the effect of job insecurity on labour supply · 2017. 2. 15. · evaluate their level of job...

187AUSTRALIAN JOURNAL OF LABOUR ECONOMICS

Volume 18 • Number 2 • 2015 • pp 187 - 204

The Effect of Job Insecurity on Labour Supply H. Xavier Jara,UniversityofEssex

Abstract The aim of this paper is to analyse the effect of job insecurity on labour supply. We propose a discrete choice model of labour supply, in which the choice alternatives are characterised by bundles of income, hours of work and job insecurity. The results show that job insecurity has a negative and significant effect on individuals’ utility. Moreover, once job insecurity is included in the discrete choice alternatives, the predictive power of the model improves significantly. Labour supply elasticities are significantly higher than those obtained with a traditional model and increase with the level of job insecurity. Finally, a decrease of job insecurity at work has a positive and significant effect on participation. Policies aimed at improving working conditions could, in this sense, be useful to create incentives in labour market.

Keywords:Discretechoice,Laboursupply,Jobinsecurity

JELclassification:C25,J22,J81

1. Introduction Aconsiderablebodyofresearchhasdocumentedtheimportanceofjobinsecurityasaworkdomain.Attheindividual level, jobinsecurityhasbeenfoundtonegativelyaffectjobsatisfaction(Clark,2010;DeWitte,2005;Hellgreen,et al.1999),aswellas physical and mental health (Burgard et al. 2009; DeWitte, 1999; Dekker andSchaufeli, 1995; Ferrie, 1998). At the organisational level, job insecurity has beenshowntoberelatedwithhigherabsenteeism(ChirumboloandAreni,2005;Probst,2002),higherturnoverandquitintentions(Clark,2001;DekkerandSchaufeli,1995)andlessorganisationalcommitment(Hellgreen,et al.1999;LordandHartley,1998;Rosenblatt,et al.1999).However,theeffectofjobinsecurityatamoreaggregatelevelhasbeen lessconsidered in the literature. In thisarticle,westudy theeffectof jobinsecurityonlaboursupply.Inparticular,weextendtraditionaldiscretechoicemodelsoflaboursupplytoincorporatejobinsecurityinthechoicealternativesandcomparetheestimatedlaboursupplyresponsestothoseofamodelwhereonlyadiscretehours’setcharacterisesjobalternatives.

Addressforcorrespondence:H.XavierJara,InstituteforSocialandEconomicResearch,UniversityofEssex,WivenhoePark,CO34SQ,Colchester,UnitedKingdom.Email:[email protected]: The author is grateful to Erik Schokkaert, AndréDecoster, ErwinOoghe,Maarten Goos, Rolf Aaberge and Peter Haan for valuable comments and remarks on earlierversionsofthepaper.Theauthorisresponsibleforallremainingshortcomings.©TheCentreforLabourMarketResearch,2015

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by University of Essex Research Repository

https://core.ac.uk/display/74373922?utm_source=pdf&utm_medium=banner&utm_campaign=pdf-decoration-v1

188AUSTRALIAN JOURNAL OF LABOUR ECONOMICSVOLUME 18 • NUMBER 2 • 2015

Discretechoicemodelsoflaboursupplyhavebecomeincreasinglypopularastheyfacilitatedealingwithnon-linearandnon-convexbudgetsetsaswellasaccountingformultiplegoodsintheutilityfunction,comparedtothetraditionalapproachbasedon a continuous set of hours. The idea behind the discrete choice approach is todefineafinitenumberofworkinghours’alternativesandtoexplicitlyspecifyautilityfunctioncharacterisingtheindividual’sutilityateachofthealternativesofthediscretehoursset.Theestimationofthediscretechoicemodelprovidesdirectlytheparametersdefiningtheshapeoftheutilityfunction.

Most studies, using the discrete labour supply approach, take income andhoursofworkastheonlychoicevariablesaffectingindividuals’decisions.However,weagreewithDagsvikandStrøm(2006)that‘hoursofworkandincomeareonlytwooutofseveraljobrelatedattributes,whichareimportantforindividualbehaviourinthelabourmarket’.Dagsvik(1994)andDagsvikandStrøm(1992)proposeamodeloflaboursupplywhichaccountsfortheimportanceofqualitativefactorsofjobs.Thismodelofdiscretechoicelaboursupplyassumesthatthealternativesarecharacterisedby‘jobpackages’whicharedefinedbyabundleofhoursofwork,wageratesandothernon-pecuniaryjobattributes.OtherstudiesonlaboursupplysuchasAaberge,DagsvikandStrøm(1995),AabergeandColombino(2013)andDagsvikandStrøm(2006)useasimilarmethodology.DagsvikandStrøm(2006)introduce,forinstance,jobsector(publicorprivatesector)intheiranalysis,assumingthatjobsindifferentsectorsmaydifferintermsofnon-pecuniaryattributes.Inarecentpaper,KunzeandSuppa(2013)investigatetheeffectof introducingjobcharacteristics indiscretechoicemodelsoflabour supply.However, alternatives are defined only over discrete hours’ choices,while job characteristics enter the utility function through interactions betweenincomeandleisure.ContrarytoKunzeandSuppa(2013),inthispaper,weproposean extendeddiscrete choicemodel,where the choice alternatives are characterisedbybundlesof income,hoursofworkand job insecurity inorder toallowformoreflexibilityinthechoicesavailabletoindividualsandtoanalysetheeffectofchangesinjobinsecurityonlaboursupplydecisions.

Ourcontribution to the literature is twofold.First,weprovideanextensionofdiscretechoicelaboursupplymodelsinordertoallowfortheintroductionofnon-pecuniaryjobattributesintheanalysis, inourcasejobinsecurity.Weanalysehowsuchextensionaffectslaboursupplyelasticities.Second,wecomplementtheliteratureontheconsequencesofjobinsecurity.Inparticular,weshowthatjobinsecurityhasa significant effecton labour supplydecisions,whichcouldbeof interest in termsoflabourmarketpoliciesbasedonnon-monetaryincentives.Thepaperisstructuredasfollows.Section2brieflydiscussedthenatureofjobinsecurityasajobattribute.Section3presentsthediscretechoicemodelstobeusedinourlaboursupplyanalysis.First,wepresentthetraditionallaboursupplymodelwhereonlyhoursofworkdefinethechoiceset.Then,theextendedconditionallogitisintroducedinordertoallowthechoicesettobecharacterisedbybundlesofincome,hoursofworkandjobinsecurity.Section4describesthedataandpresentssomesummarystatistics.Section5presentsthe estimates of the structural labour supply models. Section 6 discusses labourresponsesintermsofwageelasticitiesandchangesinthepredictedprobabilitiesfromadecreaseofjobinsecurity.Finally,section7concludes.

