransportationhub.TheDutcheconomyhasexpandedby3percentormoreinBURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) SamplesandprocedureAfterinformativemeetingswithrepresentativesofthemanagement,personneldepartment,andwork-erscouncil,allemployeesfromthefourorganizationsreceivedpaper-and-pencilquestionnairesandreturnenvelopesattheirwork.Thequestionnaireswereaccompaniedbyaletter,inwhichthegoalof eachofthepreceding4yearsandrealGDPgrowthwasabout3.6percentin2001.Thelabormarketwastight,withanunemploymentrateaslowas2.4percent.Duringthatyear,thegovernmentimplementeditsmostcomprehensivetaxreformsinceWorldWarII,designedtoreducehighincometaxlevelsandredirectthescalburdenontoconsumption.TheDutchwereamongtherst11EUcountriesestablishingtheeurocurrencyzoneonJanuary1,1999.OrganizationsEmployeesfromfourdifferentDutchserviceorganizationsparticipatedinthepresentresearch:aninsurancecompany,apensionfundcompany,anOccupationalHealthandSafetyService,andahome-careinstitution.EachoftheseorganizationshasarichhistoryintheNetherlands,anddeliversservicesforthelocalDutchmarket.Thefourorganizationsclearlydifferregardingtheiractivitiesandculture.Theinsuranceandpensionfundcompaniesarenance-related,andhaveamanagementbyobjectivesphilosophy;employeesmainlyworkwithinformation.Incontrast,theoccupationalhealthandsafetyserviceandhomecareorganizationdo‘peoplework’,andcanbecharacterizedbyastronghelpingorientation.Bothnancialorganizationswentthroughatransitionatthetimeofthestudy.DuetochangesintheDutchlaw,themarketbecamemoreopenandcompetitive.Thereforethestructureoftheorganizationwastransformedfromafunctionalorganizationtoanorganizationbasedonclientteams.Inaddition,thepensionfundcompanyisdifferentfromthethreeotherpar-ticipatingorganizationsinthatitischaracterizedbyarelativelyyoungworkforce(meanage35yearsvs.40yearsfortheremainingorganizations)withlimitedjobtenure.Thetwohumanserviceorganizationsdidnotgothroughrecentrestructurings,buthadproblemswithattractingasufcientnumberofworkers,duetotheshortageoflabourinthetightlabourmarket.Thehome-careinstitu-tionispubliclyfundedbypremiumsthatarecollectedbythetaxauthorities.DuetochangesinDutchworkingconditionslegislation,theoccupationalhealthandsafetyservicehadbeentrans-formedintheearly1990sfromapublicfundedagencyintoacompetitivecommercialorganization.ActivitiesTheservicesoftheinsurancecompanyincludelifeandfuneralinsurance,disabilityinsurance,pen-sionfunds,andhomeloans;themainactivitiesoftheemployeesinthepensionfundcompanyaretocollectpremiums,toadministrate,andtopayoutmonthlyallowancesandpensions.Incontrast,theoccupationalhealthandsafetyservicesprovideassessmentofsafetyrisks,physicalrisks,andpsy-chosocialhealthrisks;preventionofwork-relatedhealthproblems,accidents,anddisability;pro-tectionandpromotionofemployees’safety,health,andwell-being;whereasthemainactivitiesofthehome-careemployeesincludednurturingofclientswithphysicalormentalhealthimpairments.Home-careemployeesprovidedtheirclientswithsupporttoimprovetheirdailyfunctioning(e.g.,helpthemtogetoutofbedandgotothetoilet,washthem,listentotheirgrieves,andtakecareofhouseholdchores,suchaswashing,cleaningandcooking).W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) thestudywasbrieyintroduced,andthecondentialityandanonymityoftheanswerswereempha-sized.Theemployeeswerekindlyrequestedtolloutthequestionnaireinprivateandtopostitafter-wardsinaspecialboxattheirdepartments.Sample1381)consistedofemployeesofaninsurancecompany(responserate61percent).Thesampleincluded261males(69percent)and120females(31percent).Themeanagewas40years10.4).Meanorganizationaltenurewas12.5years(SD10.5years).Twelvepercentofthesamplehadamanagerialposition.Mostparticipantshadafull-timejob(64percent).Theservicesofthiscompanyincludelifeandfuneralinsurance,disabilityinsurance,pensionfunds,andhomeSample2202)consistedofemployeesofseveralbranchesofalargeOccupationalHealthandSafetyService(responserate63percent).Thesampleincluded76males(38percent)and126females(62percent).Themeanagewas40years(SD9.0)andmeanorganizationaltenurewas7years6.5).Themajorityofthesamplewasontenure(90percent)andmostemployeesworkedfull-time(75percent).Twenty-vepercentoftheparticipantshadasupervisoryposition.Occupa-tionalhealthservicesprovide:assessmentofsafetyrisks,physicalrisks,andpsychosocialhealthrisks;preventionofwork-relatedhealthproblems,accidents,anddisability;protectionandpromotionofemployees’safety,healthandwell-being.Sample3507)consistedofemployeesofapensionfundcompany(responserate83percent).Thesampleincluded173males(34percent)and334females(66percent)andtheirmeanagewas35years(SD9.5).Meanorganizationaltenurewas6.5years(SD7.5).Themajorityofthissamplehadasteadycontract(73percent)andmostemployeesworkedfull-time(91percent).Sixteenpercentoftheparticipantshadasupervisoryposition.Themainactivitiesoftheemployeesinthisorganizationweretocollectpremiums,toadministrateandtopayoutmonthlyallowancesandpensions.Sample4608)consistedofemployeesofahome-careinstitution.Comparedtotheothersam-ples,theresponserate(47percent)wasrelativelylow.Presumablythiswascausedbythefactthat,incontrasttotheothersamples,inSample4theinvestigationwasnotinitiatedbyacademicresearchersbutbymanagement,whichmighthavereducedemployees’motivationtoparticipate.Thelargemajor-ityofthesamplewasfemale(97percent),andthemeanagewas42years(SD10.0).Meanorgani-zationaltenurewas9years(SD7.0).Mostemployeesworkedpart-time:1–10hours(17percent),11–20hours(39percent),21–30hours(25percent),31–36hours(13percent),andmorethan36hours(6percent).Themainactivitiesofthehomecareemployeesincludednurturingofclientswithphysicalormentalhealthimpairments.Home-careemployeesprovidedtheirclientswithsupporttoimprovetheirdailyfunctioning(e.g.,helpthemtogetoutofbedandgotothetoilet,washthem,listentotheirgrieves,takecareofhouseholdchores,suchaswashing,cleaningandcooking).BurnoutwasassessedwiththeDutchversion(Schaufeli&VanDierendonck,2000)oftheMaslachBurnoutInventory–GeneralSurvey(MBI-GS;Schaufelietal.,1996).TheMBI-GSincludesthreesub-scales:Exhaustion(EX)(veitems;e.g.,‘Ifeelusedupattheendofaworkday’);Cynicism(CY)(veitems:e.g.,‘Idoubtthesignicanceofmywork’);ProfessionalEfÞcacy(PE)(sixitems;e.g.,‘Icaneffectivelysolvetheproblemsthatariseinmywork’).Allitemswerescoredonaseven-pointfrequencyratingscalerangingfrom0(‘never’)to6(‘always’).HighscoresonEXandCYandlowscoresonPEareindicativeofburnout(i.e.,allPEitemsarereversiblyscored).Internalconsistencies(Cronbach’saredisplayedinTable1.Afterremovingoneitem(‘Ijustwanttodomyjobandnotbebothered’),thereliabilityoftheCYscalewassubstantivelyincreasedinallsamples.ThisparticularCYitemisfoundtobenotoriouslyunsound(seealsoSchutte,Toppinen,Kalimo,&Schaufeli,2000).BURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) wasassessedwiththeUtrechtWorkEngagementScale(UWES;Schaufelietal.,2002a).TheitemsoftheUWESaregroupedintothreesubscalesthatreecttheunderlyingdimensionsofengagement:Vigor(VI)(sixitems;e.g.,‘WhenIgetupinthemorning,Ifeellikegoingtowork’);Dedication(DE)(veitems;e.g.,‘Iamenthusiasticaboutmyjob’),andAbsorption(AB)(sixitems;e.g.,‘WhenIamworking,Iforgeteverythingelsearoundme’).TheengagementitemsaresimilarlyscoredasthoseoftheMBI-GS.Inordertoavoidansweringbias,burnoutandengagementitemswererandomlymergedintoa33-itemquestionnaire.Jobdemands:Twotypesofjobdemandswereincluded:quantitative(i.e.,workload)andqualitative(i.e.,emotional)demands.WorkloadwasbasedonaDutchversion(Furda,unpublished,1995;seealsoDeJongeetal.,2000)ofKarasek’s(1985)jobcontentscale.Thescaleincludesveitemsthatrefertoquantitative,demandingaspectsofthejob(e.g.,timepressure,workinghard);including‘Myjobrequiresworkingveryhard.’Itemswerescoredonafour-pointfrequencyscale,rangingfrom1(‘never’)to4(‘always’).Emotionaldemandswereassessedbyave-itemscaledevelopedbyVanVeldhovenandMeijman(1994;seealsoVanVeldhovenetal.,2002).Anexampleitemis:‘Isyourworkemotionallydemand-ing?’Itemswerescoredinasimilarwayasthepreviousscale.Jobresources:Threejobresourcesweremeasuredthatpertaintothetask(i.e.,performancefeedback),andtheinterpersonal(i.e.,socialsupportfromcolleagues)andorganizationaljobcontexts(i.e.,supervisorycoaching),respectively.Performancefeedbackwasassessedwiththreeitems,baseduponKarasek’s(1985)jobcontentinstrument;e.g.,‘Ireceivesufcientinformationaboutthegoalofmywork.’Itemswerescoredonave-pointfrequencyscale,rangingfrom1(‘never’)to5(‘always’).Socialsupportfromcolleagueswasmeasuredwitha10-itemscaledevelopedbyVanVeldhovenandMeijman(1994);e.g.,‘Canyouaskyourcolleaguesforhelpifnecessary?’Itemswerescoredinasimilarwaytothepreviousscale(seealsoVanVeldhovenetal.,2002). Table1.Internalconsistencies(Cronbach’s)ofthestudyvariablesSample1Sample2Sample3Sample4Variable(381)(202)(507)(JobdemandsWorkoverload0.850.850.830.83Emotionaldemands0.720.780.750.72JobresourcesFeedback0.840.780.800.83Socialsupport0.810.810.800.82Supervisorycoaching0.870.910.880.89Exhaustion0.890.900.860.820.800.760.770.72Professionalefcacy0.780.790.750.74Vigor0.820.800.780.75Dedication0.900.890.900.88Absorption0.750.730.770.70Health0.840.830.820.83Turnoverintention0.890.800.780.72Afterremovingoneitem(seetext).W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) SupervisorycoachingwasmeasuredusingaDutchadaptationofGraenandUhl-Bien’s(1991)12-itemLeader–Memberexchangescale(LeBlanc,1994);e.g.,‘Mysupervisoruseshis/herinuencetohelpmesolvemyproblemsatwork.’Itemswerescoredinasimilarwayasbothpreviousscales.Healthproblemsweremeasuredwitha13-itemquestionnaireforpsychosomatichealthcomplaints,includingheadaches,cardiovascularproblems,andstomachaches(Dirken,1969).Responsecate-goriesrangedfrom1(‘neverorrarely’)to4(‘veryoften’).ThisinstrumentisusedbytheDutchCen-tralBureauofStatisticstomonitorhealthproblemsintheDutchpopulation.Turnoverintentionwasassessedwithathree-itemscale,basedonVanVeldhovenandMeijman(1994);e.g.,‘Iintendtochangejobsduringthenextyear.’Itemswerescoredonave-pointscale,rangingfrom1(‘completelydisagree’)to5(‘completelyagree’)(seealsoVanVeldhovenetal.,2002).AscanbeseenfromTable1,withoutexception,-valuesofallscalesmeetthecriterionof0.70(Nunnaly&Bernstein,1994).Structuralequationmodeling(SEM)methodsasimplementedbyAMOS(Arbuckle,1997)wereusedfordataanalyses.Modeltestingwascarriedoutinallfoursamplessimultaneouslybyusingtheso-calledmultiple-groupmethod.First,twosecond-ordertwo-factormodelsofburnoutandengagementweretestedacrosssamples.Theone-factormodel(M1)assumesoneunderlyingfactorthatincludesallburnoutandengagementdimensions,whereasthesecondmodel(M2)assumesthatthethreeburnoutdimensions(EX,CY,andPE)andthethreeengagementdimensions(VI,DE,andAB)bothloadonseparateyetnegativelycorrelatedfactors—‘burnout’and‘engagement’,respectively.InordertotestHypothesis1,thetofM1andM2wasthencomparedwiththatofanalternativemodel(M2alternativethatassumesthatinsteadofloadingon‘burnout’PEloadson‘engagement.’Next,followingTaris,Bok,andMeijer(1998)thefactorialinvarianceofthebestttingmodel—presumablyalternative—wasinvestigatedby:(1)constrainingthesecond-orderfactorloadingstobeequalacrossallsamples;(2)constraininginadditionthecovariancesbetweenbothlatentvariablesbeequalacrossallsamples.Whenthetofbothsuccessivemodelsdoessignicantlydeteriorate,maximumfac-torialinvariancehasbeendemonstrated.InordertotestHypotheses2–4,ourresearchmodelasdepictedinFigure1wasttedtothedata).Twoindicators(i.e.,workloadandemotionaldemands)wereusedtoestimatethelatentjobdemandsvariable,whereasfeedback,socialsupportofcolleagues,andsupervisorycoachingwereusedasindicatorsofthelatentjobresourcesvariable.Furthermore,latentill-healthandturnoverinten-tionvariableswerepostulatedthatwereeachmeasuredbyasingleindicator.Inordertoaccountforrandommeasurementerrorsofbothscaleswesettherandomerrorvariancesassociatedwitheachvariableequaltotheproductofitsvarianceandthequantityoneminusitsestimatedreliability,asindicatedbycoefcient(Bollen,1989).Threestructuralmodelswithincreasingconstraintsweresubsequentlyttedtothedata.First,allparameterswerefreelyestimatedacrosssamples;thatis,para-meterswereallowedtodifferineachsample.Next,theparametersofthestructuralmodel(i.e.,thoseparametersthatspecifytherelationshipsbetweenthelatentvariablesofthemodelasdepictedinFigure1)wereconstrainedtobeequalacrosssamples.Inthenalstep,inadditiontothesestructuralparameters,theparametersofthemeasurementmodel(i.e.,thoseparametersthatspecifytherelation-shipsbetweenlatentandmanifestvariables)wereconstrainedtobeequalacrosssamples.Followingthesamelogicasoutlinedabove,whenthetofthesuccessivemodelsdoessignicantlydeterio-rate,maximuminvariancehasbeendemonstrated.Thatis,inthatcasenotonlydosimilarstructuralrelationsexistbetweenthelatentvariablesineachsample,butalsotheestimatedparametersthatlinkthemanifestvariableswiththelatentvariablesareequal.BURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) FitindicesMaximumlikelihoodestimationmethodswereusedandtheinputforeachanalysiswasthecovariancematrixoftheitems.Thegoodness-of-tofthemodelswasevaluatedusingthefollowingabsolutegoodness-of-tindices(cf.Joreskog&Sorbom,1986):(1)thegoodness-of-tstatistic;(2)theRootMeanSquareErrorofApproximation(RMSEA);(3)theGoodnessofFitIndex(GFI);(4)theAdjustedGoodnessofFitIndex(AGFI).Sinceissensitivetosamplesize—theprobabilityofrejectingahypothesizedmodelincreaseswithsamplesize—theuseofrelativegoodness-of-tmeasuresisstronglyrecommended(Bentler,1990).FollowingMarsh,Balla,andHau(1996),threerelativegood-ness-of-tmeasureswerecalculated:(1)NormedFitIndex(NFI);(2)Non-NormedFitIndex(NNFI);and(3)ComparativeFitIndex(CFI).SincethedistributionoftheGFIandtheAGFIisunknown,nostatisticaltestorcriticalvalueisavailable(Joreskog&Sorbom,1986).Valuessmallerthan0.08forRMSEAareindicativeofanacceptablet,andvaluesgreaterthan0.1shouldleadtomodelrejection(Cudeck&Browne,1993).Forallthreerelativetindices,asaruleofthumb,valuesgreaterthan0.90areconsideredasindicatingagoodt(Hoyle,1995).InordertotestHypothesis,thetofthreemodelmodelswascompared.First,theone-factormodel(M1)wassimultaneouslyttedtothedataofthefoursamples.Inthenextstep,M2wasttedthatassumesthatEX,CY,andPEloadonalatentburnoutvariable,whereas,VI,DE,andABloadonalatentengagementvariable.AscanbeseenfromTable2,M1tsratherpoorlytothedatawithnoneofthetindicesmeetingitscriterion.Unfortunately,ttingM2failedbecausetheproducedcovariancematrixofbothlatentvariables—‘burnout’and‘engagement’—wasnotpositivedenite,whichmightbecausedbymodelmisspecication.Next,thealternativemodel(M2alternative)wastested,whichassumesthatinsteadofloadingonburnoutPEloadsonengagement.AscanbeseenfromTable2,alternativeshowedanacceptablettothedataandittteds1ignicantlybettertothedatathan0.001.Hence,Hypothesis1bwasconrmed.However,basedontheso-calledmodicationindices,thetofM2alternativecouldbesignicantlyimproved(184.35,0.001)byallowingtwopairsoferrorstocorrelate;namelythe Table2.Thetofthesecond-orderfactormodelsoftheMBI-GSandtheUWES;multiplegroupanalysesincludingSample1(381),Sample2(202),Sample3(507),andSample4(d.f.GFIAGFIRMSEANFINNFICFIM1605.13360.900.780.100.860.790.87alternative420.64320.970.830.080.910.840.91M3236.29240.960.860.050.950.880.95M4288.20420.950.900.060.940.920.94M5290.35450.950.910.060.940.930.94Nullmodel4477.07600.490.280.21———:GFI,Goodness-of-FitIndex;AGFI,AdjustedGoodness-of-FitIndex;RMSEA,RootMeanSquareErrorofApproximation;NFI,NormedFitIndex;NNFI,Non-NormedFitIndex;CFI,ComparativeFitIndex.Foradescriptionofthemodels,seetext. Thecovariancematricesoftheburnoutandengagementscalesofthefoursamplesareavailableonrequestfromtherstauthor.W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) errorsofEXandVI,andofCYandDE,respectively.Despitethedangerofchancecapitalization(MacCallum,Roznowski,&Necowitz,1992)wedecidedtoincludetheseerrortermsinare-speciedmodel(M3)sincebothcanbeinterpretedstraightforwardly.Namely,asindicatedabove,EXandVIaswellasCYandDEareeachassumedtobothspanadimensionthathasbeenlabeledactivityandiden-tication,respectively(seeIntroduction).Obviously,partoftheinitialerrorvarianceiscommonandmightbeexplainedbyeachofthesetwodimensions.Inordertoassesstheinvarianceofthefactorloadings(M4)andofthecovariancesofthelatentfactors(M5)acrosssamples,thetofM4andofM5wasassessedrelativetothatofM3.AlthoughthetofM4andM5remainsacceptableaccordingtomosttindices,aformal-differencetestrevealedthatthedifferencebetweenM3andM4wassignicant,0.001.Hence,thefactorloadings(aswellasthecorrelatederrorterms)areinvariantacrosssamples.LogicallyspeakingthedifferenceintbetweenM3andM5isalsosignicant,Insum:(1)burnoutandengagementdorefertooneunderlyingdimension;(2)thecoreofburnoutisconstitutedbyEXandCY,whereasinadditiontoVI,DEandAB,PEalsoloadsonthelatentengage-mentvariable—henceHypothesis1wasconrmed;(3)theestimatedcorrelationsbetweenthelatentburnoutandengagementvariablesaremoderateandrangefrom0.38to0.51acrossthefoursam-ples;(4)thefactorloadingsandcorrelationsbetweenlatentfactorsareinvariantacrosssamples.InordertotestHypotheses2–4,themodelasdisplayedinFigure1wasttedtothedataofallfoursamplessimultaneously.Table3showsthedescriptivestatisticsofthestudyvariablesinthetotalsam-ple.Consistentwiththepreviouslyobtainedresults,EXandCYwereincludedasindicatorsofthelatentburnoutvariable,whereasVI,DE,AB,andPEwereincludedasindicatorsofthelatentengage-mentvariable.Furthermore,alsoinaccordancewiththeresultsofthepreviousanalyses,theerrortermsofEXandVI,andofCYandDE,wereallowedtocorrelate.AscanbeseenfromTable4,theresearchmodeldisplayedinFigure1(M)tsmarginallywelltothedatawithonlyRMSEAmeetingitscriterionof0.08andCFIapproaching0.90.InspectionofthemodicationindicesrevealedthatthetwouldimproveiftheerrortermsofDEandAB,andofVIandAB,wouldbeallowedtocorrelate.ThisillustratesthespecialroleofABvis-a-visbothotherengagementscales,aswasindicatedintheintroduction.Indeed,re-specifyingthemodelaccordingly)improvedthetofthemodelsignicantly,0.001,withanadditionalincreaseintindices.Next,twoconstrainedmodelswerettedtothedata:(1)amodelwithallstructuralparametersconstrainedtobeequalacrosssamples(M1constrained);(2)amodelthatinadditionassumesthatallparametersofthemeasurementmodelareequalacrosssamplesconstrained).Inbothcases,comparedtoMre-sepciedthetoftheconstrainedmodelstothedatadete-rioratessignicantly:62.66,0.001and0.001,forM1andM2,respectively.Hence,relationshipsbetweenlatentvariables—andbetweenlatentandobservedvariables—areinvariantacrosssamples.InspectionoftheparameterestimatesinallfoursamplesrevealedthatparticularlytheestimatesinSample4(HomeCareInstitute)differedfromthoseintheremainingthreesamples.Therefore,themulti-groupanalyseswererepeatedwithoutSample4.Indeed,itappearedthatcomparedtoM,thetofM1tothedatadiddeteriorate,26.12,n.s.Thus,therelationshipsbetweenthelatentvariablesasspeciedinourresearchmodel(Figure1)areinvariantacrossSample1,Sample2,andSample3.Figure2pre-sentstherangesofthestandardizedparametersinSamples1–3(top),aswellasthoseofSample4(bottom).InordertotestHypotheses2and3thatassumethatburnoutandengagementplayamediatingrole,wasttedsimultaneouslytothedataofallfoursamplesagain,butnowwithadditionalpathsrunningfromjobdemandsandjobresourcestohealthproblemsandintentiontoquit,respec-tively.IfthismodelwouldtbettertothedatathanMthiswouldsuggestthatburnoutandengagementdoplayamediatingrole.However,thiswasnotthecasesincethealternativemodelBURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) Table3.Mean(),standarddeviation(SD)andcorrelations()ofthestudyvariables(totalsample:VariableSD1234567891011121Workload2.620.492Emotionaldemands2.100.560.203Socialsupport3.890.860.180.014Coaching3.090.760.110.305Feedback2.970.890.120.190.556Exhaustion1.741.050.280.237Cynicism1.761.460.140.090.318Professionalefcacy4.150.830.030.040.250.240.329Vigor3.940.910.000.060.150.250.300.160.5910Dedication4.131.120.020.100.260.290.300.290.630.6611Absorption3.710.940.070.100.070.190.210.120.470.620.6712Healthproblems3.470.370.150.180.200.560.1613Turnoverintention2.431.020.160.050.160.230.150.200.11:Correlationsof0.09orhigheraresignicantat0.001.Thecovariancematricesofthefoursamplesareavailableonrequestfromtherstauthor.W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) ttedsignicantlyworsetothedatacomparedtoM0.001.InspectionoftheparameterestimatesofthedirecteffectsrevealedthatonlyinSample3wastherelationshipbetweenjobresourcesandturnoverintentionsubstantial(i.e.,0.30),whereasintheremainingsevencasesthesedirecteffectswereeithernon-signicantormarginallysignicantwithcriticalvaluesHence,wecanconcludethatHypothesisisconrmed:burnoutfullymediatestherelationshipbetweenjobdemandsandhealthproblems.Inasimilarvein,Hypothesisassumingthatengagementmediatestherelationshipbetweenjobresourcesandturnoverintentionwasalsosupported.OnlyinSample3wasapartialmediationeffectobserved.AscanbeseenfromFigure2,jobdemandsandjobresourceswerenegativelycorrelatedinallsam-ples,butonlyinSamples3(0.30)and4(0.38)wasthisnegativeassociationwassignicant,sothatHypothesis4awaspartiallyconrmed.Furthermore,burnoutandengagementarenegativelyrelated,thusconrmingHypothesis4b.Itshouldbenoted,however,thatasaresultofourpreviousanalyses,PEismodeledasaconstitutingelementofengagementratherthanofburnout.AsexpectedinHypothesis4c,jobresourcesarenegativelyrelatedtoburnout.Initsturn,ashypothesizedinHypothesis4d,burnoutispositivelyrelatedtoturnoverintentions.Finally,andagainasexpected,healthproblemsandturnoverintentionwerepositivelycorrelatedinallsamples.However,theassociationswereratherweakandonlyreachedsignicanceinSample1(0.26),sothatHypothesis4dwasconrmedonlyinSample1. Table4.Thetoftheresearchmodel(seeFigure1),multiplegroupanalysesincludingSample1(Sample2(202),Sample3(507),andSample4(d.f.GFIAGFIRMSEANFINNFICFI1041.462240.920.860.050.870.850.89re-specied876.922160.930.880.040.890.880.91constrained939.582430.920.880.040.880.870.91constrained1265.652740.900.870.050.840.850.87Nullmodel7905.853120.490.400.12———:GFI,Goodness-of-FitIndex;AGFI,AdjustedGoodness-of-FitIndex;RMSEA,RootMeanSquareErrorofApproximation;NFI,NormedFitIndex;NNFI,Non-NormedFitIndex;CFI,ComparativeFitIndex.Fortheadescriptionofthemodels,seetext. Figure2.Theresearchmodel(standardizedpathcoefcients).Top:rangeofcoefcientsinSamples1–3.Bottom:coefcientsofSample4BURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) Inordertoruleoutthepossibilitythatothercross-linksmightexist,Mwasttedtothedataagain,butnowwithtwoadditionalpaths:jobdemandsengagementandengagementproblems.However,comparedtoM,theresultingmodelttedsignicantlyworsetothedata:Overall,itcanbeconcludedthatparticularlystrongandconsistentrelationshipsexistbetween(1)jobdemandsandburnout,(2)burnoutandhealthproblems,and(3)jobresourcesandengagement(seeFigure2).Incontrast,thecross-linksbetweenjobresourcesandburnout,andbetweenburnoutandturnoverintention,aremuchweaker,asistherelationshipbetweenengagementandturnoverintention.ThisagreeswiththetwomainprocessesthatwehavedescribedintheIntroduction:(1)anenergetic,effort-relatedprocess:jobdemandshealthproblems;(2)amotivationprocess:jobresourcesengagementlowturnoverintention.Furthermore,somepathsareremarkablysimilaracrosssampleswithrangessmallerthan0.10,whereasrangesofotherpathsaretypically0.20(theestimatesofSample4exceedtheserangesonsomeoccasions).Thethreestrongestpathsarealsothemostsimilaracrosssamples:demandsburnout,burnouthealthproblems,andresour-engagement(seeFigure2).Inaddition,theestimatedpathburnoutturnoverintentionisrathersimilaracrosssamplesaswell.Finally,morevarianceisexplainedinburnout(meanacrosssam-ples:48percentrange:29–72percent)andhealthproblems(meanacrosssamples:37percent;range:33–42percent)comparedtoengagement(meanacrosssamples:27percent;range:26–28percent)andturnoverintention(meanacrosssamples:17percent;range:9–24percent).Thissuggeststhattheeffort-driven,energeticprocessismorepervasivethanthemotivationalprocess.Theaimofthecurrentstudywastotestaresearchmodelthatspeciespossiblepredictorsandcon-sequencesofburnoutandengagement,basedonanenergeticallydrivenandamotivationaldrivenpro-cess,respectively.Indoingso,wesoughttocontributetowhatiscurrentlyknownaspositivepsychology(Seligman&Csikszentmihalyi,2000).Thatis,inadditiontoanegativepsychologicalstatelikeburnout,whichiscommonlyresearchedinoccupationalhealthpsychology,wealsoexploredtheroleofitspositiveantipode:jobengagement.Inordertoreducetheriskofchangecapitalization,amulti-sampleapproachwasusedthatincludedfoursamplesconsistingofemployeesfromcommercialorganizations(i.e.,aninsurancecompany,apensionfundcompany,andanoccupationalhealthandsafetyservice)aswellasfromanot-for-prothumanserviceorganizationforhomecare.BurnoutandengagementInsteadofloadingononesinglegeneralwell-beingdimension,theburnoutandengagementscalesloadedontwoseparate,moderatelynegativelycorrelatingdimensions.Obviously,incontrasttotheassumptionofMaslachandLeiter(1997),burnoutandengagement—whenmeasuredbydifferentinstruments—domergeintoonesingledimensionwithhighoppositefactorloadingsforeachcon-struct.Instead,itappearedthatanaprioriformulatedalternativemodeltsbesttothedatawhichassumesa‘coreofburnout’dimension(Greenetal.,1991),includingexhaustion(EX)andcynicism(CY)aswellasanenlargedengagementdimension.Thelatterincludesinadditiontothethreeengage-mentscales—vigor(VI),dedication(DE),andabsorption(AB)—thepositivelywordedburnoutscalethatassessesprofessionalefcacy(PE).Bothdimensionsaremoderatelynegativelyrelated,roughlyW.