亞開行-勞動力市場中非工資福利偏好的性別差異：來自在線自由職業(yè)平臺的實驗證據(jù)（英）

ADBIWorkingPaperSeries

GENDERDIFFERENCESIN

PREFERENCESFORNON-PECUNIARY

BENEFITSINTHELABORMARKET:

EXPERIMENTALEVIDENCEFROMAN

ONLINEFREELANCINGPLATFORM

RakeshBanerjee,TusharBharati,

AdnanM.S.Fakir,YiweiQian,

NaveenSunder

No.1376

May2023

AsianDevelopmentBankInstitute

RakeshBanerjeeisalecturerattheUniversityofExeterBusinessSchool,Devon,UnitedKingdom(UK).TusharBharatiisseniorlecturerattheUniversityofWesternAustraliaBusinessSchool,WesternAustralia.AdnanM.S.FakirisalecturerattheUniversityofSussexBusinessSchool,Brighton,UK.YiweiQianisanassociateprofessorattheSouthwesternUniversityofFinanceandEconomics,Chengdu,Sichuan,People'sRepublicofChina.NaveenSunderisanassistantprofessoratBentleyUniversity,Waltham,Massachusetts.

TheviewsexpressedinthispaperaretheviewsoftheauthoranddonotnecessarilyreflecttheviewsorpoliciesofADBI,ADB,itsBoardofDirectors,orthegovernmentstheyrepresent.ADBIdoesnotguaranteetheaccuracyofthedataincludedinthispaperandacceptsnoresponsibilityforanyconsequencesoftheiruse.TerminologyusedmaynotnecessarilybeconsistentwithADBofficialterms.

Workingpapersaresubjecttoformalrevisionandcorrectionbeforetheyarefinalizedandconsideredpublished.

TheWorkingPaperseriesisacontinuationoftheformerlynamedDiscussionPaperseries;thenumberingofthepaperscontinuedwithoutinterruptionorchange.ADBI’sworkingpapersreflectinitialideasonatopicandarepostedonlinefordiscussion.Someworkingpapersmay

developintootherformsofpublication.

Suggestedcitation:

Banerjee,R.,T.Bharati,A.M.S.Fakir,Y.Qian,andN.Sunder.2023.GenderDifferencesinPreferencesforNon-PecuniaryBenefitsintheLaborMarket:ExperimentalEvidencefromanOnlineFreelancingPlatform.ADBIWorkingPaper1376.Tokyo:AsianDevelopmentBankInstitute.Available:

/10.56506/RAPT9219

Pleasecontacttheauthorsforinformationaboutthispaper.

Emails:tushar.bharati@.au

WeacknowledgefinancialsupportfromAsianDevelopmentBankInstituteGrantR2135-1andaBentleyUniversityFACgrant.YuwenMeng,AnanyaVhavle,andTejaswiniVondivilluprovidedexceptionalresearchassistance.ThepaperhasbenefitedfromvaluablefeedbackfromSimonChang,DanielSuryadarma,andseminarparticipantsattheADBWorkshopon“TheSocialandEconomicImpactofOnlineCommerceonWomen,”theAustralianDevelopmentEconomicsWorkshop,theAustralianGenderEconomicsWorkshop,UniversityofAdelaide,UniversityofSouthernCalifornia,andUniversityofWesternAustralia.Allremainingerrorsareourown.

AsianDevelopmentBankInstitute

KasumigasekiBuilding,8thFloor

3-2-5Kasumigaseki,Chiyoda-ku

Tokyo100-6008,Japan

Tel:+81-3-3593-5500

Fax:+81-3-3593-5571

URL:

E-mail:info@

?2023AsianDevelopmentBankInstitute

R.Banerjeeetal.

ADBIWorkingPaper1376

Abstract

Weconductanexperimentonamajorinternationalonlinefreelancinglabormarketplatformtostudytheimpactofgreaterflexibilityinchoosingworkhourswithinadayonfemaleparticipation.Wepostidenticaljobadvertisements(for320jobs)coveringawiderangeoftasks(80distincttasks)thatdifferonlyinflexibilityandthewageoffered.Comparingthenumbersofapplicantsforthesejobs,wefindthatthoughbothmenandwomenpreferflexibility,theelasticityofresponseforwomenistwicethatforthemen.Flexiblejobsreceive24percentmorefemaleapplicationsand12percentmoremaleapplicationscomparedtoinflexiblejobs.Ourfindingshaveimportantimplicationsforexplaininggenderdifferencesinlabormarketoutcomesandforfirmsinterestedinattractingmorewomenemployees.

