世界銀行-按性別劃分的交通成本和女性的時間使用：來自印度Pink Slip計劃的證據（英）

PublicDisclosureAuthorizedPublicDisclosureAuthorized

PolicyResearchWorkingPaper10701

Gender-SpecificTransportationCostsandFemaleTimeUse

EvidencefromIndia’sPinkSlipProgram

YutongChen

KeremCosar

DevakiGhose

ShirishMahendru

SheetalSekhri

WORLDBANKGROUP

DevelopmentEconomics

DevelopmentResearchGroup

February2024

PolicyResearchWorkingPaper10701

Abstract

Thispaperestimatesasyntheticdifference-in-differencesspecificationontheroll-outofaprogramprovidingfreebustransitforwomeninseveralIndianstates,toexaminetheimpactonwomen’stimeallocationandlaborsupply.Householdexpendituresonbusesfallandwomensavetimeontravel.However,thereissubstantialheterogeneity.Skilledemployedwomenincreaselaborsupplyandreduce

timeonhouseholdchores.Low-skilledmarriedwomenincreasetimeonhouseholdactivitiesandreducelaborsupply.Unemployedwomenincreasejobsearchwithnoeffectonemployment.Thefindingsshowthatgenderroleswithinhouseholdsunderminetheeffectofgender-specifictravelsubsidiesonfemalelaborsupply.

ThispaperisaproductoftheDevelopmentResearchGroup,DevelopmentEconomics.ItispartofalargereffortbytheWorldBanktoprovideopenaccesstoitsresearchandmakeacontributiontodevelopmentpolicydiscussionsaroundtheworld.PolicyResearchWorkingPapersarealsopostedontheWebat

/prwp.Theauthorsmay

becontactedatdghose@.

ThePolicyResearchWorkingPaperSeriesdisseminatesthefindingsofworkinprogresstoencouragetheexchangeofideasaboutdevelopmentissues.Anobjectiveoftheseriesistogetthefindingsoutquickly,evenifthepresentationsarelessthanfullypolished.Thepaperscarrythenamesoftheauthorsandshouldbecitedaccordingly.Thefindings,interpretations,andconclusionsexpressedinthispaperareentirelythoseoftheauthors.TheydonotnecessarilyrepresenttheviewsoftheInternationalBankforReconstructionandDevelopment/WorldBankanditsaffiliatedorganizations,orthoseoftheExecutiveDirectorsoftheWorldBankorthegovernmentstheyrepresent.

ProducedbytheResearchSupportTeam

GENDER-SPECIFICTRANSPORTATIONCOSTSANDFEMALE

TIMEUSE:EVIDENCEFROMINDIA’SPINKSLIPPROGRAM*

YutongChen?KeremCo?sar?DevakiGhose§ShirishMahendru?SheetalSekhri‖

Keywords:Transport,Gender,Time-Use,FemaleLaborForceParticipation

JELcodes:J16,J22,R41

*WethanktheUVAdepartmentofeconomicsforfinancialsupport.Forhelpfulcommentsandfeedback,wethankSAnukritiandtheOfficeoftheChiefEconomist,SouthAsiaRegion.

?yc3jk@(UniversityofVirginia)

?kerem.cosar@(UniversityofVirginia,CEPR,CESifo,NBER)

§dghose@(DevelopmentEconomicsResearchGroup,TheWorldBank).TheviewsexpressedinthispaperdonotrepresenttheviewsoftheWorldBankoritspartnerorganizationsandsolelyrepresenttheauthors’personalviews.

?shirish.mahendru@giz.de(GesellschaftfürInternationaleZusammenarbeit(GIZ)India)

‖ssekhri@(UniversityofVirginia)

1

1Introduction

Womenfacesignificantcommutingbarriersnotonlyindevelopingcountriesbutalsointhedevelopedworld.Socialnormsonhouseholdchoresandtravelingaloneaffectfemalelaborsupplythroughcommutingdecisions(

FanningMadden,

1981;

TurnerandNiemeier,

1997;

LeeandMcDonald,

2003;

Abe,

2011;

McQuaidandChen,

2012).