189H. XAVIER JARA

The Effect of Job Insecurity on Labour Supply

2. Job Insecurity as a Job Attribute Jobsarecharacterisedbymultipleattributes,whichmightaffectindividuals’laboursupplydecisions.Inthispaper,wefocusonjobinsecurity,whichhasbeenconsideredoneof themost importantdomainsatwork (Clark,2001and2010).Moreover, jobinsecurityhasprovedtosignificantlyaffectimportantindividuals’outcomessuchaswell-being(Clark,2001and2010;Green,et al.2013),health(Burgard,et al.2009;Ferrie, 1998) and organizational commitment (Hellgreen, et al. 1999; Lord andHartley,1998;Rosenblatt,et al.1999).

AccordingtoGreenhalghandRosenblatt(1984)jobinsecurityisdefinedastheperceivedpowerlessnesstomaintainthedesiredjobcontinuity.Theideathatjobinsecurityreferstoindividuals’perceptionoftheirjobsituationhighlightsthefactthatbothasubjectiveandanobjectivecomponentcharacterisethisconcept.Individualsevaluatetheirlevelofjobinsecuritybasedonobjectiveinformationfromtheirjobs.Studies have shown that perceived job insecurity provides reliable informationabout objective indicators of insecure jobs. In particular, perceived job insecurityissignificantlyassociatedwithtemporaryemploymentandjobsector;publicsectorbeingconsideredmoresecurethantheprivatesector(Campbell,et al.2007;ClarkandPostel-Vinay,2009;DeloffreandRioux,2003;NäswallandDeWitte,2003).Moreover,perceivedjobinsecurityhasprovedtobeagoodpredictoroffutureunemploymentexperiences even after controlling for observed objective variables, implying thatself-perceivedjobinsecuritycontainsusefulprivateinformationaboutjobs,whichisotherwisenotdirectlyavailableinsurveys(Campbell,et al.2007;DeloffreandRioux,2003).Infact,restructurationsuchasprivatizationofformerlypubliccompanies,aswellaslayoffshavebeenshowntoincreaseself-perceivedjobinsecurity(Ferrie,et al.1995;Nelson,et al.1995).

The particular interpretation of job insecurity information is moreoverlinkedtotheformulationofthequestionsincludedinsurveys.Atleastthreedifferentformulationsareusedinsurveys,whicharerelatedtofeelingsofinsecurity(‘Doyoufeelyourjobissecure?’),satisfactionwithjobinsecurity(‘Howsatisfiedareyouwithyourjobsecurity?’)andtheprobabilityoflosingajob(‘Whatistheprobabilitythatyouloseyourjobinthenext(e.g.12months?’).Notethatthefirsttwoformulations,inparticular,confoundtwocomponentsofjobinsecuritydiscussedintheliterature:theprobabilityofjoblossandthecostofjobloss(Campbell,et al.2007).Thevariablesrelatedtothesethreetypesofquestionsarehoweverhighlycorrelatedandallthreearesignificantlyassociatedwithobjectiveindicatorsofinsecurejobs.Nevertheless,itisimportanttokeepinmindwhichtypeofinformationisavailablewheninterpretingtheresults.

While acknowledging the particularities related to the concept of jobinsecurity, in this paperweconsider it as aproxy for an indicator of theobjectiveinsecuritycharacterisingajob.Throughoutouranalysis,wediscusstheimplicationsofsuchassumptiongiventhetypeofinformationavailableintheBritishHouseholdPanelSurvey(BHPS)usedinourstudy.


3. Discrete Choice Models of Labour Supply Discrete choice models of labour supply are particularly popular in the frameworkofbehaviouralmicrosimulationof taxandbenefitreforms.Infact,manypoliciesarespecifically aimed at encouraging labour market participation of certain populationgroups.Forinstance,Brewer,et al.(2007)useastructuralmodeloflaboursupplyinordertoevaluatetheeffectoftheWorkingFamiliesTaxCreditonlabourmarketparticipationintheUK.Moreover,othertaxandbenefitreformsmightalsoaffectindividuals’laboursupplybehaviour,whichneedstobetakenintoaccountwhenevaluatingtheeffectofsuchpoliciesondifferentoutcomevariables,suchaspovertyorinequality.

Inthissectionwedescribethemodelmostwidelyusedtoestimatediscretechoice labour supply, namely the conditional logit model. The model is derivedundertheassumptionofutilitymaximisation.Considerindividual ichoosesamongafinite number of job alternatives, J. The utility obtained fromalternative j isUij,j =1,…,J.IndividualichoosesalternativejifandonlyifUij >Uik,∀k ≠j.Theutilityfunctioncanbedecomposedinadeterministicandastochasticcomponent:Uij =Vij +eij,wherethedistributionoftherandomvectorei ={ei1,…,eiJ}isgivenbyF(ei).Theprobabilitythataparticularalternativejischosenis:

Pij =Prob(Uij >Uik,∀k ≠j) =Prob(Vij +eij >Vik +eik,∀k ≠j) =Prob(eik <eij +Vij−Vik,∀k ≠j)

Dependingonthespecificationof thedistributionof therandomcomponent,

differentdiscretechoicemodelscanbeobtained.Theconditionallogitmodelisobtainedassumingthatthestochasticcomponent,eij,isindependentandidenticallydistributedoveralternativesandfollowsatype-oneextremevaluedistribution,givenby:

F(ei)=e

−e−eij

Undertheconditionallogitsetup,theprobabilitythatalternativejischosenis

givenby(McFadden,1974):

Pij =Prob(eik <eij +Vij−Vik,∀k ≠j)

=∑Jk=1e

Vik

In our basicmodel, individuals choose among a finite number ofworking

hoursalternativesinordertomaximisetheirutility,definedovernetincomeandhoursofwork.Weassumethatthegrosswageratesarefixedandindependentofthehoursofwork.Thedecisionistakengiventhegrosswageratesandthetaxandbenefitsystem.

