ESSAYSONTHEECONOMICSOFORGANTRANSPLANTATIONByKeithTeltserADISSERTATIONSubmittedtoMichiganStateUniversityinpartialoftherequirementsforthedegreeofEconomicsŒDoctorofPhilosophy2016ABSTRACTESSAYSONTHEECONOMICSOFORGANTRANSPLANTATIONByKeithTeltserChapter1:DoKidneyExchangesImprovePatientOutcomes?AddressingtheincreasinglyunmetdemandfortransplantablekidneysintheU.S.requirescreativ-ityduetolegislationthatprohibitsbuyingandsellingorgans.Toincreasethenumberofsuccessfultransplantsthattakeplace,transplantcentershavebeguntoimplementkidneyexchangeprograms.Inthischapter,Iestimatethenumberofadditionaltransplantsgeneratedbykidneyexchangesbyanalyzinghowtheprobabilityofreceivinganexchangetransplantaffectstheprobabilitythatapatientexperiencesothertransplantoutcomes,includingdeathwhilewaiting.Todothis,Icreateanovelmeasureofexchangeprevalencethatexploitsvariationinexchangeactivityacrosstimeandtransplantcenters,aswellastheimportanceofpatientproximitytocentersperformingexchanges.Ithat6.2ofevery10exchangetransplantsrepresentlivingdonortransplantsthatwouldnothaveoccurredintheabsenceofexchange.Ialsothatatenpercentagepointincreaseintheprobabilityofreceivinganexchangetransplantincreasesone-yearandtwo-yeargraftsurvivalbyroughly2.1percent,andreduceswaitinglistregistrationdurationby10percent.Chapter2:TestingforHeterogeneityintheEffectsofKidneyExchangesonPatientOut-comesInChapter1ofthisdissertation,Ithattheintroductionofexchangeimprovespa-tientoutcomeswithrespecttoquantityoflivingdonortransplants,transplantsurvival,andwaitingtime.Sinceoneofthepurportedofkidneyexchangeisthatitincreasesaccesstolivingdonortransplantsforhard-to-matchpatients,andgivenexistingconcernsinthetransplantcommu-nityaboutdemographicdisparitiesintransplantaccessandoutcomes,Ianalyzetheextenttowhichexchangeshavedifferentialimpactsontransplantoutcomesacrosspatientsofvariousbloodtypes,levelsofsensitivitytoforeignorgans,races,ages,andlevelsofeducation.Toestimatetheeffectsofinterest,Iexploittheimportanceofpatientproximitytoexchangecentersinordertogenerateexogenousvariationintheprobabilityofreceivingatransplantviaexchange.Themainingisthatrisingexchangeprevalencetendstoharder-to-matchbloodtypestheleastwithrespecttobothquantityandqualityoftransplants,thoughtheyseemtohavelargerpositiveimpactsonpatientswithhigher(comparedtolower)levelsofsenstitivitytoforeignorgans.Thesecondmainisthatblackpatientsandolderpatientsexperiencethesmallestgainsintransplantquantityasexchangeprevalenceincreasesrelativetootherraceandagegroups,respectively.Chapter3:AllocatingScarceOrgans:HowaChangeinSupplyAffectsTransplantWaitingLists(withStacyDickert-ConlinandToddElder)Becausegeographyplaysakeyroleindeceaseddonororganallocation,shockstothelocalsupplyoforganswilllikelyaffecttransplantwaitlists.WeusedataontransplantrecipientsfromtheScien-RegistryofTransplantRecipientstoassesswhethershiftsinthesupplyoforgansarisingfromchangesinmotorcyclehelmetlawsaffectthebehaviorandoutcomesoftransplantcandidates.Wethat,followingrepealsofstatewidemotorcyclehelmetlaws,thelocalsupplyoftransplantableorgansfromdonorskilledinmotorvehicleaccidentsincreasesbynearly20percent.Thesesupplyshocksinducestrongresponsesfromtransplantcandidates:wstolocaltransplantwaitlistsincreaseby12percentintheyearsfollowingtherepeal.Thesewsareespeciallypronouncedamongthosewholiveoutsidethelocalarea,implyingthatintheabsenceofaformalpricingmech-anism,waitingtimesfororgansaretherelevantfipricefldetermininglistingdecisions.Inaddition,transplantsfromlivingdonorsdeclinefollowingsupplyshocks,suggestingthattheavailabilityoftransplantsfromlivinganddeceaseddonorscandidates'decisionstoseekorgansfromlivingdonor.Finally,theaveragetransplantismoresuccessfulfollowingtheincreaseintrans-plantableorgans,asmeasuredbytheprobabilityofthetransplantedorgansurvivingatleastoneyear.Thisqualityimprovementmayexplainthedramaticincreasesinwaitliststotheareaswithmoreorgans.ACKNOWLEDGEMENTSFirstandforemost,IwouldliketoexpressmysinceregratitudetoStacyDickert-Conlinforhermentorshipasbothadvisorandco-author,forhergenerositywithhertime,forherthoroughcom-ments,forherthoughtfuladvice,andforherunwaveringsupport.IwouldalsoliketoexpressmygratitudetoMikeConlinforhisinspiringenthusiasmaboutmyresearch,forhishonestadvice,andforhisencouragement.IwouldliketothankToddElderforhismentorshipasaco-author,forhiscommentsonmyresearch,forhisexcellentworkasplacementdirector,andforbringingexcitementandhumortoourresearchmeetings.IwouldliketothankKyooilKimforservingonmydissertationcommittee,andforsharinghisvaluableinsightsontheeconometricaspectsofmyresearch.IwouldalsoliketothankAlanLeichtmanforapproachingmeafterIaskedsomequestionsataseminaronkidneyexchangeattheUniversityofMichigan,forinvitingmetositinonseveralmeetingsdiscussingkidneyexchangeresearchandpractice,andforservingonmydissertationcommittee.IamgratefultothatAlanwaswillingtosharehisknowledgeofkidneyexchange,organdonationingeneral,andthemedicalliteraturewithaneconomistfioutsider."IwouldliketothankAndrewBibler,DanielLitwok,GregWallsworth,ChristianAhlin,SorenAnderson,JohnGoddeeris,GarySolon,LeahLakdawala,andMarceloPerraillonfortheirhelpfulcommentsandfeedback.Iwouldalsoliketothankseminarparticipantsatthe2015AEAannualmeeting,the2015MEAannualmeeting,the2015APPAMFallResearchConference,the2015SEAannualmeeting,MichiganStateUniversity,andUniversityofLouisvillefortheirhelpfulcommentsanddiscussion.IwouldliketorecognizetheUnitedNetworkforOrganSharing(UNOS),theOrganPro-curementandTransplantationNetwork(OPTN),theMinneapolisMedicalResearchFoundation(MMRF),andtheRegistryofTransplantRecipients(SRTR).ThedatareportedinChap-ters1and2havebeensuppliedbyUNOSasthecontractorforOPTN.Theinterpretationandreportingofthesedataaretheresponsibilityoftheauthor(s)andinnowayshouldbeseenasanivofpolicyoforinterpretationbytheOPTNortheU.S.Government.IwouldespeciallyliketothankJenniferWainrightatUNOSforherhelpwithmydatarequestandansweringfollow-upquestions.ThedatareportedinChapter3havebeensuppliedbyMMRFasthecontractorfortheSRTR.TheinterpretationandreportingofthesedataaretheresponsibilityoftheauthorsandinnowayshouldbeseenasanofpolicyoforinterpretationbytheSRTRortheU.S.Government.WewouldliketothankGopiGoda,JoseFernandezandseminarparticipantsatMontanaStateUniversityforveryvaluableinputtotheproject.WearealsogratefultoJonathanSiegleforexcellentresearchassistance.IwouldliketoexpressmyappreciationforthesupportIreceivedfromtheGraduateSchoolandtheDepartmentofEconomicsatMichiganStateUniversity,includingtheUniversityEnrichmentFellowship,ResearchEnhancementGrant,andDissertationCompletionFellowship.IwouldalsoliketoexpressmyappreciationfortheofstaffintheDepartmentofEconomics,es-peciallyLoriJeanNichols,fortheirdedicationtothesuccessofthedepartmentandtheeconomicsgraduatestudents.Last,butcertainlynotleast,Iwouldliketothankmyfamilyfortheirunconditionalloveandsupport.Iespeciallywanttothankmywife,Christina,formovingtoMichiganandthenKentuckywithme,formakingsureIrememberedtoeat,forbeingmysidekickonthejobmarket,andforthecountlessotherwaysinwhichshehassupportedmyacademicandcareerpursuits.TABLEOFCONTENTSLISTOFTABLES.......................................viiiLISTOFFIGURES.......................................xiCHAPTER1DOKIDNEYEXCHANGESIMPROVEPATIENTOUTCOMES?.....11.1Introduction......................................11.1.1BackgroundonKidneyTransplantationandKidneyExchange.......41.1.2LiteratureReview..............................81.2ConceptualFramework................................111.3Data..........................................151.4Estimation.......................................181.4.1MeasuringActivity..............................211.5Results.........................................231.6Robustness......................................251.7Conclusion......................................29CHAPTER2TESTINGFORHETEROGENEITYINTHEEFFECTSOFKIDNEYEXCHANGESONTRANSPLANTQUANTITYANDQUALITY.....312.1Introduction......................................312.2BackgroundandMotivation.............................342.3Data..........................................392.4Estimation.......................................462.5Results.........................................492.5.1BloodType..................................492.5.2PRAScore..................................512.5.3Age......................................522.5.4Race......................................542.5.5Education...................................552.5.6HeterogeneousEffectsonDonors......................562.6Conclusion......................................58CHAPTER3ALLOCATINGSCARCEORGANS:HOWACHANGEINSUPPLYAFFECTSTRANSPLANTWAITINGLISTS(withStacyDickert-ConlinandToddElder)................................613.1Introduction......................................613.2DataandInstitutionalDetails.............................643.2.1DataonOrganDonationsandTransplants..................643.2.2TransplantWaitingLists...........................653.2.3HelmetLawsandWaitingLists........................693.3TheEffectsofMotorcycleHelmetLawsontheSupplyofOrgans.........703.4AConceptualFrameworkofRationingbyWaitlists.................743.5WaitlistResponsestotheChangeintheSupplyofOrgans.............76vi3.5.1WaitlistAdditionsResponsestotheChangeinSupply...........763.5.2SubstitutionAwayfromLivingDonorTransplants..............813.6HealthOutcomesfromtheChangeintheSupplyofOrgans.............833.7Conclusion......................................85APPENDICES.........................................87APPENDIXATABLESANDFIGURESFORCHAPTER1..............88APPENDIXBTABLESANDFIGURESFORCHAPTER2..............102APPENDIXCTABLESANDFIGURESFORCHAPTER3..............118APPENDIXDPROOFSFORCHAPTER1.......................135APPENDIXESUPPLEMENTARYINFORMATIONFORCHAPTER3.......140BIBLIOGRAPHY........................................143viiLISTOFTABLESTableA.1:FrequencyofRegistrationOutcomes........................88TableA.2:Survival,MatchQuality,WaitingTime(2000-July2014)............88TableA.3:SensitivityandAge(2000-July2014).......................89TableA.4:PatientDistancetoTransplantCenterofOperation................89TableA.5:MeasuringCrowd-Out,2SLSResults.......................90TableA.6:QualityMeasures,2SLSResults..........................90TableA.7:LeadsandLagsofIV,Crowd-Out.........................91TableA.8:LeadsandLagsofIV,Quality...........................92TableA.9:MeasuringCrowd-Out,2SLS,UsingLags.....................93TableA.10:QualityMeasures,2SLS,UsingLags.......................94TableA.11:MeasuringCrowd-Out,2SLS,30MileRadius...................95TableA.12:QualityMeasures,2SLS,30MileRadius.....................95TableA.13:MeasuringCrowd-Out,2SLS,75MileRadius...................96TableA.14:QualityMeasures,2SLS,75MileRadius.....................96TableA.15:MeasuringCrowd-Out,2SLS,Zip-LevelAggregates...............97TableA.16:MeasuringCrowd-Out,2SLS,IncludingWLRemovalsduetoDeteriorationofConditionandTransplantRefusal.......................97TableB.1:FrequencyofRegistrationOutcomes........................102TableB.2:Survival,MatchQuality,WaitingTime(2000-July2014)............103TableB.3:HeterogeneityinTransplantQuantity.......................104TableB.4:HeterogeneityinTransplantQuality........................105TableB.5:RegistrationOutcomeHeterogeneitybyLevelofKidneyExchangeActivityin2013.......................................106viiiTableB.6:BloodType,2SLSResults,Quantity........................107TableB.7:BloodType,2SLSResults,Quality........................108TableB.8:PRACategory,2SLSResults,Quantity......................109TableB.9:PRACategory,2SLSResults,Quality.......................110TableB.10:AgeCategory,2SLSResults,Quantity.......................111TableB.11:AgeCategory,2SLSResults,Quality.......................112TableB.12:Race,2SLSResults,Quantity...........................113TableB.13:Race,2SLSResults,Quality............................114TableB.14:Education,2SLSResults,Quantity.........................115TableB.15:Education,2SLSResults,Quality.........................115TableB.16:HeterogeneousEffectsonLivingDonationbyBloodType............116TableB.17:HeterogeneousEffectsonAnonymousDonationbyBloodType.........116TableC.1:WaitingListAdditions,byOrganandYear....................118TableC.2:WaitingListExitsbyYearandReasonforLeaving................119TableC.3:NumberofPersonsonWaitlists,byOrganandYear...............120TableC.4:ChangesinStateMotorcycleHelmetLaws,1988-2012..............121TableC.5:NumberofDonorsandNumberofOrgansTransplantedfromMVAandNon-MVADonors....................................122TableC.6:EstimatesoftheEffectoftheRepealofHelmetLawsonPerCapitaOrganDonors,OrganDonations,andOrganTransplants,byOrgan.......123TableC.7:EstimatesoftheEffectoftheRepealofHelmetLawsonWaitingListAdditionsbyIn-VersusOut-of-Area...............124TableC.8:EstimatesoftheEffectofHelmetLawRepealsonWaitingListAdditionsbyIn-VersusOut-of-AreaandbyMultilistingStatus...............125TableC.9:EstimatesoftheEffectofHelmetLawRepealsonPerCapitaLivingOrganDonors,byRelationtotheRecipient....................126ixTableC.10:EstimatesoftheEffectofHelmetLawRepealsonGraftSurvivalamongTransplantRecipients,byOrgan......................127xLISTOFFIGURESFigureA.1:KidneyTransplantsandWaitingListSize,Per1,000,000U.S.Residents....98FigureA.2:TransplantsbyDonorType,Per1,000,000U.S.Population...........99FigureA.3:Two-wayandThree-wayExchange,andDonorChainDiagrams........99FigureA.4:CentersPerformingAtLeastOnePairedorListExchange,orBoth.......100FigureA.5:NumberofExchangesWithin50Miles,ExcludingOwn.............101FigureB.1:KidneyTransplants/WaitingListSize,Per1,000,000U.S.Residents......117FigureC.1:DonationServiceAreaMapoftheUnitedStates................128FigureC.2:EventStudyEstimatesofHelmetLawRepealsonPerCapitaOrganDonorsandTransplants....................129FigureC.3:EventStudyEstimatesofHelmetLawRepealsonPerCapitaWaitlistAdditions...........................130FigureC.4:EstimatesofHelmetLawRepealsonFlowofPerCapitaWaitlistAdditionstoNeighboringOPOswithaLawChange(Treatment)relativetotheFlowtoNeighboringOPOswithoutaLawChange(Control):AllOrgans....................................131FigureC.5:EstimatesofHelmetLawRepealsonFlowofPerCapitaWaitlistAdditionstoNeighboringOPOswithaLawChange(Treatment)relativetotheFlowtoNeighboringOPOswithoutaLawChange(Control):KidneysandNon-Kidneys............................132FigureC.6:EstimatesofHelmetLawRepealsonFlowofPerCapitaWaitlistAdditionstoNeighboringOPOswithaLawChange(Treatment)relativetotheFlowtoNeighboringOPOswithoutaLawChange(Control):KidneyMultilistersandKidneyNon-Multilisters................133FigureC.7:EventStudyEstimatesofHelmetLawRepealsonOne-yearGraftSurvival..134xiCHAPTER1DOKIDNEYEXCHANGESIMPROVEPATIENTOUTCOMES?1.1IntroductionIn2014,4,445peoplediedwaitingforakidneytransplant.Inthesameyear,36,156peopleenteredawaitinglistforakidney,andonly17,108peoplereceivedakidneytransplant.1FigureA.1showsthedramaticgrowthinthewaitinglistforkidneysovertime,whiletransplantsgrowatamuchslowerpace.AddressingtheincreasinglyunmetdemandfortransplantablekidneysintheU.S.whilemaintainingorincreasingtransplantqualityrequirescreativityduetotheNationalOrganTransplantationAct(NOTA)of1984,whichbannedthesaleofhumanorgans.Asearlyastheyear2000,transplantcentershavebeenattemptingtoincreasethenumberandqualityoftransplantsbyfacilitatingkidneyexchangesamongpatientswithwillingbutincompat-iblelivingdonors.Inthemostbasictypeofexchange,atwo-waypairedexchange,patientsmayfiswap"theirwillingdonorswhenthedonorfromonepairisamatchforthepatientinanotherandviceversa.Pairedexchangescanbeextendedintodonorchains,whereanaltruisticdonorstartsaseriesofpairedexchangesbydonatinganonymouslytoapatientwithawillingincompatibledonor.Anothervariationislistexchange,whereawillingdonorgivesakidneytosomeoneonthewaitinglistinexchangeforthenextcompatibledeceaseddonorkidneyforhisorherintendedpatientinneed(Delmonicoetal.,2004).Rothandcoauthors(Rothetal.,2004;Rothetal.,2007;Ashlagietal.,2011)madelargecontributionstothedevelopmentofkidneyexchangesbyapplyingexistingmechanismdesignliteraturetothepatient-donormatchingproblem,simulatingandcomparingtheeffectivenessofvariousmechanisms,andaidinginthereal-worldimplementationofexchangeprograms.Simula-tionsinRothetal.(2004)demonstratethepotentialofexchangetoincreasethenumberofliving1BasedonOrganProcurementandTransplantationNetwork(OPTN)Dataasof9/4/2015,foundathttp://optn.transplant.hrsa.gov.1donortransplants,whilealsoaccountingforthepossibilityofpatientssubstitutingtoexchangesfromdirectlivingdonors-thosewhogivetoknownandcompatiblepatients.Infact,FigureA.2showsconsiderablegrowthinpairedandlistexchangesstartingaround2005;exchangesincreasedfrom0.6percentofalllivingdonortransplantsin2005to12.1percentin2014.2Thetrendsindirectliving,pairedexchanges,andlistexchangesfrom2005to2014areconsistentwiththehy-pothesisthatsomepatientsreceivingkidneysviapairedandlistexchangewouldhavereceivedadirectlivingtransplantintheabsenceofexchange.Duringthesametimeperiod,weseegrowthintransplantsfromanonymousdonors-thosewhogivetounknownpatients.Thegrowthofkidneyexchangemayexplainthisobservationaswell,sinceanonymousdonationscanfacilitatemoretransplantswhenusedtostartdonorchains.Theintroductionofexchangemaythereforecrowdoutsomedirectlivingdonationsduetosubstitutionwhilecrowdinginanonymousdonations.Rothetal.(2004)alsopredictimprovementsintissuetypecompatibilityasmeasuredbythenumberofHumanLeukocyteAntigen(HLA)mismatches.However,theauthorsobtaintheirre-sultsunderasetofnecessaryrestrictionsandstrongassumptionsaboutpatientpreferences.More-over,thesimulationscannotaccountforpotentialgrowthinanonymousdonations.Asaresult,theextenttowhichtheintroductionofkidneyexchangeaffectsthequantityandqualityofkidneytransplantsremainsanempiricalquestion.Mypaperisthetoanalyzethecausalrelationshipbetweentheintroductionofexchangeandobservedpatientoutcomes.Iestimatethenumberofadditionaltransplantsgeneratedbykidneyexchangesbyanalyzinghowtheprobabilityofreceivinganexchangetransplantaffectstheprobabilitythatapatientexperiencesothertransplantoutcomes,includingdeathwhilewaiting.IdothisusingOPTNStandardTransplantandAnalysisResearch(STAR)whichcontaintheuniverseofwaitinglistregistrationsandtransplants,alongwithanovelmeasureofexchangeprevalencebasedonapatient'sresidentiallocationandmonthofregistrationoutcome.Usingthesameapproach,Ithenestimatetheresultingimprovementsingraft(transplant)sur-vival,matchquality,andwaitingtime.Holdingallelseequal,wewouldexpecttoseeimproved2Basedonauthor'scalculationsusingOPTNdataasof9/4/2015.2overallgraftsurvivalifmorepeoplereceivelivingdonationswiththeintroductionofexchangeandlivingdonorkidneysyieldlongergraftsurvivalthandeceaseddonorkidneys.3,4Moreover,thosewhoswitchfromadirectdonortoanexchangemaydosoforreasonsthatleadtoimprovedgraftsurvivaloverall.,reductionsinHLAmismatchesandwaitinglistregistrationdurationareassociatedwithimprovedgraftsurvivalaccordingtoresearchinthetransplantationliterature(Opelz,1997;Meier-Kriescheetal.,2002;DavisandDelmonico,2005).Withtheintroductionofexchange,weexpectreducedfrictionswhensearchingforalivingdonor.Moreover,ifexchangeincreasestotallivingdonations,weexpectreducedexcessdemandfordeceaseddonorkidneys.Bothoftheseeffectscouldyieldreductionsinwaitinglistregistrationdurationfortheexchangerecipientandotherswhoarewaiting.Exchangealsoallowspatientstosearchforabettermatch,ratherthanhavingtorelyonacompatiblefriend,relative,orthedeceaseddonorkidneywaitinglist.Improvingmatchqualityisanintegralpartofsuitablelivingdonorsforhard-to-matchpatientsand,aswewillseeinSection1.3,recipientsofkidneyexchangetendtobeindividualswhoarehardertomatch.Thepaperproceedsasfollows.TheremainderofSection1providesadditionalinformationonkidneytransplantation,kidneyexchange,andrelevantliterature.Section1.2presentsaconceptualframeworkmodelingtheimpactofkidneyexchangeonthedecisiontodonate,whichmotivatestheempiricalanalysis.Section1.3discussesthedatathatIuseandprovidesdescriptivestatistics.I3Whilethereappearstobealackofcausalevidence,itisgenerallyacceptedinthetransplantcommunitythatlivingdonorkidneytransplantsaremoresuccessfulthanthosefromdeceaseddonors.CalculationsbasedonOPTNindividual-leveltransplantdata(asof12/31/2014)from1988to2008revealthat3.2percentoflivingdonorkidneygraftsfailedwithinoneyearcomparedto7.9percentofdeceaseddonorkidneygrafts.Similarly,16.8percentoflivingdonorkidneygraftsfailedwithinveyearscomparedto29.6percentofdeceaseddonorkidneygrafts.Notethatthesefailuresalsoincludedeathsofthosewithnon-functioninggraftswithinthetime-frame.Deathsofthosewithfunctioninggraftsbeforeoneyearorveyearswereexcludedfromtherespectivecalculations.4Intheirpaperprovidinganoverviewoflivingkidneydonationpracticesasof2005,DavisandDelmonico(2005)suggestthatthisispartlyduetoreducedwaitingtimeandtimespentondialysisforlivingdonorkidneyrecipientscomparedtodeceaseddonorkidneyrecipients.3thenpresentaninstrumentalvariableapproachtoestimatingtheeffectofexchangesonpatientout-comesinSection1.4,whereIusetime-varyinglocalkidneyexchangeactivityasaninstrumentforwhetherapatientreceivesakidneyviaexchange.Section1.5presentsthetwo-stageleastsquaresestimates,whereIthatroughly6.2ofevery10transplantsviaexchangerepresentlivingdonortransplantsthatwouldnothaveoccurredintheabsenceofexchange.Conditionalonreceivingatransplant,Indthatatenpercentagepointincreaseintheprobabilityofreceivingatransplantviaexchangeincreasesone-yeargraftsurvivalby2percentagepoints,two-yeargraftsurvivalby1.9percentagepoints,andreduceswaitinglistregistrationdurationby59.2days,relativetoanaverageoverallone-yeargraftsurvivalrateof93percent,two-yearrateof88percent,and597dayregis-trationduration.Iasmallnegativebuteffectofexchangeontissuetypematchquality.Section1.6presentstheresultsofvariousrobustnesschecks,andSection1.7concludes.1.1.1BackgroundonKidneyTransplantationandKidneyExchangeTherearetwomaintreatmentoptionsavailabletoapatientexperiencingkidneyfailure:trans-plantationanddialysis.Dialysisisanongoingtreatmentthatprovidessomeofthebloodthathealthykidneyswouldperform.However,forthosewithchronickidneydiseaseorend-stagerenaldisease,dialysisisnotacurenoranattractivelong-termtreatment.Thesepatientscanturntotransplantationforamorepermanentandxiblesolution.Onceapatientdecidestopursueakidneytransplant,theymayregisteronawaitinglistforadeceaseddonorkidneyand/orsearchforawillinglivingdonor.5LookingatFigureA.2,deceaseddonationsarebyfarthemostcommonsourceofkidneytrans-plants.In2014nearly68percent,or11,570of17,108,ofkidneytransplantscamefromdeceaseddonors.6Doctorsrecoverkidneysfromeligibledeceaseddonors,whicharethenallocatedbyorganprocurementorganizations(OPOs)acrosstheUnitedStates.7Whenahealthykidneyisrecovered5TheNationalKidneyFoundation,www.kidney.org,providesmoredetailedinformation.Ac-cessedJuly13,2015.6BasedonOPTNdataasof9/18/2014.7OPOs,alongwithallotherprofessionalsinvolvedwithorgantransplantationanddonation4andbecomesavailablefortransplantation,theOPOservicingtheareainwhichthekidneywasrecoveredgeneratesaprioritylistofcompatiblepatientsonthewaitinglist.Thissubsetofpa-tientsisrankedbasedonfactorsincludingtissuematch,bloodtype,lengthoftimeonthewaitinglist,immunestatus,anddistancebetweenthepotentialrecipientandthedonor.Afterthelistisgenerated,theOPOoffersthekidneytothetransplantteamofthepatient.8Bloodtypecompatibilityistheconditionthatneedstobemetfortransplantsuccess.Ingeneral,peoplewithtypeObloodcanonlyreceivefromtypeOdonors,buttheycangivetoanyotherbloodtype.PeoplewithtypeAbloodcanonlygivetotypeAorABpatients.PeoplewithtypeBbloodcanonlygivetotypeBorABpatients.Finally,peoplewithtypeABbloodcanonlygivetotypeABpatients.Additionalcompatibilityconcernsincludetissuetypematchingandpatientsensitivity.TissuetypematchisbasedonthenumberofHLAmismatchesbetweenpatientanddonor,ofwhichtherecanbebetween0and6.ThemoreHLAmismatchesbetweenthepatientandaprospectivedonor,themorelikelyitisthatthepatient'sbodywillrejectthetransplantinspiteofanyreasonableim-munosuppressivedrugtreatment.Patientsensitivityismeasuredbythelevelofantibodiespresentinthepatient'sblood,calledthePanelReactiveAntibodies(PRA)score,whichtakesonavaluebetween0to100andindicatesthepercentageofthegeneralpopulationwithwhomthepatientislikelyincompatible.Asanalternativetowaitingforadeceaseddonorkidney,patientscansearchforlivingdonorswithintheirnetworkoffamilyandfriends.Theymaybeparticularlylikelytodosoiftheyfeeltheyfacetoolongofawaitand/ortheywanttomaximizeexpectedgraftsurvivaltime.Onceapatientpotentialdonors,theyundergothesamecompatibilityteststhatareusedfordeceaseddonations.Donorsarealsoscreenedforheartandlungdisease,kidneyfunction,andpsychologicalsuchastransplantcentersanddoctors,makeuptheOrganProcurementandTransplantationNet-work(OPTN),whichisadministeredbytheUnitedNetworkforOrganSharing(UNOS).UNOSisaprivateorganizationundercontractwiththeU.S.DepartmentofHealthandHumanServices(DHHS)tomaintainthenationalorgantransplantsystem.8UNOSTransplantLiving,http://www.transplantliving.org/before-the-transplant/about-organ-allocation/matching-organs,accessedJuly13,2015.5wellness.9Anadditionalconstraintfacinglivingdonorsisthecostofdonating.Eventhoughmed-icalexpensesforlivingdonationaretypicallycoveredbytherecipient'sinsuranceoratransplantcenter'sOrganAcquisitionFund,10potentialdonorsmaynotbeabletoaffordtheassociatedtravelcosts,timeoffofwork,orriskoffuturemedicalproblemsresultingfromtheprocedure.11Mostlivingdonationsoccurbecausepatientsawillingcompatibledonor.In2014,nearly85percentoflivingkidneydonationsweredirectlivingdonations.12Amongdirectlivingdonorsin2014,about15percentwerethepatient'sspouseorlifepartner.Roughly24percentwerethepatient'sfullorhalfsibling,16percentwerethepatient'schild,10percentwerethepatient'sparent,7percenthadsomeotherbiologicalrelationship,and27percenthadsomeothernon-biologicalrelationshiptothepatient.13Exchangeoffersanadditionallivingdonationoptiontothosewhohavewillingbutincom-patibledonors.Sucharrangementsincludepairedexchanges,listexchanges,anddonorchains,asdiscussedintheintroduction.Inapairedexchange,patientsmayfiswap"theirwillingdonorswhenthedonorfromonepairisamatchforthepatientinanotherandviceversa.Thiscanbeextendedtoincludethreeormoreincompatiblepairsaswell.Pairedexchangescanalsotaketheformofadonorchain,whichisnotlimitedtoaedgroupofincompatiblepairs.Tostartadonorchain,anon-directedlivingdonor,orfiGoodSamaritan"donor(accordingtotheNationalKidneyRegistry)whoisnotgivingonbehalfofalovedoneinneed,donatesakidneytoapatientwhohasawillingincompatibledonor.Thewillingdonorofthepatientreceivingthenon-directeddonor's9UNOSTransplantLiving,http://www.transplantliving.org/living-donation/being-a-living-donor/tests/,accessedJuly13,2015.10UNOSTransplantLiving,http://www.transplantliving.org/living-donation,accessedJuly13,201511AlthoughNOTAexplicitlybannedtheexchangeofvaluableconsiderationfororgans,itisnowlegaltoreimburselivingdonorsforcertaincostsincurredincludinglostwagesandtravelcostsaccordingtoSection3oftheOrganDonationandRecoveryImprovementAct(2003).OnemajorsourceofsuchreimbursementistheNationalLivingDonorAssistanceCenter(NLDAC).Inaddition,certainstateshavebegunimplementingtaxandpaid-time-offincentivesforlivingdonors.12Basedonauthor'scalculationsusingOPTNdataasof9/4/2015.13Basedonauthor'scalculationsusingOPTNdataasof9/4/2015.6kidneywillthendonatetoapatientinanotherincompatiblepair.Thisprocesscontinuesuntilnomorematchesarefound,arecipient'swillingincompatibledonorbacksout,orthewillingdonorgivestosomeoneonthewaitinglistwhoisnotpartofapair.FigureA.3depictsdiagramsofatwo-wayexchange,three-wayexchange,anddonorchain.Exchangearrangementsarefacilitatedbymatchingincompatiblepatient-donorpairswhohavesignedupwithanexchangeregistry.Theseregistriesmaybemanagedbyasingletransplantcenter,suchastheJohnsHopkinsUniversityIncompatibleKidneyTransplantProgram,orbyaconsortiumoftransplantcenterswherecenterssharetheirregistries,suchastheNationalKidneyRegistryandAllianceforPairedDonation.Thecenterorconsortiummatchespairsintheregistryoverarangeofcharacteristicssimilartothoseusedinrankingcandidatesfordeceaseddonorkidneyswiththeobjectiveofmaximizingsomemixofquantityandqualityofmatches.Eachcenterorconsortiumimplementsitsownobjectivefunction.Forexample,thealgorithmthattheAllianceforPairedDonationusesgiveshighestprioritytopatientswithhighPRAscores,patientswhopreviouslydonated,patientsunder5yearsold,andtomatcheswithzeroHLAmismatches.14ThepairedexchangeintheU.S.occurredin1994,andthelistexchangeintheU.S.occurredin1996.15Thefitrading"oflivingdonorkidneysviaexchangegainedtractionwithinthepastdecade,beginningwiththe2003legalopinionfromUNOSthatkidneyexchangesdonotviolateNOTAandthepassageofHR707inDecemberof2007thatchangedthelawtoexplicitlyallowkidneyexchange.Intheearlyyears,participatingcentersfoundmatchesmanually-lookingatmedicalchartsandmatchingpatientstodonorsbyhand(O'BrienandKellan,2012;Hantoetal.,2010).In2005,theNewEnglandPairedKidneyExchange(NEPKE)beganusingfiacomputeroptimizedmatchingprogramdevelopedbyRoth,Ünver,andSönmez,"wheretwo-andthree-waymatcheswerefiincludingclosednon-directeddonor(NDD)andlistexchangechains"(Hantoetal.,2010).Withcomputerizedmatchingcamesizablegainsintransplantsviakidneyexchange.Figure14"OurMatchingAlgorithm",http://paireddonation.org/our-algorithm/,accessed10/14/2015.15BasedonOPTNdataasof9/18/2015.Thepairedexchangein1994wasanisolatedevent.Nonewereobservedagainuntil2000.7A.2highlightsthegrowthinpopularityofpairedandlistexchangearoundthistimeperiod(2005-2007).FigureA.4reinforcesthis,showingrapidgrowthinthenumberoftransplantcentersthatperformedatleastonepairedexchange,listexchange,orbothinagivenyear.By2014,pairedexchangesaccountedfor9.9percentofalllivingkidneytransplantsandlistexchangesaccountedfor2.1percent.16Thelongestreporteddonorchaintodateincluded68people-34patientsand34donors.171.1.2LiteratureReviewPreviousresearchineconomicslargelyfocusesonthefactorsdeceaseddonorkidneysupply,andhowchangesinthesupplyofdeceaseddonorkidneysaffectlivingkidneydonations.Infact,thisworkmayhelpexplainsomeoftheobserveddeclineindirectlivingdonationsfrom2005to2014.Dickert-Conlinetal.(2011)showthatthesupplyofdeceaseddonorkidneysisresponsivetochangesinmotorcyclehelmetlaws.Follow-upworkbyDickert-Conlinetal.(Chapter3ofthisdissertation)andFernandezetal.(2013)showthatdeceaseddonororgansupplyshockscauseadecreaseinlivingorgandonations.Usingalternativestrategies,Sweeney(2010)andAnderson(2016)alsothatincreasesinthesupplyofdeceaseddonororganscrowdoutlivingdonations.Anotherstrandofresearchineconomicsfocusesontheeffectofincentivesonlivingdonations.Laceteraetal.(2014)noeffectofstate-sponsoredtaxandpaidleaveincentivesonlivingkidneydonation,thoughtheydopositiveeffectsforbonemarrowdonation.However,Schnieretal.(2014)thattransplantcentereffortstoreimbursedonors'travel-relatedexpensesincreaselivingkidneydonations.Untilnow,researchineconomicsconcerningkidneyexchangeislimitedtotheoreticalworkonmatchingandthesimulationofpatientoutcomes.ArepresentativesetofsimulationsfromRoth16BasedonOPTNdataasof7/10/2015.17"LongestKidneyChainEverCompletedWrapsupatUWHospitalandClinics,"accessed9/10/2015.http://www.uwhealth.org/news/longest-kidney-chain-ever-completed-wraps-up-at-uw-hospital-and-clinics/455498etal.(2004)suggeststhatoneadditionaltransplantviaexchangereducesdirectlivingdonationsby0.48,18whichimpliesthat5.2ofevery10transplantsviakidneyexchangerepresentnewlivingdonortransplantsthatwouldnothaveoccurredintheabsenceofkidneyexchange.Therepresenta-tivesetofsimulationsalsoyieldsapost-introductionreductionofaverageHLAmismatchesfrom4.83to3.85forthosereceivingalivingdonortransplant.Theseresultsarebasedontheassumptionthatpatientsareficautious,"meaningthatpatientswillonlyenteranexchangeiftheirowndonorisincompatibleorifthematchwillreducethenumberofHLAmismatchesbyatleastonecom-paredtotheircompatibledonor.Ifinsteadpatientsareassumedtobefirational,"meaningthattheyonlycareaboutminimizingthenumberofHLAmismatchesanddonorage,morepatientsswitchfromdirectlivingdonorstoexchangeandthenumberofadditionallivingdonationsforevery10exchangesdecreasesto4.4andtheaveragenumberofHLAmismatchesdecreasesto3.71.However,usingthesimulatedresultstoestimatecrowd-outandmatchqualityimprovementignoressomeimportantconsiderations.First,thesimulationsuseaedsetofpatient-donorpairs,meaningthattheydonotallowforpossiblechangesinthenumberandcompositionofpatient-donorpairsinresponsetotheintroductionofexchange.Second,thoseswitchingfromcompatiblelivingdonationtoexchangeinthesimulationsaredrivenonlybyimprovementsinthenumberofHLAmismatchesanddonorage.Inreality,manyotherfactorsmayaffectthedecisiontosubstitutebetweenthesetwomethods,includingconcernfordonorwell-being.Inthetransplantationliterature,Massieetal.(2013)estimatethepotentialutilizationofkidney18Theresearcherssimulateedpoolsof30,100,and300unrelatedpatient/donorpairsran-domlygeneratedtocloselyOPTNpopulationstatistics.Theyassumethatpatients'pref-erencesaredeterminedbymaximizingtheprobabilityofasuccessfultransplant,givencertainconstraints.Therepresentativesimulationscitedhereuse100pairsandarebasedontheas-sumptionthat40%ofpatientswouldpreferwaitinglistprioritytotheirincompatiblewillingdonor'skidney,whichallowsforthepossibilityoflistexchange.ThecalculationIuseis:(OwnDonorTXsWithExchangeOwnDonorTXsWithoutExchangeExchanges),wherethetotalnumberofowndonortransplantsis22.81withexchangeand54.79withoutexchange,andthenumberofex-changesis66.16.Thiscalculationgivesusacrowdoutestimateof31:9866:16or-0.48,andyieldsthefollowinginterpretation:ofthe66.16exchangesperformed,31.98(or4.8outof10)ofthepatientsinvolvedwouldhavereceivedalivingkidneydonationotherwiseand66.16-31.98=34.18wouldnot.SeeTable3ofRothetal.(2004)forthebasenumbers.9exchange(combiningpairedandlistexchange)intheUnitedStatesifcentersachievedseveraldifferentthresholdsofexchangeutilization.19However,theircalculatedhypotheticalgainsinexchangesdonotnecessarilytranslateintoequivalentgainsinthenumberoflivingdonorkidneytransplants.Astheauthorsthemselvesnote,fisomeparticipantsinKPD[kidneypaireddonation]mayhaveexchangedakidneyfromacompatibleolderdonorforayoungerdonorkidney;othersmayhaveeventuallylocatedacompatiblelivedonorwithoutKPD,orundergonedesensitization."TheyexpressthebelieffithatmostKPDrecipientswouldotherwisenothavereceivedLDKT[livingdonorkidneytransplant],andthatinferencesaboutdisseminationofthismodalitywouldnotbebiasedbytheseissues,"butdonotprovideevidenceforthisclaim.Thereexistsadditionaldescriptiveresearchonsurvivalcomparisonsbetweenexchangeanddirectlivingtransplantsinthetransplantationliterature.Segevetal.(2008)nostatisticallydifferenceingraftsurvivalratesbetweenkidneyexchangerecipientsanddirectlivingrecipients,evenwhencontrollingforobservables.InareviewofseveralexchangeprogramsintheUnitedStates,Mierzejewskaetal.(2013)reportthesamebasedoncomparisonsofmeansurvival,despiteexchangerecipientsbeingmoresensitizedonaverage.ThisisconsistentwiththeofDelmonico(2004)andGjertsonandCecka(2000)thatthenumberofHLAmismatcheshaslittletonoeffectongraftsurvivalofacompatiblekidney,butgoesagainsttheearlierofOpelz(1997).Thesesurvivalcomparisonsimplynon-negativechangesingraftsurvivalforthosesubstitutingtowardkidneyexchangeandawayfromadirectlivingtransplant,andanincreaseingraftsurvivalforthosesubstitutingawayfromadeceaseddonortransplant.Thatsaid,thecomparisonsdonotprovidecausalestimates.Thelimitedexistingliteratureontransplantgainsandimprovementsinrecipientoutcomesinvitesfurtherresearchaddressingthesequestions.19Theauthorsestimateasetofnegativebinomialregressionoftotalexchanges,pairedandlist,onasetofcontrolsforallcenters,thetop20percentintermsofexchangeutilization,andthetop10percent.Thesecontrolsincludecenterwaitlistsize,proportionofwaitlistedpatientsthatreceivedadirectedlivingdonation,distributionofwaitlistedpatients'race,education,age,insurance,dialysisstatus,andPRA.Theygeneratethepredictednumberofexchangesforeachtransplantcenterbasedontheestimatedparametersforeachutilizationlevel.Thentheysumovercenters'max(observedexchanges,predictedexchanges)foreachutilizationlevel.101.2ConceptualFrameworkThissectiondevelopstheconceptualframeworktoguidetheempiricalworkofthispaper.Iaskwhethertheintroductionofexchangeincreasesthenumberoftransplants,andwhetherpatientsalsoexperienceincreasedgraftsurvival,increasedHLAmatchquality,andreducedwaitingtimes.Themarketfortransplantablekidneysisuniqueinthatitrevolvesaroundtheallocationoflife-savingscarceresourceswithouttheexistenceofaformal/legalpricemechanism.Thismarketreliesonindividualsmakingvoluntarycontributionsupontheirdeathorwhiletheyarealive,andlivingdonorsoftenreceivelittletonocompensationatallforthecoststheyincur.Duetothelackofapricemechanism,itisincrediblydiftoanalyzetheintroductionofexchangeasonemightanalyzetheintroductionofanewproduct.Thisismadeevenmoredifbythefactthatpatientscannotfreelychoosefromamongalloftheavailableoptionsatanygiventime;theirchoicedependsinlargepartonthechoicesandcompatibilityofpotentialdonorsandtheavailabilityofdeceaseddonorkidneys.Theintroductionofkidneyexchangeunlocksasetofpotentialdonorsavailabletoapatientinneed.Apotentiallivingdonormustonlybewillingandeligibletodonate;thecompatibilityofdonorandintendedrecipientisnolongerrequired.Theextenttowhichthispoolofpotentialdonorsexpandsreliesontheandcostsinvolvedingivingandreceivingakidneyviaexchange;theseareprimarilydeterminedbyaccesstoaparticipatingexchangeprogramandthethicknessofthemarket-thenumberofotheraccessibleincompatiblepairslookingtoparticipateinanexchange.Asthepoolofpotentialdonorsforanygivenpatientexpands,thelikelihoodofasuit-ablelivingdonorincreases.Thisexpansiongivesthepatientinneedadditionalchoiceswhenseekingoutalivingdonor,whichcouldincreasethepatient'sutilitythroughincreasedlikelihoodofreceivingakidney,improvedmatchquality,improvedoverallwell-beingamongpotentialandactualdonors,andreducedwaitingtime.Atthesametime,astheoptionsavailabletoagivenpa-tientimprove,thereisgreaterincentiveforpotentialdonorstofreeride.Whileapatientmayit11moretoseekoutatransplantviaexchangeinsteadofacompatiblelivingdonororwaitingforadeceaseddonor,apotentialcompatibledonorwhowouldhavegivenintheabsenceofexchangemaynotbewillingtogivewhencompatibilityisnolongerastrictrequirement.Considerasimplemodelofthepotentialdonor,i's,decisiontogivetoanindividual,j.Thepo-tentialdonorcaresaboutherownconsumptionaswellasthepatient'stransplantoutcome.Patientj'sexpectedtransplantoutcomeisgivenbyE[Qj(Di;Z)],whereQjrepresentsthequalityoftrans-plantforpatientj.Kidneyqualityvariesforeachperson,andthereisadistributionofqualitygivenbyF(Q).Theexpectedtransplantoutcomeofpatientjisafunctionofthemodebywhichigivestoj,Di,andabinaryvariableZforwhetherkidneyexchangeisanoption.Thepotentialdonoralsocaresaboutthewell-beingofotherpatientsinneed.Giventhissetup,supposei'sgeneralutilityisgivenbythefollowing:Ui=Xi+E[Qj(Di;Z)]+aiE[Qj(Di;Z)](1.1)whereXiisacompositegood,aiisanon-negativedonoraltruismparameter,andE[Qj(Di;Z)]representstheoutcomesofotherpatientsinneedoftransplants.SupposeihaswealthWiandthecostofidonatingtojisgivenbyC(Di;Z).Thecostofdonatingmayincludethingssuchastravelcosts,uncoveredmedicalexpenses,andtimeoffofwork.PotentialdonorifacesthebudgetconstraintXi+C(Di;Z)=Wi.Supposethatthecostofdonatingis0whenidoesnotdonate,implyingthatXi=Wiwhenidoesnotdonate.Supposefurtherthat,whengivingdirectly,qualityisknownandqualityandcostsareindependentofexchangeprevalence.Supposethatqualityisalsoknownwhengivingviaexchange,andthatthecostofgivingviaexchangeishighwhenexchangehasyettobeintroduced.Then,donori'sutilitywhenshedonatesdirectlyisgivenby:Ui(Direct)=[WiC(Direct;Z)]+Qj(Direct;Z)+aiE[Qj(Direct;Z)](1.2)Whenshegivesviaexchange,herutilityis:Ui(Exch)=[WiC(Exch;Z)]+Qj(Exch;Z)+aiE[Qj(Exch;Z)](1.3)12Whenshedoesnotgiveatall,herutilityis:Ui(None)=Wi+E[Qj(None;Z)]+aiE[Qj(None;Z)](1.4)Thereexistsaqualitythreshhold,Q,suchthatiwillbeindifferentbetweendonatingandnotdonating.ForvaluesofQ<Q,iwillchoosenottodonatetoj.Proposition1.Whenexchangeisintroduced,thequalitythresholdQwhereiisindifferentbe-tweendonatingdirectlyandnotdonatingincreases.Assuch,Prob[Ui(Direct)Ui(None)]willdecreasemeaningthatfewerindividualswilldonatedirectly.Proof.SeeAppendixD.1.Thisresultfollowsinpartfromtheassumptionthattheintroductionofexchangedoesnotaffectthecostofdonatingdirectly,northequalityofthedirecttransplant.Ialsoassumethattheexpectedoutcomesofotherpatientsimprovebythesameamountwiththeintroductionofexchangewhenidonatesdirectlyandwhenidoesnotdonateatall.Thus,sincetheexpectedoutcomeofpatientjwhenidoesnotdonateimproveswiththeintroductionofexchange,i'sdirectdonationmusthaveahigherbaselinequalityinordertojustifythedecisiontodonate.Itistrivialtoshowthattheprobabilitythatichoosestogiveviaexchangeratherthandirectlyornotatallweaklyincreaseswhenexchangeisintroduced,giventhatthecostofdonatingviaexchangeisassumedtobehighwhenexchangehasyettobeintroduced.Potentialdonoriwilldonateviaexchangeratherthannotatalliftheweightedsumoftheimprovementinj'soutcomeandtheoutcomesofotherpatientsinneedoutweighsthecostofdonatingviaexchange.Potentialdonoriwilldonateviaexchangeratherthandirectlyiftheweightedsumofimprovementsinpatientoutcomesoutweighsthecostdifferentialofdonatingviaexchangecomparedtodirectly.Wearemorelikelytoobservethissecondoutcomewheniisarelativelypoormatchforj.Asthecostdifferentialapproacheszero,iwillchooseexchangeratherthandirectdonationifthereisanygainintheweightedsumofpatientoutcomesfromdonatingviaexchange.Dependingoni's13altruismtowardothersandtheamountofgeneratedforotherpatientsinneed,imayprefertodonateviaexchangeratherthandirectlyeventhoughitmightmeanaworseoutcomeforj.Themodelcanbeextendedtoaccountforincreasesinanonymousdonationsaswell.Sup-poseanindividualdoesnothavealovedoneinneedofakidneybutcaresaboutthenumberoftransplantsobtainedbypatients,suchthatshewouldgetthefollowingutilityfromdonatinganonymously:Ui(Anon)=[WiC(Anon;Z)]+aiE[Qk(Anon;Z)](1.5)Comparethistoherutilitywhenshedoesnotdonate:Ui(None)=Wi+aiE[Qk(None;Z)](1.6)Here,Iindexpatientsusingk,ratherthanjand-j,sincetheanonymousdonorisnotdonatingonbehalfofaparticularpatientj.Proposition2.Whenexchangeisintroduced,thequalitythresholdQwhereiisindifferentbe-tweendonatinganonymouslyandnotdonatingchangesambiguously.Individualswillbemorelikelytodonateanonymouslytostartadonorchain,andlesslikelytodonateanonymouslytoasingleindividual.Proof.SeeAppendixD.2.Theintuitionforthispropositionisasfollows.Individualswillbelesslikelytodonateanony-mouslytoasingleindividualcomparedtonotatallformuchthesamereasonastheywouldbelesslikelytodonatedirectly.Inthiscase,theirdonationisnothelpingimprovetheoutcomesofotherpatients.Moreover,theoutcomeofthepatientreceivingi'skidneyshouldnotbeaffectedbytheexistenceofexchange.However,thepatient'sexpectedoutcomewhenidoesnotdonateimproveswiththeintroductionofexchange.Therefore,iwillbelesslikelytogivedirectlytoasingleindividual.Wheniisdonatinganonymouslyinordertostartadonorchain,thesurplusgeneratedbyhelpingevenoneadditionalpersonreceiveatransplantviaexchangeisverylikely14tooutweightheimprovementoftheinitialpatient'sexpectedoutcomewhenidoesnotdonateandexchangehasyettobeintroduced.Finally,weshouldnotexpecttoseeanychangeindeceaseddonortransplantsaslongasde-ceaseddonationratesdonotrespondtokidneyexchanges.Suchachangewouldrequireachangeinhowmanypeopleregisterasdonors,thenumberoffamiliesthatgiveconsentonceapotentialdonordies,and/orachangeinthenumberofrecoveredkidneysdeemedtransplantable.Noneoftheseseemlikelygiventhesizeofthewaitinglistandthatexchangeisstillasmallshareoftotaltransplantsdespiteitsrecentgrowthinutilization.Additionally,becausethewaitinglistissolong,thereshouldalwaysbesomeonewillingandabletoacceptasuitabledeceaseddonorkidney.Wecouldseetheallocationofdeceaseddonororgansnaturallyshifttowardareaswithfewerpairedexchanges,butthisshouldhavenoeffectontheoverallnumberofdeceaseddonortransplants.1.3DataInthefollowinganalysis,Ibeginwithregistration-leveldataasofDecember31,2014fromtheStandardTransplantAnalysisandResearch(STAR)whichisavailablebyrequestfromOPTN.Thedatacontain788,106observationsofkidneywaitinglistregistrationsandtransplantsthatoccurredfrom1988to2014.Oftheseobservations,398,984areregistrationsthatresultedintransplantsandtransplantsthatoccurredwithoutanassociatedwaitinglistregistration.20Ihaveinformationontheoutcomeofeachregistrationincluding:transplant,death,transfertoadiffer-entcenter,orstillwaitingasofDecember31,2014.Irestrictmyanalysistoobservationsthatresultedineitheratransplantordeath,astheseencompasstheclearandregistrationoutcomes.2120Somerecipientsoflivingdonorkidneysneverregisterforthedeceaseddonorwaitinglist.Theseobservationsareenteredasatransplantrecordandlacktheinformationthatisonlyrelevantwhenapatientregistersonthewaitinglist.21Notethatpatientsmayhavemultiplewaitinglistregistrations.Patientswithmultipleregistra-tionswhodiedwhilewaitingareonlycountedonce,andIusetheirearliestlistingcodedassuch.Fortheendingdateoftheseobservations,Iusetheearliestreportedregistrationenddateamongthosecodedasremovalduetodeath.Forpatientswhoreceiveatransplant,onlyoneregistration15DonorrelationisthekeyvariableIuseforthosewhoreceiveatransplantinordertodeterminewhetheritisadirectliving,deceased,anonymous,orexchangetransplant.Kidneytransplantsarecodedwithoneofthefollowingdonorrelationships:sibling,twin,child,parent,otherrela-tive,other,miscellaneousunrelateddonor,pairedexchange,listexchange,anonymous,ordeceased.22Thesedataarerich;Iobservevariablesincludingbloodtype,PRA,race,educa-tion,previoustransplantstatus,age,gender,registrationdate,transplantdate,HLAmismatches,additionalmedicalinformation,donorcharacteristics,andtransplantfollow-upinformation(fromwhichgraftsurvivaliscalculated).Additionally,Iobtainedzipcodeinformationforpatientsandtransplantcenters.Irestrictmyanalysistodatafrom2000toJuly2014.Thereisalargeincreaseinthequalityofreportingforthedonorrelationshipvariablein2000.From1988to1999,thereareanaverageof77unreporteddonorrelationshipsperyearcomparedtooneperyearfrom2000to2013.Thedatashowonlyonepairedexchangeand8listexchangesoccurringbefore2000,sothereislittlereasontobeconcernedabouttheexclusionofpre-2000data.Duetolagsindataprocessing,AugustthroughDecember2014areincompletewithrespecttodonorrelationshipatthetimeofmyextract.TableA.1showsthefrequencyofobservedtransplantoutcomesbydonorrelationshipinclud-ing:directedliving,pairedandlistexchange,anonymousdonation,deceaseddonation,anddeathwithoutatransplant.Weseethatexchangesaccountfor1.5percentofobservedoutcomes,anony-mousdonationsaccountfor0.5percent,directedlivingdonationsaccountfor28percent,deceaseddonationsaccountfor49percent,anddeathswhilewaitingaccountfor21percent.Almost26percentofdirectliving,6percentofexchange,and3percentofanonymoustransplantsoccurredwithoutanassociatedwaitinglistregistration.23Comparingthe2013snapshot,thelastcompletewillshowthetransplantoutcome,whichistheobservationIuse.22Note:Icanonlyconnectdonorstotheiractualrecipients.Therefore,withrespecttoex-changes,Icannotconnectdonorstothelovedoneonwhosebehalftheyaredonating.Also,Icannotobservewhetherananonymousdonor'skidneyisusedtostartadonorchain.23BasedonOPTNdataasof12/31/2014.16yearofdata,tothe2007snapshot,weseethattransplantsviadeceaseddonation,exchange,andanonymousdonationincreasedinlevelandpercentofoutcomes.Wealsoseeadecreaseindirectedlivingdonationsandaslightdecreaseindeathswhilewaiting.TableA.2presentsthefractionofkidneygraftsthatsurviveatleastoneyear,atleasttwoyears,thenumberofHLAmismatchesbetweendonorandrecipient,andthedurationofthewaitinglistregistration.Weseethatgraftsurvivalisslightlyhigherforexchangesbutverysimilaracrossalllivingdonortransplantsforbothoneandtwoyearsat96-97percentand93-94percent,respec-tively.24One-andtwo-yeargraftsurvivalislowerforrecipientsofdeceaseddonorkidneysat90percentand86percent,respectively.LookingatthenumberofHLAmismatchesbyobservedoutcome,weseethatdirectlivingdonationshavethelowestnumberofHLAmismatchesat3.16followedbydeceaseddonationsat3.88.ThereisnodifferenceintheaveragenumberofHLAmismatchesacrossexchangesandanonymousdonationsat4.31.Recipientsofdirectlivingdona-tionshavetheshortestwaitinglistregistrationdurations,roughly237days,followedbyexchangeat468days,anonymousat654days,anddeceaseddonationat813days.25Bothregistrationdura-tionandnumberofHLAmismatchesappearuncorrelatedwithgraftsurvivalamonglivingdonorkidneyrecipientsafteroneandtwoyears.TableA.3presentsthreedifferentmeasuresofmeanPRA,26patientageatthetimeoflisting,24Graftsurvivalisforobservationswithanon-missinggraftsurvivaltimeintheOPTNdata.Ittakesonavalueof0ifthepatientdiedbeforeoneortwoyears,orifthereisareportedgraftfailurewithinoneortwoyears.Ittakesonavalueof1ifthegraftsurvivaltimeexceedsoneortwoyears,orifthepatients'lastknownstatusisalivewithafunctioningkidney.Theassumptionisthatthesefilost"individualswouldhavereturnedtothesystemortheywouldhaveadeathdatereportedthroughtheSocialSecurityDeathMasterFileiftheirgraftfailedortheydied.Inordertoensureadequatetimehaspassedforfollow-updata,Irestricttheanalysisofone-yeargraftsurvivaltotransplantsoccurringonorbeforeDecember