Arecent

OECD

(2016)

reportfindsthatmeninOECDcountrieshaveanaveragecommutingtimeof33.4minutesperday,whilewomenhaveanaverageof21.9minutes,resultinginagendercommutinggapof31.1%.Thepatternsareevenstarkerindevelopingcountries.Forexample,accordingtodatafromtheNationalTime-useSurvey(2020),onaverage,Indianwomenspendonly8minutesperdayonemployment-relatedtravel,whilemenspend36minutes.

ILO

(2017)

reportsthatlackoftransportationdecreaseswomen’sprobabilityofparticipatinginthelabormarketby16.5percentagepointsamongdevelopingcountries.Inshort,thereisampleevidencethatcommutingbarriersdistortwomen’slaborsupply.Thisnaturallybegetsthe

questionofwhetherreducingcommutingbarrierscanincreasewomen’slaborsupply.

Inthispaper,wefirstexamineifwomenareresponsivetothecostoftransportationindeterminingtheirtraveldemand,orwhethergendernormsaresoentrenchedthatdemandisinelastic.Second,ifwomentravelmorefrequentlyoroptforfastermodesoftransportationduetodecreasedcommutingcosts,howdoesthisaffecttheallocationofwomen’stimeamonghouseholdchores,commuting,andlaborsupply?Tothisend,weexploittherolloutofafreebusingscheme,thePinkSlipProgram,intwostatesofIndia.Indiaisapertinentsettingtostudythisquestion.Thereisanoverwhelminggendergapincommutingtimeandmodesused.The2011CensusofIndiarevealsthat30.2%ofwomentraveltoworkonfoot,andonly24.6%useanykindoftransportation,highlightingthelimitedaccesstotransportationoptionsformanywomen.1Thereisalsoevidencethatwomenuseslower

modesoftransportationtocommutetoworkasfastermodesareusuallymoreexpensive

1Incontrast,only20%ofmentravelonfootandmorethan50%ofmenuseanykindoftransportation.

2

(AnandandTiwari,

2006

).Concurrently,femalelaborforceparticipationislowinIndiawithalaborforceparticipationrateof24%in2019,significantlylowerthantheaverageof

46%inlow-andmiddle-incomecountries(

WorldBank,2019

).2

WeleveragethePinkSlipprograms’state-wideroll-outinPunjab,andTamilNaduinAprilandMay2021,respectively.WhileDelhialsointroducedthisprograminNovember2019,datalimitationsprecludeusfromincludingitinoursample.3Reportsfromthegroundindicatethatwomen’sresponsetotheinitiativewasoverwhelminglypositive.Forexample,fromJuly2021toMarch2022,thepercentageofwomentravellingbybusinTamilNadu

increasedfrom40%to61%(Sundaram,

2022).Accordingto

Goswami

(2021),womenmade

upthemajorityofridersonDelhiTransportCorporationbusesbyMarch2021.

Toidentifythecausalimpactoftheprogramonwomen’slaboroutcomes,wecollateddatafromanumberofsources.OurmainempiricalanalysisisbasedontheConsumerPyramidsHouseholdSurvey(CPHS)datamaintainedbytheCentreforMonitoringtheIndianEcon-omy(CMIE).TherichCPHSdataisapanelofabout160,000householdsacrossallIndianmajorstatesafter2014.Itincludescomprehensiveinformationnotonlyonhouseholdex-pendituresandmembers’demographiccharacteristicsandemploymentstatusbutalsoontheirtimeusepatternsandallocationoftimeonvariousactivities.Webolsterthefindings

withalargeprimarysurveyofwomenconductedinDelhi.

Ouridentificationapproachcompareswomenintreatedstates(i.e.,PunjabandTamilNadu)tothoseintheirgeographicalneighborswhichwewillrefertoascontrolstateshenceforth.Theimplementationofthepolicyin2021providestemporalvariation.Toaddressconcernsaboutendogeneity,weimplementasyntheticdifferences-in-differencesstrategy(SDID)pro-

posedby

Arkhangelskyetal.