Moreformally,lethibethenumberofhoursworkedbyindividuali.WedefineJ discrete hours alternatives so that hij represents the number of hoursworked byindividualiunderalternativej,withj =1,…,J.Inourbasicmodel,fouralternativesaredefined,J =4:inactivity,part-time,full-time,overtime.Letyijbeindividuali’snetincomegiventhehourschoicehijandxiavectorofindividualcharacteristics.Thenetincomeyij,whenhi =hijischosen,isdefinedas:

eVij

191H. XAVIER JARA


yij =wi hij +mi +G(wi,hij,mi,xi),

where wi are gross hourly wage rates, mi is non-labour income and the functionG(wi,hij,mi,xi)representsthetax-benefitruleswhichdependongrosswages,hoursofwork,non-labourincomeandindividualcharacteristics.Severalfunctionalformscanbeusedtospecifythedeterministicpartoftheutilityfunction.FollowingKeaneandMoffitt(1998)andBreweret al.(2007),wedefineitasasecondorderpolynomial.Inourbasicconditionallogitmodel,thedeterministicpartoftheutilityfunctionisgivenby:

V(yij,hij,xi)=ayy y2ij +ahhh

2ij +ayh yij hij +ay yij +ah(xi)hij ,

whereweaccountforobservedheterogeneityinpreferencesforhoursofworkthroughinteractionswithpersonalcharacteristics:

ah(xi)=ah0+a′hx xiInourextendedmodel, jobinsecurity is introducedasanon-pecuniaryjob

attributeaffectinglaboursupply.Threejobinsecuritylevelsaredefinedcharacterisinglow,middleandhighjobinsecurity.Atotaloftenalternativesareavailablerepresentinginactivityandcombinationsofhoursofworkandjobinsecuritylevels.Moreformally,let sij, represent the level of job insecurity of individual i under alternative j. Thedeterministicpartoftheutilityfunction,inourextendedmodelisgivenby:

V(yij,hij,xi)=ayy y2ij +ahhh

2ij +asss

2ij +ayh yij hij +ays yij sij +ahs hij sij +ay yij

+ah(xi)hij +as(xi)sij ,

wherewe allow for observed preferences heterogeneity for hours ofwork and jobinsecurity:

ah(xi)=ah0+a′hx xi

as(xi)=as0+a′sx xi

Unobserved heterogeneity in preferences could also be accounted for byintroducing random terms in ah(xi) and as(xi) (see, Train, 1998 and Train, 2003).However, in our empirical analysis, unobservedheterogeneity in preferences is notaccountedforgiventhesmallsizeofoursample.Thesamplelikelihoodfunctionfortheconditionallogitmodelisgivenby:

L =∏∏[Pij(yij,hij,xi)]

dij

wheredijisadummyequaltooneifindividualichoosesalternativejandzerootherwise.Notethatinordertoconstructthediscretechoicealternativesinlaboursupply

N

i = 1

J

j = 1


modelsitisusuallyassumedthatgrosshourlywagesarefixedandindependentofhoursofwork.1Whileindependencebetweenwagesandhoursofworkisgenerallyacceptedin the structural labour supply literature, in our extended models the relationshipbetweenwagesandjobinsecuritydeservesmoreattention.Infact,ifjobinsecurityisconsideredadisamenitysomesortofcompensatingwagedifferentialsmightexistfor jobswith higher insecurity,whichwould need to be taken into account in ourlaboursupplymodel.2Forinstance,consideranindividualcurrentlyinajobwithlowjob insecurity, inorder toconstructallherpossiblechoicealternatives,weneed todefinewhatwouldbeherwageunderthealternativeofmiddleandhighinsecurity.Inthisstudy,wetakeintoaccounttherelationshipbetweenwagesandjobinsecuritybyrandomlyassigningwagesbyeducationgroupsfromthewagedistribution,aswillbeexplainedinsection4.

4. Data OuranalysisusesdatafromwavetenoftheBritishHouseholdPanelSurveycontaininginformationforyears2000and2001.TheBHPSisanationallyrepresentativesurveyfortheUnitedKingdom,whichprovidesinformationaboutindividualandhouseholdcharacteristics,wages, other income sources andworking conditions.We limit ouranalysis towave tenof theBHPSbecauseof theneedofdevelopingadetailed taxandbenefitmicrosimulationmodeltocalculatedisposableincomeforeachdiscretehouralternative.Ourmicrosimulationmodel isbasedonEUROMODversion21A,whichsimulatestax-benefitrulesfortheUKin2001(see,SutherlandandGutierrez,2004).Wavetenof theBHPScontains15,603individuals,however,werestrictouranalysistosinglefemales,whogavefullinterview.Thisrestrictionismadefortworeasons.First,focusingonsingleindividualsenablesustoneglectinteractionswithinthehouseholdinthecontextoflaboursupply.Second,thesampleofsinglemalesistoosmallfortheestimationofthemodelsandajointestimationofmalesandfemalesmightbias theresultsforwomen.Asit isusuallydonein the literature,wefurtherexclude individuals in self-employment because their labour supply decisionsmaydifferconsiderablyfromthoseofsalariedworkersandtheirincomeinformationfromsurveysisconsideredlessreliablethanforemployees.Disabledindividuals,full-timestudents and pensioners are also excluded in order to keep only those individualsavailableforthelabourmarket.Thisleavesuswithasampleof750females.

Beforerestrictingouranalysistooursampleofinterestweneedtotreattheproblemofnon-observedwagesfornon-workers.Wedothisbyestimatingatwo-stepHeckmanselectionmodelforwomen,usingthewholesampleoffemales(N=8,035).Thewageequationdependsonvariables related tohumancapital, suchasageandeducationaswellasregiondummiestocontrolfordifferencesinlabourmarkets.Theselectionequationisbasedontheusualexclusionrestrictionsforidentificationwhere,inadditionaltotheprevious,non-labourincome,beingmarriedandhavingchildrenofdifferentagesareincludedasvariables(see,VanSoest,1995;Haan,2010;Bargain,et al.2014).Theresultsoftheestimationareshownintable1.

1 Some exceptions are studies by Aaberge and co-authors. See, for instance, Aaberge andColombino(2013).2Foradiscussionaboutcompensatingwagedifferentialssee,Rosen(1987).

193H. XAVIER JARA


Table 1 - Heckman Selection Model for Females

coef. st. errorlog hourly wage equationage 0.0590*** (0.0037)age² -0.0007*** (4.82e-5)cse 0.146*** (0.0321)o-levels 0.224*** (0.0253)a-levels 0.340*** (0.0285)higherdegree 0.455*** (0.0251)universitydegree 0.818*** (0.0284)constant 0.368*** (0.0787)selection equationage 0.148*** (0.0073)age² -0.002*** (8.61e-5)cse 0.203*** (0.0682)o-level 0.442*** (0.0537)a-level 0.437*** (0.0637)higherdegree 0.601*** (0.053)universitydegree 0.573*** (0.0655)non-labourincome -0.0014*** (6.28e-5)married -0.188*** (0.0414)child(0-2) -0.464*** (0.0613)child(3-4) -0.381*** (0.0614)child(5-11) -0.183*** (0.0436)child(12-15) -0.0334 (0.0456)constant -1.911*** (0.152)lambda 0.0729*** (0.0254)Observations 8,035

Note:RegiondummiesincludedinthewageandselectionEquations.