Keywords:workplaceflexibility,onlinefreelancingjobs,femalelaborforceparticipationJELClassification:J22,O14,J16,L86

1

1Introduction

Dowomenvaluenon-pecuniaryjobattributesmorethanmen?Theanswertothisques-tionhasimportantimplications.First,thevaluemenandwomenattachtovariousnon-pecuniarybenefitscouldexplainpartofthegenderwagegap(

Petrongolo

2019

).Womenmayhavehighervaluationsfornon-pecuniarybenefitslikeflexibility,becauseofsocialnormsaroundwhoshouldshouldertheresponsibilityofhouseholdwork.Flexibilitymightallowwomentobalancehouseholdandwagework(

Sullivan

2019

).

1

Menandwomenmaysortintodifferentjobsbasedonnon-pecuniarybenefits,andfirmsmaymakelowerwageofferstoemployeesthatdemandexpensivenon-pecuniarybenefits(

Penneretal.

2022

).Second,thelimitedprovisionofthesenon-pecuniarybenefitsmaycausewomentostayoutofthelabormarket.Femalelaborforceparticipationinmanydevelopingcoun-tries,particularlyinAsia,remainslowevenafteraccountingforthelevelofeconomicdevelopment.Oneexplanationcouldbethatfrictionsinthelabormarketoftenleadtolimitedprovisionofthenon-pecuniarybenefitsthatwomenprefer(

Gupta

1993

;

Macpher-

sonandHirsch

1995

;

DeLeireandLevy

2004

;

GrazierandSloane

2008

;

Kleinjans

2009

;

Borker

2018

).Third,theincreaseintheuseofInternettechnology,theriseofthegigecon-omy,andthechangesbecauseofCOVID-19haveledtoanincreasedprovisionofflexibility,openingupnewdebatesaroundflexibleworkingarrangements.Akeyquestioniswhetherfirmsmustprovideflexibleworkingarrangementstoattractbetteremployeesortoretainexistingones.Theanswerdependsonhowstrongthepreferenceforflexibleworkingarrangementsis.Genderdifferencesinthesepreferenceswillhaveimplicationsforthecompositionandthediversityoftheworkforce.Thus,firmsandpolicy-makersinterestedintheoptimalresponsetothesechangesmustthereforeunderstandgenderdifferenceinpreferencesfortheseworkarrangements(

Cooketal.

2021

;

Gottliebetal.

2021

).

2

Despitethefar-reachingimplications,answeringthequestionisempiricallychalleng-

1Thatsaid,alargeliteraturehasdocumentedgenderdifferencesinvariousattributes,suchascompeti-tiveness,riskpreference,andwillingnesstonegotiate,thatarerelevanttowagedetermination(

Crosonand

Gneezy

2009

;

AzmatandPetrongolo

2014

;

ExleyandKessler

2019

).However,theexactsourcesofthesedifferencesareoftenunknown.Thesedifferencescanresultfromsocialnorms,likethedifferenceincom-petitivenessbetweenmatrilinealandpatrilinealsocieties,orhaveevolutionaryroots.Genderdifferencesinpreferenceforflexibilitycouldalsobeaproductofsocialnormsorhaveotherroots.Inthisstudy,weremainagnosticaboutthesourcesofthesedifferences.

2Somestudiesarguethatincreasedflexibilitymayincreasegenderdisparitiesbyreinforcingexistinggendernorms(

LottandChung

2016

;

Chung

2019

).Menmightuseittoworkandearnmorewhilewomenmightbeexpectedtocontributemoretohouseholdworknowthattheirworkarrangementsareflexible.