(2021)atthestatelevel.Thisapproachcombinesthesynthetic

2Thefemalelaborforceparticipationratehasincreasedto37%accordingtothemostrecentfemalelaborforcesurveybytheMinistryofStatisticsandProgramImplementation(Source:

/mu2vx7m3)

3TheCMIE,ourdatasourceinthispaper,startedtocollectinformationonindividualtimeusageafterDelhistartedtoimplementthePinkSlipscheme.So,weareunabletoincludeDelhiinoursampledirectly.

3

controlmethod(

AbadieandGardeazabal,

2003

)withthedifference-in-differencesstrategytotakeadvantageofthebenefitsofboth.Eventstudymodelsrevealparallelpre-trends.TheweightsfromtheSDIDapproachareusedinallsurveydata-basedanalysesattheindi-viduallevel.WefindevidenceconsistentwithelasticdemandfortransportationforwomenusingtheCPHSdata.Overallexpendituresontravel,specificallyonpublicbuses,trams,andferries,werereducedforhouseholdswithwomenintreatedstatescomparedtocontrolstates.FindingsfromtheDelhisurveyalsocorroboratetheseresultsattheindividuallevel.Newfemaleusersofbusesreportnegligibletransportationcostsafterthepolicychangeas

opposedtosubstantialtransportcostspriortoitusingothermodes.

Wefindheterogeneouseffectsofthepolicychangeonemployedandunemployedwomenwhichareinoppositedirections.Consequently,thesecanceleachotherandthepolicyap-pearstohaveanulloveralleffect.4Wefocusontimespentonhouseholdchores,traveling,andlaborsupply.Skilledemployedwomenusethetimesavedfromcommutingtoincreasehoursoflaborsuppliedplausiblyduetothesubstitutionfromslowermodesoftransporta-tiontofasterfreebuses.Aninterestingfindingofouranalysisisthatprovidingfreebusestowomenhasthemostconsiderableeffectonthetime-usepatternsofunemployedwomen,especiallythosewhoarenotmarried.Thesewomenspendmoretimeoutsidethehouse,andmoretimetraveling,partiallyspendingthistimesearchingforjobsmoreintensivelyandfartherawayfromhome.However,wedonotfindanincreaseinthelikelihoodofthemfind-ingemploymentuptofourmonthsafterthescheme,indicatingthatwomenfaceadditional

hurdlestofindingemploymentintheshortrun.

Insharpcontrasttoskilledemployedwomen,low-skilledemployed,especiallymarriedwomen,spendthetimesavedfromcommutingtosubstituteforhouseholdchores.Infact,married

womenwithlowskillsreducedtheirhoursoflaborsupplied.Thisreductionisconsistent

4Weshowevidencethattheprogramdoesnotleadtochangesintheemploymentormaritalstatusofwomen,asaresultofwhichtheemploymentandmaritalstatusarepre-determinedrelativetothepolicychange.Thus,heterogeneityalongthesemarginsisnotconfoundedbychangesinthemarginsthemselves.

4

withourfindingsofintrahouseholdsubstitutionintimeuseanditsallocationtoactivities.Whilelow-skilledemployedmarriedwomenincreasedthetimespentonhouseholdchoresandreducedtheirlaborsupply,marriedmen’sbehaviorchangedintheoppositedirection:employedmarriedmenreducedtheirtimespentonhouseholdchoresandincreasedtheirworkhours.Unemployedmarriedmenincreasedthetimespentsearchingforjobs.Thisreducedlaborsupplyfromlow-skilledmarriedwomencouldbedrivenbyashiftfrommentowomenofhouseholdchoresthatrequireanindivisible,discreteamountoftime.Forexam-ple,alongcommutebymothersmayhavepreviouslymadeitoptimalforthefathertotaketheirkidstoschool,perhapsonabus.Aftertheroll-outoffreebusridesforwomen,thisresponsibilitycouldbereallocatedtothemotherwhocouldavailofapotentiallyfastermodeoftransport(bus)forfree,allowingthehusband’sworkhourstoincreaseattheexpenseofthemother’stime.Thegenderwagegapincreasesthelikelihoodofsuchre-optimizationatthehouseholdlevel.Insum,whilethefreebusprovisionbenefitedskilledandunmarriedwomenbyimprovingtheirlabormarketoutcomes,low-skilledmarriedwomenrespondedby

reducingtheirworkhoursanddoingmorehouseholdchores.