Mostvariablespresenttheexpectedsigns,bothintheselectionandinthe

wageequations.Inparticular,wagesandtheprobabilityofparticipationincreasewithageatadecreasingrate.Thehigherthelevelofeducation,thehighertheprobabilityofparticipationandthehigherthewage.Beingmarrieddecreasestheprobabilityofparticipationforwomen,asexpected.Participationislowerwiththepresenceofyoungchildreninthehouseholdandtheseeffectsaresignificant.Non-labourincomehastheexpectednegativeandsignificanteffectonparticipation.Finally, thecoefficient fortheinverseMill’sratio(lambda)ispositiveandsignificant,implyingaselectivityandthereforethattheobservedwagesarehigherthanthewageoffersofarandomsample.

UsingtheresultsobtainedfromtheHeckmanselectionmodel,grosshourlywages are imputed for non-workers. Once the information on gross hourly wagesisavailableforall individuals,weneedtocalculatethedisposableincomeforeachdiscretehouralternative.Aspreviouslymentioned,disposable income iscalculatedusing our own tax and benefit microsimulation model for the BHPS, based onEUROMODversion21A.Eleventaxandbenefitrulesaresimulated:minimumwage,nationalinsuranceemployeecontributions,contributoryjobseekersallowance,winterfuelallowance,incometax,children’staxcredit,childbenefit,workingfamiliestax


credit,incomesupport,housingbenefitandcounciltaxbenefit.Otherbenefitsarenotsimulatedbutareincludedinthecalculationofdisposableincome.3

Consider now the distribution of weekly hours of women in our sample,presented in figure 1. Important peaks are observed for inactivity and full-timework(around40hoursperweek),aswellasasmallpeakforpart-timework(around20 hours per week). Taking this into consideration, we define four discrete hourspoints,characterisinginactivity,part-timework,full-timeworkandovertimework:h={0,20,40,55} which correspond to the intervals {0−5,6−34,35−45,>45}. Thesediscretehourspointsrepresentthesetofalternativesinourbasicmodel.

Figure 1 - Distribution of Female Weekly Hours of Work

Intheextendedmodel,jobinsecurityisusedasanon-pecuniaryjobattributetobeincludedinthejobchoicebundle.TheBHPSprovidesinformationconcerningsatisfactionwithjobsecurityatwork.Jobsecuritytakesvaluesbetween1and7with1 representing that the individual is ‘not satisfied at all’with job security atworkand7,thattheindividualis‘completelysatisfied’.Asdiscussedinsection2,despitethesubjectivenatureofthisvariableself-perceivedjobinsecurityisassociatedwithobjective indicators of insecure jobs, such as temporary contracts (Campbellet al.2007;ClarkandPostel-Vinay,2009;DeloffreandRioux,2003;NäswallandDeWitte,2003). Moreover, satisfaction with job insecurity is correlated with other type ofinformationaboutself-perceivedjobinsecurity,suchasthelikelihoodofbecomingunemployed. For our extended labour supply model, we generate a job insecurityvariable takingvalues2,4and6,where2represents‘lowjob insecurity’(values5to7fromtheoriginalvariable;satisfiedwithjobsecurity),4represents‘middlejob3AdetaileddescriptionofthetaxandbenefitmicrosimulationmodelfortheBHPScanbemadeavailableonrequestfromtheauthor.

�.�1

�0

�.�0�5

�W�e�e�k�l�y� �h�o�u�r�s� �o�f� �w�o�r�k

�0 �2�0 �4�0 �6�0 �8�0 �1�0�0

195H. XAVIER JARA


insecurity’(value4fromtheoriginalvariable;neithersatisfiednordissatisfied)and6‘highjobinsecurity’(values1to3fromtheoriginalvariable;dissatisfiedwithjobsecurity).Byregroupingtheoriginalvaluesinsuchway,weexpecttocapturebetterthoseindividualsininsecurejobs(thosedissatisfiedwiththeirjobsecurity)andatthesametimethisallowsustosavecomputationaltimebyreducingthenumberofchoicealternatives.Ten discrete choice alternatives are therefore defined for the extendedmodel,representingbundlesofhoursofworkandjobinsecurity:(hours, insecurity),wherehours = {0,20,40,55} and insecurity = {2,4,6}.4Descriptive statistics of thevariablesusedinourlaboursupplymodelarepresentedintable2.

Table 2 - Descriptive Statistics of the Labour Supply Sample

mean std. dev.Netincome(£perweek) 250.41 113.75Hoursofwork(perweek) 21.86 16.93Age 37.71 10.57Noqualification 0.179 0.383CertificateofSecondaryEducation(CSE) 0.095 0.293O-levels 0.207 0.405A-levels 0.097 0.297Higherdegree 0.279 0.449Universitydegree 0.144 0.351Childrenaged0-2 0.065 0.247Childrenaged3-4 0.103 0.304Childrenaged5-11 0.333 0.472Childrenaged12-18 0.256 0.437Jobinsecurity 2.651 1.392lowjobinsecurity(2) 0.579 0.494mid.jobinsecurity(4) 0.044 0.205highjobinsecurity(6) 0.095 0.293Numberofobservations 750

Asmentionedintheprevioussection,inordertoconstructthediscretechoicealternativesinourextendedlaboursupplymodels,weneedtoconsidertherelationshipbetweenwagesandjobinsecurity.Inthedata,foreachindividualweobserveagrosshourlywageandaparticularlevelofjobinsecurity.Inordertodefinethewageraterelatedtootherlevelsofjobinsecurity,wefirstimputewagesbyrandomlyassigningfromthewagedistributionofindividualswiththesamelevelofeducationbutwhichareobservedinadifferentjobinsecuritygroup.Inthisway,foreachindividualinoursampleweobtainthreegrosshourlywageratesrelatedtolow,middleandhighjobinsecurity.Theaveragehourlywagesforeachlevelofinsecurityarepresentedinthetable3.Weobservethatonaveragewagesincreasewiththelevelofjobinsecurity,whichisinlinewiththeideaofcompensatingwagedifferentials(see,Rosen,1987).

4Theresultsarerobustfordifferentvaluesofdiscretehoursofworkandjobinsecurity.