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) sharingonetenthtoaquarteroftheirvariance.Thus,putdifferently,thealternativemodelsuggeststhatPEloadsonthe‘wrong’factor;ratherthanbeingaburnoutcomponent,PEappearstobeanele-mentofengagement.However,itcannotcompletelyberuledoutthatthisndingreectsanartifactsince,liketheengagementitems,allPEitemsarepositivelyformulated,whereasallEXandCYitemsarenegativelyformulated.Asnotedbefore,intryingtominimizeansweringbiasthatcouldleadtosuchanartifactwerandomlymergedallpositiveandnegativeitemsintoonequestionnaire.Ourresultsagreewithcumulatingevidencethatpointstothedivergentrolethatlackofprofessionalefcacyplaysascomparedtoexhaustionandcynicism,whichseemtoconstitutetheessenceoftheburnoutsyndrome(Maslachetal.,2001).Furthermore,ourresultsconcerningthestructureofburnoutandengagementagreewithearlierndingsobtainedbySchaufelietal.(2002a)intwoSpanishsam-plesofstudentsandemployees,respectively.Despitethissuccessfulreplication,thepresentstudywasunabletoshowthatthefactorloadingsandthecorrelationsbetweenburnoutandengagementareinvariantacrossallfoursamples.Thusitseemsthat,althoughthemodeltswellacrosssamples,thecontributionofthescalescorestobothfactorsaswellastherelationshipbetweenthemdiffersfromsampletosample.Inaddition,thecurrentstudyshowedthatthetofthetwo-factormodelimprovessignicantlywhencorrelationsareallowedbetweentheerrortermsofVIandEX,andofDEandCY.ThisagreeswithourtheoreticalreasoningintheintroductionthatVIandEX,andDEandCY,aretheend-pointsofatwounderlyingdimensionsthathavebeenlabeled‘activation’and‘identication,’respectively.Insum,resultsofthecurrentstudyreplicatethendingsofSchaufelietal.(2002a),whoshowedthatburnoutandengagementaremoderatelynegativelyrelated,wherebyinsteadofaburnoutcomponentPEseemstobeanelementofengagement.Moreover,someindicationswerefoundfortheexistenceoftwounderlyingdimensions(i.e.,activationandidentication),whichhintstothecentralroleofVIandEX,andDEandCY,attheexpenseofPEandAB.TheresearchmodelThehypothesizedmodel(seeFigure1)tsreasonablywelltothedataofthefoursamplesunderstudy—albeitaftertwocorrelationsbetweenerrortermsofengagementscaleswereallowed.Thesetwocorrelatederrorterms,whichbothinvolveAB,illustratethespecialrolethatthisdimensionseemstoplay,ascomparedtobothcorecharacteristicsofengagement(i.e.,VIandDE).Furthermore,thestructuralrelationshipsbetweenthelatentvariablesoftheresearchmodelareinvariantacrossthreeofthefoursamples;onlytheestimatesfromSample4(ahomecareorganization)differsystematicallyfromthoseoftheotherthreesamples.Inordertoinvestigatethepossibilitythatthisdifferencereectstheover-representationof(part-timeworking)femalesinSample4ascomparedtotheotherthreesamples(97percentvs.61percent),themodelasdepictedinFigure2(M)wasttedsepa-ratelyinthemale(529)andthefemale(1169)subsamples.Sincecomparedtothefemalesample,Mttedslightlybetterinthemalesample,itcannotberuledoutthatthedivergentresultsobservedinSample4areduetothefactthatthissamplealmostexclusivelyconsistsoffemales,ofwhich—inaddition—53percentwork20hoursperweekorless.Thetwomostcharacteristicdif-ferencesinestimatedmodelparametersthatareobservedamongwomenworkinginhomecarearethathighjobdemandsarerelativelystrongly(0.73)relatedtoburnoutsymptoms,whichintheirturnarerelativelystrongly(0.48)relatedtotheintentiontoleavetheorganization(seeFigure2).Itcanbespeculatedthatthefactthatwomenwhoareemployedinhomecareandwhoexperiencehighlevelsofjobdemandsand(thus)burnoutareparticularlylikelytoleavetheorganizationbecausetheyper-ceiveavailablealternativeopportunities(Rusbult&Farrell,1983).Namely,insteadofworkingfull-timeastheprincipalbreadwinner,mostwomeninhomecareworkpart-timetosupplementthefamilyBURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) budget.Accordingly,thebarriertoleavetheworksituationisratherlow,particularlywhenjobdemandsaswellaslevelsofburnoutarehighthetraditionalroleashouse-wifemaybecomemoreappealing.Notonlydoesthehypothesizedresearchmodelttothedata,butitissuperiortothetoftwoalternativemodelsaswell:(1)amodelthatinadditionassumesdirectpathsfromjobdemandstohealthproblems,andfromjobresourcestoturnoverintention;(2)amodelthatinadditionassumescross-linksbetweenjobdemandsandengagement,andbetweenengagementandhealthproblems.Takentogether,theresultsfromtheseadditionalanalysessuggestthat:(1)burnoutandengagementseemtofullymediatethejobdemands—healthproblemsrelationshipandthejobresources—turn-overintentionrelationship,respectively;(2)burnoutratherthanengagementseemstoplayacentralroleinthemodel;itisrelatedtoallvelatentvariablesinthemodelwhereasengagementisonlyrelatedtothreelatentvariables.Alsothemodelexplainsmorevarianceinburnoutthaninengagement.Themainassumptionoftheproposedmodel,namelythattwounderlyingprocessesexist,remainsunchallenged;(1)aneffort-drivenenergeticprocessinwhichburnoutplaysakeyroleandthatmightleadtonegativehealthoutcomes;(2)amotivationalprocessthatisdrivenbytheavailabilityofjobresourcesandinwhichengagementplaysakeyrole.Itseemsthat—intermsofexplainedvariance—theformerprocessismorepervasivecomparedtothelatterprocess.Moreover,comparedtotherela-tionshipsbetweentheelementsofthesetwoprocesses,bothhypothesizedcross-links(i.e.,jobresour-burnoutandburnoutturnoverintention)arerelativelyweak.Finally,itappearedthatjobdemandsandjobresourcesarenegativelyrelated;thecorrelationisnon-signicantinSamples1and2andsignicantinSamples3and4.Obviously,highlevelsofjobdemandsconcurwithlowlevelsofjobresourcesandviceversa.Thismakessensebecausejobresourceshavebeendenedasthoseaspectsofthejobthatmayreducejobdemands.Sowhenrecoursesarelacking,demandsarenotlikelytobediminishedandwillremainhigh.Onamoregenerallevel,ourstudyillustratesthatnegativepsychologicalstates(i.e.,burnout)andpositivepsychologicalstates(i.e.,engagement)playsimilarrolesinquitedifferentprocesses.Thefor-merplaysamediatingroleinaneffort-basedenergeticprocessthatisdrivenbyhighjobdemandsandthateventuallymightleadtohealthproblems,whereasthelatterplaysamediatingroleinamotiva-tionalprocessthatisdrivenbyavailableresourcesandthatmightleadtoorganizationalattachment(i.e.,alowturnovertendency).Hence,includingapositivepsychologicalstateincreasesourunder-standingofemployeefunctioning.StudylimitationsanddirectionsforfurtherresearchThemainlimitationofthecurrentstudyisitscross-sectionalnature.AlthoughtheuseofarrowsintheresearchmodelasdepictedinFigures1and2suggestscausality,thissuggestionisclearlywrong.Forinstance,itcannotberuledoutthattherelationshipbetweenjobdemandsandburnout,orbetweenburnoutandhealthcomplaints,isinuencedbya‘thirdvariable,’mostnotablynegativeaffectivity(NA;Watson&Clark,1984).Research,however,seemstojustifytheomissionofthispotentialcon-founderinjobstressresearchusingself-reports.Forinstance,Moyle(1995)inastudyofpossibleinuencesthatNAcouldhaveonthestressor–strainrelationshipconcludedthatNAcannotgenerallyaccountfortheobservedcorrelationsbetweenworkenvironmentmeasuresandstrains.InasimilarveinmorerecentstudieshavecorroboratedthatNAdoesnotoverlydistortrelationshipsbetweenself-reportmeasuresofstressorsandstrains(e.g.Schoneld,1996;Williams,Gavin,&Williams,1996;Dollard&Wineeld,1998;DeJongeetal.,2001).Theseresultsledsomeauthorstovoicecriti-cismsagainstthepreviouslyrecommendedpracticetocontrolforperson-basedmeasuresofNA(Briefetal.,1988).Forinstance,Karaseketal.(1998)arguedthatthecurecouldbeworsethantheproblemW.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) andcouldeasilybeoverdone,leadingtoTypeIIstatisticalerrors:truevarianceinstrainmeasurescouldberemovedwithNA.Likewise,afterreviewingarangeofpossiblemechanismsbywhichNAcouldaffectthestressor–strainrelationship,Spector,Zapf,Chen,andFrese(2000)warnnotto‘throwoutthebabywiththebathwater.’TheyarguethatpartiallingNAoutisthewrongapproach,insteadabetterqualityofthedataisthewayforward.Atanyrate,ourresearchmodelneedstobetestedlongitudinally(sinceNAissupposedtobeastablepersonalitytraititisautomaticallyaccountedforinlongitudinalpanelstudies).Thatis,itshouldbeinvestigatedwhetherjobdemandsandjobresourcesatTime1predictburnoutandengagementatTime2,andwhetherburnoutandengagementintheirturnpredicthealthproblemsandorganizationalwithdrawalatTime3.Asfarasburnoutisconcerned,fewlongitudinalstudieswithmixedresultshavebeenconductedregardingitsantecedentsandconsequences(foranoverviewseeSchaufeli&Enzmann,1998,pp.93–98).Sofar,nolongitu-dinalstudiesonengagementhavebeencarriedout.Secondly,ourstudyexclusivelyreliesonself-reportmeasures.AsnotedintheIntroduction,parti-cularlytheeffort-relatedenergeticprocesslinkingjobdemands,burnout,andhealthproblemshasphy-siologicalconcomitantsthathavetobeincludedinfutureresearchonourmodel.Todate,onlyfewstudieshavebeencarriedoutonthepsychophysiologyofburnout(e.g.,Melamed,Ugarten,Shirom,&Kahana,1999),whichyieldedconictingresults.Inasomewhatsimilarvein,thesecondmotivationalprocessincludesbehavioralindicatorsthatmaybemeasuredmoreobjectivelybyusingcompanyles(e.g.,regardingactualpersonnelturnover,sicknessabsenteeism,performance).Sofar,studiesonengagementthatincludesuchobjectivelymeasuredbehavioralindicatorsarelacking.Thirdly,despitethedifferentrolethattheburnoutandengagementseemtoplayintheenergeticandmotivationalprocesses,respectively,theirantecedentsandconsequencespartlyoverlap;thatis,likeengagementburnoutisalsorelatedto(lackof)recourses—albeitmuchweaker—andtoturnoverintention.Theformerrelationshipismostlikelyduetothefactthatdemandsandresourcesarenotindependent—lackingresourcesmayalsoincreasedemands—whereasthelattermightreectthenat-ureburnoutwhichischaracterizedby(mental)withdrawalfromthestressfuljobenvironment.Inordertomakeanevenmoreconvincingcaseforthediscriminantvalidityofburnoutandengagement,itfollowsthatfutureresearchshouldincludeotherresourcesthataremoredistalfromjobdemands(e.g.,careerperspective,skilldevelopment,andlearningopportunities)aswellasotheroutcomesthathavelessconceptualoverlapwithburnout(e.g.,jobperformanceandextra-rolebehavior).Althoughwetestedourmodelinfourindependentsamplesthatweredrawnfromdifferenttypesoforganizations,thereisstillanapparentneedforreplicationinothersamplesusingdifferenttypesofindicatorsforjobdemandsandjobresourcesaswellasfordifferenttypesofindividualandorganiza-tionaloutcomes.Infact,ourmodelisindifferentasfarasspecicjobdemands,jobresources,healthoutcomes,ormotivationaloutcomesareconcerned.PracticalimplicationsSincejobdemandsplayacentralroleinthehypothesizedenergeticprocessthatmightleadtoburnoutandhealthproblems,butalsotopotentialnegativeorganizationaloutcomessuchastheintentiontoleavetheorganization,reducingthosedemandsseemstobewarranted.Manypreventiveorganiza-tionalbasedstrategiesexisttotacklehighjobdemands,suchasjobredesign,exibleworkschedules,andgoalsetting(foranoverviewseeQuick,Quick,Nelson,&Hurrell,1997;pp.163–206).Increasingjobresources(e.g.,throughparticipativemanagement,increasingsocialsupport,andteambuilding),ontheotherhand,wouldeventuallyleadtomoreengagementatthejob,butitsindirecteffectonturn-overintentionisrathersmall;andsoisitsdirecteffectonburnout.Hence,fromapreventivepointofview,decreasingjobdemandsistobepreferredaboveincreasingjobresources.BURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) Sinceburnoutplaysakeyroleinthemodel,individual-basedinterventionstoreduceburnoutsymp-tomsmightalsobeanavenuetoexplore(foranoverviewofsuchinterventionsseeSchaufeli&Enzmann,1998,pp.146–168).Recently,VanderKlink,Blonk,Schene,andVanDijk(2001)haveshowninameta-analysisof48studiesthatparticularlystressmanagementprogramsthatuseacog-nitive-behavioralapproachareeffectiveinreducingstressreactions,includingburnout.Clearly,suchindividual-basedprogramsshouldbesupplementedbyorganization-basedprogramsinordertobeeffectiveinthelongrun.Forinstance,basedonanin-depthanalysesof11casestudies,KompierandCooper(1999)identiedacombinationofwork-directedandworker-directedmeasuresasoneofthevecriticalsuccessfactorsforthepreventionofjobstressinorganizations.Wehopethatwiththeproposedmodelwehavecontributednotonlytoabetterunderstandingofburnout—atraditionalissueinoccupationalhealthpsychology—butalsototherelativelynew‘posi-tive’conceptofjobengagement.AuthorbiographiesWilmarB.SchaufelireceivedhisPhDinpsychologyfromGroningenUniversityintheNetherlands.HeisfullProfessorofClinicalandOrganizationalPsychologyatUtrechtUniversity,TheNetherlands,andscienticdirectoroftheresearchinstitutePsychologyandHealth.Hisresearchareaisoccupa-tionalhealthpsychologyandmoreparticularlyjobstress,workengagement,andburnout.ArnoldB.BakkerreceivedhisPhDinPsychologyfromGroningenUniversityintheNetherlands.HeisAssociateProfessorofSocialandOrganizationalPsychologyatUtrechtUniversity,theNetherlands,anddirectorofHCMGHumanCapitalManagementGroup.Hisresearchinterestsincludeworkengagement,owatwork,andburnout.Hespecializesinweb-enabledresearchinorganizations.ReferencesArbuckle,J.L.