Ourpapercomplementsagrowingbodyofworkstudyingtheeffectsofbarrierstowomen’smobilityandaccesstopublictransportsystemsonfemalelaborforceparticipation(

Field

andVyborny,

2022;

Martinezetal.,

2020;

Alametal.,

2021;

Leietal.,

2019;

ILO,

2017;

PetrongoloandRonchi,

2020).

Farréetal.

(2022)showthata10-minuteincreaseincom

-mutingdecreasesthelikelihoodofmarriedwomenparticipatinginthelabormarketby4.6percentagepoints.Usingajobsearchmodelwherecommutematters,

LeBarbanchonet

al.

(2021)estimatedthatapproximately10%ofthewagegapbetweenmenandwomenin

re-employmentinFrancecouldbeattributedtodifferencesinthewillingnesstocommutebetweengenders.

Blacketal.

(2014)showthatmetropolitanareas’commutingtimesare

oneexplanationforthelargevariationacrossUScitiesinmarriedwomen’slaborforcepar-

ticipation.Inacloselyrelatedpaper,

FieldandVyborny

(2022)showthatwomen-onlybuses

5

increasefemalejobsearchinPakistan.5Weextendthisliteraturebyhighlightingthatde-mandfortransportationiselasticforwomenbutthereisheterogeneitybyskillandmaritalstatus.Freebusesdoincentivizeskilled,unmarriedandemployedtoincreasetheirlaborsupply,andunemployedwomentosearchmoreintensivelyforjobs.Butintrahouseholdgendernormsseemtoprecludeunskilledemployedmarriedwomenfromdirectlyimprovinglabormarketoutcomes.Surprisingly,low-skilledmarriedwomenusethetimesavedfromcommutingusingfreebusestodomorehouseholdwork,replacingsomeofthehouseholdworkpreviouslydonebytheirspouses,whointurnincreasetheirworkhours.Thus,inthepresenceofrestrictivegendernorms,reducingcommutingcostsalonemaynotbeenoughfor

womentoincreaseworkhoursorincreaseparticipationinthelabormarkets.

Thesecondstrandofliteratureweconnecttoanalyzestheeffectsofreductionsincommutetimes,oftenbyprovidingtransitsubsidiesinrandomizedcontrolexperiments,onjobsearch

andemploymentcreation(Franklin,

2018;

Abebeetal.,

2016;

Phillips,

2014;

Moreno-Monroy

andPosada,

2018

).Thegeneralconsensusinthisliteratureisthatreductionsincommutingcostsincreasejobsearchintensityandemployment.Wedemonstratethatsocialnormsindevelopingcountriescanunderminetheeffectsofpoliciesthatreducecommutingcostsforwomen.Whileunmarriedskilledandunemployedwomenincreasejobsearchefforts,wedonotdetectanincreaseinemploymentpossiblyduetootherformsofdiscriminationanddisparities,consistentwithrestrictivegendernormsdiscussedby

Jayachandran

(2021)and

DinkelmanandNgai

(2022).

Low-skillmarriedwomen’slaboroutcomesbecomeworseif

anything.

Ourfindingshaveimportantpolicyimplications:Ifthegoalofprovidingfreetransporta-

tiontowomenistoincreasewomen’slaborforceparticipation,onlyasmallshareofthe

5

DasguptaandDatta

(2023)useacross-sectionaltime-usesurveyandcomparemenandwomenacross

statestoassesshowPinkSlipschemeinDelhiaffectedwomen’stimeusepatterns.Theydocumentanincreaseof30to50minutesinthetimewomenspentonworkduringthefirsttwomonthsaftertheintroductionofthescheme.Wefindanullcausaleffectontimeuseinourpanelestimationmaskedbyheterogeneitybyemploymentstatus.