2

ing.Weonlyobservethegenderdistributionofemployeesandthefinalpackageofpe-cuniaryandnon-pecuniarybenefitsthatemployeesreceiveinequilibrium.Besidesprefer-encesfornon-pecuniarybenefits,severalobservedandunobserveddemand-andsupply-sidefactorsplayaroleindeterminingtheequilibrium.Forexample,inequilibrium,wemayobserveahigherproportionofwomeninflexible,low-payingdeskjobscomparedtolessflexible,high-payingconstructionjobs.However,thissortingcouldalsoresultfromahigherproductivityofmeninjobsthatrequirephysicalstrength.Anempiricallyob-servedassociationbetweencertainnon-pecuniarybenefitsandsharesoffemaleemployeeacrossfirms,industries,orsectorsoftheeconomy,therefore,doesnotnecessarilyimplythatwomenpreferthesebenefits.Anothercomplicationinidentifyingapreferenceforaparticularnon-pecuniarybenefitfromobservablereal-worldjobchoicesisthatjobstypi-callyvaryalongseveraldimensionsofnon-pecuniarybenefits.Jobsthatprovidegreaterworkplaceflexibilitymayalsohavelowertravelrequirementsandcouldalsobelocatedinsaferworkplaces.Itis,therefore,evenmorechallengingtoinferapreferenceforaspecificnon-pecuniarybenefitfromobservedchoices(

WiswallandZafar

2018

;

Wasserman

2019

;

Adams-Prassl

2020

;

Heetal.

2021

;

MasandPallais

2017

).

Inthispaper,weaddresstheseempiricalchallengesbyusingarandomizedauditstudythatfocusesonaspecificnon-pecuniarybenefit:theflexibilityinchoosingworkhoursdur-ingtheday.Weconductedourexperimentonamajoronlinefreelancelabormarketplat-form.Wepostedfourotherwiseidenticaljobadvertisementsforeachof80distincttasksthatvaryonlyintheirflexibilityandthewageoffered.

3

Flexiblejobs(“high-flexibility”)al-lowthefreelancertochooseanytwo-hourwindowduringthedayonapre-specifieddatetocompletethetask.Inflexiblejobs(“l(fā)ow-flexibility”)requiretheworktobecompletedwithinapre-specifiedtwo-hourperiodofourchoosingonthepre-specifieddate.Thejobsalsodifferinthewageoffered:a“high-wage”jobpostingoffersalump-sumonetimepaymentofUSD40anda“l(fā)ow-wage”jobpostingoffersalump-sumonetimepaymentofUSD30.Thus,wehave320jobpostingsfor80distincttasks.

4

Wecollectedinformationaboutthenumberofmaleandfemaleapplicantsforeachofthejobpostings,aswellasseveralapplicant-levelcharacteristics.Sincethejobpostingsforeachspecifictaskvaried

3Thesetaskscoverawiderangeofactivities,suchasproofreading,writing,andcoding.

4Thefourjobscorrespondingtoeachtaskwerepostedatthesametimeandonthesamedayoftheweek,butindifferentweeks,usingthesameuseraccount.Werandomizedthe80tasksacrossdaysoftheweekandacrossuseraccounts.Theorderinwhichwepostedthefourjobswithinataskwasalsorandom.Eachjobpostingwasopenforoneday,afterwhichwehiredoneapplicantatrandomtodothejobandpaidthepromisedwage.

3

onlyalongthedimensionofflexibilityorwageoffered,wecanattributeanydifferencebetweenmaleandfemaleapplicationresponsestoadifferenceinthevalueattachedtothesedimensionsofpecuniaryandnon-pecuniarybenefits.

Webelievethatthecontextofonlinefreelancelabormarketsisparticularlyrelevantforansweringthisquestion.First,theonlinefreelancemarketgeneratessizablelevelsofemployment.Estimatessuggestthatthereare14millionactiveonlineworkers.Asub-stantialamountoftherecentgrowthhascomefromdevelopingcountriesofSouthAsia(

Stephanyetal.

2021

).

5

Second,onlinelabormarketsarelikelytobecomemoreimportantinthenearfuture.Firmshavemadeinvestmentsinadaptingtoremoteworkingduringthepandemic.Theseinvestmentsmayhavecreatednewknowledge(possiblyinmanage-mentskills)indealingwithonlineremoteworking.Thefixednatureoftheseinvestmentsalongwiththenewknowledgeislikelytocreateincentivesforfirmstoworkinanonlineremoteenvironmentparticularlybyhiringonlinefreelancers(

Umaretal.

2021

).