Table 3 - Average Gross Hourly Wages by Job Insecurity Level (£ /hour)

mean std. dev.Lowjobinsecurity 8.1525 5.5656Middlejobinsecurity 8.1737 5.1319Highjobinsecurity 8.2691 4.8355

Thegrosshourlywagesrelatedtodifferentlevelsofjobinsecurityarethen

constructedwithrespecttotheaveragegrosshourlywagesofthewholepopulation.Considerfor instance that the jobof individual i ischaracterisedbyagrosshourlywagewiandalowlevelofjobinsecurity.Thegrosshourlywageofindividualiundermiddlejobinsecuritywillthenbegivenbywi +wi (8.1737−8.15250)/8.1525andunderhighinsecuritybywi +wi (8.2691−8.15250)/8.1525.

Table 4 provides summary statistics for each discrete hours alternativeof our sample of interest.As previouslymentioned, the twomain groups are full-time employment and inactivity. Average age is slightly higher in the part-timeand overtimework groups. The inactivity group presents the lowest percentage ofindividualswithhighereducation.Inparticular,only24.26percentoftheinactiveshavehighereducation,comparedto68.75percent,inovertimework.Finally,intermsofjobinsecurity,thepercentageofwomendissatisfiedwiththeirjobsecuritysituationisthehighestforovertimejobs.

Table 4 - Discrete Employment Statistics

Hours Share Higher Net income Job Insecurityper week (%) Age Education (%) per week (% dissatisfied)0 31.33 36.49 24.26 203.4 -20 27.87 38.92 39.71 221.66 11.9640 36.53 37.67 56.57 304.3 13.8755 4.27 39.13 68.75 322.08 18.75

5. Estimation of the Labour Supply Models This section presents the results of the structural labour supply estimation. Twomodelsareestimatedandcompared.Thefirstmodel is theconditional logitmodeltraditionallyusedinthediscretechoiceliteraturetoestimatelaboursupply,inwhichonlydiscretehoursalternativesdefinethechoiceset.Thesecondmodelconsistsofanextensiontothebasicconditionallogitmodel,wherethechoicesetisnowdefinedbycombinationsofhoursofworkandjobinsecurityandwherejobinsecurityissettozerofortheinactives.5Inbothmodels,age,highereducationdummiesanddummiesforchildrenofdifferentagesareusedasregressorstoaccountforobservedheterogeneityinpreferences.Table5presentstheestimatedparametersforthesemodels.

5Anestedlogitmodelwasalsoestimated,inwhichalternativesaregroupedintotwonestsrepresentinginactivityandparticipation.Theinactivitynestcontainsasinglealternativewhiletheparticipationnest is formedbyalternativescharacterisedbybundlesofhoursofworkand job insecurity.Theresultsobtainedwiththenestedlogitmodelaresimilartothosepresentedinthepaper.

197H. XAVIER JARA


Table 5 - Estimated Parameters of the Structural Model

Conditional Logit: Conditional Logit: Hours InsecurityVariable Coef. St. Error Coef. St. ErrorIncome² -5.46 (4.048) -6.094 (4.243)Income 12.244*** (2.121) 9.992*** (2.17)Hours² -0.146*** (0.186) -0.548*** (0.407)Insecurity² - - 0.317*** (0.031)Income2hours -0.941** (0.465) -0.428 (0.509)Income2insecurity - - 0.062 (0.283)Hours2insecurity - - 0.003 (0.032)Hours 1.013*** (0.149) 3.941*** (0.273) 2age -0.087*** (0.025) -0.133*** (0.031) 2highedu. 0.262*** (0.054) 0.249*** (0.064) 2child0-2 -0.783*** (0.146) -0.681*** (0.208) 2child3-4 -0.544*** (0.103) -0.567*** (0.134) 2child5-11 -0.326*** (0.059) -0.426*** (0.073) 2child12-18 -0.100* (0.058) -0.091 (0.071)Insecurity - - -3.309*** (0.258) 2age - - 0.067** (0.03) 2highedu. - - -0.008 (0.054) 2child0-2 - - -0.062 (0.203) 2child3-4 - - 0.078 (0.122) 2child5-11 - - 0.126* (0.068) 2child12-18 - - -0.05 (0.069PseudoR-squared 0.1308 0.1772

Notes:Themainvariableshavebeenrescaledasfollows:income/1000;hours/10;age/10.Standarderrorsinparentheses.*p<0.05,**p<0.01,***p<0.001

Ingeneralourresultsareinlinewithourexpectationsandcoefficientsacross

themodelsestimatedgointhesamedirection.Forbothmodels,marginalutilityofincomeispositiveforover99percentofobservationsandasthecoefficientofincomesquareisnegative,concavityinincomefortheutilityfunctionisrespected.Marginalutilityofhoursofworkisnegativeforaround78percentofobservationsinthebasiclaboursupplymodel,whileundertheextendedmodelthisholdsforaslightlylowerpercentageofcases,ofaround70percent.Turningtojobinsecurity,marginalutilityisnegativeforaround90percentofobservationsinourextendedmodel.Moreover,theestimatedeffectsofjobinsecurityaresignificant,confirmingtheimportanceofaccountingfornon-pecuniaryjobattributesonlaboursupply.

Inlinewithourexpectations,singlewomenwithyoungchildrenhavelowerpreferences forworkand these effects are significant inbothmodels estimated. Inparticular,thestrongestnegativeeffectisforwomenwithveryyoungchildrenagedunderthreeyearsold.Similarly,forboththebasicandextendedmodel,individualswithhighereducationhavehigherpreferencesforworkandtheinteractionbetweenhours of work and age presents a negative and significant coefficient. Turning tojob insecurity, we remark that only the interactionwith age presents a significant


coefficient, with preferences for job insecurity decreasing with age. Women withchildrenagedlessthanthreeyearsoldpresentlowerpreferencesforjobinsecurity,asexpected,however,theeffectisnotsignificant.

Theabilityofourmodelstofitthedatacanbetestedbycomparingpredictedandobservedfrequencies.Predictedfrequenciesareobtainedbyaveragingupindividualprobabilitiesforeachdiscretehoursalternativeoverthewholesample,whileobservedfrequenciesaresimplythefrequenciesofeachobservedchoiceoverthewholesample.Table6showsthat thebasicconditional logit,whereonlyhoursofworkdefinethechoice set, performs poorly in terms of fitting the data. Full-time is considerablyunderestimatedandpart-timeisstronglyoverestimated.Theseresultsareinlinewiththe literature,wheresuchproblemshavebeen treatedmainlybyaddingalternativespecific dummies (Van Soest, 1995) or fixed costs ofwork (Blundell et al. 2000).Hereweconsiderwhethertheadditionalinformationintermsofjobcharacteristicsservestoimprovethepredictivepowerofthemodel.Thisseemstobethecase,astheintroductionofjobinsecurityintothemodelimprovesconsiderablythefitofthedata.Additionally,asshownintable5,thepseudo-R-squaredorLikelihoodRatioIndexofMcFadden(1974)ishigherforourextendedlaboursupplymodel,confirmingthatitprovidesabetterfitthanthebasicmodel.