Borkeretal.

(2020)haveanongoingexperimentwheretheywanttocomparethepartial

equilibriumresultsoffreebuspassestothegeneralequilibriumeffectsofDelhi’spolicy.

6

workforce,primarilyskilledemployedwomen,benefitfromthisprogramintheimmediate

short-run.

Therestofthepaperisorganizedasfollows.Section

2

describesthestudysetting.Section

3

describesthedatasets.Section

4

outlinestheempiricalmethodologytoestimatetheim-pactsofreductionsincommutingcosts.Section

5

presentstheresults.Section

6

provides

concludingremarks.

2Background

InIndia,limitedinfrastructureandtransportservicesrestrictmobilityforbothmenandwomen,butwomenfrequentlyexperienceextrasocio-culturalandeconomicfactorsthatnegativelyaffecttheircommutepatterns(

SrinivasanandRogers,

2005;

Tripathietal.,

2017;

Alametal.,

2021).Giventhesizeablegenderwagegap(

DuraisamyandDuraisamy,

2016;

Deshpandeetal.,

2018)andtheadditionalbarriersinaccessingthefinancialsystemcom

-paredtomen(

Khera,

2018

),paidaccesstotransportationisplausiblyharderforIndianwomen.Besides,thelowrateoffemaleusageofpublictransportmightraiseaperceptionproblemsincemorefemalepresenceinpublictransportationmakeswomenfeelsafer(

Saj-

jadetal.,

2017).

Inasurveyof3,800studentsatDelhiUniversity,

Borker

(2021)found

thatwomenarewillingtotravel27minutesmoreperdayor40%morethantheirdailytraveltimeiftheycanuseaperceivedsafertransportroute.Thesefactorsputwomenatadisadvantageregardingaccesstotransportservicesandinfrastructure(

Astropetal.,

1996;

DominguezGonzalezetal.,

2020),potentiallyaffectingtheirparticipationinlabormarkets

(PatacchiniandZenou,

2005;

ILO,

2017;

Sajjadetal.,

2017;

Martinezetal.,

2020).

Inlightofthesechallenges,theDelhigovernmentintroducedaschemeofferingfreebusridestoallwomeninthecityfromNovember2019onward(

Kejriwal,

2019).Theschememakes

bustravelfreeforwomeninallDelhiTransportCorporation(DTC)andClusterbuses.On

eachride,busoperatorsprovideapinktickettoeachwoman.Afterward,Delhi’sgovernment

7

compensatesthebusoperatorswith|10—equivalentof$0.14,allcurrencyconversionsusethe11/2019exchangerate—perpinkticketride(

TheEconomicTimes,

2019;

Durai,

2021

).Theprogramshowedanearlyresponse:just20daysafterthescheme’slaunch,femaledaily

ridershipinDTCandclusterbusesincreasedfrom33%to44%(

Sengar,

2019).

SpurredbythegoodreceptionoftheinitiativeinDelhi,onApril1andMay7,2021,thestatesofPunjabandTamilNadu,respectively,implementedfreebusrideschemesforwomenintheirstates,allowingfreetravelingovernment-ownedpublicbuses.6FromJuly2021toMarch2022,thepercentageofwomencommutingbybusinTamilNaduincreasedfrom40%

to61%(Sundaram,

2022).ThisincrementledtheGovernmenttoincreasethebudgetofthe

programfrom|12million($168mn),allocatedinthefirstyear,to|15.2million($212.8mn)

inthesecondyear(

Durai,

2021;

Sundaram,

2022

).