6

Third,despitetherecentgrowth,theparticipationofwomeninonlinelabormarketscontinuestolagbehind.DatafromtheOnlineLaborObservatoryshowsthatonly39percentoftheworkersarefemale(

Stephanyetal.

2021

).Inaddition,therearesignificantdifferencesacrosscountriesandoccupations.IntheUS,41percentofworkersarefemaleswhileonly28percentofallonlineworkersinIndiaarefemales.

Theresultsfromtheexperimentsuggestagenderdifferenceinthepreferenceforflex-ibility.Flexiblejobsattractahighernumberofapplicationsfrombothmenandwomen.However,comparedtoinflexiblejobs,flexiblejobsleadtoa24percentriseinthenumberoffemaleapplicantsasopposedtoa12percentriseinthenumberofmaleapplicants.Thus,comparedtomen,alargerproportionofwomen(oftheworkersintheplatform)findflexibilityabindingconstraint.Flexibilityalsomaketheapplicantpoolmoregenderdiverse,leadingtoa2percentriseintheproportionoffemaleapplicants.Womenarealsomorelikelytoputeffortintogettingaflexiblejob.Comparedtoinflexiblejobs,womenaremorelikelytomakeanapplicationbeforemenandincludetheirpreviousworksampleintheapplicationforaflexiblejob.Ourresultsalsosuggeststhatthevaluationofflexibilityissufficientlyhigh:anincreaseinthewagebyonly10USDwillnotattractthesamesetofworkersthatvalueflexibility.

5Forexample,theshareofIndiaintheonlinelabormarkethasgrownfrom25percentin2017to33percentin2021(Stephanyetal

.

2021

).

6Foramoredetaileddiscussion,pleasesee

HarvardBusinessSchool

(

2020

).

4

Wecontributetotheliteratureongenderandnon-pecuniarybenefits.Alargeliteraturehashighlightedtheimportanceofnon-pecuniarybenefits,particularlyforwomen(

Goldin

andKatz

2011

;

FlabbiandMoro

2012

;

Goldin

2014

;

SullivanandTo

2014

;

Bronson

2014

;

LavettiandSchmutte

2016

;

Sorkin

2018

).However,mostpapersfacethekeychallengetoempiricallydisentangletheroleofpreferencesfromotherunobservedcharacteristicsoftheworker,firmandjoblevel

7

.Inaddition,manypapersintheliteraturefacethedatachallengeofidentifyingtheroleofaspecificnon-pecuniarybenefit.Inthispaper,weover-comethesechallengesbyusinganexperimentthatallowscausallyidentifyingtheroleofpreferencesandatthesametimewefocusonaspecificnon-pecuniarybenefit.

Themorerecentliteraturehasusedexperimentsthatelicitstatedpreferences(andwillingnesstopay)forvariousjobcharacteristics(

WiswallandZafar

2018

;

Maestasetal.

2018

;

MasandPallais

2017

),broadlyfindingthatwomenhaveahigherwillingnesstopayfornon-pecuniaryjobbenefits.

WiswallandZafar

(

2018

)usesasampleofstudentsfromatopUSuniversityandfindsthatwomenarewillingtogiveupahighersalaryforjobstabilityandjobflexibility.

Maestasetal.

(

2018

)usestheAmericanWorkingConditionsSurveyandfindsthatwomenhaveahigherpreferenceforjobswithlessphysicalworkandmorepaidleave.Thesepapersvalidatethestatedpreferencesbylookingatrealjobattributes,soastocheckthatthestatedpreferencesoftherespondentsmatchtheactualjobcharacteristics.Evenifthestatedpreferencesmatchtherealjobattributes,wedonotobservethesetofjobsfromwhichtherespondentsarechoosinginthereallabormarket.Thus,atleastpartially,theconcernremainsthatthestatedpreferencesarenotincentivecompatible.Ourexperimentaddstothisbyfocusingontherevealedpreferencesofwork-ersforflexibility.Inthis,ourpaperismostcloselytorelatedto

Heetal.

(

2021

).TheyconductafieldexperimentusingaChinesejobboardandfindthatmarriedfemaleshaveastrongerpreferenceforflexiblejobsthandomarriedmales.Weaddtothefindingsof

Heetal.

(

2021

)byfocusingontheworldwideonlinefreelancelabormarketandonappli-cationsforarangeof80distinctjobtypesthatvaryacrossseveraldimensions,includingbeingmaleorfemaledominated.