(1997).AmosusersÕguideversion4.0.Chicago,IL:SmallwatersCorporation.Bakker,A.B.,Schaufeli,W.B.,Sixma,H.,Bosveld,W.,&VanDierendonck,D.(2000).Patientdemands,lackofreciprocity,andburnout:ave-yearlongitudinalstudyamonggeneralpractitioners.JournalofOrganizationalBehavior,21,425–441.Baumeister,R.,&Leary,M.R.(1995).Theneedtobelong:desireforinterpersonalattachmentsasafundamentalhumanmotivation.PsychologicalBulletin,117,497–529.Bentler,P.M.(1990).Comparativetindexesinstructuralequationmodels.PsychologicalBulletin,107,238–Bollen,K.A.(1989).Anewincrementaltindexforgeneralstructuralmodels.SociologicalMethodsandResearch,17,303–316.Brief,A.P.,Burke,M.J.,George,J.M.,Robinson,B.S.,&Webster,J.(1988).Shouldnegativeaffectivityremainanunmeasuredvariableinthestudyofjobstress?JournalofAppliedPsychology,73,207–214.Cohen,L.,&Holliday,M.(1982).Statisticsforthesocialsciences.London:Harper&Row.Csikszentmihalyi,M.(1990).Flow:Thepsychologyofoptimalexperience.NewYork:Harper.Cudeck,R.,&Browne,M.W.(1993).Alternativewaysofassessingmodelt.InK.A.Bollen,&J.ScottLongTestingstructuralequationmodels(pp.1–9).NewburyPark,CA:Sage.DeCharms,R.(1968).Personalcausation.NewYork:AcademicPress.Deci,W.L.,&Ryan,R.M.(1985).Intrinsicmotivationandself-determinationinhumanbehavior.NewYork:W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) DeJonge,J.,Dollard,M.F.,Dormann,C.,LeBlanc,P.M.,&Houtman,I.L.D.(2000).Thedemand–controlmodel:specicdemands,speciccontrol,andwell-denedgroups.InternationalJournalofStressManagement,7,269–287.DeJonge,J.,Dormann,C.,Janssen,P.M.P.,Dollard,M.F.,Landeweerd,J.A.,&Nijhuis,F.J.N.(2001).Testingthereciprocalrelationshipsbetweenjobcharacteristicsandpsychologicalwell-being:acrosslaggedstructuralequationmodel.JournalofOccupationalandOrganizationalPsychology,74,29–46.Demerouti,E.,Bakker,A.B.,DeJonge,J.,Janssen,P.P.M.,&Schaufeli,W.B.(2001a).Burnoutandengagementatworkasafunctionofdemandsandcontrol.ScandinavianJournalofWorkandEnvironmentandHeath,27Demerouti,E.,Bakker,A.B.,Nachreiner,F.,&Schaufeli,W.B.(2001b).Thejobdemands:resourcesmodelofburnout.JournalofAppliedPsychology,86,499–512.Demerouti,E.,Bakker,A.B.,Vardakou,I.,&Kantas,A.(2003).Theconvergentvalidityoftwoburnoutinstruments:amultitrait–multimethodanalysis.EuropeanJournalofPsychologicalAssessment,18,296–Diener,E.(1999).Introductiontothespecialsectiononthestructureofemotion.JournalofPersonalityandSocialPsychology,76,803–804.Dirken,J.M.(1969).Arbeidenstress:Hetvaststellenvanaanpassingsproblemeninwerksituaties[Workandstress:determiningproblemsinadaptingtowork-settings].Groningen:Wolters-Noordhoff.Dollard,M.F.,&Wineeld,A.H.(1998).Atestofthedemand–control/supportmodelofworkstressincorrectionalofcers.JournalofOccupationalHealthPsychology,3,243–263.Geurts,S.,Schaufeli,W.B.,&DeJonge,J.(1998).Burnoutandintentiontoleaveamonghealth-careprofessionals:asocialpsychologicalapproach.JournalofSocialandClinicalPsychology,17,341–362.Graen,G.B.,&Uhl-Bien,M.(1991).Thetransformationintoprofessionalsintoself-managingandpartiallyself-designingcontributors:towardatheoryofleadershipmaking.JournalofManagementSystems,3,25–39.Green,D.E.,Walkey,F.H.,&Taylor,A.J.W.(1991).Thethree-factorstructureoftheMaslachburnoutinventory.JournalofScienceBehaviorandPersonality,6,453–472.Hackman,J.R.,&Oldham,G.R.(1980).Workredesign.Reading,MA:Addison-Wesley.Hobfoll,S.E.(2002).Socialandpsychologicalresourcesandadaptation.ReviewofGeneralPsychology,6,307–Hockey,G.J.(1993).Cognitive–energeticalcontrolmechanismsinthemanagementofworkdemandsandpsychologicalhealth.InA.D.Baddeley,&L.Weiskrantz(Eds.),Attention,selection,awarenessandcontrol:AtributetoDonaldBroadbent(pp.328–345).Oxford:OxfordUniversityPress.Hockey,G.J.(1997).Compensatorycontrolintheregulationofhumanperformanceunderstressandhighworkload:acognitive–energeticalframework.BiologicalPsychology,45,73–93.Houkes,I.,Janssen,P.P.M.,DeJonge,J.,&Nijhuis,F.J.N.(2001).Specicrelationshipsbetweenworkcharacteristicsandintrinsicworkmotivation,burnoutandturnoverintention:amulti-sampleanalysis.EuropeanJournalofWorkandOrganizationalPsychology,10,1–23.Hoyle,R.H.(1995).Thestructuralequationmodelingapproach:basicconceptsandfundamentalissues.InR.H.Hoyle(Ed.),Structuralequationmodeling,concepts,issues,andapplications(pp.1–15).ThousandOaks,CA:Jamal,M.(1990).TherelationshipofjobstressandType-Abehaviourtoemployees’jobsatisfaction,organizationalcommitment,psychosomatichealthproblems,andturnovermotivation.HumanRelations,43Jones,F.,&Fletcher,B.C.(1996).Jobcontrolandhealth.InM.J.Schabracq,J.A.M.Winnubst,&C.L.CooperHandbookofworkandhealthpsychology(pp.33–50).Chichester:Wiley.reskog,K.G.,&Sorbom,D.(1986).LISRELuserguideversionVI(4thed).Mooresville,IL:ScienticSoftwareKarasek,R.A.,Brisson,C.,Kawakami,N.,Houtman,I.,&Bongers,P.(1998).TheJobContentQuestionnaire(JCQ):aninstrumentforinternationallycomparativeassessmentsofpsychosocialjobcharacteristics.JournalofOccupationalHealthPsychology,3,322–355.VanderKlink,J.J.L.,Blonk,R.W.B.,Schene,A.H.,&VanDijk,F.J.H.(2001).Thebenetofinterventionsforworkrelatedstress.AmericanJournalofPublicHealth,91,270–276.Karasek,R.A.(1985).JobcontentquestionnaireanduserÕsguide.Lowell:UniversityofMassachusettsLowell,DepartmentofWorkEnvironment.Knardahl,S.,&Ursin,H.(1985).Sustainedactivationandthepathophysiologyofhypertensionandcoronaryheartdisease.InJ.Orlebele,G.Mulder,&L.VanDoornen(Eds.),Psychophysiologyofcardiovasculaircontrol(pp.151–167).NewYork:PlenumPress.BURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) Kompier,M.A.J.,&Cooper,C.L.(Eds.).(1999).Preventingstress,improvingproductivity:Europeancasestudiesintheworkplace.London:Routledge.LeBlanc,P.(1994).Desteunvandeleiding:EenonderzoeknaarhetLeaderÐMemberExchangemodelindeverpleging[Leader’ssupport:AstudyoftheLeader–MemberExchangemodelamongnurses].Amsterdam:Lee,R.T.,&Ashforth,B.E.(1993).Alongitudinalstudyofburnoutamongsupervisorsandmanagers:comparisonbetweentheLeiterandMaslach(1988)andGolembiewskietal.(1986)models.OrganizationalBehaviorandHumanDecisionProcesses,54,369–398.Lee,R.T.,&Ashforth,B.E.(1996).Ameta-analyticexaminationofthecorrelatesofthethreedimensionsofjobburnout.JournalofAppliedPsychology,81,123–133.MacCallum,R.,Roznowski,M.,&Necowitz,L.B.(1992).Modelspecicationsincovariancestructureanalysis:theproblemofcapitalizationonchance.PsychologicalBulletin,111,490–504.Marsh,H.W.,Balla,J.R.,&Hau,K.T.(1996).AnevaluationofIncrementalFitIndices:aclaricationofmathematicalandempiricalproperties.InG.A.Marcoulides,&R.E.Schumacker(Eds.),Advancedstructuralequationmodeling,issuesandtechniques(pp.315–353).Mahwah,NJ:Erlbaum.Maslach,C.(1993).Burnout:amultidimensionalperspective.InW.B.Schaufeli,C.Maslach,&T.Marek(Eds.),Professionalburnout:Recentdevelopmentsintheoryandresearch(pp.19–32).Washington,DC:Taylor&Maslach,C.,&Jackson,S.E.(1986).MaslachBurnoutInventory:Manual(2nded.).PaloAlto,CA:ConsultingPsychologistsPress.Maslach,C.,Jackson,S.E.,&Leiter,M.(1996).MaslachBurnoutInventory:Manual(3rded.).PaloAlto,CA:ConsultingPsychologistsPress.Maslach,C.,&Leiter,M.P.(1997).Thetruthaboutburnout:Howorganizationscausepersonalstressandwhattodoaboutit.SanFrancisco,CA:Jossey-Bass.Maslach,C.,&Schaufeli,W.B.(1993).Historicalandconceptualdevelopmentofburnout.InW.B.Schaufeli,C.Maslach,&T.Marek(Eds.),Professionalburnout:Recentdevelopmentsintheoryandresearch(pp.1–16).Washington,DC:Taylor&Francis.Maslach,C.,Schaufeli,W.B.,&Leiter,M.P.(2001).Jobburnout.AnnualReviewofPsychology,52,Meijman,T.F.,&Mulder,G.(1998).Psychologicalaspectsofworkload.InP.J.D.Drenth,&H.Thierry(Eds.),HandbookofworkandorganizationalpsychologyWorkpsychology(Vol.2,pp.5–33).Hove,U.K.:PsychologyMelamed,S.,Ugarten,U.,Shirom,A.,&Kahana,L.(1999).Chronicburnout,somaticarousalandelevatedsalivarylevels.JournalofPsychosomaticResearch,46,591–598.Moyle,P.(1995).Theroleofnegativeaffectivityinthestressprocess:testsofalternativemodels.JournalofOrganizationalBehavior,16,647–668.Myers,D.G.(2000).Thefunds,friends,andfaithofhappypeople.AmericanPsychologist,55,56–67.Nunnaly,J.C.,&Bernstein,I.H.(1994).Psychometrictheory(3rded.).NewYork:McGraw-Hill.Quick,J.C.,Quick,J.D.,Nelson,D.L.,&Hurrell,J.J.(Eds.).(1997).Preventivestressmanagementinorganizations.Washington,DC:AmericanPsychologicalAssociation.Rusbult,C.E.,&Farrell,B.(1983).Alongitudinaltestoftheinvestmentmodel:theimpactonjobsatisfaction,jobcommitment,andturnoverofvariationsinrewards,costs,alternatives,andinvestments.JournalofAppliedPsychology,68,429–438.Russell,J.A.,&Carroll,J.M.(1999).Onthebipolarityofpositiveandnegativeaffect.PsychologicalBulletin,,3–10.Ryan,R.M.,&Frederick,C.(1997).Onenergy,personality,andhealth:subjectivevitalityasadynamicreectionofwell-being.JournalofPersonality,65,529–565.Schaufeli,W.B.,&Bakker,A.B.(2001).Werkenwelbevinden:naareenpositievebenaderingindeArbeids-enGezondheidspsychologie[Workandwell-being:towardsapositiveapproachinOccupationalHealthGedrag&Organisatie,14,229–253.Schaufeli,W.B.,&Enzmann,D.(1998).Theburnoutcompaniontostudyandresearch:AcriticalanalysisLondon:Taylor&Francis.Schaufeli,W.B.,&VanDierendonck,D.(2000).HandleidingvandeUtrechtseBurnoutSchaal(UBOS)(UBOS)UtrechtBurnoutScale].Lisse:SwetsTestServices.Schaufeli,W.B.,Leiter,M.P.,Maslach,C.,&Jackson,S.E.(1996).MaslachBurnoutInventory–GeneralSurvey.InC.Maslach,S.E.Jackson,&M.P.Leiter(Eds.),TheMaslachBurnoutInventory:Testmanual(3rded.,pp.22–26).PaloAlto,CA:ConsultingPsychologistsPress.W.B.SCHAUFELIANDA.B.BAKKER2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004) Schaufeli,W.B.,Taris,T.,LeBlanc,P.,Peeters,M.,Bakker,A.,&DeJonge,J.(2001).Maaktarbeidgezond?Opzoeknaardebevlogenwerknemer[Workandhealth:thequestoftheengagedworker].DePsycholoog,36Schaufeli,W.B.,Salanova,M.,Gonza,V.,&Bakker,A.B.(2002a).Themeasurementofengagementandburnoutand:aconrmativeanalyticapproach.JournalofHappinessStudies,3,71–92.Schaufeli,W.B.,Martnez,I.,MarquesPinto,A.,Salanova,M.,&Bakker,A.B.(2002b).Burnoutandengagementinuniversitystudents:acrossnationalstudy.JournalofCrossCulturalPsychology,33,464–481.Schoneld,I.S.(1996).Relationofnegativeaffectivitytoself-reportsofjobstressorsandpsychologicalJournalofOccupationalHealthPsychology,1,397–412.Schutte,N.,Toppinnen,S.,Kalimo,R.,&Schaufeli,W.B.(2000).ThefactorialvalidityoftheMaslachBurnoutInventory–GeneralSurveyacrossoccupationalgroupsandnations.JournalofOccupationalandOrganiza-tionalPsychology,73,53–66.Seligman,M.E.P.,&Csikszentmihalyi,M.(2000).Positivepsychology:anintroduction.AmericanPsychologist,,5–14.Spector,P.E.,Zapf,D.,Chen,P.Y.,&Frese,M.(2000).Whynegativeaffectivityshouldnotbecontrolledinjobstressresearch:don’tthrowoutthebabywiththebathwater.JournalofOrganizationalBehavior,21,79–95.Taris,T.W.,Bok,I.A.,&Meijer,Z.Y.(1998).Assessingstabilityandchangeofpsychometricpropertiesofmulti-itemconceptsacrossdifferentsituation:ageneralapproach.JournalofPsychology,123,301–316.Toppinen-Tanner,S.,Kalimo,R.,&Mutanen,P.(2002).Theprocessofburnoutinwhite-collarandblue-collarjobs:eightyearprospectivestudyofexhaustion.JournalofOrganizationalBehavior,23,555–570.VanVeldhoven,M.,&Meijman,T.F.(1994).Hetmetenvanpsychosocialearbeidsbelastingmeteenvragenlijst:DeVragenlijstBelevingenBeoordelingvandeArbeid[Themeasurementofpsychosocialstrainatwork:thequestionnaireexperienceandevaluationofwork].Amsterdam:NIA.VanVeldhoven,M.,DeJong,J.,Broersen,S.,Kompier,M.,&Meijman,Y.(2002).Specicrelationshipsbetweenpsycehosocialjobconditionsandjobrelatedstress:athree-levelanalyticapproach.WorkandStress,16,207–Watson,D.,&Clark,L.A.(1984).Negativeaffectivity:thedispositiontoexperienceaversiveemotionalstates.PsychologicalBulletin,96,465–490.White,R.W.(1959).Motivationreconsidered:theconceptofcompetence.PsychologicalReview,66,297–333.Williams,L.J.,Gavin,M.B.,&Williams,M.L.(1996).Measurementandnon-measurementprocesseswithnegativeaffectivityandemployeeattitudes.JournalofAppliedPsychology,54,87–94.BURNOUTANDENGAGEMENT2004JohnWiley&Sons,Ltd.J.Organiz.Behav.,293–315(2004)