Table 6 - Observed vs. Predicted Frequencies

Predicted Conditional logit: Conditional logit:Alternatives Observed Hours InsecurityInactivity 31.33 28.19 32.00Part-time 27.87 37.19 25.77Full-time 36.53 25.97 38.93Overtime 4.27 8.66 3.30Inactivity 31.33 - 32.00Part-time lowinsecurity 22.93 - 20.11 mid.insecurity 1.60 - 2.28 highinsecurity 3.33 - 3.38Full-time lowinsecurity 29.2 - 30.76 mid.insecurity 2.27 - 3.36 highinsecurity 5.07 - 4.81Overtime lowinsecurity 3.33 - 2.64 mid.insecurity 0.13 - 0.28 highinsecurity 0.80 - 0.38

Theintroductionofjobinsecurityasanadditionaljobattributeinthelabour

supplymodelraisestheissueofomittedvariablesbias.Asshownintheresults,jobinsecurityhasasignificanteffectand improves thefitof themodel.Other relevantjob attributes couldpotentially influence labour supplydecisions and leaving themoutcouldbiastheparameterestimates.Itisclearthattheextensionofthemodelis

199H. XAVIER JARA


limited to theavailabilityof information in thedata. Inourcase, thechoiceof jobinsecuritywasnotonlyrelatedtothis limitationbutalsototheliteraturereportingtheimportanceofjobsecurityasaworkdomain(Clark,2001and2010).Therefore,despiteincorporatingonlyanadditionaldimensionofworkintothemodel,weconsiderjobinsecurityisoneoftheimportantdimensionstoaccountfor.Inthesameline,intermsofmodelspecification,theintroductionofadditionaljobattributesinthelaboursupplymodelraisesthequestionoftherelevanceoftheassumptionofindependenceofirrelevantalternatives(IIA),underlyingtheconditionallogitmodel.InordertotesttherestrictionimposedbytheIIAassumption,weestimatedanestedlogitmodelforbothourbasic andextended labour supplymodels.Alternativesweregrouped intotwonestsrepresentinginactivityandparticipation.ThenestedlogitmodelpartiallyrelaxestheIIAassumptionbyallowingcorrelationbetweenthechoicesinsideeachnest(see,Train,2003).Theresultsofthenestedlogitmodelsaresimilartothoseoftheconditionallogitmodelspresentedinthepaper.6

6. Labour Supply Elasticities and Responses to Changes in Job Insecurity Theparameter estimates obtained in the previous section canbeused to calculatelaboursupplyelasticitiesandtoanalysetheeffectsofpolicyreformsonparticipationandlaboursupply.Theaimofthissectionistwofold.First,wageelasticitiesobtainedwithourmodelsarecompared.Then,usingourextendedmodel,weanalysetheeffectofachangeinjobinsecurityonlaboursupply.

Laboursupplyelasticitiesindiscretechoicemodelsarecalculatednumericallyusingtheestimatedparametersoftheutilityfunction(see,CreedyandKalb,2005).First,weincreasegrosshourlywagesbyonepercentandcomputethenewdisposableincomeforeachalternativeusingourtaxandbenefitmicrosimulationmodel.Then,withtheparametersfromtheutilityfunction,obtainedinthepreviousestimation,wecalculatetheaveragesimulatedprobabilityofbeingateachalternativeforboththeoldandthenewvalueofdisposableincome.Theseprobabilitiesarethenusedtocomputetheexpectedvalueoflaboursupplybeforeandafterthewageincrease,following:

E[h|y,x]=∑Pij hj

Finally,laboursupplyelasticitiesarecomputednumericallybydividingthe

percentagechangeinexpectedlaboursupplybythepercentagechangeinwages,oneper cent in this case.Table7 shows the elasticities derived fromour laboursupplymodels.

6Resultsarenotreportedherebutcanbemadeavailableuponrequestfromtheauthor.

J

j = 1


Table 7 - Labour Supply Elasticities

Conditional logit: Conditional Logit: Hours InsecurityTotal 0.126 0.153*Lowjobinsecurity - 0.141*Middlejobinsecurity - 0.158*Highjobinsecurity - 0.158*

*significantlydifferentthantheelasticityofthebasicconditionallogit

Thecalculatedlaboursupplyelasticitiesarequiteinlinewithpreviousstudies.

Infact,elasticitiesforsinglefemalesareingeneralbetween0and0.7(Bargainet al.2012).Ourbasicconditionallogitmodelprovidesalaboursupplyelasticityof0.126,while our extended labour supplymodel provides a higher elasticity of 0.153. Thestatisticalsignificanceofthedifferenceinelasticitiesisconfirmedusingbootstrappingtechniqueswith1,000repetitions.Inourextendedconditionallogitmodel,afurtherdistinctioncanbemadebycalculatingelasticitiesfordifferentlevelsofjobinsecurity.Labour supply elasticities are lower under low job insecurity compared tomiddleand high insecurity. This result suggests that wage increases would have weakereffectsforwomenfacinggoodconditionsatwork,intermsofjobsecurity.However,bootstrappingtechniquesshownosignificantdifferencesbetweenelasticitiesforlowjobinsecurityandhighjobinsecurity.Thismightberelatedtothesmallsampleofindividualsinhighinsecurityjobs.

Inadditiontothecalculationoflaboursupplyelasticities,ourextendedmodelallowsustoanalysetheeffectofjobinsecurityonlaboursupply.However,becauseofthequalitativenatureofourjobinsecurityvariable,thesamemethodologyusedtocalculatewageelasticitiescannotbeapplied.Herewesimulatetheeffectofadecreaseofjobinsecuritybyobservingthechangeinpredictedprobabilitiescalculatedbyourmodel.Wedecreaselevelsofjobinsecuritybytheequivalentofonestandarddeviationofjobinsecurity(std.dev.1.392)forindividualswithmiddleandhighinsecurity.Table8presentsthepredictedprobabilitiescalculatedwithourmodelbeforeandafterthedecreaseinjobinsecurity.