Ourpaperalsocontributestoalargeliteraturethatinvestigatesdifferencesinpref-erencesbetweenmenandwomen,particularlytheirimplicationsforthelabormarket

7Foraliteraturereviewofthistopiccoveringstudiesfromseveraldisciplines,see

ChungandVander

Lippe

(

2020

)

5

(

CrosonandGneezy

2009

;

AzmatandPetrongolo

2014

;

ExleyandKessler

2019

).Broadly,theliteraturedocuments,usingbothfieldandlabexperiments,thattherearesignificantgenderdifferencesinvariousattributes,suchasriskpreferencesandcompetitiveness,thathaveanimpactonlabormarketoutcomes.Weaddtothatliteraturebydocumentinggenderdifferenceinpreferencesforflexibilityinjobs.Lastly,ourpaperalsoaddstoarecentandgrowingmultidisciplinaryliteraturethatfocusesonvariousaspectsofthegigeconomyandtheonlinefreelancelabormarket(

StantonandThomas

2016

,

2020

;

Cook

etal.

2021

;

StantonandThomas

2021

).Ingeneral,thisliteraturenotesthatthereisonlylimiteddataaboutonlinefreelanceworkers.Weaddtothisliteraturebycollectingarichsetofdataaboutapplicantsandtheirapplications.Inaddition,wealsofocusontherolethatflexibilitymayplayinlimitingtheparticipationofwomeninonlinelabormarkets.

8

2ConceptualFramework

Webeginwithasimpleconceptualframeworktohelpinterprettheresultsfromtheex-periment.Assumethattherearentwo-hourtimeslotsduringthedayduringwhichafreelancercancompletethetaskweadvertise.Inourinflexiblejobads,wespecifythetwo-hourslotinwhichthehiredfreelancermustwork.Intheflexiblejobs,theapplicantscanchoosetoworkduringanytwo-hourwindowduringtheday.LetusdenotethesetofpossibletimeslotsbyS=(1,2,3,,n).

Workershaveanopportunitycostofworkingduringthesetimeslots.Suchanopportu-nitycostcapturesthepecuniarycostsofworking,suchasforgonewagesfromalternativeoccupations,andnon-pecuniarycosts,suchasdelaysinchildcareorotherfamilyobliga-tions.Thereisnouncertaintyaboutthepotentialrealizationoftheseopportunitycosts.Workerscanfullyandcorrectlypredicttheseopportunitycosts.Weindexworkersbyi∈I,whereIistheuniverseoffreelancersontheplatformwhoseeouradvertisement.Letus

8Acrucialaspectoftheonlinefreelancelabormarketisthatitallowsworkerstochoosejobsthatbestmatchtheirconstraintsandrequirements.Thisaffordsworkersgreaterflexibilityinchoosingtheirworkschedule.However,asignificantnumberofonlinejobscomewithstrictdeadlines.Whileworkershavetheoptiontochoosebetweenjobs,thesestrictdeadlineslimittheabilityofworkerstoallocatetheirworkflexiblywithintheday.Thislackofflexibilityinallocatingthejobwithinthedaycanbeonefactorthatlimitsfemalelaborforceparticipationintheonlinelabormarket,bothattheintensiveandextensivemargins.Moreover,ifwomenvalueflexibilityinonlinejobs,theymaybewillingtoacceptlowerwagesforgreaterflexibility.However,women’spreferenceforjobflexibilityintheonlinelabormarkethas,largely,remainedempiricallyunverified,agapthatthispaperseekstoaddress.

6

denotetheopportunitycostofworkingduringtimeslots∈Sforworkeribycis.

Forsimplicity,weassumethattheapplicationcostsarezero(orminimal)andworkersapplytoalljobsthattheywilltakeifoffered.Thisisnotanunrealisticassumptioninourcontext.Theworkersusuallyaddminordetails(likeashortcoverletter)totheirexistingprofileontheplatformtomakeanapplication.Therearealsonointerviewsforthesejobs.

9

Workeriwillapplyforaninflexiblejobofferingawagewtobedoneduringtimeslotif

w?ci>0.

However,ifthesamejobwithawagewallowstheworkertochoosetheirworkhours∈S,thenaworkeriwillapplyif

w?ci>0,

whereci=min(ci1,ci2,ci3,,cin).