3Data

3.1MainData:ConsumerPyramidsHouseholdSurvey

OurmainsourceofdataistheConsumerPyramidsHouseholdSurvey(CPHS).Itisahousehold-levellongitudinalsurveyconductedbytheCentreforMonitoringIndianEcon-omy(CMIE).StartingwiththefirstwaveinJanuary-April2014,theCMIErunssurveysthreeroundsayear(January-April,May-August,andSeptember-December).Eachwavecoversabout160,000householdsfromallmajorIndianstates,maintainingaconsistentlyhighhouseholdresponserateofover80%.Amulti-stagestratifiedsurveydesignisde-ployed.Thebroadestlevelofstratificationisahomogeneousregion(HR),whichisdefinedasasetofneighboringdistrictswithinastatethatiscomparableinthefollowingcharac-

teristics:climate,urbanization,femaleliteracyrate,andpopulation.InAppendix

TableC1,

6InPunjab,theseincludePEPSURoadTransportCorporation(PRTC),PUNBUS,PunjabRoadwaysBuses,andCityBusServices,butdidnotincludeACbuses,VolvoBuses,andHVACBuses(

Express,

2021)InTamil

Nadu,thefreerideschemeincludesticketsfortheTamilNaduStateTransportCorporation(TNSTC)ordinarycitybuses.

8

welistthetwotreatedstatesofPunjabandTamilNadu—whichimplementedfreebusrideschemesforwomen—andtheircontrolstates,i.e.,thestatesthatareadjacenttothetreatedstatesanddidnotdistributefreebusticketstowomen.7SincetheCPHSdataisrepresen-tativeatthelevelofHRs,ouranalysisonlyincludesneighboringHRsinthecontrolstates,i.e.,weexcludeHRsincontrolstatesthatarenotadjacenttotreatmentstates.Appendix??displaysthemapoftreatedandcontrolHRs.Intotal,wehave20HRsintwotreatmentand

sevencontrolstates.

TheCPHShasfoursections:ConsumptionPyramids(CP),PeopleofIndia(PoI),AspirationIndia(AsI),andIncomePyramids(InP).Inthisstudy,weuseCP,InP,andPoIdata.TheCPisahousehold-levelmonthlysurveyreportinghouseholdexpendituresonvariouskindsofgoodsandservices,withoutabreakdownofindividualmembersofmulti-personhouseholds.Specifically,itaskshouseholdsabouttheirmonthlyexpenditureonallkindsoftransportincludingexpensesonacombinedcategoryof“buses,trains,andferries”(BTF).OurstudyperiodfortheCPisfromNovember2020toSeptember2021.TheInPisamonthlysurveythattrackstheincomeofeachhouseholdmember.WeusethesamestudyperiodasintheCPdata.ThePoIisanindividual-levelsurveyconductedeveryfourmonths.Therearethreewavesinayear:January-April,May-August,andSeptember-December.ThePoIdatahasinformationonone’semploymentstatus,timeusage,anddemographiccharacteristicslikegender,educationlevel,andmaritalstatus.FromthePoIdata,weknowhowmuchtimeapersonspendsonhouseholdactivities,atwork,andtraveling.Reportedtimeontravelisthetimespentbyapersontravelingfromoneplacetoanotherforallkindsofpurposesincludingwork-relatedactivities.TheCPsurveydoesnotaskspecificquestionsabouttime

spentcommutingtowork,searchingforajob,oronleisure.

WeusesixwavesofPoIfromMay-August2020toJanuary-April2022(orfromthe20th

7TheCMIEstartedtocollecthouseholdmembers’timeusageinformationinthewaveofSeptember-December2019,whilethegovernmentofDelhistartedthepinkslipschemeinNovember2019.WedonotincludeDelhiintheanalysissincewedonothavepre-periodinformationontimeusagethere.

9

wavetothe25thwave).WealsomatchthehouseholdsthatappearedinthePoIdatatohouseholdsintheCPdata.Appendix

TableC2

liststhestudyperiodsforthetwosectionaldatasets.Werestrictoursampletowomen(orhouseholdshavingwomen)agedbetween15and65attheirfirstappearanceinthedata.Appendix

TableC3

liststhevariablesusedintheanalysisandtheirdefinitions.Appendix

TableC4

displaysthesummarystatisticsofourstudysample.PanelAdisplaysthehouseholdcharacteristicsinDecember2020.Differencesbetweenhouseholdsintreatedvs.controlHRsintermsofruralresidence,numberofpeople,andper-capitaincomeandexpendituresarerelativelysmall.InpanelB,wecomparewomenintreatedHRstothoseincontrolHRsinMay-August2020.Thedistributionsofage,maritalstatus,andeducationarecomparableforthetwogroupsofwomen.Womenintreatedareasarelesslikelytoparticipateinthelabormarketbutconditionalonparticipation,theyaremorelikelytobeemployed.Theyalsotendtospendmoretimeonhouseholdactivitiesand

workbutlesstimeontravelthanwomenincontrolareas.

3.2DelhiPrimarySurvey

TheCMIEstartedtocollecthouseholdmembers’timeusageinformationinthewaveofSeptember-December2019,whilethegovernmentofDelhistartedthePinkSlipschemeinNovember2019.Sincewedonothavepre-periodinformationontimeusageinDelhi,wedonotincludeitintheanalysis.Tocomplementandbolsterourbaselineanalysis,however,weuseprimarydatacollectedviaasurveyinDelhiinFebruary2020bytheGesellschaft

fürInternationaleZusammenarbeit(GIZ)India(Mahendru,

2022).

Thesurveycollecteddatafrom1,525femalebususers(1,294continuoususersand231newusers)and500nonusers.8Thesampleisrandomlyselectedatmajorattractionsandgenerationpoints

acrossDelhi(Appendix

FigureB1

).9Toconstructacomparablesampleofnonusers,new

8ContinuoususerswerewomenwhotookbusesbothbeforeandaftertheimplementationofthePinkSlipProgram.Newuserswerewomenwhobeganusingbusesaftertheprogram’simplementation.Nonuserswerewomenwhodidnotusebusesbeforeoraftertheprogram.

9Thegenerationpointsareallmajorlocationswithinthecitywheretripseitheroriginateorareattractedto.Theseincludemajorworkareas,shoppingdistricts,majorschools,andothers.

10

users,andcontinuoususers,weemploypropensityscorematching.Wematchwomenonthefollowingvariables:age,occupation,10totalaveragemonthlyhouseholdincome,totalaveragemonthlyhouseholdexpenditureontravel,ownershipofprivatevehicles,andknowl-edgeofhowtodrive.OurstudysampleoftheDelhiprimarysurveyisthematchedsample(n=1,290)consistingof184neverusers,182newusers,and924alwaysusers.ComparedtothetreatedsampleintheCPHSdata,thematchedsampleischaracterizedbyhigherlevelsofeducationandannualhouseholdincome.Specifically,approximately68%ofthebususersinthematchedsampleholdagraduatedegree,and85%ofthemreportanannualhousehold

incomeofover|240,000.

Wecompareperceptionsofbusesbetweennonusersandusers(panelAof

Table1

),aswellasbetweennewusersandcontinuoususers(panelBof

Table1

)inthepost-schemeperiod.Wespecificallyexaminetheirdifferencesintheperceptionsoffiveaspects:1)affordability

andavailabilityofbustransit;2)safetyregardingaccidents,crashes,threats,andthefts;

3)connectivity;4)busfrequency,waitingtime,travelduration,andunnecessarystops;5)accessibilitytobusstops.InpanelA,wecanseethatnonusersconsistentlygivelowerrat-ingsacrossallfiveperspectives;thatis,theyfindbustravellesssatisfactoryacrossallfivedimensionscomparedtousers.Theyareparticularlyconcernedaboutsafetyissues.Theaverageratingconcerningsafetyamongneverusersis1.63,indicatingalevelofsatisfac-tionthatfallsbetweenhighlyunsatisfactoryandunsatisfactory.InpanelB,wefindthatcomparedtocontinuoususers,newuserstendtogivealowerratingonsafetybutahigherratingregardingtheaffordabilityandavailabilityoffreecommutes.Theseresultssuggestthatwhenitcomestotransportation,womenprioritizenotonlysafetyconcernsbutalsoaffordability.Reducingcostsinpublictransportationhasthepotentialtoencouragewomen

tochoosebusesasaviableoptionforjobsearchandcommuting.

10Theoccupationvariableconsistsofthefollowingcategories:service,business,informalworker,dailywager,homemaker,andstudent.

11

3.3AuxiliaryData

Sinceourstudyperiod