Table 8 - The Effect of a Decrease in Job Insecurity

Predicted Probabilities Conditional Logit: Insecurity Before AfterInactivity 32.00 29.52Part-time 25.77 27.02Full-time 38.93 40.09Overtime 3.30 3.37

201H. XAVIER JARA


Our results show that a decrease in job insecurity has a positive effect onparticipation.Infact,theprobabilityofinactivitydecreasesbyaround2.5percentagepoints. Allworking alternatives present an increase, themost important being forpart-timeandfull-timeemployment.Thisresultisparticularlyinterestingintermsofpolicybecauseobjectivesaimedatprovidingincentivesforparticipationcouldalsobeachievedthroughthechannelofimprovingnon-pecuniaryjobattributes,andnotonlythroughmonetaryincentives.Inordertohaveanideaofthemagnitudethatthedecrease in inactivity represents,wecalculated the increase inoverallgrosswagesnecessarytoobtainanequivalentdecreaseininactivity.Anincreaseinoverallgrosswagesofaround20percentwouldbeneededinordertoobtainasimilardecreaseintheprobabilityofinactivity.

Theseresultsprovideaninterestinginsightintotheeffectofnon-pecuniaryjob attributes on labour supply, however, it is important to remark that this laboursupplymodeldoesnot take into account the reactionoffirms topolicies aimedatimprovingworkingconditions.Infact,fromthedemandside,providingbetterworkingconditionsmight represent additional costswhichcouldbe linked toadecrease inwages.Thiswouldresultinanegativeeffectoflaboursupplyandthereforethetotaleffectwouldbeambiguous.Theincorporationoflabourdemandwithinoursettingrepresents,inthissense,animportantstepforfutureresearch.

7. Conclusion Theaimofthispaperwastoprovideaninsightintotheeffectofnon-pecuniaryjobattributes on labour supply. Two models were estimated and compared. First, weestimatedaconditionallogitmodelwherethechoicesetisdefinedonlyintermsofdiscretehoursalternatives.Thisistheapproachmostwidelyusedtoestimatediscretechoicelaboursupply.Then,weproposedanextensiontothemodelinwhichthechoicesetischaracterisedbybundlesofincome,hoursofworkandjobcharacteristics;jobinsecurityinourcase.Theestimationofthesestructurallaboursupplymodelswasdoneusingmaximumlikelihood.

Differentobservationscanbedrawnfromourresults.Firstofall,asexpected,jobinsecurityhasanegativeeffectonindividuals’utility,withacalculatedmarginalutility which is negative for around 90 per cent of the observations. Second, thepredictivepowerofthemodelimprovesconsiderablywhenjobinsecurityisincludedas an attribute of choice, compared to the basic conditional logit model withoutalternativespecificdummiesorfixedcostsofwork.Third,laboursupplyelasticitiescalculated with the extended labour supply model are higher than those of thetraditional conditional logitmodel and these differences are significant.Moreover,wageelasticitiesforwomenworkingunderlowjobinsecurityarelowerthanthoseoffemalesinmiddleandhighinsecurityjobsimplyingthatindividualsworkingundergoodjobsecurityconditionswouldrespondlesstowagechanges.Finally,ourresultscontributetotheliteratureontheeffectsofjobinsecurity,showingthatadecreaseofjobinsecuritydecreasestheprobabilityofinactivitybyaround2.5percentagepoints.Thisresultisparticularlyinterestingasitsuggeststhatpoliciesaimedatimprovingworkingconditionscouldbeusedtocreateincentivesforlabourmarketparticipation.


An importantaspectbehindour results is theuseof self-reportedworkingconditions,inourcaseself-perceivedjobinsecurity.Thisisthetypeofinformationavailableinmosthouseholdsurveys.Whileperceivedjobinsecurityhasbeenshowntobeassociatedwithobjectiveindicatorsofinsecurejobs,itisimportanttokeepinmindthatasubjectivecomponentcharacterisessuchmeasures.Inparticular,itcouldbethecase,forsomeindividuals,thatanimprovementinobjectivejobsecurityfactorswouldnotbereflectedinabettersubjectiveperceptionofjobsecurity.Forthisreason,thefindings obtained in this study shouldbe contrastedwith future analysis usingobjectivemeasuresofworkingconditions.Inthecaseofjobinsecurity,forinstance,anindicatorofobjectiveinsecuritycouldbeconstructedbasedoninformationconcerningtemporaryemployment,sectorofwork,presenceofjobsecurityguaranteesatthefirmlevel(availablefromlinkedemployer-employeedata),amongothers.

To conclude, we believe that the incorporation of working conditions inthe analysis of labour supply offers potential opportunities for future research. Inparticular, multiple factors characterise jobs therefore a better understanding ofthemain attributes influencing labourmarket participation could be useful from apolicy perspective.Moreover, it should be possible to consider towhich extent (ifany) incorporating jobattributes inbehaviouralmicrosimulationmodelsaffectsex-anteevaluationsofpolicyreforms.Finally,fromamethodologicalpointofview,theincorporationofadditionaljobattributeshighlightstheimportanceofaccountingforpossiblecorrelationsbetweenwagesandjobcharacteristicsaspartoftheanalysisoflaboursupply.

References Aaberge, R. and Colombino, U. (2013), ‘Using a Microeconometric Model of

HouseholdLabourSupplytoDesignOptimalIncomeTaxes’,Scandinavian Journal of Economics,115(2),255-594.

Aaberge,R.,Dagsvik,J.K.andStrøm,S.(1995),‘LaborSupplyResponsesandWelfareEffectsofTaxReforms’,Scandinavian Journal of Economics,97(4),635-59.

Bargain,O.,Orsini,K.andPeichl,A.(2014),‘ComparingLaborSupplyElasticitiesinEuropeandtheUnitedStates:NewResults’,Journal of Human Resources,49(3),723-838.

Blundell, R., Duncan,A.,McCrae, J. andMeghir, C. (2000), ‘The LabourMarketImpactoftheWorkingFamilies’TaxCredit’,Fiscal Studies,21(1),75-103.

Brewer,M.,Duncan,A.andShephard,A.(2007),‘DidWorkingFamilies’TaxCreditWork?TheImpactofIn-WorkSupportonLabourSupplyinGreatBritain’,Labour Economics,13(3),699-720.

Burgard, S., Brand, J. andHouse, J. (2009), ‘Perceived Job Insecurity andWorkerHealthintheUnitedStates’,Social Science & Medicine,69,777-785.

Campbell,D.,Carruth,A.,Dickerson,A.andGreen,F. (2007), ‘JobInsecurityandWages’,Economic Journal,117(518),544-566.