Now,letusassumethatthedistributionofciacrossindividualshasaprobabilityden-sityfunctionf(ci)andacumulativedistributionfunctionF(ci).Next,assumethedistri-butionofciisgivenbytheprobabilitydensityfunctiong(ci)andacumulativedistributionfunctionG(ci).Forajobthatoffersawagewbbutnoflexibilityinchoosingworkhours,theshareofallapplicantsapplyingforthejobwillbegivenby:

G(w)=\0wg(ciˉ)dciˉ

Similarly,forflexiblejobswithawagew,theshareofallapplicantswhowillapplyforthe

jobwillbegivenby:

F(w)=\0wf(ci~)dci~.

BasedonourfindingsfromTables

2

and

3

,wehave

F(w)<G(w),Aw∈{wL,wH},

9However,therearesomelimitstothemonthlynumberofunsuccessfulapplicationsaworkercanmakeontheplatformforfree.

7

wherewL=30andwH=40inourexperiment.Thisimpliesthattheremustbeatleastoneindividualisuchthat

ci<ci≤w.

Or,F(.)first-orderstochasticallydominatesG(.).Theestimatedeffectofflexibilityin

Ta-

ble2

isproportionaltoG(w)?F(w).Inotherwords,thecoefficientof5.99isproportionaltotheshareofallapplicantsforwhomci<ci.Thehigher(lower)thenumberofappli-cantswithci<ci,thehigher(lower)willbetheestimatedeffectofflexibility.

Next,letusdifferentiatethedistributionofciandciformalesandfemales.Formales,letusdenotethecumulativedistributionfunctionsbyFM(ci)andGM(ci).Forfemales,wedenotethembyFF(ci)andGF(ci).Toconstructamappingthatwillhelpuscomparetheeffectsofflexibilityacrossthetwogenders,letusassumeFM(ci)=FF(ci).Thatis,thedistributionofminimumopportunitycostforthetwogendersisthesame.

10

Alargereffectofflexibility(inpercentageterms)onwomen,asweobserveinTables

2

and

3

,implies:

GF(w)<GM(w)<FM(w)=FF(w),Aw∈{wL,wH}.

Inotherwords,ourfindingsofahigherpercentageeffectofflexibilityonfemalesthanmalesimply

ci<ci≤w

istrueforalargershareoffemaleapplicantsthanmaleapplicants.Thismeansthattheopportunitycostofworkingduringthe8to10amslotis,onaverage,higherforfemalesthanformales.

3ExperimentalDesignandDataCollection

Weconductedourexperimentononeofthelargestonlinefreelancelabormarketplatform,whichattractsclientsandfreelancersfromaroundtheworld.Theprocessofmatchingafreelancerwithaclientstartswithaclientpostingadescriptionoftheirjobandawagethattheywillpayafreelancertocompleteit.Theclientmayinvitespecificfreelancers

10Thissimplifyingassumptionisnotentirelyimplausible.Considerascenariowhereallfemalesandmaleshaveatleastonetwo-hourwindowintheentiredaywhentheiropportunitycostofworkingontheplatformiscountingstarsduringthedaytime,whichtheyallvalueequallyand,unfortunately,minimally.

8

toapplyforthejoborpostthejobforanyfreelancerwhomaybeinterested.Candidatesapplywithacoverletter,theirproposedwage(acounteroffer),andotherdetails,suchaspastexperiencewithsimilarwork,thatmayindicatetheircompetenceandinterestinthejob.Theclientcanthenchooseoneormorefreelancerstoperformthetask.Next,theclientsendsthechosenfreelanceracontractspecifyingtheagreednumberofhours,afixedoranhourlywage,andadeadlinefortheworktobecompletedby.Atthisstage,thechosenfreelancercanacceptthecontract,renegotiatewiththeclient,orrejecttheoffer.

Ourexperimententailspostingseveraljobsonthisplatformasclientsandstudyingtheresponseswereceivefromthefreelancers.Specifically,wepostfourvariations(‘jobs’)of80distincttasks,whichcoverawiderangeofactivities.Ourjobadvertisementsresemblethejobadvertisementstypicallypostedontheplatform.Withfourvariationsforeachofthe80tasks,thee