Chirumbolo, A. and Areni, A. (2005), ‘The Influence of Job Insecurity on JobPerformance and Absenteeism: The Moderating Role of Work Attitudes’,Journal of Industrial Psychology,31(4),65-71.

Clark,A.(2001),‘WhatReallyMatters inaJob?HedonicMeasurementusingQuitData’,Labour Economics,8(2),223-242.

203H. XAVIER JARA


Clark,A.(2010),‘Work,JobsandWell-BeingAcrosstheMillennium’,inDiener,E.,Halliwell,J.andKahneman,D.(ed.),International Differences in Well-being.OxfordUniversityPress.

Clark, A. and Postel-Vinay, F. (2009), ‘Job Security and Job Protection’, Oxford Economic Papers,61(2),207-239.

Creedy,J.andKalb,G.(2005),‘DiscreteHoursLabourSupplyModelling:Specification,EstimationandSimulation’,Journal of Economic Surveys,19(5),697-734.

Dagsvik, J.K. (1994), ‘Discrete andContinuousChoice,Max-StableProcesses, andIndependencefromIrrelevantAttributes’,Econometrica,62(5),1179-1205.

Dagsvik, J.K. and Strøm, S. (1992), ‘Labor SupplywithNon-ConvexBudget Sets,Hours Restrictions and Non-Pecuniary Job Attributes’, Technical Report,StatisticsNorway.

Dagsvik,J.K.andStrøm,S.(2006),‘SectoralLaborSupply,ChoiceRestrictionsandFunctionalForm’,Journal of Applied Econometrics,21(6),803-826.

DeWitte,H. (1999), ‘Job Insecurity and PsychologicalWell-Being:Review of theLiteratureandExplorationofSomeUnresolvedIssues’,European Journal of Work and Organizational Psychology,8(2),155-177.

De Witte, H. (2005), ‘Job Insecurity: Review of the International Literature onDefinitions, Prevalence, Antecedents and Consequences’, SA Journal of Industrial Psychology,31(4),1-6.

Dekker, S.W. and Schaufeli, W.B. (1995), ‘The Effects of Job Insecurity onPsychologicalHealthAandWithdrawal:ALongitudinalStudy’,Australian Psychologist,30(1),57-63.

Deloffre,A.andRioux,L.(2003),‘DoWorkersCorrectlyAssessTheirJobSecurity?AEuropeanComparison’,TechnicalReport,CERCandCREST-INSEE.

Ferrie,J.E.(1998),‘AnUncertainFuture:TheHealthEffectsofThreatstoEmploymentSecurity in White-Collar Men and Women’, American Journal of Public Health,88,1030-1036.

Ferrie,J.E.,Shipley,M.J.,Marmot,M.G.,Stansfeld,S.andSmith,G.D.(1995),‘HealthEffects ofAnticipationof JobChange andNon-Employment:LongitudinalDatafromtheWhitehallIIStudy’,BMJ,311(7015),1264-1269.

Green,C.P. andLeeves,G.D. (2013), ‘JobSecurity,FinancialSecurity andWorkerWell-Being:NewEvidenceontheEffectsofFlexibleEmployment’,Scottish Journal of Political Economy,60(2),12-138.

Greenhalgh,L.andRosenblatt,Z.(1984),‘JobInsecurity:TowardConceptualClarity’,The Academy of Management Review,9(3),438-448.

Haan, P. (2010), ‘A Multi-state Model of State Dependence in Labor Supply:Intertemporal Labor Supply Effects of a Shift from Joint to IndividualTaxation’,Labour Economics,17(2),323-335.

Hartley,A.andLord,J. (1998), ‘OrganizationalCommitmentandJobInsecurity ina Changing Public Service Organization’,European Journal of Work and Organizational Psychology,7(3),341-354.

Hellgren,J.,Sverke,M.andIsaksson,K.(1999),‘ATwo-dimensionalApproachtoJobInsecurity:ConsequencesforEmployeeAttitudesandWell-Being’,European Journal of Work and Organizational Psychology,8(2),179-195.


Keane,M.andMoffitt,R.(1998),‘AStructuralModelofMultipleWelfareProgramParticipation and Labor Supply’, International Economic Review, 39(3),553-589.

Kunze,L.andSuppa,N.(2013),‘JobCharacteristicsandLabourSupply’,RuhrEconomicPapers0418,Rheinisch-WestfälischesInstitutfürWirtschaftsforschung,Ruhr-UniversitätBochum,UniversitätDortmund,UniversitätDuisburg-Essen.

McFadden,D.(1974),‘ConditionalLogitAnalysisofQualitativeChoiceBehaviour’,inZarembka,P.(ed.),Frontiers in Econometrics,105-142.AcademicPress.

Nelson, A., Cooper, C.L. and Jackson, P.R. (1995), ‘Uncertainty Amidst Change:TheImpactofPrivatizationonEmployeeJobSatisfactionandWell-Being’,Journal of Occupational and Organizational Psychology,68(1),57-71.

Näswall,K.andDeWitte,H.(2003),‘WhoFeelsInsecureinEurope?PredictingJobInsecurityfromBackgroundVariables’,Economic and Industrial Democracy,24(2),189-215.

Probst, T.M. (2002), ‘The Impact of Job Insecurity on EmployeeWork Attitudes,JobAdaptation,andOrganizationalWithdrawalBehaviours’,inBrett,J.andDrasgow,F. (ed.),The Psychology of Work: Theoretically Based Empirical Research,141-168.LawrenceErlbaum.

Rosen, S. (1987), ‘The theory of equalizing differences’, inAshenfelter,O. andR.Layard(ed.),HandbookofLaborEconomics,Volume1,Chapter12,641-692.Elsevier.

Rosenblatt,Z.,Talmud,I.andRuvio,A.(1999),‘AGender-BasedFrameworkofTheExperienceof Job Insecurity and itsEffectsonWorkAttitudes’,European Journal of Work and Organizational Psychology,8(2),197-217.

Sutherland,H.andGutierrez,R.(2004),‘EUROMODCountryReport–UnitedKingdom(2001Tax-BenefitSystem)’,TechnicalReport,UniversityofCambridge.

Train,K.(1998), ‘RecreationDemandModelswithTasteDifferencesOverPeople’,Land Economics,74(2),230-239.

Train,K. (2003),Discrete Choice Methods with Simulation.CambridgeUniversityPress.

VanSoest,A.(1995),‘StructuralModelsofFamilyLaborSupply:ADiscreteChoiceApproach’,The Journal of Human Resources,30(1),63-88.

the effect of job insecurity on labour supply · 2017. 2. 15. · evaluate their level of job...

Documents