31,2012,andDecember31,2011fortwo-yeargraftsurvival.25Note:roughly1/3oflivingdonorkidneyrecipientsneverregisterforthewaitinglist.Inthesecases,registrationdurationissettozero.26RecallthatthePRAscoretakesonavalueof0to100,andisinterpretedtobethepercentageofthegeneralpopulationwithwhomthepatientislikelytobeincompatible.InformationonmostrecentPRA,bothClassIandClassIIwhichmeasuredifferentHLAantibodies,onlydatesbackto2004duetochangesinPRAusageanddatacollection.InformationonendingcalculatedPRA(C-PRA)beginsin2007.C-PRAisnowthedefaultmeasureusedfordeceaseddonorkidney17anddonorage.PatientswhodiewhilewaitingforakidneyhavethehighestPRAscoresonaverageat23(C-PRA),followedbyexchangeat22,deceaseddonationat20,anonymousdonationat17,andrecipientsofdirectlivingkidneysatamuchlower9.Thissuggeststhatthosereceivingexchangesareex-antehardertomatch.Itispossiblethatexchangesarepursuedaftertransplantcandidatesexploreand/orexhaustmoreconventionaloptions.Recipientsofdeceaseddonationsandexchangestendtobeolderatthetimeofregistrationat47-48yearsoldcomparedtodirectlivingandanonymousrecipientsat44-45,andsurpassedonlybypatientswhoeventuallydieonthewaitinglistwhoregisterat53onaverage.Donorsoftransplantsviaexchangeandanonymousdonationaretheoldestonaverageat43-44yearsold,followedbydirectlivingdonorsanddeceaseddonorsat37-40,whichreinforcesthenotionthatexchangerecipientsareinamoredesperatepositioncomparedtothosereceivingdirectlivingdonations.Iftrue,thissuggeststhattheremaynotbemuchsubstitutionoccurringbetweendirectlivingdonationsandexchanges.1.4EstimationThegoalofthispaperistocausallyestimatethenumberofadditionaltransplantsgeneratedbykidneyexchanges.Tothisend,Iestimatetheeffectofanincreaseintheprobabilityofreceivinganexchangetransplantontheprobabilityofexperiencingoneoftheotherregistrationoutcomes.Iwillrefertotheseasthecrowd-outestimates.Irestrictestimationtoobservationscorrespondingtooneofthefollowingveoutcomes:directliving,anonymous,exchange,ordeceaseddonortransplant,ordeathonthewaitinglist.Eachobservationintheestimationsampletakesonavalueof1foroneandonlyoneofthevecorrespondingbinaryvariables.Usingthismodel,ratherthanadurationmodelinwhichIcouldincludepatientsstillwaiting,allowsmetoaccountfortheroughly33percentofpatientswhoreceivelivingdonortransplantswithoutregisteringandavoidproblemscausedbyselectionbiasinwhoregistersonthewaitinglistbeforetransplantordeath.Iuseatwo-stageleastsquaresapproachtoobtaintheparametersofinterest.Themostimportantallocation(Cecka,2010).18reasonfordoingsoisthatregressinganon-exchangebinaryregistrationoutcomeindicatorontheindicatorfortransplantviaexchangeyieldsestimatescompletelydrivenbytheobservedsharesofregistrationoutcomes,andtheestimatesmathematicallycannotbegreaterthanzero.Therearealsopossiblereversecausalityandselectionissues;thosewhoaresickerandlesslikelytoalivingdonortransplantmaybemorelikelytoseekoutexchangesandmayalsoexperienceworsetransplantqualityoutcomes.Considerthefollowingsetofreducedformindexedbyrforeachregistrationoutcomeofinterest:Yrizt=qrActivityzt+Xibr+arz+grt+erizt(1.7)whereYrepresentstheregistrationoutcomeofinterestforindividualiinzipcodezinmontht.Activityisameasureoflocalexchangeactivitytozipcodezinthemonthofoutcomet.Xisavectorofobservablesforpatienti,includingrace,gender,bloodtype,education,previoustransplantstatus,PRAscore,andageatlisting.27Iincludeazipcodeedeffect,arz,tocontrolforanyunobservedheterogeneityacrosszipcodeswherepatientslivethatarecorrelatedwithlocalkidneyexchangeactivity.Suchfactorsmightincludeaverageafqualityofnearbyhealthcareinstitutions,andproximitytohighereducationorotherresearchinstitutions.Ialsoincludeamonth-yearedeffect,grt,tocontrolfornationwidetransplantationtrendsandnational-levelpolicyshocks.eriztistheidiosyncraticerrorterm.qrcapturestheeffectoflocalexchangeactivityonYr,ortheeffectontheprobabilityofexpe-riencingtheregistrationoutcomecorrespondingtoYrconditionalonexperiencinganytransplantordeathoutcome.ThemainfocusoftheestimationresultswillbetheratioofqkqE,whereqkisthe27PRAisincludedasacategoricalvariable.First,IusethemaximumreportedPRAscoreamongseveralavailablemeasuresincludingcurrent,mostrecent,peak,andendingPRAscores.Ithenconvertthisvalueintoa0ifitisbetween0and10,1ifitisbetween10and20,...,and9ifitisbetween90and100.Then,IconcatenatethiswithanindexvaluecorrespondingtothePRAvariablefromwhichthemaximumPRAscoreisobtained.So,forexample,avalueof21forthecategoricalvariablewouldindicatethattheindividual'smaximumPRAscorecomesfromthesecondPRAvariableandisbetween10and20.ObservationsmissingaPRAscorearegiventheirowncategorysothattheyarenotexcludedfromtheanalysis.19effectofActivityonanynon-exchangeoutcomekandqEistheeffectofActivityontheprobabilityofobservingatransplantviaexchange.Aonepercentagepointincreaseintheprobabilityofre-ceivinganexchangetransplantresultsinchangeofqkqEintheprobabilityofexperiencingoutcomek;thesetoftheseratiosyieldthecrowd-outestimatesofinterest.Iestimatetheratiosofinterestdirectlyusingtwo-stageleastsquares,wherethestageis:YEizt=qEActivityzt+XibE+aEz+gEt+eEizt(1.8)Thesecondstageis:Ykizt=lk^YEizt+Xibk+akz+gkt+ekizt(1.9)whereYkiztisthenon-exchangeoutcomevariable,^YEiztisthepredictedprobabilityofobservingatransplantviaexchangeobtainedfromestimatingthestageand^lkissimply^qk^qE.Next,Iestimatehowanincreaseintheprobabilityofreceivinganexchangeaffectsthequalityoftransplantoutcomes,conditionalonreceivingatransplant.Iusethesamemethodasoutlinedabove.ThestageisthesameasEquation(1.8).Thesecondstageis:Yqizt=lq^YEizt+Xibk+aqz+gqt+eqizt(1.10)whereqnowindexesthefollowingqualityoutcomes:one-yeargraftsurvival,two-yeargraftsur-vival,thenumberofHLAmismatches,andwaitinglistregistrationduration.AonepercentagepointincreaseintheprobabilityofreceivinganexchangetransplantresultsinchangeofqqqE=lqinqualityoutcomeYq.,aonepercentagepointincreaseintheprobabilityofreceivinganexchangetransplantresultsinachangeofq1yrqE=l1yrpercentagepointsintheprobabilityofexpe-riencingone-yeargraftsurvival,andsimilarlyfortwo-yeargraftsurvival.AonepercentagepointincreaseintheprobabilityofreceivinganexchangetransplantresultsinachangeofqHLAqE=lHLAinthenumberofHLAmismatchesofatransplant.Finally,aonepercentagepointincreaseintheprobabilityofreceivinganexchangetransplantresultsinachangeofqdurationqE=ldurationinreg-istrationdurationmeasuredindays.201.4.1MeasuringActivityTheestimationstrategyexploitsvariationinkidneyexchangeactivityacrosslocationandtimeinordertoestimatetheeffectofkidneyexchangesonpatientoutcomes.Activityismeasuredusingthenumberofexchangesthatoccurredattransplantcenterswithin50milesofapatient'szipcodeofresidenceinthemonthofthepatient'stransplantordeath.IadjustthismeasuredownwardbyoneifIobservethatthegivenpatientreceivedanexchangewithin50miles.28Thismeasureboththepotentialoflocaltransplantcenterstoperformexchangesandtherealizationofthatpotential.Sincepatientsanddonorsmustbeabletotraveltotransplantcentersfortestingandeventualtransplantprocedures,Activitypatientaccesstoexchangeand,consequently,iscorrelatedwiththeprobabilityofapatientreceivinganexchangetransplant.Iuse50milesastheradiusbasedonthepercentilespresentedinTableA.4.Mostpatientswhoreceivetransplantsdosowithin50milesoftheirhomezipcode-between60and70percentoverall.29FigureA.5showstheaveragemonthlynumberofexchangesperformedwithin50milesforindividualsinthesampleovertimeandobservedregistrationoutcome.Weseethatthisnumberrangesfrom0to2.75,andtendstobehigherforrecipientsofexchangesandanonymousdonations.Tobevalid,Activitymustbeexogenoustothedependentvariablesofinterest;itcannotaffectnon-exchangeoutcomesdirectlyorthroughanomittedvariable,norhaveareversecausalrela-tionshipwiththethenon-exchangeoutcomes.Theinclusionofzipcodeandmonthedeffectscontrolforanynationaltrendsandtime-invariantdifferencesacrosslocationsinthedependentvariables.Themainthreatstotheexogeneityoftheinstrumentarewhethertransplantcentersadoptandpromoteexchangeasatransplantoptioninresponsetolocaldemandforexchangeandidiosyncraticpre-trendsinthedependentvariablesofinterest.Forexample,wemaybeconcerned28IuseGISmappingsoftwarealongwiththezipcodesofpatientsandtransplantcenterstodeterminewhichtransplantcentersarewithin50milesofthecentroidofeachobservedpatientzipcode.Ithenaggregateoverthesenearbycenterstodeterminehowmanytransplantsviakidneyexchangeoccurredeachmonthwithinthe50mileradius.29ItesttherobustnessofthefollowinganalysistotheuseofdifferentradiiandpresenttheresultsinSection1.6.21thattransplantcentersinareasexperiencingworseningoutcomes,suchasdecliningdirectlivingdonations,aremorelikelytoadoptandpromoteexchange.However,thepresenceofafichampion"-aparticularleaderorsmallgroupofindividualsworkingatagiventransplantcenterwhowanttoimplementthenewmethodology-appearstobethedrivingforcebehindeffectiveexchangeadoptionandpromotion.Forexample,GaretHilcreatedtheNationalKidneyRegistry,apairedexchangeconsortium,afterhisdaughterhadtoendureafidifandextensivedonorsearch."30Inadditiontohavingachampion,centerswithactiveexchangeprogramsappeartobethosewiththeresourcesnecessaryforsuchanundertaking.ConversationswithAlanLeichtman,aprominentnephrologistwhoinitiatedthepairedkidneydonationprogramattheUniversityofMichiganin2008,31supporttheimportanceofthesetwocriteria.Universityhospitalsarelikelycandidatestohaveachampionandtheresourcesneeded;13ofthe20mostactivekidneyexchangecentersareatuniversitymedicalcentersbasedonthenumberofexchangesperformedbetween2000andJuly2014.ThemostactivecentersincludeJohnsHopkinsHospital,MethodistSpecialtyandTransplantHospital,UniversityofMichiganMedicalCenter,NorthwesternMemorialHospital,UniversityofMarylandMedicalSystem,andUCLAMedicalCenter.32Includingzipcodeedeffectsalleviatestheconcernthatcenterswithsufresourcestostartorparticipateinexchangeprogramstendtobeinareasthatexperiencesystematicdifferencesinpatientoutcomes.Recallthatexchangeappearstobeatransplantoptionpeoplepursueafterrealizingtheirweakprospectsofatimelydeceaseddonationordirectlivingdonation.Iftransplantcentersadoptandpromoteexchangesinresponsetolocaldemand,wewouldexpecttoseeworseningoutcomesleadtoincreasesinfutureexchangeactivity.ItestthisandpresenttheresultsinSection1.6.Decliningratesindeathswhilewaiting,whichisconsideredanimprovingoutcome,arecorrelated30NationalKidneyRegistry,AboutUs,accessed10/27/2015.http://www.kidneyregistry.org/about_us.php.31http://med.umich.edu/intmed/nephrology/news/index.htm,accessed8/26/2015.32BasedonOPTNdataasof12/31/2014.22withincreasesinfutureexchangeactivity.However,Inoevidencethatworseningoutcomesarecorrelatedwithfutureexchangeactivity.1.5ResultsTableA.5presentstheestimatesofhowtheprobabilityofreceivingakidneyexchangeaffectsone'sprobabilityofreceivingadirectliving,anonymous,ordeceaseddonorkidney,aswellasdeathonthewaitinglist,conditionalonexperiencingoneoftheseveoutcomes.Thecolumnpresentsthestageresults,whichcorrespondto^qEintheprevioussection:theeffectofnearbyexchangeactivityinthemonthofapatient'sregistrationoutcomeontheprobabilitythatthepatientreceivesakidneyviaexchange.Theestimateof^qEishighlyfurthertheimportanceofproximitytoexchangeactivitywithrespecttoobtainingatransplantviaexchange.Thesecondthroughcolumnspresentthesubstitutionestimates,whichcorrespondtothedifferent^lks.Thesecondcolumnshowsstatisticallyreductionof0.5percentagepointsintheprobabilityofreceivingadirectlivingtransplantassociatedwithaonepercentagepointincreaseintheprobabilityofreceivingakidneyexchange.Thisestimateimpliesthat5ofevery10transplantsviaexchangewouldhavebeendirectlivingdonortransplantsintheabsenceofexchange.Thethirdcolumnshowsastatisticallyincreaseof0.12intheprobabilityofreceivingananonymousdonationwithaonepercentagepointincreaseintheprobabilityofreceivinganexchange.Thissuggeststhatforevery10transplantsviaexchange1.2additionalpeopledecidetodonateanonymously.Takentogetherwiththesecondcolumnestimate,thissuggeststhat105+1:2=6:2ofevery10transplantsviaexchangerepresentlivingdonortransplantsthatwouldnothaveoccurredintheabsenceofexchange.Thefourthcolumnshowsthattheprobabilityofreceivingatransplantviaexchangehasastatisticallynegativeeffectof-0.09ontheprobabilityofreceivingadeceaseddonorkidney.AsdiscussedinSection1.2,itislikelythatanyrelationshipbetweenthesetwoprobabilities23isbasedontheallocationofdeceaseddonorkidneysshiftingtoareaswithlesskidneyexchangeprevalence,ratherthanareductioninthenumberofdeceaseddonororgantransplants.Finally,thecolumnshowsastatisticallyrelationshipbetweenexchangeanddeathonthewaitinglist;aonepercentagepointincreaseintheprobabilityofreceivingakidneyexchangedecreasestheprobabilityofdyingonthewaitinglistbya0.54percentagepoints.Thissuggeststhatevery10exchangetransplantsreducedeathsamongthosewaitingby5.4.FromTableA.5,itisclearthatexchangesincreasethenumberoflivingdonortransplants.TableA.6presentswhattheintroductionofexchangemeansforgraftsurvival,thenumberofHLAmismatches,andthedurationofwaitinglistregistrationsendingintransplant.Thecolumnshowsthatatenpercentagepointincreaseintheprobabilityofreceivinganexchangestatisticallyincreasestheprobabilityofone-yeargraftsurvivalby2percentagepoints,or2.2percentrelativetoaverageone-yearsurvivalof93percent.Thesecondcolumnshowsasmallerstatisticallyincreaseintwo-yeargraftsurvival:1.9percentagepoints,or2.2percentrelativetoaveragetwo-yearsurvivalof88percent.ThethirdcolumnshowstheeffectofexchangeintroductiononHLAmismatches,onepossiblemechanismthroughwhichgraftsurvivalcouldimprove(Opelz,1997).TheestimateshowsthattheintroductionofexchangehasapositivebutstatisticallyeffectonthenumberofHLAmismatches.Whileexchangemayenablepatientstoclosermatches,patientsmayalsobewillingtoacceptaslightlyworsematchinsteadofhavingtorelyonacompatiblefamilymember,forexample,whomaybeamuchclosermatch.Moreover,ifmostofthosereceivingkidneyexchangesareharder-to-matchindividualsandwouldnothavereceivedatransplantotherwise,thenwewouldexpecttheirtransplantstoincreasetheoverallaverageofHLAmismatches.Finally,thefourthcolumnshowstheeffectofexchangeintroductiononregistrationduration,anothermechanismthroughwhichgraftsurvivalcouldimprove(Meier-Kriescheetal.,2002).Atenpercentagepointincreaseintheprobabilityofreceivinganexchangereducesregistrationdurationby59.2days,or10percentrelativetotheregistrationdurationaverageof597days.Theresultisclear;patientstransplantsmorequicklywhenexchangesareintroduced.Partofthis24islikelyduetotheincreaseinlivingdonortransplants,wheretheregistrationdurationsaremuchshorterthanthewaitforadeceaseddonorkidney.Itisalsolikelythatreducedsearchfrictionsandreducedexcessdemandfordeceaseddonorkidneyscontributetothiseffect,thoughthereisnocleanwaytoisolatethesedifferentcomponents.1.6RobustnessInthissection,IaddressconcernssurroundingthevalidityofActivityasaninstrumentandestimateotherinordertotesttherobustnessofthetwo-stageleastsquaresresults.First,wemaybeconcernedthattransplantcentersadoptandpromoteexchangeasatransplantoptioninresponsetolocalchangesinthedependentvariables.Second,wemaybeworriedaboutreversecausalityinthestageregression.Sincepairedexchangesofteninvolveatleasttwopatientsbeingtransplantedsimultaneously,sometimesatthesametransplantcenter,apatientreceivinganexchangewillsometimesimplytheoccurrenceofatleastoneotherexchangeinthesamemonthatanearbycenter.Third,theremaybeissuescausedbyreversecausalityintheanonymousdonationregression.Sinceanonymousdonationscanfacilitatedonorchains,anadditionaltransplantviaanonymousdonationmaycauseanincreaseinthelevelofnearbyexchangesperformedinthesameperiod.Fourth,wemaybecuriousaboutthesensitivityoftheresultstousingdifferentmileageradiiindeterminingthelevelofnearbyactivity.Fifth,wemaywonderiftheresultsarerobusttozip-codeaggregation,whichovercomestheweaknessinherentinexcludingthosewhoarestillwaitingforakidneyfromthecrowd-outanalysis.Finally,wemayquestiontherobustnessofthecrowd-outanalysistotheinclusionofpatientsremovedfromthewaitinglistduetobeingtoosicktotransplant,orthosewhorefusedtransplantsofferedtothem.Itestwhethercurrentoutcomemeasuresappeartoaffectfuturelevelsoflocalexchangeactivity.TablesA.7andA.8presenttheresultsfromthedatatoequation(1.7)whilein-cludingseveralleadsandlagsofActivity.Iinclude12monthsofleadsandlags,andconsolidatethemintotwo-monthaveragestohelpsmoothoutthesemeasures.IthenperformjointF-testsof25thestatisticalofalloftheleadsineachDeathonwaitinglististheonlyoutcomeforwhichIevidenceofanoutcomevariableaffectingfuturevaluesofActivity.Theresultsuggeststhatarelativedeclineindeathsonthewaitinglistcorrespondstofutureincreasesinlocalexchangeactivity,andsuggeststhattheestimatedreductionindeathswhilewaitingduetoexchangecouldbeoverstated.However,thisresultistheoppositeofwhatwewouldexpectifcenterswereadoptingandpromotingkidneyexchangesinresponsetoworseningoutcomes.Com-binedwiththeresultsfromthetestsofthequalityoutcomemeasures,itdoesnotappearthatlocalexchangeactivityisendogenoustotrendsinpatientoutcomes.Toaddressconcernsofreversecausalityresultingfromtheuseofcurrent-periodnearbyex-changeactivitymeasures,IreplicatetheresultsfromTablesA.5andA.6usingone-andtwo-monthlagsofnearbyexchangeactivityasinstrumentsinstead.Anonymousdonationsandexchangescanonlyfacilitatecurrentorfuturetransplantsviaexchange,hencetheuseoflagstoavoidthereversecausalityissue.However,lagsalsoreducetheprecisionoftheestimatesofinterest.TablesA.9andA.10presenttheresultsfromthisexercise.Weseethattheestimateofdirectlivingcrowd-outisveryclosetotheoriginal2SLSestimatebutisnowstatisticallytheestimateofanonymousdonorcrowd-inistwicethemagnitudeoftheoriginal,thedeceaseddonorcrowd-outestimateturnspositivebutisstillstatisticallyandthestatisticallyreduc-tionofdeathsonthewaitinglistnearlydoublesinmagnitude.Theestimatesofdirectlivingcrowd-outandanonymouscrowd-inarethemostimportantfordeterminingthechangeintheoverallnumberoftransplants.Together,theseestimatesimplythat5.3ofevery10transplantsviaexchangerepresentanincreasetothenumberoftransplantsper-formed.Thisislowerthanouroriginalestimateof6.2,butthetwoarenotstatisticallydistinguish-ablefromoneanother.Asforthequalityresults,weseethatthegraftsurvivalestimatesdecreaseandthemagnitudesoftheHLAmismatchandregistrationdurationestimatesincrease.Thegraftsurvivaleffectsarenolongerstatisticallythetwo-yeargraftsurvivalestimatebecomesnegative,andtheestimatedeffectonregistrationdurationmorethandoubles.However,theresultsusinglaggedexchangeactivityarestillencouraginginthattheone-yeargraftsurvivalestimateis26stillpositive,thereisstillasmallpositivebutinsigneffectonHLAmismatches,andthereisstillareductioninregistrationduration.Next,ItestwhethertheresultsfromSection1.5aresensitivetotheuseofdifferentmileageradiiindeterminingtheleveloflocalexchangeactivity.TableA.11presentsthecrowd-outresultswhenusinga30mileradiusinsteadof50miles.Theresultsshowasmallreductioninthecrowd-outestimatefordirectlivingtransplants,largercrowd-inofanonymousdonations,largercrowd-outofdeceaseddonortransplantsbutstillstatisticallyandalmostnochangeinthenegativeeffectondeathsonthewaitinglist.ThechangesintheeffectofActivityonanonymousanddeceaseddonationsarethemostnotable,whilethestageestimatebarelychanges.TableA.12presentstheeffectsongraftsurvival,HLAmismatches,andregistrationdurationwhenusinga30mileradius.Theseresultsarevirtuallyidenticalcomparedtotheoriginal2SLSestimates,exceptforaslightattenuationoftheeffectonregistrationduration.Theobserveddifferencessuggestthattheprobabilityofreceivingananonymousdonortransplantismoreheavilybyexchangeactivitywithin30milesthanactivity30to50milesaway.Itispossiblethatthesupplyofdeceaseddonorkidneysismuchmoreresponsivetoimmediatelocalexchangeactivity.However,itismorelikelythattheestimatepicksupregionalshiftsinallocationwhenasmallerradiusisused,whichisamorecoherentexplanationgiventhesimilarityoftheresultsinTableA.12andtheoriginal2SLSresults.TableA.13presentsthecrowd-outresultswhenusinga75mileradiusinsteadof50miles.Again,theseresultsaresimilartotheoriginal2SLSresults.Theeffectofexchangeondirectlivingdonationsbecomesslightlysmallerinmagnitude,andathereisaslightincreaseinthemagnitudeoftheeffectsondeceaseddonortransplantsanddeathsonthewaitinglist.TableA.14presentstheeffectsofexchangethequalitymeasureswhenusinga75mileradius.Again,theseresultsareveryclosetotheoriginal2SLSresults,withslightattenuationintheeffectsongraftsurvivalandregistrationduration.Next,Itestwhethertheoriginal2SLScrowd-outresultsaresensitivetoaggregationtothezipcodemonthlevel.Theoriginal2SLSresultsrelatetheprobabilityofreceivinganexchangetothe27probabilityofexperiencingtheotherregistrationoutcomes.Notethatallthecrowd-outeffectsofexchangeoneachotheroutcomesumto-1whenusingregistration-leveldata;thistheassumptionthateveryexchangewouldhaveresultedinoneoftheotherfouroutcomesintheabsenceofexchange.Aggregatingyieldsamoredirectunit-measureofthenumberofadditionaltransplantsperformedasaresultoftheintroductionofkidneyexchange,ratherthaninferringitfromchangingprobabilities.Suchaggregationallowsmorexibilityintheestimatesandcanprovideinsightintohowmanyindividualsreceivinganexchangewouldhavecontinuedtowaitforatransplant.Zipcodemonthisthemostnaturalaggregationlevelgiventheprecedinganalysis,andstillallowsfortheuseofnearbyexchangeactivitybasedonzipcode.Analogoustothemadewhenusingregistration-leveldata,Isubtractthenumberofwithin-50-milesexchangetrans-plantsreceivedbypatientsinaparticularzipcodemonthfromthenearbyactivitymeasureforthatparticularzipcodemonthobservation.TableA.15presentstheresultsfromthistest,whichareincrediblysimilartotheoriginal2SLSresults.Noticethatthecrowd-outcoefnolongersumto-1,buttheydocomeverycloseat-0.93.Thissuggeststhat0.7ofevery10exchangerecipientswouldhavecontinuedtowaitforakidneyratherthanexperienceoneoftheotherfouroutcomes.Thecrowd-outofdirectlivingdonorsisnearlyidenticalat-0.52comparedto-0.5,thecrowd-inofanonymousdonorsis0.13comparedto0.12,thecrowdoutofdeceaseddonortransplantsissmallerinmagnitudeandstillstatisticallyat-0.039,andthereductionindeathsonthewaitinglistisslightlysmallerinmagnitudeat-0.5comparedto-0.54.Theseestimatesimplythat105:2+1:3=6:1ofevery10exchangesarelivingdonortransplantsthatwouldnothaveoccurredintheabsenceofexchange,whichisagainverysimilartotheoriginalestimateof6.2.Finally,Itesttherobustnessoftheoriginal2SLScrowd-outresultstotheinclusionofpatientsremovedfromthewaitinglistasaresultofbeingtoosicktotransplant,andthoseremovedasaresultofrefusingtransplantoffers.Inmyestimationsample,25,952patientswereremovedfromthewaitinglistforthesereasons,15percentofwhomrefusedatransplantand85percentofwhom28weredeemedtoosicktotransplant.Again,theresultsarenearlyunchanged:5.3ofevery10ex-changeswouldhavebeenadirectlivingtransplant,1.3additionalanonymousdonationsoccurforevery10exchanges,thereisastatisticallychangeindeceaseddonortransplants,andathereisareductionindeathswhilewaitingof6.2forevery10exchanges.Similarly,thegaininlivingdonationsisrelativelyunchanged:105:3+1:3=6ofevery10transplantsviaexchangerepresentnewlivingdonortransplantsusingthis1.7ConclusionAsofOctober9,2015,101,160candidatesarewaitingforakidney.33Thegrowingshortageoftransplantableorganshasdriveneconomists,transplantpractitioners,andlawmakerstodevelopcreativesolutions.Theinnovationoftransplantationamongpatientswithincompatiblewillingdonorsviaexchangehasgrowninpopularityinrecentyears,facilitatedbysingle-centerregistriesandconsortiaoftransplantcentersusingcomputer-optimizedmatchingmechanisms.Evaluatingtheextenttowhichexchangeimprovesobservedpatientoutcomesisthemostdirectwayofevaluatingeffortstointroduceandpromotethismodeoftransplantation.Suchananalysisismorefeasiblenowwiththegrowingprevalenceofexchange.Thispaperisthetoundertakesuchanevaluationbeyondtheuseofsimulations.Iuseadministrativedataonwaitinglistregistra-tionsandtransplants,whichcontainarichsetofpatientandtransplantcentercharacteristics.Toidentifytheeffectsofinterest,Iconstructanovelmeasureoftime-varyinglocalexchangeactivityusingpatientandtransplantcenterlocationdata.Theevidencepresentedinthispapersuggeststhatkidneyexchangeprogramsarehighlyef-fectiveinincreasingthenumberoftransplants,suchthat6.2ofevery10transplantsviaexchangerepresentlivingdonortransplantsthatwouldnothaveoccurredintheabsenceofexchange.Thisisespeciallyimportantgiventhatmanypatientsreceivingexchangetransplantsarehardtomatchandfaceriskofdyingwhilewaitingforacompatibledeceaseddonorkidney.Of33BasedonOPTNdataasof10/9/2015.29the6.2newtransplants,1.2comefromanonymousaltruisticdonorswhoarenowabletofacili-tatemanymoretransplantswithasinglekidneydonationandmorewillingtodonateasaresult.BasedonOPTNdataasof9/11/2015,therehavebeen4,782transplantsperformedviapairedandlistexchange.Theresultsofthispapersuggestthat2,965ofthoserepresentlivingdonorkidneytransplantsthatwouldnothavehappenedintheabsenceofexchange.Inadditiontocreatingmanyadditionallivingdonortransplants,theevidencepresentedinthispapersuggeststhattheincreasingprevalenceofexchangehasaimpactontransplantqualityoutcomes.Conditionalonreceivingatransplantofanykind,atenpercentagepointincreaseintheprobabilityofreceivingatransplantviaexchangeisshowntoincreaseaverageone-yeargraftsurvivalby2percentagepoints,two-yeargraftsurvivalby1.9percentagepoints,andreducesregistrationdurationby59.2days,relativetoanaverageoverallone-yeargraftsurvivalrateof93percent,two-yearrateof88percent,and597dayregistrationduration.Theseresultsareinlinewithwhatwewouldexpectgiventhatexchangesincreasethenumberoflivingdonortransplantsperformed.Dependingoncost-effectiveness,theofthispapershouldbeencouragingtotransplantcentersconsideringoralreadyperformingkidneyexchanges.TheresultsshouldalsoencouragelawmakersandtheUnitedNetworkforOrganSharingtofurtherpromoteexchangeasatransplantoption.ThispolicyimplicationisbolsteredbyresearchfromRothandcoauthorsdemonstratingtheincreasingreturnstoscaleassociatedwithlargerregistriesofincompatiblepairs.Thatsaid,wemustalsoconsidertheequityimplicationsofkidneyexchange.RecentworkbySönmezandÜnver(2015)suggeststhroughsimulationsthatpatientswithbloodtypesABandOtheleastfromtheintroductionofexchangewhilepatientswithbloodtypeBthemost.ThisisespeciallyimportanttoconsidergiventhatpatientswithtypeObloodarealreadythehardesttomatchonbloodtype.Futureworkisneededtoevaluateboththecost-effectivenessandequityimplicationsofexchange.30CHAPTER2TESTINGFORHETEROGENEITYINTHEEFFECTSOFKIDNEYEXCHANGESONTRANSPLANTQUANTITYANDQUALITY2.1IntroductionIn2014,4,445peoplediedwaitingforakidneytransplant.Inthesameyear,36,156peopleenteredawaitinglistforakidney,andonly17,108peoplereceivedakidneytransplant.1FigureB.1showsthedramaticgrowthinthewaitinglistforkidneysovertime,whiletransplantsgrowatamuchslowerpace.AddressingtheincreasinglyunmetdemandfortransplantablekidneysintheU.S.whilemaintainingorincreasingtransplantqualityrequirescreativityduetotheNationalOrganTransplantationAct(NOTA)of1984,whichbannedthesaleofhumanorgans.Sincetheyear2000,transplantcentershaveincreasedthequantityandqualityofkidneytrans-plantsbyfacilitatingexchangesamongpatientswithwillingbutincompatiblelivingdonors(Chap-ter1).Exchangeisanattractivealternativeforpatientswhoareunabletoacompatiblelivingdonorduetobloodtypeincompatibilityand/orsensitivitytoforeignorgans.Inthemostbasictypeofexchange,atwo-waypairedexchange,patientsmayfiswap"theirwillingdonorswhenthedonorfromonepairisamatchforthepatientinanotherandviceversa.Pairedexchangescanbeextendedintodonorchains,whereanaltruisticdonorstartsaseriesofpairedexchangesbydonatinganonymouslytoapatientwithawillingincompatibledonor.Anothervariationislistexchange,whereawillingdonorgivesakidneytosomeoneonthewaitinglistinexchangeforthenextcompatibledeceaseddonorkidneyforhisorherintendedpatientinneed(Delmonicoetal.,2004).Withtheabsenceofaformalprice/marketmechanism,kidneyallocationandtransplantationaredeterminedbywaitingtime,bloodtype,sensitivitytoforeignorgans,accesstofacilities,ac-cesstodonors,andaccesstosuitablealternatives.Theimportanceofequityinorganallocation1BasedonOrganProcurementandTransplantationNetwork(OPTN)Dataasof9/4/2015,foundathttp://optn.transplant.hrsa.gov.31ishighlightedinthesenatereportassociatedwithNOTA:fiAnequitablepolicyandsystemisnec-essarysothatindividualsthroughoutourcountrycanhaveaccesstoorgantransplantationwhenappropriateandnecessary"(SenateReport98-382).Despitethisguidingprinciple,thereexistsig-disparitiesinaccesstolivingkidneydonationacrossbothimmunologicanddemographiccharacteristicsofpatients.Foranypatienttofromtheintroductionofexchangeasatrans-plantalternative,thatpatientmustatleasthaveaccesstoawillinglivingdonorandaparticipatingtransplantcenter.Oneofthepurportedofkidneyexchangeisthatexchangemitigatesim-munologicdisparitiesinaccesstolivingdonortransplants.Inotherwords,exchangeissupposedtoincreaseaccessforpatientswithhighsensitivitytoforeignorgansand/orharder-to-matchbloodtypes.Atthesametime,however,theofintroducinganewtransplantoptionthataimstosolvecompatibilityissuesthroughthechanneloflivingdonationmayreinforceexistingdisparitiesinaccesstolivingdonorkidneytransplants.Thegoalofthispaperistoestimatethedifferentialeffectsofkidneyexchangesonout-comesacrosspatientsofseveralimmunologiccategories.TheimmunologiccharacteristicIanalyzeispatientbloodtype,whereimbalancesexistduetotheasymmetricalnatureofbloodtypematchingbetweendonorsandrecipients.Ingeneral,typeOpatientscanonlyreceiveakidneyfromatypeOdonor,meaningtheyfacethesmallestpoolofpotentialdonors.Meanwhile,typeApatientscanreceivefromtypeOandtypeAdonors,typeBpatientscanreceivefromtypeOandtypeBdonors,andtypeABpatientscanreceivefromadonorofanybloodtype.SönmezandÜnver(2015)developatheoreticalframeworkthat,whenusedtosimulatepatientoutcomesbe-foreandaftertheintroductionofexchange,suggestskidneyexchangesdisproportionatelybloodtypeAandbloodtypeBpatients.Thisoutcomeisproblematicifoneofthegoalsofkidneyexchangesistofacilitatetransplantsforhard-to-matchpatients,suchasthosewithtypeOblood.ThesecondimmunologiccharacteristicIstudyissensitivitytoforeignorgans,theimportanceofwhichishighlightedbyanallocationrulethatawardsadditionalprioritytopatientsonthedeceaseddonorwaitinglistwhohaveaPanelReactiveAntibody(PRA)scoreof80orhigher(OPTN,2016;Sweeney,2011).ThePRAscore,whichtakesonavaluebetween0and100,32measuresthelevelofantibodiesagainstforeignproteinspresentinthepatient'sbloodandindicatesthepercentageofthegeneralpopulationwithwhomthepatientislikelyincompatible.Again,ifexchangesaimtofacilitatetransplantsforhardtomatchpatients,wewouldhopetoseethelargesttransplantquantityandqualitygainsaccruingtopatientswithhigherPRAscores.Afteraddressingimmunologicconcerns,thesecondgoalofthispaperistoanalyzewhetherexchangeresultsindifferentialimprovementsinpatientoutcomes.Tothisend,Istudytheeffectsoftheintroductionofexchangeacrosspatientsofdifferentraces,educationlevels,andagegroups.Ichosethesethreecharacteristicsbasedonexistingstudiesinthetransplantationliterature.First,studiessuggestthatblackpatientsexperiencelowerlivingdonorkidneytransplantationrates,aswellaslowertransplantsurvivalrates,comparedtowhitepatients(Goreetal.,2009;Gordonetal.,2010;Watermanetal.,2013;Patzeretal.,2012).Studiesalsosuggestthatolderpatients,andless-educatedpatients,arelesslikelytoreceivelivingdonorkidneytransplants(Goreetal.,2009).Iftheintroductionofexchangeexacerbatesuchdifferences,itmaywarrantthealterationofmatchingalgorithmsinawaythatplacesgreaterweightonmatchingdisadvantagedpatients.Similarly,suchmayencourageincreasedemphasisontheuseoflistexchangeasitallowsparticipatingcenterstoallocatelivingkidneydonationsfromincompatiblepatient-donorpairstodisadvantagedpatientsonthedeceaseddonorwaitinglist.Despitewidespreadconcerninthetransplantcommunityaboutthedisparitiesaddressedabove,thereisanapparentlackofempiricalresearchontherelationshipbetweenkidneyexchangeandexistingimmunologicanddemographicdisparities.Tothisgapintheliterature,IextendtheestimationapproachdevelopedinChapter1inordertoconductaseriesofheterogeneityanalysesfocusingonbloodtype,PRA,race,education,andageusingtheOPTNStandardTransplantandAnalysisResearch(STAR)2Iconstructatime-varyingmeasureoflocalexchangeactivityandinteractitwithpatientcharacteristicsofinterestinordertocausallyidentifyheterogeneityintheeffectsoftheintroductionofkidneyexchangeonthequantityoflivingdonorkidneytransplants.I2TheOPTNSTARthatIusecontainindividual-leveldataincludingtheuniverseofwaitinglistregistrationsandorgantransplantsintheUnitedStatesfrom1988through2014.33thenanalyzetheextenttowhichdifferentsubgroupsofpatientsexperienceachangeintransplantqualityfollowingachange(orlackthereof)intransplantquantity,wherequalityismeasuredbygraftsurvival,waitinglistregistrationduration,andtissue-typematchquality.3,42.2BackgroundandMotivationInChapter1,usingindividual-levelwaitinglist,transplant,andfollow-updata,Iestimatetheprobabilitythatapatientreceivesatransplantviaexchange,asopposedtoatransplantviadirectlivingdonation,anonymousdonation,deceaseddonation,ordyingwhilewaiting,basedonthenumberofexchangesperformedwithina50mileradiusofthepatient'szipcodeofresidence.Ithenestimatethecounterfactualtransplantoutcomes,aswellasoutcomesmeasuringqualityoftransplant,usingvariationinthepredictedprobabilityofreceivingatransplantviaexchange.Ithatforevery10transplantsviaexchange,6.2representlivingdonortransplantsthatwouldnothaveoccurredintheabsenceofexchange.Ofthe6.2newlivingdonortransplants,1.2aretheresultofanonymousdonationcrowd-in,presumablyduetotheuseofdonorchains.IalsoinChapter1thata10percentagepointincreaseintheprobabilityofreceivingatransplantviaexchangeresultsina2.2percentincreaseinone-andtwo-yeargraftsurvival,anda10percentreductioninwaitinglistregistrationduration.However,InostatisticallyeffectontissuetypematchqualityasmeasuredbythenumberofHumanLeukocyte3Whilethereappearstobealackofcausalevidence,itisgenerallyacceptedinthetransplantcommunitythatlivingdonorkidneytransplantsaremoresuccessfulthanthosefromdeceaseddonors.CalculationsbasedonOPTNindividual-leveltransplantdata(asof12/31/2014)from1988to2008revealthat3.2percentoflivingdonorkidneygraftsfailedwithinoneyearcomparedto7.9percentofdeceaseddonorkidneygrafts.Similarly,16.8percentoflivingdonorkidneygraftsfailedwithinveyearscomparedto29.6percentofdeceaseddonorkidneygrafts.Notethatthesefailuresalsoincludedeathsofthosewithnon-functioninggraftswithinthetime-frame.Deathsofthosewithfunctioninggraftsbeforeoneyearorveyearswereexcludedfromtherespectivecalculations.4Intheirpaperprovidinganoverviewoflivingkidneydonationpracticesasof2005,DavisandDelmonico(2005)suggestthatthelivingdonationsurvivaladvantageispartlyduetoreducedwaitingtimeandtimespentondialysisforlivingdonorkidneyrecipientscomparedtodeceaseddonorkidneyrecipients.34Antigen(HLA)mismatches.ThemoreHLAmismatches(ofwhichtherecanbebetween0and6)betweenthepatientandaprospectivedonor,themorelikelyitisthatthepatient'sbodywillrejectthetransplantinspiteofanyreasonableimmunosuppressivedrugtreatment.Theresultsaremostlyconsistentwithearliersimulationsanddescriptivestatistics,which,whileunabletoaccountforthepotentialanonymousdonationcrowd-in,estimatedthat5ofevery10transplantsviaexchangerepresentnewlivingdonortransplants(Rothetal.,2004;Mierzejewskaetal.,2013).Whileexchangesappeartoincreasethenumberoftransplantsonthewhole,studyingtheextenttowhichexchangesdifferentiallyimpacttheoutcomesofpatientsinvariousimmunologicanddemographicsubgroupscanhelpguidepoliciesgoverningkidneyexchangesmovingforward.Wallisetal.(2011)discussearlyconcernswithkidneyexchangesregardingdisparitiesacrossbloodtypes;namely,thatpairswithtypeO(universal)donorsortypeAB(universal)recipientswouldnotparticipatetherebyleavingpairscontainingOpatientsandABdonorsunmatched.How-ever,duetotissue-typeincompatibility,pairscontainingOdonorsorABrecipientsmaystillittoparticipate.AccordingtoWallisetal.,therearealsopossiblefairnessissuesin-troducedbylistexchanges,whereacommonarrangementinvolvesanon-typeOlivingdonationbeingexchangedfordeceaseddonorkidneywaitinglistpriorityforatypeOpatient.Suchar-rangementswoulddisproportionatelyincreaseexcessdemandforbloodtypeOdeceaseddonorkidneys.SönmezandÜnver(2015)theoreticallyanalyzetheefyandequityofincentivizingcompatible-pairparticipationinkidneyexchanges,andconductsimulationsusingaggregatepa-tientdatafromthe2011calendaryear.Ofparticularrelevancetothispaperistheiranalysisofchangesinwaitingtimeandnumberoftransplantsacrosspatientsofdifferentbloodtypeswhenmovingfromasystemwithonlydirectlivingdonations(noexchange)toasystemoffiregularex-change"(thecurrentexchangesystem).ThispaperismotivatedbysimilarconcernstoWallisetal.andthispaper.TheintuitiveexplanationgivenbySönmezandÜnveristhat"itismuchharderforblood-typeOpatientstofromlivedonationorlivingdonororganexchangethanpatientsofotherbloodtypes."35Theirsimulationsimplythattheintroductionofexchangeyieldsanincreaseinlivingdonortransplantsby22percentforbloodtypeOpatients,comparedto30percentfortypeA,64percentfortypeB,and12percentfortypeAB.Theyalsoyieldareductioninwaitingtimefordeceasedandlivingdonortransplantrecipientsby4percentforbloodtypeO,11percentfortypeA,18percentfortypeB,and4percentfortypeAB.WhiletheofSönmezandÜnversuggestthatexchangesmakeeveryonebetteroff,theyalsosuggestthattypeAandtypeBpatientsenjoylargertransplantquantitygainscomparedtotheharder-to-matchtypeOpatients.Inadditiontoresearchonkidneyexchanges,priorresearchineconomicslargelyfocusesonthefactorsdeceaseddonorkidneysupply,howchangesinthesupplyofdeceaseddonorkidneysaffectlivingkidneydonations,andtheeffectsofincentivesonlivingorgando-nation.Dickert-Conlinetal.(2011)showthatthesupplyofdeceaseddonorkidneysisresponsivetochangesinmotorcyclehelmetlaws.Follow-uppapersbyDickert-Conlinetal.(2015)andFernandezetal.(2013)thatthesesupplyshockscauseadecreaseinlivingorgandonations.Usingalternativestrategies,Sweeney(2011)andAnderson(2016)alsoshowthatdeceaseddonorkidneytransplantscrowdoutlivingdonations.Onthesubjectofincen-tives,Laceteraetal.(2014)noeffectofstate-sponsoredtaxandpaidleaveincentivesonlivingkidneydonation,thoughtheydopositiveeffectsforbonemarrowdonation.However,Schnieretal.(2014)thattransplantcentereffortstoreimbursedonors'travel-relatedexpensesincreaselivingkidneydonations.Whenweturntothemedicalliteratureontransplantation,wealargebodyofworkfocusingonawiderangeofdisparitiesintransplantaccessandoutcomes.Inadditiontobloodtype,PRAscorerepresentsanotherimmunologiccharacteristicacrosswhichdisparitiesmayexist.PossibleexplanationsfordisparitiesacrossPRAscoresincludelongerwaitsforcompatiblekidneysandincreasedwillingnesstoacceptlowerqualitycompatiblekidneys.Nassiretal.(2015)thatpatientswithhigherPRAsfacedhighertransplantcosts.Brayetal.(2006)andFaravardehetal.(2013)thatpatientswithhigherPRAsexperiencedhigherriskofgraftfailure.Interestingly,Brayetal.thatpatientswithPRAgreaterthan80hadidenticalve-yeargraftsurvivalto36unsensitized(PRAequalto0)patients;apossibleexplanationforthisisthedeceaseddonorwaitinglistprioritygiventopatientswithPRAgreaterthan80(OPTN,2016;Sweeney,2011).Similartobloodtypeinequality,exchangescaninequalityacrossPRAgroupsduetomatchqualityandallocationrules.PatientswithalowPRAscoreareeasiertomatchconditionalonablood-typematch,whilepatientswithahighPRAarethehardesttomatch.Again,notethatpatientswithPRAscoresover80receivepriorityonthedeceaseddonorwaitinglist,suchthatapatientwithaPRAscoreof80receivesalowermarginalofalivingdonor(relativetowaitingforadeceaseddonor)comparedtoapatientwithaPRAscoreof79.Forsimplicity,IsplitpatientsintoPRAcategoriesof0to40,40to80and80to100.PRA0to40patientshavethemostaccesstocompatibledonorsoverall,andalsohavethehighestlikelihoodofamatchthroughexchangeconditionalonbeingunabletoacompatibledeceasedorlivingdonor.PRA40to80patientshavetheleastaccesstodeceaseddonors,andlessaccesstolivingdonorsthanPRA0to40patients.PRA80to100patientshavethemostaccesstodeceaseddonorsduetotheallocationpriorityrules,andtheleastaccesstolivingdonors.ThesedifferencesimplythatPRA0to40patientswillbetheleastlikelytopursueexchange,buttheoneswhodowillhavethemostsuccess.TheyalsoimplythatPRA40to80patientswillbethemostlikelytopursueexchange,andtheoneswhodowillhavemoresuccessthanPRA80to100patientsbutlessthanPRA0to40patients.Finally,PRA80to100willbebetween0-40and40-80inpropensitytopursueexchange,andwillthosewhodowillbetheleastsuccessful.Whilethepredictedneteffectsaresomewhatambiguous,itseemsthatPRA40to80patientsshouldbetheoneswhomostasawholefromexchangeintermsofquantityandqualityimprovements.Sincetheprocessofalivingdonor,andlivingdonationitself,issimilarbetweenex-changeanddirectdonation,wewouldexpecttoseeexchangeshavedifferentialimpactsonpatientoutcomesinapatternconsistentwithexistingdemographicdisparitiesacrosspatientrace,age,andeducation.Goreetal.(2009)that,holdingobservabledemographiccharacteristicsconstant,whitetransplantrecipientsarealmosttwiceaslikelytoreceivealivingdonorkidneythanblack37recipients.5Theauthorsarguethatblackpatientsmaybelesslikelytoaskfriendsandrelativestobedonorsduetodifferencesinperceivedmedicalurgencyandageneraldistrustofthemed-icalsystem.Ifthisargumentiscorrect,blackpatientswillalsobelesslikelytoaskfriendsandrelativestodonateviaexchangemeaningtheintroductionofexchangewillblackpatients(withrespecttolivingdonortransplants)muchlessthanitwouldwhitepatients.Racialdisparitiespervadepost-transplantoutcomesaswell;Gordonetal.(2010)that,foralltypesofkidneytransplants,blackpatientsexperiencethelowestgraftsurvivalratesafterone,three,andveyears.Ifexchangesyieldarelativelysmallgaininlivingdonortransplantquantityforblackpatients,theyarealsolikelytoyieldarelativelysmallgainintransplantqualitysince,onaverage,livingdonortransplantssurvivelongerthandeceaseddonortransplants.Olderagemayleadtodecreasedtransplantaccessbasedoncomparisonsofperceivedbeneandcostsbydoctors,patients,andprospectivelivingdonors.Segevetal.(2009)thatolderpatientshaveloweraccesstotransplants,bothlivinganddeceased.PartoftheinequalitybetweenagegroupsindeceaseddonortransplantationstemsfromShare35,apolicyintroducedin2005thatgivesminorsprioritytodeceasedorgansfromdonorsage35orless.Goreetal.(2009)thatolderpatientsexperienceareductionintheprobabilityofreceivingalivingdonorkidneytrans-plant.Fuggleetal.(2010)thatpatientsaged60andolderexperiencedincreasedprobabilityofearly(withinthreeyears)graftfailureafterreceivingalivingdonortransplant.Ingeneral,theliteraturesuggeststhatexchangewouldfurtherincreaseinequalityacrossage.Thatsaid,despitethelackofastandardagelimitacrosscenters,somecentersimposeagerestric-tionsforreferraltothedeceaseddonorwaitinglist.6,7Thisimpliesthatelderlypatientsfaceamore5Goreetal.alsothatwhitepatientsarealmost1.5timesmorelikelytoreceivealivingdonorkidneythanfiother"(non-white,non-black,non-Hispanic)recipients.6UNOSTransplantLiving,http://www.transplantliving.org/community/patient-resources/frequently-asked-questions/#age,accessedMay5,2016.7Forexample,theUniversityofWisconsintransplantcenterstatesthatitisun-commonforthemtooffertransplantationtopatientsolderthan75yearsold.http://www.uwhealth.org/transplant/kidney-transplant-frequently-asked-questions/10573,ac-cessedMay5,201638limitedsetofinitialoptionsonaverage.Therefore,theintroductionofexchangepotentiallyrepre-sentsalargerrelativegainintransplantoptionsforelderlypatientscomparedtoyoungerpatients,especiallyminorsthathavedisproportionatelylargeaccesstodeceaseddonortransplantation.Less-educatedindividualsmayalsosufferfromdecreasedtransplantaccessduetoitsrelation-shipwithoverallsocioeconomicstatus,includingincomeandhealthinsurancecoverage.Thoseintheirsocialnetworkmayalsohaveloweraveragesocioeconomicstatusandthereforearelessabletobeartheeconomiccostsofbecomingalivingkidneydonor.Suchpatientsmayalsobelesssavvyinnavigatingthetransplantationsystem,includingoutaboutandpursuingexchanges.Inlinewiththis,Segevetal.(2009)that,amongelderlypatients,patientswithcollegeedu-cationhavehigheraccesstotransplants.Goreetal.(2009)thatcollege-educatedpatientshaveahigherprobabilityofreceivingalivingdonortransplant.Therefore,wewouldexpectthattheintroductionofkidneyexchangewouldmaintainorfurtherincreaseinequalityinlivingdonortransplantationacrosspatientsofdifferenteducationlevels.2.3DataInthefollowinganalysis,Ibeginwithindividual-leveldataextractedonDecember31,2014fromtheStandardTransplantAnalysisandResearch(STAR)whichisavailablebyrequestfromOPTN.Thedatacontain788,106observationsofkidneywaitinglistregistrationsandtransplantsthatoccurredfrom1988to2014.Oftheseobservations,398,984arewaitinglistregistrationsthatresultedintransplantsandtransplantsthatoccurredwithoutanassociatedwaitinglistregistration.8Thedatacontaininformationontheoutcomeofeachregistrationincluding:transplant,death,transfertoadifferentcenter,orstillwaitingasofDecember31,2014.Irestrictmyanalysisto8Somerecipientsoflivingdonorkidneysneverregisterforthedeceaseddonorwaitinglist.Theseobservationsareenteredasatransplantrecordandlacktheinformationthatisonlyrele-vantwhenapatientregistersonthewaitinglist.Almost26percentofdirectliving,6percentofexchange,and3percentofanonymoustransplantsoccurredwithoutanassociatedwaitinglistreg-istration,basedonOPTNdataasof12/31/2014.Suchlimitationsrenderdurationmodelsunattrac-tivewhensettingouttoanswerthequestionsraisedinChapter1aswellasthepresentChapter.39observationsthatresultedineitheratransplantordeath,astheseencompasstheclearandwell-registrationoutcomes.9fiDonorrelation"isthekeyvariableIuseforthosewhoreceiveatransplantinordertodeter-minewhetheritisadirectliving,deceased,anonymous,orexchangetransplant.Kidneytransplantsarecodedwithoneofthefollowingdonorrelationships:sibling,twin,child,parent,otherrelative,other,miscellaneousunrelateddonor,pairedexchange,listexchange,anonymous,ordeceased.10,11Thesedataarerich;Iobservevariablesincludingbloodtype,PRA,race,educa-tion,previoustransplantstatus,age,gender,registrationdate,transplantdate,HLAmismatches,additionalmedicalinformation,donorcharacteristics,andtransplantfollow-upinformation(fromwhichgraftsurvivaliscalculated).Additionally,throughaspecialrequest,Iobtainedzipcodeinformationforpatientsandtransplantcenters.IrestrictmyanalysistodatafromJanuary2000toJuly2014.Thereisalargeincreaseinthequalityofreportingforthedonorrelationshipvariablein2000.Forexample,outofanaverageof4,710yearlylivingdonorkidneytransplants,thereareanaverageof77unreportedlivingdonorrelationshipsperyearfrom1988to1999comparedtooneperyearfrom2000to2013.Thereislittlereasontobeconcernedabouttheexclusionofpre-2000data,asthedatashowonlyonepairedexchangeand8listexchangesoccurringbefore2000.Duetolagsindataprocessing,AugustthroughDecember2014areincompletewithrespecttodonorrelationshipatthetimeofmyextract.TableB.1showsthefrequencyofobservedtransplantoutcomesbydonorrelationshipinclud-9Notethatpatientsmayhavemultiplewaitinglistregistrations.Patientswithmultipleregistra-tionswhodiedwhilewaitingareonlycountedonce,andIusetheirearliestlistingcodedassuch.Fortheendingdateoftheseobservations,Iusetheearliestreportedregistrationenddateamongthosecodedasremovalduetodeath.Forpatientswhoreceiveatransplant,onlyoneregistrationwillshowthetransplantoutcome,whichistheobservationIuse.10NotethatIcanonlyconnectdonorstotheiractualrecipients.Therefore,withrespecttoexchanges,Icannotconnectdonorstothelovedoneonwhosebehalftheyaredonating.Also,Icannotobservewhetherananonymousdonor'skidneyisusedtostartadonorchain.11Notethatdonorchains,exchangesthatbeginwithanon-directedanonymousdonor,oftenendinatransplanttothedeceaseddonorwaitinglist.Inthesecases,thetransplantiscountedasalivingdonortransplantvialistexchange.40ing:directedliving,pairedandlistexchange,anonymousdonation,deceaseddonation,anddeathwithoutatransplant.Weseethatexchangesaccountfor1.5percentofobservedoutcomes,anony-mousdonationsaccountfor0.5percent,directedlivingdonationsaccountfor28percent,deceaseddonationsaccountfor49percent,anddeathswhilewaitingaccountfor21percent.Comparingthe2013snapshot,thelastcompleteyearofdata,tothe2007snapshot,weseethattransplantsviade-ceaseddonation,exchange,andanonymousdonationincreasedinlevelandpercentofoutcomes.Wealsoseeadecreaseindirectedlivingdonationsandaslightdecreaseindeathswhilewaiting.TableB.2presentsthefractionofkidneygraftsthatsurviveatleastoneyear,atleasttwoyears,thenumberofHLAmismatchesbetweendonorandrecipient,andthedurationofthewaitinglistregistration.Weseethatgraftsurvivalisslightlyhigher(butnotstatisticallydifferent)forexchangesbutverysimilaracrossalllivingdonortransplantsforbothoneandtwoyearsat96-97percentand93-94percent,respectively.12One-andtwo-yeargraftsurvivalislowerforrecipientsofdeceaseddonorkidneysat90percentand86percent,respectively.LookingatthenumberofHLAmismatchesbyobservedoutcome,weseethatdirectlivingdonationshavethelowestnumberofHLAmismatchesat3.16followedbydeceaseddonationsat3.88.ThereisnodifferenceintheaveragenumberofHLAmismatchesacrossexchangesandanonymousdonationsat4.31.Recipientsofdirectlivingdonationshavetheshortestwaitinglistregistrationdurations,roughly237days,followedbyexchangeat468days,anonymousat654days,anddeceaseddonationat813days.13BothregistrationdurationandnumberofHLAmismatchesappearuncorrelatedwithgraftsurvivalamonglivingdonorkidneyrecipientsafteroneandtwo12Graftsurvivalisforobservationswithanon-missinggraftsurvivaltimeintheOPTNdata.Ittakesonavalueof0ifthepatientdiedbeforeoneortwoyears,orifthereisareportedgraftfailurewithinoneortwoyears.Ittakesonavalueof1ifthegraftsurvivaltimeexceedsoneortwoyears,orifthepatients'lastknownstatusisalivewithafunctioningkidney.Theassumptionisthatthesefilost"individualswouldhavereturnedtothesystemortheywouldhaveadeathdatereportedthroughtheSocialSecurityDeathMasterFileiftheirgraftfailedortheydied.Inordertoensureadequatetimehaspassedforfollow-updata,Irestricttheanalysisofone-yeargraftsurvivaltotransplantsoccurringonorbeforeDecember31,2012,andDecember31,2011fortwo-yeargraftsurvival.13Note:roughly1/3oflivingdonorkidneyrecipientsneverregisterforthewaitinglist.Inthesecases,registrationdurationissettozero.41years.TablesB.3andB.4showsummarystatisticshighlightingtheheterogeneityintransplantquan-tityandqualityoutcomesacrossthepatientcharacteristicsofinterest.LookingatTableB.3,weseethatbloodtypeAandOpatientsarethemostlikelytoreceivelivingdirecttransplants-30per-centand27percent,respectively.BloodtypeBandOpatientsareamongtheleastlikelytoreceiveadeceaseddonorkidneytransplant,at47percentand48percent,respectively.TypeOpatientsaretheleastlikelytoreceiveatransplantviaexchangeandanonymousdonationat12percent,andareamongthemostlikelytodiewhilewaitingat23percentalongwithtypeBpatientsat25percent.TypeABpatients,theuniversalrecipients,aretheleastlikelytoreceivealivingdirecttransplant(24percent),mostlikelytoreceiveadeceaseddonortransplant(60percent),andleastlikelytodiewhilewaiting(14percent).TableB.3alsoshowsthatpatientswithhigherPRAscoresarelesslikelytoreceivealivingdirecttransplant,morelikelytoreceiveadeceaseddonortransplant,andmorelikelytodiewhilewaiting.Interestingly,patientswithPRAsinthemiddlerangeof40to80arethemostlikelytoreceivetransplantsfromanonymousdonorsandviaexchangeat0.7percentand2.4percent,respectively.Thismaysuggestthatthesepatients,whilenotsensitizedenoughtoreceiveadiscretejumpinpriorityonthedeceaseddonorwaitinglist,arestillsensitizedenoughtosearchforoptionsbeyondlivingdirectanddeceaseddonation.ContinuingwithTableB.3,weseethatolderpatientsarelesslikelytoreceivelivingdirecttransplants,lesslikelytoreceiveexchangetransplantsamongpatientsofadultage,lesslikelytoreceiveananonymousdonation,andmorelikelytodiewhilewaiting.Minorsareaninterestingagecategory;they,alongwith18to35yearolds,havethehighestratesoflivingdirecttransplantationataround39percent,arethemostlikelytoreceiveadeceaseddonortransplantat57percent(likelyduetoShare35),aretheleastlikelytoreceiveanexchangetransplantat0.9percent,themostlikelytoreceiveananonymoustransplantat0.7percent,andtheleastlikelytodiewhilewaitingat2.6percent.Whitepatientsarethemostlikelytoreceivealivingdirecttransplantat36percent;thenexthighestarehispanicpatientsat26percent.Whitepatientsarealsotheleastlikelytoreceiveadeceaseddonortransplantat44percent,mostlikelytoreceiveatransplantvia42exchangeat1.8percent,andtheleastlikelytodiewhilewaitingat18percent.Ontheotherhand,blackpatientsaretheleastlikelytoreceivealivingdirectdonationat15percent,mostlikelytoreceiveadeceaseddonortransplantat58percent,amongtheleastlikelytoreceiveanexchangeoranonymousdonationat0.8and0.3percent,respectively,andthemostlikelytodiewhilewaitingat26percent.Finally,adultpatientswithahighschooldiplomaorlessarelesslikelytoreceivealivingdirecttransplant(24vs.33percent),morelikelytoreceiveadeceaseddonortransplant(52vs.46percent),lesslikelytoreceiveanexchangeoranonymousdonation(1vs.2percentand0.4vs.0.6percent,respectively),andmorelikelytodiewhilewaiting(23vs.18percent).MovingontoTableB.4,weseealmostnodisparityingraftsurvivalafteroneandtwoyearsacrossbloodtype-roughly92.5percentafteroneyearand88.5percentaftertwoyears.However,wedoseethatbloodtypeBandOpatientsspendthelongesttimeonthewaitinglistbeforereceivingatransplantat705and677days,respectively.PatientswithtypeAandOblood,however,enjoythefewestnumberofHLAmismatchesat3.58and3.62,respectively.LookingatPRAscores,weseethatgraftsurvivalappearstodecreaseasPRAincreases.Asexpected,weseethatregistrationdurationincreaseswithPRAscore.ThosewithhigherPRAscoresreceivetransplantswithfewerHLAmismatches;thisseemscounterintuitiveuntillearningthatPRAscoresarerelatedtothepresenceofHLAantibodies-highPRApatientscannotreceiveakidneywithcertainHLAmismatchesortheirbodywillrejectthekidneyimmediately.ContinuingwithTableB.4,weseethatelderlypatientsexperiencelowergraftsurvivalrates(90.5percentforpatientsover55vs.93-94percentafteroneyear),faceshorterwaitsforatrans-plantwhencomparedtootheradultpatients(542daysvs.around660days),andreceivetrans-plantswithmoreHLAmismatchescomparedtootheradultpatients.Turningtorace,weseethatblackpatientsexperiencetheworstgraftsurvivalrates(90.5percentvs.theaverage92.5percent),highestwaitingtimes(815daysvs.theaverage601days),andthehighestHLAmismatchav-erage4.16vs.theaverage3.64).WhitepatientshavegraftsurvivalratesinlinewithHispanic,andpatientsinthefiOther"racecategory(includingAsian,NativeAmerican,PIslander,andMultiracialpatients),buttheyexperiencetheshortestwaitingtimesbyalargemarginaswellas43thelowestHLAmismatchaverage.LowerwaitingtimesandHLAmismatchesmaybeattributableinparttothehigherfrequencyoflivingdirecttransplantsamongwhitepatients,manyofwhicharefromblood-relatives.Finally,patientswithahighschooldiplomaorlessexperienceslightlylowergraftsurvivalratesandslightlylongerwaitingtimesonaverage,andtheredoesnotappeartobeamedicallydifferenceinthenumberofHLAmismatches.Extendingthesummarystatisticsfurthertomorecloselymirrorthestrategyem-ployedinthispaper,TableB.5showsdifferencesinheterogeneityinregistrationoutcomesamongpatientslivinginmoreandlessexchange-activezipcodes.Forthepurposesofthistable,exchangeactivityismeasuredusingthemonthlyaverageofthenumberoftransplantsviapairedandlistex-changethatoccurwithin50milesofapatient'szipcodeofresidenceinthe12monthsleadinguptothatpatient'sregistrationoutcome(transplantordeathwhilewaiting).Iwilldiscussthecausalimpactsofexchangeonheterogeneoustransplantoutcomesintheregressionanalysesthatfollowthissection.First,TableB.5showsthatbloodtypeABpatientsinthemostexchange-active,comparedtotheleast-active,zipcodesexperiencea39percenthigherrateoflivingdonortransplantation,but16percentlowerratesofdeceaseddonortransplantationand20percenthigherratesofdeathwhilewaiting.Comparethesemostversusleast-activedifferencestopatientsofotherbloodtypes,whichareonly4-9percenthigher,3-5percentlower,and-3to5percenthigher,respectively.InadditiontoshowingthatABpatients,thosewhohavetheeasiesttimecompatibledonors,receivelivingdonationsmuchmorefrequentlyinmoreexchange-activeareas,thesenumbersshowthatthedistributionofABpatients'outcomesaremuchmoreinlinewiththeotherbloodtypedistributionsinmoreexchange-activeareas.LookingatpatientssplitupbyPRAscores,weseethatpatientsinmoreexchange-activeareaswithhigherPRAscores(40-100)experiencethelargestpositivedifferencesinlivingdonationat23percentforPRA40-80and41percentforPRA80-100.Thereisalargedisparityinlivingdo-nationsbetweenlowandhighPRApatientsinareasthatarelessexchange-activeduetoincreasedmatchingdifsasPRAscoreincreases;thesenumbersshowthatthisparticulardisparityis44lowerinareaswithmorelocalexchangeactivity.Turningtothedemographiccharacteristic,age,weseethatlivingdonationis17percenthigherfor18-35yearoldpatientsinmoreexchange-activeareas,comparedto3percent,9percent,and4percentforpatientsaged0-17,36-55,andover55,respectively.Whilepatientsover55seemtoreceivelivingdonationsmorefrequentlyinmoreexchange-activeareas,inrealtermsthisincreaseisoutweighedbyalowerfrequencyofdeceasedtransplantsandhigherfrequencyofdeathswhilewaitingintheseareas.Elderlypatientsexperiencethelowestcombinedtransplantratesofallagegroupsinlessexchange-activeareas;basedonthesedescriptivestatistics,higherlevelsofexchangeactivityappeartoyoungerpatientsattheexpenseoftheelderly.Nextispatientrace.Weseethatwhitepatientshavethehighestratesoflivingtransplanta-tioninlessexchange-activeareas,buthavethelowestpercentagedifference(14percent)whencomparingmoretolessexchange-activeareas.Comparethistoa26to27percentdifferenceforblack,Hispanic,andotherracepatients.Interestingly,mostofthedifferenceinlivingtransplantsbetweenmore-andless-activeareasseemstobeoffsetbydeceasedtransplantsforwhiteandotherracepatients,whilemostofthedifferenceinlivingtransplantsforHispanicpatientsseemstobeoffsetbydeathsonthewaitinglist.Finally,lookingatthedescriptivestatisticsforeducationrevealslittletonoheterogeneityinthedistributionofregistrationoutcomesinmoreversuslessexchange-activeareas.Notethattheremaybeunaccounted-forsystematicdifferencesinthecompositionofpatients,aswellasthetransplantenvironment,inareaswithvaryinglevelsofexchangeactivity.Forthisreason,itwillbeparticularlyusefultocontrolforrelevantobservablepatientcharacteristicsandfortime-invariantheterogeneityacrosscounties.Idiscussmyestimationapproachaccountingfortheseconcernsinthefollowingsection.452.4EstimationThegoalofthispaperistocausallyestimatethenumberofadditionaltransplantsgeneratedperkidneyexchangetransplantfordifferentgroupsofpatients,andthentestwhethersuchestimatesdifferfromoneanother.Tothisend,Iestimatetheeffectofanincreaseintheprobabilityofreceivinganexchangetransplant,interactedwiththevariouspatientcharacteristicsofinterest,ontheprobabilityofexperiencingoneoftheotherregistrationoutcomes.Irestrictestimationtoobservationscorrespondingtooneofthefollowingveoutcomes:directliving,anonymous,exchange,ordeceaseddonortransplant,ordeathonthewaitinglist.Eachobservationintheestimationsampletakesonavalueof1foroneandonlyoneofthevecorrespondingbinaryvariables.Usingthismodel,ratherthanadurationmodelinwhichIcouldincludepatientsstillwaiting,allowsmetoaccountfortheroughly33percentofpatientswhoreceivelivingdonortransplantswithoutregistering.Excludingthesepatientswouldintroducebiasduetoselectionofpatientsontothewaitinglistbeforetheireventualtransplantordeathoutcome.Inthispaper,Iestimateheterogeneouseffectsbyinteractingpatientcharacteristicswithlocalkidneyexchangeactivityandabinaryvariableindicatingtransplantviaexchange.Formypur-poses,themainadvantageofthisinteractionapproachoverasplit-sampleapproachistheeaseoftestingwhethertheestimatedeffectsamongdifferentsubgroupsarestatisticallydifferentfromoneanother.Themaindisadvantageisthatthisapproachrestrictsthecoefofthenon-interactedcontrolvariablestobethesameforallsubgroups.Forexample,whenanalyzingracialheterogeneity,theestimatedeffectofreceivingatransplantinthepast(oneoftheincludedcontrolvariables)onanoutcomeofinterestisrestrictedtobethesameforpatientsofallraces;asplit-sampleapproachwouldallowthisestimatedeffecttovarybyrace.Thefollowingallowsmetoestimatetheheterogeneouseffectsofinterestdirectly:Ykizt=JXj=1qEjCijYEizt+XibE+aEz+gEt+eEizt(2.1)Ykrepresentsthetransplantoutcomeofinterest,includinglivingdirect,deceased,anonymous,anddeathonthewaitinglist.YErepresentstheindicatorforwhetherpatientiinzipcodezinmonth46treceivesatransplantviaexchange.Xisavectorofobservablesforpatienti,includingrace,gender,bloodtype,education,previoustransplantstatus,PRAscore,andageatlisting.14Cijisthecategoricalpatientcharacteristicofinterestforagivenheterogeneityanalysis,includingbloodtype,PRA,age,race,andeducation,andwherejindexestheparticularcategorywithinthegivencharacteristic.NotethatthecharacteristicCiisalsoincludedinthesetofcontrols,Xi,withameasureofthatcharacteristicatleastaspreciseasCi.Forexample,inmeasuringheterogeneityacrosspatientagegroupsIuseagegroupsof0-17,18-35,36-55,andover55,whileincludingacontinuousmeasureofpatientageinthesetofcontrolsXi.Ialsoincludeazipcodeedeffect,arz,tocontrolforanyunobservedheterogeneityacrosszipcodeswherepatientslive,especiallyunobservablescorrelatedwithexchangeactivityandtrans-plantoutcomes.Suchfactorsmightincludeaverageafqualityofnearbyhealthcareinstitu-tions,andproximitytohighereducationorotherresearchinstitutions.Ialsoincludeamonth-yearedeffect,grt,tocontrolfornationwidetransplantationtrendsandnational-levelpolicyshocks.eriztistheidiosyncraticerrorterm.However,sinceYEiztisendogenous,soaretheinteractionsCijYEizt.Therefore,ImustestimatetheparametersofinterestfromEquation2.1usingtwo-stageleastsquares.IuseActivityzt,amea-sureoflocalexchangeactivityinzipcodezinthemonthofoutcomet,asaninstrumentforYEizt.15Here,Imustavoidthefiforbiddenregression"problem,whichwouldarisefromregressingYEiztonActivityztalongwiththefullsetofcontrolsasthestageregressionandtheninteractingdYEiztwithCi1;:::;CiJforthesecondstageregression.Toproperlyestimatetheeffectsoftheseendoge-14PRAisincludedasacategoricalvariable.First,IusethemaximumreportedPRAscoreamongseveralavailablemeasuresincludingcurrent,mostrecent,peak,andendingPRAscores.Ithenconvertthisvalueintoa0ifitisbetween0and10,1ifitisbetween10and20,...,and9ifitisbetween90and100.Then,IconcatenatethiswithanindexvaluecorrespondingtothePRAvariablefromwhichthemaximumPRAscoreisobtained.So,forexample,avalueof21forthecategoricalvariablewouldindicatethattheindividual'smaximumPRAscorecomesfromthesecondPRAvariableandisbetween10and20.ObservationsmissingaPRAscorearegiventheirowncategoryso,whiletheyareexcludedfromthePRAheterogeneityanalysis,theyarenotexcludedfromtheotheranalyses.15PleaserefertoSection1.4.1inChapter1ofthisdissertationfordetailsonthisinstrument.47nousinteractionterms,IuseCijActivityztasinstrumentsforCijYEizt.Therefore,eachendogenousvariable(eachinteractionterm)requiresitsownstagewheretheendogenousinteractionisregressedonalloftheinstrumentsandthefullsetofcontrolvariables.Therefore,thesecondstageis:Ykizt=JXj=1lkjYEiztCijV+Xibk+akz+gkt+ekizt(2.2)whereYkiztisthenon-exchangeoutcomevariable,YEiztCijVisthepredictedinteractionbetweenreceivingatransplantviaexchangeandsubgroupidentityCj,thesetofwhichareobtainedfromestimatingthetstageandlkjisthecausaleffectoftheprobabilityofreceivingatransplantviaexchangeontheprobabilityofexperiencingregistrationoutcomekforpatientsinsubgroupj.Next,Iestimatehowanincreaseintheprobabilityofreceivinganexchangeaffectsthequalityoftransplantoutcomes,conditionalonreceivingatransplant.Iusethesamemethodasoutlinedabove.Thesecondstageisnow:Yqizt=JXj=1lqjYEiztCijV+Xibk+aqz+gqt+eqizt(2.3)whereqnowindexesthefollowingqualityoutcomes:one-yeargraftsurvival,two-yeargraftsur-vival,thenumberofHLAmismatches,andwaitinglistregistrationduration.AonepercentagepointincreaseintheprobabilityofreceivinganexchangetransplantforapatientinsubgroupjresultsinchangeoflqjinqualityoutcomeYq.,aonepercentagepointincreaseintheprobabilityofreceivinganexchangetransplantresultsinachangeofl1yr;jpercentagepointsintheprobabilityofexperiencingone-yeargraftsurvival,andsimilarlyfortwo-yeargraftsurvival.AonepercentagepointincreaseintheprobabilityofreceivinganexchangetransplantresultsinachangeoflHLA;jinthenumberofHLAmismatchesofatransplant.Finally,aonepercent-agepointincreaseintheprobabilityofreceivinganexchangetransplantresultsinachangeoflduration;jinregistrationdurationmeasuredindays.482.5Results2.5.1BloodTypeTableB.6showsthequantityestimatesbybloodtype.Inthecolumn,weseethatpatientswithtypeABbloodexperiencethelargestabsoluteincreaseintheprobabilityofreceivinganexchangetransplantforeachadditionalexchangetransplantoccurringwithin50miles(0.8per-centagepoints,or50percent,foreveryadditionallocalexchangetransplant),andpatientswithtypeObloodexperiencethesmallestgain(0.5percentagepoints,or42percent),withAandBinthemiddle.16Notably,typeA,AB,andBbloodtypesallexperienceastatisticallylargergainintheprobabilityofreceivinganexchangethanbloodtypeOpatients.TypeOpatientsexperiencethelargestcrowd-outoflivingdirecttransplants,whiletypeABpatientsexperiencethelowest.TypeABpatientsexperiencethelargestcrowd-ineffectofanonymousdonation,followedbytypeOpatients,andthentypesAandBpatients,thoughnoneoftheseestimatesarestatisticallydistinguishablefromoneanother.Lookingattheeffectsofexchangeonlivingdirectandanonymousdonationscombined,weseethattypeOpatientsexperiencethesmallestgainsinlivingdonortransplants-only3.4ofevery10exchangesrepresentnewlivingdonortransplants(10exchanges-8.1reductioninlivingdirect+1.5additionalanonymous).Incomparison,theestimatesare8.3fortypeApatients,10.6fortypeABpatients,and6.9fortypeBpatients.FortypeOpatients,the3.4per10rateimpliesthattheexistenceofexchangeincreasedthenumberoflivingdonortransplantsby4percentin2013.Toobtainthisestimate,IusedthetotalnumberoflivingdonortransplantsforbloodtypeOpatientsin2013,2,502,alongwiththetotalnumberofexchangetransplantsforthesesamepatientsin2013,306.Usingtherateofnewlivingdonortransplants,3.4ofevery10exchanges,IcalculatedNewLivingDonorTransplantsLivingDonorTransplantsAbsentExchange.Usingthegivennumbers,thatexpressionyields16NotethatthefiFirstStage"resultscolumninthequantityregressionsdonotdisplaythetruestageresultsasdescribedintheestimationsectionofthispaper.Rather,theseresultsareincludedtodisplaythepresenceofheterogeneityintherelationshipbetweentheinstrumentalvariableandtheprobabilityofreceivingatransplantviaexchange.490:343062;502(0:34306).Usingthesamemethodofcalculation,Ithattheexistenceofexchangein2013increasedtotallivingdonationsby11percentfortypeApatients,12percentfortypeBpatients,and33percentfortypeABpatients.Notethatitmaybepossibletousetheregressionresultstogaininsightonhowanincreaseinlivingdonationsforonepatientsubgroupduetorisingexchangeprevalencecanaffectotherpatientsubgroups.Weseenegative,thoughstatisticallyeffectsofexchangeondeceaseddonortransplantsfortypeA,B,andABpatients,alongwithapositivestatisticallyeffectondeceaseddonortransplantsfortypeOpatients.Thispotentiallysuggeststhatthenon-OpatientsareonnetsubstitutingawayfromdeceaseddonortransplantsandshiftingdeceaseddonorallocationtowardtypeOpatients.ThisisreinforcedbytheobservationthattypeOpatientsexperiencereductionsindeaths(5.6per10exchanges)onthewaitinglistinlinewiththereductionsexperiencedbynon-Opatients(5.9per10fortypeA,6.6per10fortypeAB,and3.7per10fortypeB).Lookingatheterogeneousqualityeffectsacrossbloodtypes,foundinTableB.7,weseesta-tisticallybutindistinguishablegainsingraftsurvivalacrossbloodtypes.Moreover,sincegraftsurvivalratesareonaveragestatisticallyindistinguishableacrossbloodtype(seeninTableB.4),thisimpliesthattherearenotdifferentialeffectsofexchangeongraftsurvivalacrossbloodtype.However,itappearsthata10percentagepointincreaseintheprobabilityofgettinganexchangeincreasesthenumberofHLAmismatchesfortypeOpatientsby0.11(2.9percent)anddecreasesHLAmismatchesfortypeABpatientsby0.15(3.7percent).SinceABpatientsarefiuniversalrecipients"ofotherbloodtypes,suchamaybeduetosomenumberoftypeABpatientswithcompatibledonorsenteringintoexchangesinsearchofclosermatches.17Finally,itappearsthereductionsinregistrationdurationduetotheintroductionofexchangeismostlyattributabletotypeAandABpatients.Theyexperienceroughlya100dayreductioninregistrationdurationforatenpercentagepointincreaseintheprobabilityofreceivinganexchange17Inclusionofcompatiblepairshasbeendiscussedandadvocatedinthetransplantcommunity.SeeGentryetal.(2007),Reesetal.(2009),andWallisetal.(2011).50transplant,implyinga20percentreductionfortypeApatientsanda24percentreductionfortypeABpatients.TheseestimatesarestatisticallydifferentthantheestimatesfortypeBandOpatients,whodonotexperienceastatisticallyreductioninregistrationduration.2.5.2PRAScoreTableB.8showstheheterogeneousquantityeffectsacrossPRAscorecategories.PatientswiththelowestPRAs,theeasiesttomatchallelseequal,experiencethesmallestgainintheprobabilityofreceivinganexchangeasaresultofincreasinglocalexchangeprevalence(0.5percentagepointsforeveryadditionallocalexchangetransplant).However,thisisnotstatisticallydifferentthantheeffectonpatientswithPRAscoresof80to100-thepatientswhoarereceivingaddi-tionalpriorityonthedeceaseddonorwaitinglist.Theeffectforthe0to40patientsis,however,statisticallylowerthanforpatientswithPRAscoresfrom40to80(0.9percentagepoints),whoseemtorespondthemosttolocalexchangeactivity.Inlinewiththis,weseethatPRA40-80patientsexperiencethelargestgainsinlivingdonortransplantquantitywhere8.1ofevery10exchangesrepresentadditionallivingdonortransplantsforthissubgroup(10-3.6+1.7).Fromtheresults,itappearsthatmostofthesepatientsaresubstitutingawayfromthealternativeofwaitingforandreceivingadeceaseddonorkidney.Asfortheothertwogroups,thequantitygainestimatesare5.2forPRA0-40patientsand3.7forPRA80-100patients.Lookingatlivingdirectcrowd-outandanonymouscrowd-inindividually,weseethatPRA40-80patientsexperiencethelowestlivingdirectcrowd-outeffectaswellasthehighestanonymousdonorcrowd-in,whilePRA80-100patientsexperiencethelargestlivingdirectcrowd-outeffectandthelowestanony-mousdonorcrowd-in.Usingtheseestimates,Icalculatethattheexistenceofexchangein2013increasedtotallivingdonationsby6percentforPRA0-40patients,21percentforPRA40-80patients,and18percentforPRA80-100patients.ThequalityresultsforthedifferentPRAcategories,foundinTableB.9,arenotquitewhatwewouldexpectgiventhecorrespondingquantityestimates.ThoughpatientswithPRAscoresof80-100donotthemost,intermsofquality,fromanincreaseintheprobabilityofreceiving51anexchange,itseemstheyenjoythehighestqualityviaincreasedgraftsurvival.A10percentagepointincreaseintheprobabilityofreceivinganexchangetransplantincreasesoneyeargraftsurvivalby4percentagepointsforthePRA80-100patients,or4.4percent.Bycomparison,whilenotstatisticallydifferentfrom0.4,thisestimateis1.6percentagepoints,or1.7percent,forthePRA0-40and40-80patients.Thetwo-yeargraftsurvivalgainsfollowthesamepattern,wherethegainforPRA80-100patients(6.3percent)isstatisticallyhigherthantheestimatesforPRA0-80patients(1.6percentfor0-40and1.4percentfor40-80).Atthesametime,however,thePRA80-100patientsalsoexperiencethelargestgaininthenumberofHLAmismatches:0.2additionalHLAmismatches(6.3percent)forevery10percentagepointincreaseintheprobabilityofreceivinganexchangetransplant,comparedto2.7percentforPRA40-80and0.2percentforPRA0-40.PRA0-40patientsthemostwithrespecttowaitingtime;theyexperienceareductioninregistrationdurationofalmost90days(16percent)fora10percentagepointincreaseintheprobabilityofreceivinganexchangetransplant,whileweseenoeffectonregistrationdurationforPRA40-100patients.2.5.3AgeTableB.10presentstheresultsofaheterogeneityanalysisacrosspatientagecategory.Thethingthatjumpsouthereisthestrangeresultsobtainedforpatientsaged0to17.Whilethestage"effectappearsreasonableat0.4percentagepointsperadditionallocalexchangetransplant,thetwo-stageleastsquaresresultssuggestthat,amongpatientsaged0to17,eachadditionalex-changetransplantreduceslivingdirectdonations3.3andincreasesdeceaseddonortransplantsby3.1.Theseresultsareextremegivenalloftheotherquantityresultsshowingthecrowd-out/ineffectsreasonablydistributedamongthenon-exchangealternatives.LookingatTableB.3,itispossiblethatthismaybeattributedtothelowsamplesizeofpatientsaged0to17(12,797of293,003)andtheevenlowerincidenceofexchangetransplantationamongthem(1percentcom-paredtoanoverallaverageof1.4percent).Someofthesestrangeresultsextendintotheresultsforpatientsaged18to35,whereweseethatexchangesappeartocrowdinbothlivingdirectand52anonymousdonations.However,theresultsappearmuchmorereasonableforpatientsovertheageof35.Forpatientsaged36to55,weseethat5.5ofevery10exchangeswouldhavebeenlivingdirecttransplantsotherwiseand1.3additionalanonymousdonationsaremadeforevery10additionalexchangetransplants.Thisimpliesthat5.8ofevery10exchangesinthisagegrouprepresentnewlivingdonortransplants.Theintroductionofexchangedoesnotappeartohelppa-tientsovertheageof55asmuch;thereweseethat3.3ofevery10exchangesrepresentnewlivingdonortransplants(10-8.3+1.6).Thesecalculationsshowthatpatientsover55experiencethelargestlivingdirectcrowd-outeffect,butalsoa(statisticallyindistinguishably)largeranonymouscrowd-ineffectthanpatientsaged36-55.Usingthequantitygainestimatesforpatients36andolder,Ithattheexistenceofexchangein2013increasedtotallivingdonationsby10percentforpatientsaged36-55and4percentforpatientsover55.Turningtoheterogeneousqualityeffectsacrossage,resultsfoundinTableB.11,Ithatpatientsaged36to55experiencethelargestgainsingraftsurvivalassociatedwithincreasingexchangeprevalence(3.5and4.8percentagepointsor3.8percentand5.4percent,respectively,givena10percentagepointincreaseintheprobabilityofreceivinganexchangetransplant).Nootheragegroupappearstoexperiencestatisticallygainsingraftsurvival.Theresultsshowthatpatientsaged18to35experiencethemostfavorableeffectsonHLAmismatches-areductionof0.3(7.3percent)duetoa10percentagepointincreaseintheprobabilityofreceivinganexchangetransplant.Theresultsshowthatpatientsaged0to17experiencetheleastfavorableeffects-anincreaseof0.5HLAmismatches(14.2percent).Patientsovertheageof35alsoexperiencestatisticallyincreasesinthenumberofHLAmismatches,thoughtheeffectsizesarecantlysmallerthantheeffectsizefor0to17yearoldpatients(2.7percentfor36to55and3.6percentforover55).Theseresultsalsoshowthat0to17yearoldpatientsexperiencethelargestreductioninregistrationdurationassociatedwithanincreaseinexchangeprevalence,thoughtheestimatedmagnitudeappearstobeunreasonablylarge.Aswiththequantityresults,theresultsforpatientsaged0to17maybeplaguedbyissuesrelatedtosmallsamplesize.Patientsovertheageof55appeartoexperiencea200dayreduction(37percent),whilepatientsaged3653to55appeartoexperiencea60day(9.3percent)increaseinregistrationduration.2.5.4RaceTableB.12presentstheresultsofaheterogeneityanalysisacrosspatientsofdifferentraces-white,black,Hispanic,andother.Inthistable,weseethatwhitepatientsaremuchmorelikelythanpatientsofotherracestoreceiveatransplantviaexchangeaslocalexchangeactivityincreases.Experiencinganincreaseof0.9percentagepointsperadditionallocalexchange,thiseffectisslightlymorethanthreetimestheeffectforblackpatientsandtwotimestheeffectforpatientsofotherraces.Lookingatthetwo-stageleastsquaresestimates,Ithatblackpatientsexperiencethelargestlivingdirectcrowd-outeffect,followedbywhitepatientswiththesecondhighestcrowd-outeffect.Ialsothatwhitepatientshavethehighestanonymouscrowd-ineffect,followedbyblackpatients.Meanwhile,Hispanicandotherpatientsexperiencenostatisticallylivingdirectcrowd-outoranonymouscrowd-in.Combiningthecrowd-outandcrowd-inestimatesforwhitepatients,weseethat4.7ofevery10exchangetransplantswouldhavebeenlivingdirecttransplantsandthereare1.9additionalanonymousdonortransplantsforevery10exchangetransplants.Thisimpliesthat7.2ofevery10exchangetransplantsgoingtowhitepatientsrepresentnewlivingdonortransplants.Comparethisto0.4(10-10.6+1)forblackpatients,thoughitappearsthatincreasesinlivingdonationamongwhite,Hispanic,andotherracepatientsledtoblackpatientsexperiencinganincreaseindeceaseddonationrates.Includingdeceaseddonationsyieldstheestimatethat9.5ofevery10exchangesrepresentnewtransplantsforblackpatients.Theseestimatesforwhiteandblackpatientsimplythattheexistenceofexchangein2013increasedtotallivingdonortransplantsby11percentforwhitepatientsand1percentforblackpatients,andincreasedtotaltransplantsforblackpatientsby3percent.Interestingly,thereisnostatisticallycrowdout(in)oflivingdirect(anonymous)transplantsforHispanicorotherracepatients,suggestingthatthesegroupsexperiencethegreatestquantitygainsinlivingdonortransplantationfollowingtheintroductionofexchange.Moreover,54theresultssuggestthatalmostallHispanicpatientsreceivingexchangeswouldhavereceivedde-ceaseddonortransplants,whileotherracepatientswouldhavediedwhilewaitingintheabsenceofexchange.UsingtheestimatesforHispanicandotherracepatients,despitethestatisticalinsignif-icance,impliesthattheexistenceofexchangein2013increasedtotallivingdonortransplantsby12percentforHispanicpatientsand14percentforotherracepatients.Lookingatheterogeneouseffectsinqualityacrossrace,presentedinTableB.13,weseethatblackpatientsexperiencea4.5percentagepoint(5percent)increaseinoneyeargraftsurvivaland6percentagepoint(7.2percent)increaseintwoyeargraftsurvivalgivena10percentagepointincreaseintheprobabilityofreceivinganexchangetransplant.Thisislargerthantheeffectsexperiencedbywhite(1.1percentforboth),Hispanic(2.1percentforone-yearandastatistically-1percentfortwo-year),andotherracepatients(statisticallyreductionsingraftsurvival).OtherracepatientsexperiencereductionsinthenumberofHLAmismatches:0.3(7.5percent)fora10percentagepointincreaseintheprobabilityofreceivinganexchangetransplant.WhitepatientsexperienceanincreaseinthenumberofHLAmismatches,andblackandHispanicpatientsexperiencenostatisticallychangeinHLAmismatches.WhiteandHispanicpatientsexperiencethelargestreductionsinregistrationduration:144days(21percent)fora10percentagepointincreaseforHispanicpatients,and72days(15percent)forwhitepatients.However,thereductioninregistrationdurationforwhitepatientsisnotstatisticallydifferentthantheeffectonblackpatients'registrationdurations.Additionally,Ithatblackandotherracepatientsexperiencenostatisticallyeffectsofanincreaseintheprobabilityofexchangeonregistrationduration.2.5.5EducationFinally,weturntotheeducationresultsinTablesB.14andB.15.Notethatthisanalysisislimitedtopatients18yearsofageorolder.First,weseethatpatientswithmorethanahighschooleduca-tionare0.8percentagepointsmorelikelytoreceiveatransplantviaexchangeforeachadditionallocalexchangetransplantthatoccurs,comparedto0.4percentagepointsforpatientswithahigh55schooleducationorless.Followingthis,weseethat6.6ofevery10exchangesrepresentnewliv-ingdonortransplantsforpatientswithmorethanahighschooleducation(10-4.4+1),comparedto5.3forpatientswithahighschooleducationorless.Note,though,thatneitherthelivingdirectcrowd-outnortheanonymouscrowd-inestimatesarestatisticallysignidifferentfromoneanother.However,itappearsthatpatientswithmorethanahighschooleducationarestatisticallymorelikelytoreceiveadeceaseddonorkidney,asopposedtodyingwhilewaiting,intheabsenceofexchange.Theestimatesimplythattheexistenceofexchangein2013increasedlivingdonortransplantsby8percentforpatientswithuptoahighschooleducationand10per-centforpatientswithatleastsomepost-highschooleducation.Theresultssuggestthatpatientswithmorethanahighschooleducationaresubstitutingtowardexchangesawayfromdeceaseddonationsatarateofabout2per10exchanges.Thismaybefreeingupdeceaseddonorkidneysforthosewithahighschooleducationorless,whoexperienceanestimated,thoughstatisticallyincreaseof2.4deceaseddonationsper10exchanges.LookingatTableB.15,weseethatpatientswithahighschooleducationorlessexperiencealargerone-yeargraftsurvivalincrease(3.3percentcomparedto1.4percent)butasmallerregistra-tiondurationreductionthanpatientswithmorethanahighschooleducationastheprobabilityofreceivinganexchangeincreases(7.3percentcomparedto15.6percent).However,neitherofthesedifferencesarestatisticallyPatientswithahighschooleducationorlessdoexperienceastatisticallyincreaseinthenumberofHLAmismatchesastheprobabilityofreceivinganexchangeincreases,whilepatientswithmorethanahighschooleducationexperiencenoeffect.2.5.6HeterogeneousEffectsonDonorsInadditiontoheterogeneityintheeffectsofexchangeonpatientoutcomes,wemayalsoexpecttheintroductionofexchangetoaffectthecompositionoflivingkidneydonors-particularlywithrespecttobloodtype.Sincebloodtypecompatibilityisextremelyimportantfortransplantsuccess,bloodtypeOdonorsareinhighdemand.However,theintroductionofexchangelikelylessenstherelativedemandfortypeOdonorsbecauseexchangecanreduceinefinmatching.For56example,considerthecircumstancesoftwopatients:atypeBpatienthastwowillingdonors,oneofwhichistypeAandtheotheristypeO,andatypeApatientwithwillingtypeBandtypeOdonors.Forsimplicity,assumethatbloodtypeistheonlycriterionforcompatibility.Withoutexchange,bothpatientsmustrelyontheirwillingtypeOdonorsiftheyaretoreceivealivingdonortransplant.However,withexchange,theycanreceivealivingdonortransplantbyrelyingoneitheroftheirwillinglivingdonors.Toanalyzethechangesinlivingdonorcompositionbybloodtype,foreachbloodtype,Icalculatetheaggregatenumberoflivingkidneydonationsreceivedbypatientsinaparticularzipcodeandmonthfromadonorofthatbloodtype.Iregressthisonthemonthlyaverage(usingthecontemporaneousmonthalongwiththeelevenpreviousmonths)ofthenumberofexchangetransplantsoccurringwithin50milesofthatzipcode.Idonotincludemonthlyzipcodeaveragepatientcharacteristicsascontrolsinthisanalysisduetoconcernsofsimultaneity,sincedonorbloodtypehasaroleindeterminingthecompositionoftransplantrecipients.TableB.16showstheresultsfromthisanalysis.Ithatoneadditionalexchangetransplantwithin50milesincreasesthenumberoftypeOlivingdonorsby2.2percent,typeAdonorsby3.2percent,typeBdonorsby5percent,andtypeABdonorsby7.9percent(thoughthiseffectisstatisticallyTheseareconsistentwithmyargumentthatexchangesrelievetheburdenontypeOdonors,andincreasingthemarginaloflivingdonationbynon-Olivingdonors.Interestingly,theresultsalsoshowthattheintroductionofexchangeincreaseslivingdonationbyalargerpercentageamongdonorswithrarerbloodtypes.18Followinguponthethattheintroductionofexchangecrowdsinanonymousdonation,Ialsoanalyzewhetherpotentialdonorsofcertainbloodtypesaremorelikelytogiveanonymouslyasexchangebecomesmoreprevalent.WemightexpectthattypeOanonymousdonorsarerelativelymorelikelytogivebecausetheirkidneyscanbeusedtostartanysortofdonorchainWemightexpecttypeABdonorsarerelativelylesslikelytogivebecausetheirkidneyscanonlystartachainwherethebeginningpairisatypeABpatient;apatientwho,becausetheyarethe18Inthegeneralpopulation,Oismostrepresented,followedbyA,B,andAB.57universalrecipients,hasablood-type-compatibledonorwhoisincompatibleforsomeotherreason.Toanswerthisquestion,IemploythesameapproachthatIuseinproducingtheTableB.16results.Theonlydifferenceisthatthedependentvariableisnowtheaggregatemeasureofanonymousdonations.IpresenttheseresultsinTableB.17.Interestingly,IthattheeffectontypeO,typeB,andtypeABdonorsareallsimilaratarounda32to35percentincreaseinanonymousdonationforeachadditionalexchangetransplantwithin50miles.TheeffectontypeAdonorsisroughlyhalfofthat,at17.4percent.Theseresultsareinconsistentwiththehypothesis.Onepossibleexplanationisthattherisingprevalenceofkidneyexchangesanddonorchainsmainlyleadstoincreasedpromotionandawarenessofanonymousdonation.Atthesametime,anonymousdonorsmayputrelativelylittleweightonthenumberoftransplantstheyfacilitatebeyondtheone,suchthatatypeABanonymousdonorderivesalmostasmuchutilityfromdonatingtoanunknownwaitinglistcandidateasatypeOdonorderivesfromfacilitatingtwoormoretransplantsviadonorchain.2.6ConclusionAsofApril25,2016,100,102candidatesarewaitingforakidney.19Thegrowingshortageoftransplantableorganshasdriveneconomists,transplantpractitioners,andlawmakerstodevelopcreativesolutions.Theinnovationoftransplantationamongpatientswithincompatiblewillingdonorsviaexchangehasgrowninpopularityinrecentyears,facilitatedbysingle-centerregistriesandconsortiaoftransplantcentersusingcomputer-optimizedmatchingmechanisms.WhileChapter1thattheintroductionofexchangeincreasesthequantityandqualityofkidneytransplants,itisalsoimportanttoconsidertheequityimplicationsofkidneyexchanges.RecentworkbySönmezandÜnver(2015)suggeststhatexchangesarepronetohelppatientsofcertainbloodtypesmorethanothers,whichprovidesmotivationforanalyzingtheextenttowhich19BasedonOPTNwebsite,accessed4/25/2016.58exchangeshelphard-to-matchpatients.Inaddition,thereiswidespreadconcerninthetransplantcommunityaboutdisparitiesintransplantaccessandqualityacrossdimensionsincludingrace,age,andeducation,whichprovidesmotivationforanalyzingthedifferentialeffectsofexchangeonvariousdemographicsubgroupsofpatients.Theevidencepresentedinthispaper,obtainedbyextendingtheestimationapproachandiden-strategydevelopedinChapter1,showsthatthereareindeeddifferentialeffectsofthein-troductionofexchangeonpatientoutcomesacrossimmunologicanddemographicpatientgroups.Lookingatbloodtypeheterogeneity,itappearsthatexchangestendtoimprovethequantityoftransplantsmostfortypeABpatients,leastfortypeOpatients,andsomewhereinthemiddlefortypeAandtypeBpatients.ThisisincontrasttoSönmezandÜnver(2015),whothattypeABpatientsexperiencethelowestgaininquantity,followedbytypeO,typeA,andtypeBpatients.Intermsofquality,typeOpatientsexperienceimprovedgraftsurvivalonparwiththeimprovementsexperiencedbyotherbloodtypes,butfareworsewithrespecttoHLAmismatchesandchangesinregistrationduration.SönmezandÜnverthelargestreductioninwaitingtimeamongtypeAandtypeBpatients,andthelowestfortypeOandtypeABpatients.However,Ithelargestre-ductionsinwaitingtimeamongtypeAandABpatients.Whenlookingatbloodtypeheterogeneityinlivingdonorresponsestotheintroductionofexchange,Ianlargerpercentageincreaseinthenumberoflivingdonorswithmorenarrowly-compatiblebloodtypes(typeBandAB)relativetothepercentageincreaseforthosewithmorewidely-compatiblebloodtypes(typeOandA).However,Idonotsimilarresultswhenfocusingonthenumberofanonymousdonations.LookingatPRA,ameasureofhoweasyapatientistomatchindependentofbloodtype,ex-changesappeartodifferentiallyimpactdifferentPRAgroupsintermsofbothquantityandquality.PatientswithhigherPRAs,butnothighenoughtoreceivewaitinglistpriority,experiencethelargestquantitygainsinresponsetotheintroductionofexchange.WhileIcouldnotunambigu-ouslypredictthisobservedresult,itisindeedconsistentwiththeanalysispresentedinSection2.2.Withrespecttoquality,lowPRApatientsexperiencethelargestreductionsinregistrationdura-tion,whilehighPRApatientsexperiencethelargestimprovementsingraftsurvival.Eventhough59PRA40to80patientsexperiencethelargestpercentagegaininlivingdonations,andexperienceimprovements(thoughstatisticallyingraftsurvivalandregistrationduration,theydonotmoreinanyofthesedimensionscomparedtoanyotherPRAgroup.Turningtodemographicheterogeneity,itappearsthatexchangesimprovetransplantquantitymoreforyoungerpatients(excludingminors)ratherthanolderpatients.Whilepatientsofallagesexperiencequantityimprovements,exchangesappeartoexacerbatetheinequalityinlivingdonationquantityacrossagegroups.Theeffectsonqualitydisparitiesacrossagearelessclear,however.Elderlypatientsexperiencegreaterreductionsinregistrationdurationthanyounger(non-minor)patientsbutless-favorableeffectsongraftsurvivalandtissue-typematchquality.There-sultsshowthatracialquantitydisparitiesinlivingdonationbetweenwhiteandblackpatientsareexacerbatedbytheintroductionofexchange,butracialquantitydisparitiesbetweenwhitepatientsandHispanicorotherracepatientsaremitigatedbytheintroductionofexchange.However,IthatblackpatientsexperiencegainsindeceaseddonortransplantsfromkidneysfreedupbyHis-panicandwhitepatientssubstitutingtowardexchangeawayfromdeceaseddonation.Exchangesmitigatetheracialdisparityingraftsurvival(withblackpatientsexperiencingthelargestgainsingraftsurvival),whileexacerbatingtheracialdisparitiesinregistrationduration(withwhiteandHispanicpatientsexperiencingthelargestreductions).Finally,theintroductionofexchangeap-pearstohavenoeffectonquantitydisparitiesineducation,thoughitdoesappeartounfavorablyimpactrelativetissue-typematchqualityamonglesseducatedpatients.Thispaperservestoinformthedesignofkidneyexchangematchingalgorithms.Fortransplantcentersandexchangeconsortiaoperatingwiththegoalofimprovingpatientoutcomesforall,aswellasmitigatinginequalitybetweendifferentpatientgroups,theresultspresentedheresuggestthatexchangeprogramsmaywanttoplacegreateremphasisondisadvantagedpatients.Addition-ally,listexchangeshavebecomelesspopularinrecentyears.Thispapermayalsoservetobreathenewlifeintothepracticeoflistexchange,sincetheycouldfacilitatetheallocationoflivingdonorkidneysfromincompatiblepairstodisadvantagedpatientsonthedeceaseddonorwaitinglist.60CHAPTER3ALLOCATINGSCARCEORGANS:HOWACHANGEINSUPPLYAFFECTSTRANSPLANTWAITINGLISTS(withStacyDickert-ConlinandToddElder)3.1IntroductionTheNationalOrganTransplantActof1984decreedthatitisfiunlawfulforanypersontoknow-inglyacquire,receive,orotherwisetransferanyhumanorganforvaluableconsiderationforuseinhumantransplantation.flIntheabsenceofapricingmechanismforthisscarceresource,vastorganshortageshavedeveloped,withroughly122,000personsawaitingorgantransplantsintheU.S.1Thisnumbergrowsdramaticallyeveryyear,inspiteofnumerouseffortstoincreasethesupplyoftransplantableorgans,includingeducationalcampaigns(Siminoffetal.,2009;Rodriguezetal.,2007),socialmediaoutreach(Cameronetal.,2013),andcoordinationofpairedkidneyexchanges(Rothetal.,2004,2005;AusabelandMorrill,2014).Additionalreformproposalsincludemovingtoasystemofpresumedconsentfordonors(AbadieandGay,2006;Bilgel,2012),allowingcialexchangesfororgans(BeckerandElias,2007;Laceteraetal.,2014;WellingtonandSayre,2011)andalteringtheorganallocationrulestoinducemoredonations(KesslerandRoth,2012;Lietal.,2012).Theevidenceonthesuccessoftheseeffortstoincreasethesupplyoforgansislimited,andweknowverylittleabouthowashiftinsupplyoforgansmayaffecttransplantcandidates'behaviorandoutcomes.2Withoutapricingmechanisminplace,theeffectofanincreaseinthesupplyoforganswilldependonthenatureofthealternativesystemforallocatingthescarceresource.TheUnitedStatesgovernmentoverseesasystemforallocatingorgansthatattemptstoaddressabalanceofequityandefyŒasthenationwideOrganProcurementandTransplantationNetwork(OPTN)de-it,abalanceoffijustice(fairconsiderationofcandidates'circumstancesandmedicalneeds),1http://optn.transplant.hrsa.gov/,accessed11/09/2015.2Fernandez,HowardandStohr(2013)areanexception,inthattheyconsidertheeffectofanincreaseindeceasedkidneydonorsonlivingkidneydonations.61andmedicalutility(tryingtoincreasethenumberoftransplantsperformedandthelengthoftimepatientsandorganssurvive).fl3Thesystemiscomplexandvariesbyorgan,butitgenerallybeginsbygeneratingawaitlistofmedicallycompatibletransplantrecipientsinageographicarea.Ge-ographicproximityplaysacentralrolebecauseorganshavealimitedtimewhentheyareviablebetweenprocurementandtransplantation.4Asaresult,shockstothelocalsupplyoforganswilllikelyaffecttheoutcomesandcompositionoflocaltransplantwaitlists.Wemodeleachofthesegeographicareasasamarket,wherewaitingtimeactsasasignalofthescarcityoforgansgeneratedbythesupplyofdeceasedorgansandthedemandfordeceasedorgansinthefuture(LindsayandFeigenbaum,1984).Theeffectofashocktothesupplyoforgansdepends,inpart,onthebehavioralresponsestothatshock.Inthesimplestmodel,apositiveshocktothesupplyofdeceasedorgansinonemarketislikelytoreducewaitingtimeandincreasethequantityoftransplantcandidatesjoiningthewaitlistinthatmarket.Themarginalentrantsintothemarketfordeceasedorgansarelikelytobethosewithhighercostsand/orlowerofenteringawaitlist,ascomparedtopre-existingwaitlistregistrants.Withthisbasicmodelinmind,weusedataonorgandonorsandtransplantrecipientsfromtheRegistryofTransplantRecipients(SRTR)toconsiderwhethershiftsinthesupplyoftransplantableorgansaffectthebehaviorandoutcomesoftransplantcandidatesandtheirphysi-cians.Webeginbyestablishing,asinDickert-Conlin,ElderandMoore(2011;DCEMhereafter),thatvariationinstate-levelmotorcyclehelmetlawsgeneratesshockstothesupplyofdeceasedorgans.Becauseeachdeceaseddonorcanpotentiallycontributemultipleorganstopersonsonmultiplewaitlists,weextendDCEM'sanalysisbyquantifyinghowhelmetlawsaffectthesupplyofindividualorgans.Weestimatethatrepealsofstatewidehelmetlawsincreasethelocalsupplyoftransplantableorgansfromdonorskilledinmotorvehicleaccidentsbynearly20percent.Theseshocksareparticularlylargeforlungs,kidneys,andlivers.3http://optn.transplant.hrsa.gov/learn/about-transplantation/how-organ-allocation-works/4OPTNreportsthemaximumpreservationtimesforheartsandlungsat4to6hours;liverat8to12hours;pancreasat12to18hoursandkidneyat24to36hours(fromhttp://optn.transplant.hrsa.gov/learn/about-transplantation/how-organ-allocation-works/).62Next,wethattransplantcandidatesrespondstronglytolocalsupplyshocks,withwstolocaltransplantwaitlistsincreasingbyroughly12percentfollowinghelmetlawrepeals.ThesewsarelargelydrivenbythosewhohavehighermarginalcostsoflistinginthatareathanthosealreadylistedasmeasuredbythedisproportionatelylargenumberofmarginaljoinerswholiveoutsidetheDSA,relativetocandidatessigningupfortheirfihomeflwaitlists.Moreover,wethatcandidateswhoarelistedonmultiplewaitlistshavebyfarthelargestresponsetohelmetlawrepeals,withwsontowaitlistsincreasingbyover40percentrelativetobaselineamongthesecandidates.Third,wethatcrowdoutoflivingdonationmayaccountforsomeoftheincreaseinwaitlistfordeceasedorgans.5Theseeffectsaremostpronouncedforpotentialtransplantsfromlivingdonorswhoarenotbloodrelativesorspousesofthecandidate,suggestingthatthesearedisproportionatelythefimarginalflcaseswheretherelativecostsoflivinganddeceaseddonorsaremostOuronbothwaitlistwsandlivingdonorssuggestthatincreasesinthesupplyoftransplantableorgansgeneratesbehaviorthatatleastpartiallyoffsetsthedirecteffectsofahighersupplyoforgans.Inordertoestimatetheoveralleffectoncandidateoutcomes,weesti-matehowincreasesinorgansupplyeffectshealthoutcomesfortransplantcandidatesfollowingahelmetrepeal.Weevidencethatanincreaseinthesupplyofdeceasedorgansincreasestheprobabilitythatatransplantissuccessful(knownasfigraftsurvivalfl)forone,three,andveyearspost-transplant.Finally,ourraisequestionsaboutthebalanceofjusticeandmedicalutilityinthecurrentallocationmechanism,whichreliesheavilyongeographicboundaries.Thosetransplantcandidateswhohaveinformationaloradvantagesmightbemostlikelytobeabletocapitalizeonadvantageousshockstosupplyandwaitingtime.Forexample,severalarticlesinthepopularpressalludedtothelackoffifairnessflintheorganallocationmechanismin2009when5TheseareconsistentwiththoseofFernandez,HowardandStohr(2013),whoestimatethatanincreaseinthesupplyofdeceasedkidneydonorsnearlycompletelycrowdsoutkidneydonationsamongnon-biologically-relatedlivingdonors.63SteveJobs,wholivedinCaliforniaatthetime,obtainedalivertransplantinMemphis,Tennessee,whichhadamedianwaittimeroughly85shorterthanthenationalaverage.6Inthefollowingsectionweexplaintheinstitutionalsettingunderlyingorgantransplantsanddescribeourdatasourcesonorgandonationsandtransplants.Section3.3estimatesacausalre-lationshipbetweenhelmetlawsandorgandonations.InSection3.4wedescribeasimplemodelwherewaitingtimeservesasapriceinmarketwherebuyingandsellingthegoodisprohibited.InSection3.5weestimatetransplantcandidates'responsestothesupplyshocksestimatedinSection3.3,andSection3.6considershowthesupplyanddemandfororganscombinetoaffecttransplantcandidates'outcomes.Section3.7concludes.3.2DataandInstitutionalDetails3.2.1DataonOrganDonationsandTransplantsThecurrentsystemofallocatingorgansfromdeceaseddonorsoriginatedwiththe1984NationalOrganTransplantAct.Inadditiontoprohibitingtheexchangeoforgansformonetarycompen-sation,theActcreatedtheOPTNastheoverseerofasingleorganization,theUnitedNetworkforOrganSharing,chargedwithensuringanequitableallocationsystem(seeLeppkeetal.,2013,formoredetails).7ThisstudyusesdatafromtheRegistryofTransplantRecipients(SRTR).TheSRTRdatasystemincludesdataonalldonors,waitlistedcandidates,andtransplantrecipientsintheUS,submittedbythemembersoftheOrganProcurementandTransplantationNetwork(OPTN).TheHealthResourcesandServicesAdministrationoftheU.S.DepartmentofHealthandHumanSer-6AsubstantialpartofthecriticismwasbasedontheargumentthatJobsusedhismeanstoobtainanorganthatmightbefibetterservedflbybeingtransplantedtoacan-didatewithoutmetastaticpancreaticcancer,whicheventuallyledtoJobs'deathin2011.Seehttp://www.cnn.com/2009/HEALTH/06/24/liver.transplant.priority.lists/index.htmlforanexampleofthissortofresponsetoJobs'situation.7Seehttp://history.nih.gov/research/downloads/PL98-507.pdfformoredetailsonthecreationofOPTNanditsrelationshiptoUNOS.64vicesprovidesoversighttotheactivitiesoftheOPTNandSRTRcontractors.TheSRTRdatacomefromhospitalsandimmunologylaboratories,andincludedetailedinformationonpersonsontheorgantransplantwaitlists,includingtimespentonthewaitlist,transplantcentersatwhicheachpotentialrecipientisregistered,healthmarkers,limiteddemographicsincludingzipcodeofresi-dence,andreasonforleavingthewaitlist.EachobservationintheSRTRrepresentsaregistration,sowecanalsoobserveindividualswhoarelistedatmultipletransplantcentersandwhotransfertodifferenttransplantcenters.Thesedatacanbematchedtodetaileddonationdatatoviewthecircumstancesofthedonor'sdeathineachtransplantrecipient'scase.Patientswhoreceivetrans-plantsfromlivingpersonswerenotrequiredtoregisteronwaitinglists,butSRTRdataidentifylivingdirecttransplantrecipientsandtheirtransplantoutcomes.3.2.2TransplantWaitingListsPatientsneedingatransplantfromadeceaseddonormustregisterononeormoreofOPTN'sorganwaitinglists.Todoso,theymustobtainauthorizationfromaphysicianwhoisassociatedwithoneofroughly300transplantcentersintheUnitedStates.8Atransplantcoordinatoratthetransplantcenteroverseestheprocessofmedicaltestingtodeterminemedicaleligibilityandliststhecandidateonthewaitinglistforanorganfromadeceaseddonor.Eachtransplantcenterislocatedinoneof58donationserviceareas(DSAs),whicharecrucialorganizationalunitsintheorganallocationprocess.AnOrganProcurementOrganization(OPO)isthelocalmonopolywithinitsDSA,exclusivelyresponsibleforcoordinatingandfacilitatingdonationservicesbetweendonorsandtransplantcenters.9Thisincludesevaluatingpotentialdonorsandarrangingforsurgicalremovaloforgansaswellaspreservingorgansandarrangingfortheirdistributiontocandidatesonorganwaitinglists.AsFigureC.1shows,thebordersoftheDSAsbroadlyfollowstateboundaries,althoughsomelargestateshavemultipleDSAsandsomeDSAscrossborderstoincludemultiple8Thefollowinglistsadirectoryoftransplantcenters:http://optn.transplant.hrsa.gov/converge/members/search.asp9Forconsistency,wewilluseDSAthroughouttherestofthepapertorefertothegeographicareas.65statesorportionsofstates.TransplantcandidatesmayalsoregisteronmultiplewaitinglistsindifferentDSAs,aprocessdescribedasfimultilistingflhenceforth.Beyondthecostsofrequiringtransplantcandidatestobeabletophysicallyarriveintimetoreceiveanorgandonationwhiletheorganisviable,theremayalsobetransplantcenterrulesabouttheprocessrequiredforamultilistingŒacandidatemayhavetogothroughaseparateevaluationforeachlist,whichmaynotbecoveredunderin-surance,andsometransplantcentersmayrefusepersonswhoarewaitlistedatothertransplantcenters(UnitedNetworkforOrganSharing,2014).10CalculationsusingtheSRTRdatashowthatmultilistingisnotcommon,withonly6percentofallcandidateschoosingtomultilist,buttheseregistrationsthatarepartofmultilistedspellsrepresentapproximately12percentofallregistra-tions.Thehighestincidenceisforkidneysandpancreases,inpartbecausewetreatpersonswhoarewaitingforbothkidneysandpancreasesasamultilisting,butvariationinmultilistingacrossorganmayalsotheviabilityoforgansoncetheyareprocuredandtheabilityofkidneypatientstostayhealthyenoughtotravel(seeAppendixE.1forhowwemultilistedcandidatesandspellsinthedata).TableC.1showsaggregatewaitlistadditionsbyorganandyear,whichincludemultilistings.Inrecentyearsmorethan50,000newcandidatesjoinedtransplantwaitlists,withkidneysaccountingforanincreasinglylargeshareofallthewsŒ65percentin2012and2013.Additionstothewaitlistforliversaccountfor10,000to12,000oftheadditionsinthelasttwodecades,withoutmuchchangeoverthistimeincomparisontothedramaticgrowthforkidneys.Waitlistadditionsforheartsandlungsarealsorelativelysteadyovertime,whileadditionstothepancreaswaitlisthavefallensincetheirpeakintheyear2000,representingbetween3and5percentoftheoveralladditions.Fewerthan300personsenterthewaitlistforintestinesinmostyears.TableC.2showscountsofwsfromtransplantwaitlistsbyyear.AsacomparisonofthelastcolumninthetabletothelastcolumnofTableC.1shows,thenumberofpersonswhoexit10Apatient'saccruedwaittimemayormaynottransfertothenewlisting.Seehttp://optn.transplant.hrsa.gov/learn/about-transplantation/transplant-process/formoredetailsonthewaitinglistprocess.66waitlistseachyearisconsistentlybelowthenumberofnewadditions.Receivingatransplantfromadeceaseddonoristhemostcommonrouteoffthewaitinglist,accountingfor22,935ofthe52,480exitsin2013.WhenadeceaseddonorbecomesavailablewithinagivenDSA,acomputersystemgeneratesapoolofeligiblerecipientsfromthewaitlistbasedonbloodtype,othercompatibilitymeasuresandcandidates'willingnesstoacceptthequalityoftheorganoffered(OPTN,2015).11Withinthepoolofpotentialmatches,thecomputergeneratesarankingofcandidatesbasedontimeonthewaitinglist,urgencystatus,anddistancefromthedonororgan.Theweightgiventoeachofthesecharacteristicsdependsontheorgan,andothercharacteristicsmayalsoplayarole.Forexample,a2005rulechangeknownasfiShare35flgaveprioritytokidneytransplantcandidatesunderage18toreceiveorgansfromdeceaseddonorsunderage35.Ingeneral,theOPOoffersthedeceaseddonor'sorgantothecandidatewiththebestmatchintheDSA'spoolofmatches,makinggeography,aswellasorgantype,akeycomponentoftheallocationprocess.12Ifthecandidate'sphysiciansaccepttheorgan,thetransplantoccurs;otherwise,theorganisofferedtothenextpersononthelist.ThenextoffermaybemadewithintheDSAortoagoodmatchoutsidetheDSA(totheregionandthennationally,inthecaseofkidneys).13Apersonmayalsoleavetheorgantransplantwaitlistforotherreasons.TableC.2showsthat,incombination,deathanddeterioratinghealthisthesecondmostcommonreasonforexitingawaitlist(thecategorylabeledfimedicallyunsuitableflwasusedpriorto1996andwaseventuallydividedintothefideteriorated,toosickflandfiimproved,transplantnotneededflcategories).Inothercases,thecandidate'shealthimprovesandtheyareremovedfromthelistbecausetheyareno11Since1999,UNetisthecomputersystemthatgeneratespotentialmatches.AnadditionalsystementitledDonorNetwasaddedin2003anditsusewasmandatedin2007.12Thereareexceptionstothisgeographicallocationprocess:sharingarrangementsexistbe-tweenOPOsinter-orintra-regionally,althoughOPTN'sBoardofDirectorsmustapprovesucharrangements.13IntheSRTRdata,weestimatethatabout2/3ofallorgansaretransplantedinthesamegeo-graphicareainwhichtheyareprocured.Thisnumberhasgrownovertime,withthehighestshareforkidneysandkidney/pancreastransplants.OneexplanationfortransplantsoutsidetheregionisthataDSAmaynothaveatransplantcenterforallorgansinitsboundaries.67longerunhealthyenoughtoqualifyforanorgan.Inthecaseofkidneysand,rarely,otherorgans,apersonmightleavethewaitlistbecausetheyreceivedanorganfromalivingdonorŒapproximately10percentofcandidatesleavewaitlistsviathisroute.Somecandidatesexitawaitlistbecausetheytransfertoothercenters.14Inaddition,ifacandidateisawaitingamulti-organtransplant,suchasakidneyandpancreas,andreceivesoneoftheorgans,theyexitfromthemulti-organwaitlist.Notalltransplantrecipientsareontransplantwaitlists:between1987and2014,forexample,roughlyone-thirdofallkidneytransplantsinvolvedalivingdonationtoapatientwhoneverregisteredonawaitlist.Waitlistwsandwsgeneratevariationovertimeinthenumberofcandidatesonwait-lists.TableC.3showsthisvariationoverallandbyorgan,illustratingthescaleoftheorganshort-ages.Thesearecountsatapointintimeduringtheyearandincludeactiveandinactivemembersofthewaitlist,whereinactivewaitlistregistrantsarenotcurrentlyeligibleforatransplant,typicallybecauseofpoorhealth.Thenumberofregistrationsexceedsthenumberoftransplantcandidatesbecausepersonscanbemultilistedacrosstransplantcentersorformultipleorgans.Waitinglistlengthsvarydramaticallyacrossorgans,soshiftsinthesupplyoforgansresultingfromanyshockhavepotentiallyverydifferentimplicationsforaffectingthewaitlists.Whilethenumberofpersonswaitingforlungsandheartshasbeeninthethousandsinrecentyears,morethan10,000patientswerewaitingforliversandnearly100,000wereawaitingakidneyin2013.Timeonthewaitlistalsovariesdramaticallybyorgan,DSA,andmedicalfactorssuchasmatchprobability.TheOPTNAnnualreporthighlightsthevastdifferencesbygeographicregion:fitheproportionofadultsreceivingdeceased[liver]donororganswithin5yearsoflistingrangedfrom30.5Inthecontextofthedramaticgrowthinwaitlists,aswellassubstantialgeographicdisparityintheexpectedwaitingtimes,weconsiderhowanexogenousshiftintheavailabilityoforgansaffectsbehaviorandoutcomesamongtransplantcandidates.14Generallywaitingtimetransfersifacandidatetransferstoanothertrans-planthospital,buthospitalsarenotrequiredtoacceptexistingwaitingtime.http://optn.transplant.hrsa.gov/about/transplantation/transplantProcess.asp683.2.3HelmetLawsandWaitingListsWeusechangesinmotorcyclehelmetlawstoidentifyanexogenousshiftinthesupplyoforgandonors.TableC.4thetimingoflawchangesbetween1988and2013,forwhichwehavedata.Theproposedmechanismisthatthattherepealofauniversalhelmetlawincreasesthenumberofpeopleridingmotorcycleswithoutahelmet,whichinturnincreasestheprobabilityofbraindeathfromamotorcycleaccident.Braindeathistheprincipalcriteriaforbecomingadeceasedorgandonor;almostalldeceasedorgandonorsarebraindeadatthetimeoforganrecovery,despitebraindeathoccurringinlessthan1percentofalldeathsintheU.S.TableC.5showstheannualnumberoforgandonorswhodiedinmotorvehicleaccidents(MVA)andinallothercircumstances.Thetoppanelalsoshows,byorgan,thenumberoforganstransplantedperMVAdonor,whilethebottompanelshowstheanalogousnumbersfornon-MVAdonors.15MVAsaccountforanaverageof1300organdonorsperyear,althoughthisnumberhasfallensincethepeakyearof2006,withtheshareofalldeceaseddonorsthatdiedinMVAsfallingfromroughly24percentinthe1990stoonly16percentin2013.AmongMVAdeaths,thedatadonotdistinguishbetweenmotorcyclistsandnon-motorcyclists.Usingpublished,state-levelOPTNdatafrom1994to2007,DCEMusesixstate-levelrepealsandoneenactmentofauniversalhelmetlawtoestimatethatrepealinguniversalhelmetlawsin-creasesthesupplyoforgandonorswhodieinmotorvehicleaccidentsbyroughly10percentoveralland31percentamongmenaged18to34.Moreover,theirestimatesimplythateverymo-torcyclistdeathduetothelackofauniversalhelmetlawproduces0.124additionalorgandonors.16Becausetheentireeffectisdrivenbymaledonorsaged18-34,whoaredisproportionatelylikely15MotorvehicleaccidentswererecordedasacircumstanceofdeathforallyearsintheSRTRdata.BeforeApril1994,circumstanceofdeathwascodedaseitherMVAorfiallothercircum-stancesfl,butamorecompletecodingofcircumstancesofdeathbeganinApril1994(thesecir-cumstancesincludesuicide,otheraccidents,stroke,homicide,andchildabuse).16DCEM(2011)showthathelmetlawrepealsinducegradualincreasesinbothmotorcyclistdeathratesandorgandonationrates;ridersapparentlydonotabandontheirhelmetsimmediatelyfollowingarepealedhelmetlaw.Incontrast,behaviorrespondsessentiallyimmediatelytotheintroductionofhelmetlaws.69todieinmotorcycleaccidents,theirresultssuggestthattheresponseinMVAdonorsisduetomotorcyclists.Weexpectthattherelationshipbetweenthechangeindonorsandoutcomesfortransplantcan-didatesmaydifferbyorgan.First,asTableC.3shows,thesizesofwaitlistsdifferdramaticallyacrossorgans,soashocktothenumberoforgansavailablewillpotentiallyaffectcandidatesdiffer-entlydependingontheorganthattheyneed.Second,adonor,asfiapersonfromwhomatleastoneorganwasprocuredforthepurposeoftransplantfl(OPTN,2013),cancontributemultipleorganstothedeceaseddonorwaitlist(includingtwoeachofkidneysandlungs),buttheproba-bilityofaorganbeingtransplanteddiffersdramaticallyacrossorgans.TableC.5showsthatthenumberoforganstransplantedperdonorvariesovertime,butinallyearsitishigherforMVAdonorsthanfornon-MVAdonors;forexample,in2013,eachMVAdonorcontributed3.85organsthatwereeventuallytransplanted,comparedto2.93organsforeachnon-MVAdonor.Onaverage,in2013eachMVAdonorcontributed1.81kidneysthatareeventuallytransplanted.Sim-ilarly,approximately81percentofMVAdonorscontributeusablelivers.However,onlyhalfoftheMVAdonorscontributeheartsandevenfewercontributeapancreasorlung.Thenumberoflungstransplantedperdonorisstrikinglylow,giventhateachdonortypicallyhasthepotentialtodonatetwolungs.Medicaltreatmentofbraindeadvictimscandamagesomeorgansmorethanothers,lungsincluded,butthetreatmenttechnologyischangingovertimesothatthefractionoflungstransplantedfromdonorsisalsoincreasingovertime(Marceloetal.,2011).3.3TheEffectsofMotorcycleHelmetLawsontheSupplyofOrgansAllanalysesbelowusetheDSAastheunitofobservationbecausethisistheprimarygeographicunitwhenallocatingdeceasedorgans.TheCenterforMedicareandMedicaidServices(CMS)assignscountiestoDSAsandtheOPOintheDSAcoordinatesalldonationsandtransplants.Weusethemostrecentcounty-DSAdesignationprovidedbySRTR,whichisimperfect,butappearsrobusttoalternativechoicesdescribedinAppendixE.2.WeexcludePuertoRico,resultinginthe70inclusionof57DSAsinouranalysis.WeestimateDSA-andyeardonationratesasafunctionoftheshareoftheDSA'spopulationlivinginastatewithoutauniversalhelmetlawinthatyearusingthefollowingmodel:Ddt=ad+st+g(nolawshare)dt+edt(3.1)Weestimateequation(3.1)separatelyforseveraldifferentdependentvariables.Intheregres-sion,Ddtisameasureofthenumberofdeceasedorgandonorsduetomotorvehicleaccidents(MVA).Inthesecondsetofregressions,Ddtmeasuresofthenumberoforgans(kidney,liver,hearth,lung,pancreasandintestine)thataretransplantedfromMVAdonors.Inallcases,wemeasureDdtpermillionDSAresidentsusingNationalCancerInstitute(2015)countypopulationestimates.Ineachregression,dindexestheDSA,tindexestheyear,rangingfrom1988to2013,andnolawsharedtistheshareoftheDSA'spopulationnotcoveredbyauniversalhelmetlawforatleastsixmonthsinyeart.AllincludeafullsetofDSAandyearindicators(adandst,respectively).WeweighteachobservationbytheDSA'spopulationinthatyear.Estimatesofgbasedon(3.1)capturetheassociationbetweenwithin-DSAvariationovertimeinmandatedhelmetlawsandwithin-DSAvariationinthesupplyoforgandonorsandorganstransplanted.Asanexampleofourkeyindependentvariable,considertheDSAthatincorporatescountiesinwesternPennsylvania,WestVirginiaandonecountyinNewYork.Allofthosecountiesareinstatesthathaduniversalhelmetlawsuntil2003,whenPennsylvaniarepealedtheiruniversalhelmetlaw.ThelawchangewentintoeffectinAugustof2003,soin2004,nolawsharedtincreasesfrom0toabout0.75,whichrepresentstheshareoftheDSA'spopulationlivinginPennsylvania.Theestimatesofgfromequation(3.1),reportedinTableC.6,suggestthatrepealsofuniversalhelmetlaws,whichincreasetheshareofpersonsnotcoveredbyahelmetlaw,areassociatedwithmoreMVAdonorsandtransplantsoforgansrecoveredfromMVAdonors.Column(1)showsthatrepealingauniversalhelmetlawincreasesthenumberofMVAorgandonorsby0.906permillionpersons,withastandarderrorof0.234(allstandarderrorsarerobusttowithin-DSAclusteringovertime).Thisrepresentsa19percentincreaserelativetoasampleaverageof4.886donorspermillion71persons,asshowninbrackets.Column(2)showsthatthisincreaseinthesupplyofdonorsresultsinanaverageof3.429moreorganstransplantedpermillionDSAresidents,consistentwiththesummarystatisticsonorganstransplantedperMVAdonorshowninTableC.5.Columns(3)and(4)showanalogousresultsfornon-MVAdonorsandtransplantsasaplacebotest.Theestimatesinallcasesarestatisticallyandsmallrelativetotherelevantsamplemeans.TableC.6alsoshowsthattheeffectsofhelmetlawsontransplantsvarysubstantiallybyorgan.FocusingonMVAdonors,themoststrikingestimateisamonglungtransplants:repealinghelmetlawsincreasesthenumberoflungstransplantedby0.434permillionpersons,whichisa33percentincreaserelativetothesamplemeanof1.299.Kidneydonationsincreaseby1.577permillionpersons(19percentrelativetothesamplemean),heartdonationsincreaseby0.421permillionpersons(17percent),andlivertransplantsincreaseby0.774permillionpersons(21percent).Theestimatedeffectsofhelmetlawrepealsarepositiveforpancreaticandintestinaltransplantsbutarenotstatisticallyatstandardlevels,perhapsthelowrateoftransplantationperdonoramongthoseorgans.Wealsoestimatefieventstudyflmodelstohighlightthedynamicresponsesoforgandonorsandtransplantstorepealsofhelmetlaws.,westudytheresponsesovertimeinMVAorgandonorsandtransplantsinDSAsthatareheadquarteredinstateswithhelmetlawrepealsduringthe1987-2013period,focusingonhowthesequantitiesevolveovertimeincomparisontothoseinDSAsinstatesthatdonotrepealhelmetlaws.17InallDSAs,themajorityofthepopulationresidesinthestatewheretheDSA'sOPOisheadquartered,soweuseasetofdummyvariablesindicatingfiyearssincerepealflfor5yearsbeforeandaftertherepeal,withtheomittedcategorybeingtheyearbeforetherepeal:Ddt=ad+st+5Xt=5rd;ttgtt+edt(3.2)whererd;ttisabinaryvariableequaling1ifarepealoccurredinDSAdinyeartandthe17Samplesizesaretoosmalltoestimatesimilareventstudiesforhelmetlawenactmentsbecausealloftheenactments(exceptforLouisiana'sin2004)occurredinthefewyearsofourdatacoverage,precludingreliableestimatesofthepre-existingtrendsinthosestates.72observationoccurredinyeart,andzerootherwise.Inpractice,wegroupobservationsatleast5yearsbeforetherepealintoasinglecategory,andwegroupobservationsatleast5yearsafterwardintoanothersinglecategory.FigureC.2plotstheseriesofgttestimatesforMVAdonorsandMVAtransplants,bothag-gregatingacrossorgansandseparatelyforkidneys,livers,hearts,lungs,pancreasesandintestines.Theestimatesshowthat,relativetotheyearbeforetherepeal,boththenumberoforgandonorsandthenumberoftotaltransplants(bothnormedbytheDSApopulation,inmillionsofpersons)increasedintheyearoftherepealandinthefollowingyear.Afterward,thenumberofdonorsandtransplantsdonotappeartoincreasefurther,buttheyalsodonotreturntothebaselinelevel,suggestingthattherepealofhelmetlawsresultsinanewlong-runlevelofbothdonorsandtrans-plants.Theshiftisespeciallydramaticforkidneysandlivers.ConsistentwiththeestimatesfromTable6,theredoesnotappeartobeanyshiftforintestines.Apartfromshowingthedynamiceffectsofhelmetlaws,FigureC.2isusefulinillustratingthekeythreattothevalidityofestimatesfrommodelslikeequation(3.1)ŒthatDSAsinstateswithchangesinhelmetlawsmighthavedifferenttrendsindonorsandtransplantscomparedtostateswithnochangesinhelmetlaws.Themosttransparentmethodforshowingwhetherthesedifferencesintrendsexististopresentthemgraphically,andjudgingfromthelinesinthefrom5+yearsbeforerepealstotheyearbeforerepeals,thereisverylittleevidencethatpre-repealtrendsweredifferentinrepealstatesthaninotherstates.1818Asanalternativeapproach,weestimatedmodelsincludingtrendsineitherdonorsortransplantspercapita.Theestimatesfromthosemodelsareremarkablysimilartothosefrom(3.1).Forexample,theinclusionoflineartrendsincreasesthepointestimateinthecolumnofTable6from0.906to0.986(andincreasesthestandarderrorfrom0.234to0.236)anddecreasesthepointestimateshownintherowofthesecondcol-umnfrom3.429to3.409(againincreasingthestandarderrorslightly,from0.785to0.820).Theorgestimatesaresimilarlyinsensitivetotheinclusionoflineartrends.Includinghigher-ordertrendsreducesprecisionbutdoesnotmarkedlychangethepointestimatesinTable6.Therobustnessoftheestimatestothesechangesinpro-videsfurthersuggestiveevidencethattheestimatesinTableC.6arenotlikelytobedrivenbyunderlyingdifferencesintrendsbetweenDSAsthatexperiencedhelmetlawchangesandDSAsthatdidnot.733.4AConceptualFrameworkofRationingbyWaitlistsLindsayandFeigenbaum(1984)proposeamarketclearingmodelforwaitinglists.TheyapplytheirmodeltotheBritishHealthcaresystemwherethepriceofthemedicalserviceiszeroandanindividual'sdemandforthehealthprocedureisunpredictableratherthanaccuratelypredictable.Weadapttheirmodelforrationingbywaitinglistsinthecontextoforgantransplants.Theirmodelassumesthefollowing:vi(u;p;g;t)=vi(u;p)egt(3.3)Wherevisthepresentvalueofrightsobtainedfromjoiningawaitinglistforcandidatei,pisthedeliveryprice(constrainedtobezerobytheNOTA),anduisavectorofunknownattributes.Thevalueofthetransplantdependsonwhenthetransplantoccursinthefuture,whichischaracterizedbyanexponentialdecayrate,g,andanexpecteddelayindeliveryofthetransplant,t.Patientsjoinawaitlistiftheofjoiningexceedsacost,ci.Thecostsincludemedicalevaluations,transportation,andmarketinformation.Eachpatienthasacriticaldelay,^tisuchthattheywilljointhewaitlistifandonlyift^ti.Variationingandvigivesrisetoadistributionof^ti,whichgeneratesanindividualdemandfunctionforjoiningthewaitlistwhereexpectedwaitingtimetfunctionslikeaprice.Aggregatingallpotentialwaitlistcandidatesbasedontheirwillingnesstojointhewaitlistateachtgeneratesadownwardslopingfijoinerfunctionflthatissimilartoatypicaldemandfunctionthatdependsontheconsumer'swillingnesstopaygivenprices.Assumingsupplyisinelasticwithrespecttothewaitingtime19givesrisetoanequilibriumwaitingtime,t,thatclearsthemarketsuchthatthenumberoftransplantsequalsthenumberofwaitlistjoiners.Thepersonsonthewaitlistatthavehighervaluationorlowerdecayratesthanthosewhoneverjointhelist.Inthismodel,anincreaseinthesupplyoftransplantableorganswilllowertheequilibrium19LindsayandFeigenbaumdonotmakethisassumption,butinthecaseoforgans,thisseemsreasonable.74waitingtime.Themagnitudeofthechangeinwaitingtimewilldependontheresponsivenessofwaitlistjoinerstowaittime.Thesemarginaljoinersarelikelytohavelowervaluationorhigherdecayratesthanthepreviouswaitlistjoiners.Thissimplemodelprovidesasetofempiricallytestablecomparativestaticsinanequilibriumsetting.However,thesettingfororgantransplantsdiffersfromwhatthisbasicmodelcaptures.First,thenotionofawaitinglistdiffers:whenatransplantcandidateaddstheirnametoawaitlistinaDSA,theirplaceonthelistisnotAsSection3.2describes,thereisnotanactualwaitinglist,butratherapoolofcandidatesthatamatchingprogramgenerateseachtimeadeceasedorganbecomesavailable.20Dependingonhealthstatus,transplantcompatibilitymeasures,and,insomecases,waitingtimeacandidateaccruedandtransferredfromanotherlist,21apersonwhoisarecentadditiononthewaitlistmaybehigheronthelistgeneratedforaparticularorganthanthosewhoregisteredearlier.Thiscomplicatestheofthemeasureoftherelevantexpectedwaitingtime,t,forthepotentialjoiners.Forexample,theaveragewaitingtimeinaDSAmaybeuninformativetosomeonewhohasaparticularbloodtypeandtransferswaitingtimefromanothercenter.AmoregeneralsimplifyingassumptionintheLindsayandFeigenbaummodelisthatallpoten-tialwaitlistcandidateshaveequalcosts.However,inthecaseoforgantransplants,thereislikelytobelargevariationacrosscandidatesincoststolistoutsideofacandidate'sfihomeflDSAdrivenbytransportationcostsofhealthinsurancerestriction,aswellasininformationcostsaboutthesupplyoftransplantableorgansanditsvariationacrossDSA.Finally,themodelproposedbyLindsayandFeigenbaumholdsthequalityofthemedicalpro-ceduresconstantsothatincreasesinthesupplyofproceduresresultinsimplyamovementalongthedemandcurve.Therearetworeasonswhythismaynotholdinthecontextoftheorganwait-inglist.First,thesourceofthesupplyshiftmaychangethequalityofthedonationsŒifthe20AsofJune9,2016(https://www.unos.org/data/)36percentofthe120,617personsonatrans-plantlistareinactive,whichmeansthatevenifanorgancameavailable,theywouldnotbeeligibletoreceiveitbecausetheyaretoosickornotsickenough.21SeeUnitedNetworkforOrganSharing,2014fordetailsontransferringwaitlisttime.75marginaldonorsareprovidinghigherqualityorgans,morepersonsmayjointhewaitlistevenifexpectedwaittimedoesnotchange.Thisisempiricallyrelevant,asDCEM's(2011)resultsshowthattheincreaseinthesupplyoforgansfromhelmetrepealsisconcentratedinmenaged18to35,whichislikelyagroupinbetterpre-donationhealththanthetypicaldonorfromotherdeathcircumstances.Second,becauseofthematchingprocessforallocatingdeceasedorgans,alargerpoolofdeceasedorgansmayincreasetheefyoftheallocationprocessandimprovematchquality,evenholdingthequalityoftheoverallpooloforgansconstant(seeRothetal.,2004,foradiscussionofexpandingtheorganpoolinlivingkidneyexchanges).Asaresult,onceagainmorecandidatesmayjointhewaitlistevenifexpectedwaitingtimeremainsconstant.Thesequalityimprovementsresultinashiftinthejoiningfunction,ratherthanamovementalongthejoiningfunction,followingachangeinsupply.Intheremainderofthispaper,weturntotestingtheimplicationsofthisconceptualframework.Weassesshowwsontowaitlistsrespondtochangesinthesupplyoforgan,andthenweconsiderhowlivingdonationsarecrowdedoutbyincreasesindeceaseddonations.3.5WaitlistResponsestotheChangeintheSupplyofOrgans3.5.1WaitlistAdditionsResponsestotheChangeinSupplySection3.3establishedthatrepealsofhelmetlawsincreasethesupplyoforganswithinthelocalDSA.Theincreaseinsupplyislikelytoinduceanincreaseinthequantityofpersonsjoiningthewaitlistasthesupplycurvemovesalongthejoinerfunctionforthewaitlist.Ourapproachheremirrorsthatofouranalysisofdonorsandtransplants,inthatweestimateaDSA-andyearwaitlistwrateasafunctionoftheshareoftheDSA'spopulationlivinginastatewithoutauniversalhelmetlawinplaceinthatyear:Additionsdt=ad+st+g(nolawshare)dt+edt(3.4)InthisAdditionsdtisameasureofthenumberofpersonsaddedtothewaitlistin76DSAdandyeart,permillionDSAresidents.Weestimatewaitlistadditionsintheaggregateandseparatelybyorgan(kidney,liver,hearts,lungs,intestine,andpancreas).Dataonwaitlistsbeganin1987,butthestatisticsonwaitlistsizesareunreliableintheearlyyearsastransplantprogramsestablishedthemselves,soweusedatafrom1992to2013.TableC.7presentsestimatesofgfrom(3.4).Theestimatessuggestthatrepealinghelmetlaws-more,therepeal'sactualorperceivedeffectsonthenumberoftrans-plantableorgans-increasesthenumberofcandidatesjoiningwaitlistsby18.665permillionper-sons,withastandarderrorof8.733.Thisisa12percentincreaseinwaitlistwsrelativetothesamplemeanof147.717.Theorgestimatesshowthatrepealsincreasewsontowaitlistsforkidneys(10percent),livers(15percent),lungs(27percent),withsmallereffectsonhearts,intestines,andpancreases(negativeinthiscase).Theorgestimatesare,atmost,borderlineatconventionallevels.22Columns2and3inTableC.7revealadditionalinformationaboutwhichcandidatesrespondtotheperceivedsupplyshockinducedbychangesinhelmetlaws.Usingzipcodedataforcandidatesandthetransplantcentersatwhichtheyhaveregistered,wegenerateacountofthenumberofwaitlistadditionsthatcomefromwithintheDSAandthenumberofnewadditionsinvolvingcandidateswhoresideoutsidetheDSA.Fromthesamplemeans,weestimatethatroughly22percent(=32.832/147.717)ofallwaitlistwsconsistofcandidateswholiveoutsidetheDSA'sboundaries.However,waitlistwsinducedbyrepealsofhelmetlawsaredisproportionatelyconcentratedamongthoselivingelsewhere,with45percent(=8.343/18.665)ofthemarginalwscomingfromoutsidetheDSA.Thedisparityisespeciallypronouncedforkidneys,forwhich64percent(=5.395/8.481)oftheowsduetorepealscomefromoutsidetheDSA,22AsmoreevidencethatcandidatesmovetoregisterinalternativeDSAs,inunreportedspec-weincludeameasureoftheshareoftheDSA'sborderingpopulationlivinginastatewithoutauniversalhelmetlawasanindependentvariable.Generally,themagnitudeoftheco-efonownpopulationwithnolawisslightlylargerinmagnitude.ThecoefonthevariablemeasuringtheshareofborderingDSA'swithoutahelmetlawisnegative,butimpreciselyestimated.TheseresultsareconsistentwithtransplantcandidatesseekingDSAswiththeleastrestrictivehelmetlaws.77comparedtoabaselineofroughly18percent(=15.942/87.803).23Inotherwords,themarginalentrantsontowaitlistsappeartohavehighercosts,intheformoftravelcosts,thanthepre-existingwaitlistedcandidates.InFigureC.3,weplottheseriesofestimatesfromevent-studymodelsofwaitlistws,analogoustothoseshowninFigureC.2fordonorsandtransplants.AsinFigureC.2,theredoesnotappeartobeanyevidencethatDSAsinstatesthatrepealtheirhelmetlawsexhibitdifferentpre-existingtrendsinwaitlistadditionscomparedtoDSAsinstateswithoutrepeals.Again,thisismosteasilyseenbycomparingadditionsintimeperiodsveormoreyearsbeforerepeals(la-beledfi-5flinthetotheyearbeforetherepeal(labeledfi-1flintheOverallandforkidneys,wsriseimmediatelyintheyearoftherepealandsteadilyincreaseinthefollowingyears.Interestingly,within-DSAwsincreaseimmediatelybutthenfallalmostbacktobase-linelevels,whileoutside-DSAwsadjustmoregradually.Wehavenocompellingevidencewhythisoccurs,butperhapsthegradualincreasesforoutside-DSAwsstemfromgradualspreadofinformationaboutwaitlistsize.Forexample,ifPennsylvaniarepealsitshelmetlaw,individualslivinginDSAsheadquarteredinPennsylvaniamightlearnthatinformationimmedi-ately,whileindividualslivinginotherstatesmightlearnwithalag.Finally,wenotethatmanyoftheindividual-yearestimatesshowninthearestatisticallyforallorgansandforkidneys(foryearsfollowingtherepeal).ThewaitlistwsshowninTableC.7andFigureC.3representnewregistrationsbutnotnecessarilynewtransplantcandidatestotheextentthatcandidatesaremultilisting.TableC.8differentiatestheeffectsofhelmetlawrepealsonwaitlistwsseparatelyforcandidateswhomultilistandthosewhodonot.Thethreecolumns,underthefiNoMultilistingsflheading,showestimatesforthosecandidateswhoonlylistinonetransplantcenterduringtheirwaitlist23TheseestimatesdonotaccountforpersonswhowouldhaveregisteredoutsidetheirownDSAwhenahelmetlawisinplace,butdoregisterintheirownDSAwithoutahelmetlaw.InrelatedregressionsweaggregateadditionsbyDSAofresidenceandnoevidencethathelmetlawsencouragepersonswhowouldnothaveregisteredintheirhomeDSAtodoso.ThismaybebecausemostpersonsareregisteredintheirhomeDSA.78spell.Thesecandidatesarerelativelyinsensitivetoashiftinthesupplyoforgans;theestimateinthetoprowofthefiAllAdditionsflcolumnsis6.378,whichisroughly5percentofthesamplemeanofwsamongthisgroup(118.406).Incontrast,theanalogousestimateforthesubsampleofMultilistingsis11.995,whichis29percentofthesamplemeanofws(41.081).ThesedifferencesareparticularlypronouncedinthefiOut-of-DSAflestimatesŒrelativetothesamplemeanof14.114,wsontowaitinglistsincreaseby44percent(6.252)followingrepealsofhelmetlaws.TableC.8alsoshowsthattheresponseofwaitlistadditionstochangesinthesupplyoforgansvariesdramaticallyacrossorgans.Forkidneys,wheremultilistingismostcommon,annualwsofmultilistersincreaseby7.739followinghelmetlawrepeals,whichis30percentofthemeanan-nualwof26.144.Again,thiseffectismostpronouncedamongthosewhoarecomingfromoutsidetheDSA.Incontrast,repealingahelmetlawhasnoeconomicorstatisticallyef-fectonwaitlistwsamongkidneytransplantcandidateswhoonlyregisteratasingletransplantcenter,asshownincolumns(1)-(3)ofthetable.Theseestimatessuggest,perhapsnotsurprisingly,thatshockstothesupplyoforgansencouragepersonswhoarealreadylistedinatleastonelocationtolistagain.Thatis,candidatesrespondtosupplyshocksononemarginŒthedecisionofwheretolist,conditionalonlistingŒmuchmoresothananotherŒthedecisionofwhethertolistatall.Thisseemsplausible,asitisdiftoimaginethatcandidatesinneedofalife-savingorganwouldrespondalongthefilistornotflmargintochangesinorgansupplygeneratedbyvariationinhelmetlaws.Tofurtherconsiderthequestionofwhetherthemarginalregistrantsdifferinpredictedwaysfrominframarginalregistrants,weconsideraslightlydifferentsetofcomparisonDSAs.cally,weidentify13DSAsthatfisendflthemostcandidatestoeachoftheDSAslocatedinastatewithahelmetlawrepeal.Forexample,amongcandidateslistingbutnotresidinginthewesternPennsylvaniaDSA,morelivewithinthebordersofthenortheastOhioDSAthaninanyotherDSA'sborders.ThefitreatmentflgroupisthewfromthosefisendingDSAsfl(northeastOhioforexample)totheDSAsinthestateswiththerepeals(westernPennsylvania,forexample).The79ficontrolflgroupisthewsfromthosefisendingDSAsfltootherDSAs,wheretheysendtheircandidatestothemost,butdidnotexperienceahelmetlawchange(southeastOhio,forexample).FiguresC.4andC.5showtheresultsforallorgansandbyKidneyandNon-Kidney,respec-tively.TheshowthemeanpercapitawofwaitlistadditionsfromtheneighboringDSAtotheDSAwherethelawchanged(fitreatmentflw)andthemeanpercapitaofwaitlistadditionsfromtheneighboringDSAtotheDSAwherethelawdidnotchange(ficontrolflw).ThearedramaticŒthereisalargeincreaseintheaveragewofwaitlistadditionstotheDSAswherethehelmetlawsrepealoccurred,whilethewtotheotherneighboringDSAsremainsunchanged.Theresultsareconcentratedamongkidneywaitlistadditions,asshowninFigureC.5.Additionally,FigureC.6showsthatthispatternisespeciallypronouncedamongmul-tilistingwaitlistcandidates:veyearsafterahelmetlawrepeal,wstothetreatmentDSAsincreasetothreetimestheirlevelintheyearoftherepeal,whilewstothecontrolDSAsstayroughlyconstant(inallofthesewsacrossDSAsarenormedtoequal1intheyearofthehelmetlawrepeal).ThesethatasupplyshockinoneDSAdrawswaitlistadditionsfromnearbyDSAsandthatmultilistingdrivesmanyoftheseadditions.Inotherwords,DSAswherelawsdidnotchangeareaffectedbylawchangesinneighboringDSAs.Becauseofthenatureofthewaitlist,itisunclearwhotheofthesupplyshockareŒthepersonswhomovetotheDSAwiththesupplyshockand/orthepersonswhodonotbecausepersonsonthelistinfrontofthemmayhavemoved.2424Wehavemoretodoconcerningthecharacteristicsofthemarginalmover.Theconceptualframeworkpredictsthatthemarginalwaitlistjoiners(inresponsetoasupplyshock)willbeper-sonswithlowervalueandhigherdecayratesthanpersonsontheexistingwaitlist.Weestimateeventstudies,likeinequation(3.2),wherethedependentvariableisasetofwaitlistregistrantcharacteristicsincluding:whetherthecandidateismultilisted,hasreceivedaprevioustransplant,isreceivingdialysis(forkidney),isbloodtypeO,ismale,haseducationgreaterthanhighschool,isminoritystatus,hasMedicareasprimaryinsurance,andhasprivateprimaryinsurance.Theseestimatesshowfewstatisticallyoreconomicallypatterns.Oneexceptionisthattheprobabilitythatapersonhasreceivedaprevioustransplantisgenerallylowerfollowingashocktothesupplyoftransplantableorgans.Anotherexceptionisthatkidneytransplantpatientsaremorelikelytobeondialysisfollowingashocktothesupplyoftransplantableorgans.803.5.2SubstitutionAwayfromLivingDonorTransplantsTheprevioussubsectionshowsagrowthinwaitlistsinresponsetoalocalshocktothesupplyoforgans.OurevidenceoncandidateswhoaremultilistedsuggeststhatmuchofthisgrowthcomesfrompersonswhoarealreadylistedinadifferentDSAthatdidnotexperiencethesupplyshock.Formostorgans,thedeceaseddonorwaitinglististheonlyoptionforthoseinneedofatransplant.However,kidneysareanobviousexception,as34percentofallkidneytransplantscomingfromlivedonors.Mostlivingdonorsarebloodrelatives(69percent),spouses(11percent),orfriends(16percent).Theprincipalmechanismbywhichashocktothesupplyofdeceasedorganscouldaffecttheprevalenceoflivingdonortransplantsinvolvesshiftingtherelativecostofoptingforlivingversusdeceaseddonortransplants.Theprincipalcostsofjoiningawaitlist,relativetooptingtoaskarelativeorclosefriendforadonation,arethoseassociatedwithhavingtowaitforacompatibleorgan.25ThesecostsarepotentiallysubstantialŒasshownaboveinTableC.2,over10percentofthoseexitingwaitlisteachyeardosoviatheirdeath.However,livingdonationsimposeobviouscoststothedonor,andtotheextentthatcandidatesinternalizethesecosts,somecandidateswhowereclosetothemarginofoptingforjoiningthewaitlistwillbeinducedtodosoifexpectedoutcomesonthewaitinglistimprove.Asaresult,crowd-outoflivingdonationmayoccurifthesupplyofdeceaseddonororgansincreases.Inordertoinvestigatetheextentofthecrowd-outdescribedabove,weestimateavariantof(3.1)above,usingtransplantsfromlivingdonorsasthedependentvariable.TableC.9presentstheresultingestimates.Thetopestimateinthetableaggregatesalllivingdonortrans-plants,whiletheremainingestimatesdisaggregatesbythedonor'srelationshiptotheintendedrecipient.Overall,therepealofahelmetlawreduceslivingdonortransplantsby3.445(withastandarderrorof1.403),whichisconsistentwiththehypothesisthatanincreaseinthesupplyofdeceaseddonorscrowdsoutlivingdonors.TableC.9furthershowsthattheeffectofrepealinga25Thereissomeevidencethatdonationsfromlivingdonorsaremoresuccessfulthandonationsfromdeceaseddonors.81helmetlawresultsinfewerlivingdonationsfromallrelationshiptypes.Siblings,whodonatemoreorgansthananyotherrelationshiptype,showalargeabsolutedeclineindonationsof0.683permillionpersons,representinga15percentdeclinerelativetothebaselinedonationrate.Similarly,donationsfromparents,children,otherrelatives,andspousesdeclineby17,20,23,and29per-cent,respectively,relativetobaseline.Thelargestestimateinthetableisforfiallotherdirecteddonationsfl,whichcapturesdonationsinwhichthedonorandrecipientareacquaintedbutnotfamilymembers.Thiscategoryprimarilycapturesdonationsbetweenclosefriends.Helmetlawrepealsdecreasetheprevalenceofthesedonationsby1.202permillionDSAresidents,whichisslightlyover50percentofthemeantransplantationrateforthisrelationship.Finally,anonymousdonations,whicharerelativelyrare,fallby0.125,adramatic83percentofthebaselinedonationrate.26Ourestimatesoftheeffectsofhelmetlawrepealsonwaitlistwsandontheprevalenceoflivingdonationsuggestthatthelocaldemandfordeceaseddonororgansincreasesinresponsetorealorperceivedincreasesinthesupplyoftheseorgans.ThesetwomechanismsarepresumablyrelatedŒthecrowd-outoflivingorgantransplantsispartofthemechanismbywhichwaitlistwsincreasefollowingapositivesupplyshock.Infact,theestimatedoveralldeclineinlivingdonorsshowninTableC.9issimilarinmagnitudetotheestimatedincreaseinwaitinglistadditionsfromthefiwithinDSAflgroupamongthosenotmultilistinginTableC.8,lendingfurthersuggestiveevidencethatamongthoseatthemarginoflistingatallornot,acomparisonofthecostsoflivinganddeceaseddonortransplantsplaysarole.However,giventhattheestimatedincreaseinoverallwsinTableC.7greatlyexceedstheestimateddeclineinlivingdonationsinTableC.9,variationatthemarginsofwheretolist,ratherthanthecrowd-outoflivingtransplants,isapparentlythedominantmechanismatwork.26Fernandezetal.(2013)disaggregatethecrowd-outoflivingdonorsusinghelmetlawsasoneinstrumentforachangeinthesupplyofcadavericdonors.Atthestate-yearlevel,theynostatisticallyeffectofanincreaseincadavericdonorsonbloodrelateddonorsoranonymousdonors.Theyconsistentlystatisticalintheircrowd-outestimatesonlyforspousesandfriends.82Finally,wenotethattheshockstothedemandfordeceasedorganssuggestedbyTablesC.7-C.9areallatleastaslargeasthemagnitudeofthesupplyshockshowninTableC.6.Forexample,thecentralestimatesimplythatahelmetlawrepealincreasesthenumberoforgantransplantsbyroughly3.4permillionDSAresidents.However,theensuingincreaseinadditionstotransplantwaitlistsof18.665ismorethan5timesaslargeasthesupplyshockitself,raisingthequestionofwhethertheselocalshocksareeffectiveatimprovingoutcomesforcandidatesonlocalwaitlists.Weturntothisquestioninthenextsection.3.6HealthOutcomesfromtheChangeintheSupplyofOrgansWeestablishedshiftsinthenumberofdonorsandtransplantsthatresultinlargeincreasesinwaitlistadditions.Thereareatleasttworeasonstoexpectthatahighersupplymightresultinimprovedoutcomes.First,previousresearchshowsthatmoretransplantscanleadtomoreefentoutcomes(Roth,2004),withoutchangesindonorquality.Second,thisshockarisingfromamotorcyclehelmetrepealincreasesthenumberofdonorswhoareyoungmen(DCEM,2011)whomaybe,onaverage,healthierbeforedeaththandonorsofotherdeathcircumstances.Counteringthesepredictions,forkidneysatleast,istheevidencethatincreaseddeceaseddonorscrowdoutlivingdonorsandlivingdonorshavebeenshowntoresultinbetteroutcomes.Further,weknowthatthewaitlistschangesafterthehelmetrepeals,withmoreandpossiblydifferentpersonsenteringthelistandchangesinhealthoutcomesmaybeattributabletochangesinthehealthstatusofthetransplantrecipientsafterthesupplyshock.Thepredictionsoveraveragehealthoutcomesareambiguousandweturntoourstandardempiricaltoaddresstheissue.Toconsidertheeffectofahelmethasontheoutcomesofthetransplantrecipients,weestimatethefollowingmodel:GraftSurvivalXyearsTransplantdt=ad+st+g(nolawshare)dt+edt(3.5)Weestimateequation(3.5)separatelyforthreedifferentdependentvariableswhereXinthetermGraftSurvivalXyearsrepresents1,3and5yearsgraftsurvival.Estimatesfromthesespec-83canbeinterpretedastheoveralleffectofhelmetlawrepealsontheprobabilityofthetransplantedorgan(andthetransplantrecipient)survivingXyears,conditionalonreceivingatransplant.Wealsoestimateaneventstudywiththesamesetofdependentvariables.OurresultsareinTableC.10andFigureC.7.TableC.10showsthattheoneyeargraftsurvivalrateishigheroverallandforallorgansex-ceptkidneys.Theaveragechangesareabout3percentincreasesforlivers,heartsandlungs,butincreasesto6percentforlungs,8percentforintestinesandmorethan17percentforpancreas.Weknowthatthedenominatorofourdependentvariable,thenumberoftransplants,isincreasingfollowingthelawchange,sotheseestimatessuggestthatthenumberoftransplantpatientswhoseorganslastmorethanayearisalsoincreasing.Anincreaseinthesupplyoforgansfromhelmetlawrepealsalsoimprovesthreeandveyeargraftsurvivalrates,althoughthecoefarelesslikelytobestatisticallyatstandardlevelsforthesemeasures.FigureC.7showstheincreaseingraftsurvivalratesforliversandhearts(thecoefintheeventstudyforotherorgansareimpreciselyestimated)andnoeffectonkidneysfollowingthehelmetlawrepeal.Furtherempiricalevidence,availableuponrequest,showsthatstatehelmetlawrepealsreducethelikelihoodofawaitlistedcandidatedyingby0.440percentagepointsperyear,whichisroughlya9percentdecreaserelativetothesamplemean.Byindividualorgan,theresultsareconsistentwiththediscussionaboveabouttheincreaseintransplantrelativetothesizeoftheorgan'swaitlists.,thelargestpositiveeffectsondeceaseddonortransplantsareforcandidatesawaitinglivers,anddeathratesdecreaseforthisgroup:repealingahelmetlawdecreasesthenumberofpatientsonwaitlistswhodieawaitinglivertransplantsby13.654,astatisticallysandlargeeffectrelativetothesamplemeanof65.119.Takenliterally,repealingahelmetlawdecreasestheoddsofdyingonthewaitlistbyroughly21percentperyear.Again,theestimateforkidneyssuggeststhattheentireeffectofanincreasedsupplyofkidneysfromdeceaseddonorsisoffsetbyanincreaseinws(someofwhomareficrowdedinflbyadecreaseintheprevalenceoflivingdonortransplants).Asaresult,thereisnoeffectonthedeathratesofthoseawaitingkidneys.84Asofnow,wedonotidentifythemechanismfortheimprovedoutcomesformostorgans:moreefmatchesinalargerpoolofdonors,healthierdonors,andhealthierrecipientsareallcandidatesforexplainingtheimprovements.27Futureworkwilladdressthese,buttheseimprovedoutcomescouldcertainlybeareasonweseesuchdramaticresponsesinthewaitlistadditionsfollowinganincreasesinthesupplyoforgans.3.7ConclusionRepealsofmotorcyclehelmetlawscausemeasurableincreasesinthesupplyofdeceasedorgandonorsanddeceasedorgantransplantsforallorgans.Theeffectsarelargestforkidneysandlivers,buttheyareforlungs,heartsandpancreastoo.Intheabsenceofaformalpricingmech-anism,theincreaseinthesupplyoforgansappearstoactasasignalthatwaitingtimedecreasesinthosegeographicalregions.Thedemandfordeceaseddonororgansrespondsintworelatedways.First,additionalpersonschoosetoenterthewaitlistsinareaswherethesupplyincreased,andtheseappeartobelargelypersonswhoarealreadywaitlistedinothergeographicregions.Second,waitlistadditionsfordeceasedorgansappeartobebytransplantcandidateswhowouldhavepursuedalivingdonationfromfriendsorfamily.Theincreaseinthesupplyofdeceaseddonorsappearstocrowdoutlivingdonors,particularlyforkidneys,wheremorethanone-thirdofalltransplantsarefromlivingdonors.Thewaitlistresponsesareasmuchasvetimeslargerthantheincreasesindeceasedorgandonors,whichraisesthequestionofhoweffectiveanincreaseinsupplyisforthoseonthewaitlistifthewaitlistgrowsenoughtolowertheprobabilityofthoseonitofexitingwithapositiveoutcome.Perhapsreassuringly,weestimatethattheaveragehealthoutcomesoftransplantrecipientsandwaitlistedpersonsimprovefollowinganincreaseinsupplyofdeceasedorgans,suggestingtheincreasedwaitlistadditionsarenotsuftooffsetgains27Asacaveat,recallthatourestimatesarerelativetoDSAsthatdidnothavelawchanges,butbeknowthatlawchangesaffectneighboringDSAsaswell.However,ifthickermarketsorhealthierdonorsexplaintheimprovementswewouldnotexpecttoseeimprovementsingraftsurvivalinthoseDSAsthatdidnotexperiencelawchanges.Morewillbedoneonthisfront.85fromtheincreaseinsupply.Theabilityoftransplantcandidatestooffsetsupplyshocksraisesquestionsabouttherolethatgeographicboundariesplayinanallocationssystem'sgoalsofbalancingfijustice(fairconsid-erationofcandidates'circumstancesandmedicalneeds),andmedicalutility(tryingtoincreasethenumberoftransplantsperformedandthelengthoftimepatientsandorganssurvive).flThecrowd-outoflivingdonorsisarealincreaseinthedemandfordeceaseddonorsandreducestheeffectivenessofanincreaseinthesupplyofdeceasedorgansonthevastshortagesoftransplantableorgans.Likewise,ifpersonswhoarerespondingtothefipriceflsignalsarethosewhoaremulti-listing,thenitisunclearthatthepersonswiththehighestmedicalneedsarefromtheallocationsystem.Multilistersmaintaintheirplaceonwaitlistswheresupplyshocksdidnotoccur,raisingabroaderquestionofhowasupplyinoneregionbroadlyaffectstransplantcandidatesinallregions.Ourinitialevidencesuggeststhattransplantcandidatesinregionswherethesupplyincreasedare,onaveragebetteroff,asmeasuredbyanincreaseingraftsurvival.Futureworkwilldomoretoconsiderwhotheoftheseshocksare.Inparticular,aretheythemultilistersinducedtojointhelistbytheincreaseinthesupplyoforgans?Or,dothepersonsonthelistwhenthesupplyshockoccursgainmost?HowarewaitlistparticipantswhoremaininDSAsthatsendmanypersonstotheDSAswithsupplyshocksaffected?Evidencethattheelasticityofdemandwithrespecttosupplyshocksissolarge,combinedwithotherchangesintheenvironment,suchasimprovedinformationaboutthefipriceflintheformofwaitlisttimechanges(see,e.g.,http://www.txmultilisting.com/home.htm,awebsitededicatedtotheDSAswiththeshortestwaitlists),bringintoquestionwhetherthegeographicalbound-ariesfortheroundofallocationareefPerhapsamovementtowardamorenationalallocationsystem,suchastheoneusedinSpain,wouldbeequity-andefy-improving(Def-fainsandYthier,2010).86APPENDICES87APPENDIXATABLESANDFIGURESFORCHAPTER1TableA.1:FrequencyofRegistrationOutcomesObserved2000-July201420072013OutcomeNPercentNPercentNPercentExchange4,1031.392020.967533.52Anonymous1,5280.52970.461760.82DirectLiving82,84427.995,72027.044,78622.37Deceased145,40849.1310,59150.0811,16452.19DiedonWL62,06420.974,54021.474,51221.09Total295,94721,15021,391Source:OPTNSTARDataasof12/31/2014.Note:Includesalltransplantswheredonorrelationshipisobserved,anddeathsofthoseregisteredonthedeceaseddonorwaitinglist.TableA.2:Survival,MatchQuality,WaitingTime(2000-July2014)ObservedGraftSurvivalGraftSurvival#ofHLARegistrationOutcome>1year*>2years*MismatchesDuration(Days)**Exchange0.970.944.31467.68SD(0.18)(0.23)(1.25)(523.69)N2,9952,3004,0304,099Anonymous0.960.934.31654.10SD(0.20)(0.26)(1.26)(606.73)N1,2611,0971,4971,528DirectLiving0.960.933.16237.47SD(0.20)(0.26)(1.66)(346.70)N75,34870,61082,09782,698Deceased0.900.863.88813.43SD(0.29)(0.35)(1.75)(717.70)N127,580116,714144,482145,153Total0.930.883.64602.31SD(0.26)(0.32)(1.74)(667.25)N207,184190,721232,106233,478Source:OPTNSTARDataasof12/31/2014.1)Notethatsurvival,HLAmismatches,andregistrationdurationareonlyfortransplantrecipients.*One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.**Durationis0forlivingdonorkidneyrecipientswhodidnotregisteronthewaitinglist.88TableA.3:SensitivityandAge(2000-July2014)ObservedMostRecentMostRecentEndingAgeatDonorOutcomePRA(ClassI)*PRA(ClassII)*C-PRA**Listing***AgeExchange15.0214.5022.1046.5643.05SD(27.64)(28.02)(34.29)(14.47)(11.56)N3,6363,4813,5164,1034,103Anonymous8.578.6616.6844.5643.67SD(20.69)(22.28)(30.22)(15.29)(11.82)N1,1931,1061,0171,5281,528DirectLiving5.715.288.9844.1540.68SD(16.85)(16.98)(22.33)(16.05)(11.23)N52,16746,05330,04882,84482,843Deceased10.939.6319.6647.8237.49SD(24.54)(23.68)(33.23)(15.25)(16.71)N99,78789,32172,104145,407145,408DiedonWL--23.3252.85-SD--(37.35)(12.25)-N0032,20864,8250Total9.278.3118.2447.8638.76SD(22.48)(21.94)(32.67)(15.19)(14.98)N156,783139,961138,893298,707233,882Source:OPTNSTARDataasof12/31/2014.*Thesevaluesaremissingforthosewhowerenottransplanted.Duetochangesindatacollection,valuesonlyreportedfrom2004onward.**Duetochangesindatacollection,valuesonlyreportedfromlate2007onward***Forthosetransplantedwithouteverhavinglisted,thisistheageattimeoftransplant.TableA.4:PatientDistancetoTransplantCenterofOperationPercentileAllExchangeAnonymousDirectLivingDeceased104.125.634.384.853.74207.309.797.348.496.653011.0913.7910.7612.6210.164016.0521.0014.6618.0714.825024.0129.7521.1126.1022.496036.0743.7131.8938.4534.537056.2165.3649.5958.7454.458086.93103.2181.1093.0583.4090146.22193.37182.88163.57137.81N231,7964,0701,51482,019144,193Source:OPTNSTARDataasof12/31/2014.89TableA.5:MeasuringCrowd-Out,2SLSResultsVARIABLESExchangeDirectLivingAnonymousDeceasedDiedonWL(FirstStage)TotalExchangesNearby0.0056***(ExcludingOwnifRelevant)(0.00034)[0.014]ExchangeIndicator-0.50***0.12***-0.086-0.54***(2SLSEstimates)(0.11)(0.031)(0.13)(0.093)[0.28][0.0052][0.49][0.21]Observations291,995285,601285,601285,601285,601NumberofZipCodes26,25019,85619,85619,85619,856Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.TableA.6:QualityMeasures,2SLSResultsGraftSurvivalGraftSurvivalHLARegistrationVARIABLES>1year>2yearsMismatchesDuration(Days)ExchangeIndicator0.20***0.19*0.40-592***(2SLSEstimates)(0.066)(0.097)(0.38)(170)[0.93][0.88][3.64][597]Observations198,296181,901223,184224,536NumberofZipCodes17,68517,16518,42818,455FirstStageResults:DependentVariableisExchangeIndicatorTotalExchangesNearby0.0073***0.0073***0.0072***0.0072***(ExcludingOwnifRelevant)(0.00050)(0.00060)(0.00044)(0.00044)[0.015][0.012][0.017][0.018]Observations204,849188,456229,657231,015R-squared0.0260.0230.0260.027NumberofZipCodes24,23823,72024,90124,934Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.90TableA.7:LeadsandLagsofIV,Crowd-OutVARIABLESDirectedLivingAnonymousDeceasedDiedonWL12-13MonthLag0.00017-0.00045*0.0015-0.00082(0.0010)(0.00024)(0.0012)(0.00080)10-11MonthLag0.00049-0.00011-0.00120.0012(0.0011)(0.00028)(0.0012)(0.00083)8-9MonthLag0.0010-0.0000290.000044-0.00082(0.0010)(0.00027)(0.0012)(0.00080)6-7Monthlag-0.00150.00073***0.0013-0.0011(0.0010)(0.00026)(0.0012)(0.00082)4-5MonthLag-0.000200.000230.00091-0.0015*(0.0010)(0.00025)(0.0012)(0.00079)2-3MonthLag-0.0015-0.000160.000720.00075(0.00099)(0.00022)(0.0012)(0.00078)TotalExchangesNearby-0.0026***0.00040**-0.0021**-0.0012**(ExcludingOwnifRelevant)(0.00074)(0.00020)(0.00088)(0.00058)1-2MonthLead0.000930.00012-0.00015-0.00100(0.00098)(0.00023)(0.0011)(0.00076)3-4MonthLead-0.000780.00014-0.000720.0018**(0.00098)(0.00023)(0.0011)(0.00075)5-6MonthLead-0.000280.00045*0.00092-0.0019**(0.00097)(0.00024)(0.0011)(0.00074)7-8MonthLead0.00019-0.0000570.00070-0.00096(0.00096)(0.00022)(0.0011)(0.00071)9-10MonthLead0.00063-0.0000500.00022-0.00053(0.00092)(0.00022)(0.0011)(0.00075)11-12MonthLead0.0015*0.000180.0013-0.0027***(0.00088)(0.00021)(0.0011)(0.00072)JointF-testonLeads1.041.350.676.77P-valueofJointF-test0.400.230.670.00Observations263,563263,563263,563263,563R-squared0.1500.0040.1220.446NumberofZipCodes25,54125,54125,54125,541Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.2)Leadsandlagsaretwo-monthaverages.91TableA.8:LeadsandLagsofIV,QualityGraftSurvivalGraftSurvivalHLARegistrationVARIABLES>1Year>2YearsMismatchesDuration(Days)12-13MonthLag-0.0010-0.00018-0.0031-0.69(0.00096)(0.0014)(0.0048)(2.18)10-11MonthLag0.000460.00160.00583.74*(0.00087)(0.0012)(0.0048)(2.23)8-9MonthLag0.000180.0011-0.0030-6.19***(0.00089)(0.0012)(0.0048)(2.05)6-7Monthlag0.0010-0.00082-0.00040-3.15(0.00089)(0.0013)(0.0049)(2.13)4-5MonthLag-0.000780.00095-0.0074-0.26(0.00084)(0.0012)(0.0048)(2.11)2-3MonthLag-0.00031-0.00130.000064-3.26(0.00085)(0.0013)(0.0047)(2.06)TotalExchangesNearby0.0015***0.0016*0.00080-1.86(ExcludingOwnifRelevant)(0.00056)(0.00082)(0.0035)(1.58)1-2MonthLead0.000110.000360.0092**-0.89(0.00077)(0.0011)(0.0046)(2.04)3-4MonthLead-0.00033-0.000110.00480.15(0.00078)(0.0011)(0.0046)(1.93)5-6MonthLead0.00058-0.00037-0.0027-1.78(0.00076)(0.0011)(0.0046)(1.93)7-8MonthLead-0.00054-0.00090-0.0000802.69(0.00074)(0.0010)(0.0044)(1.98)9-10MonthLead0.000270.00021-0.00035-0.056(0.00075)(0.00098)(0.0044)(1.89)11-12MonthLead-0.00019-0.000840.000112.28(0.00069)(0.00092)(0.0042)(1.85)JointF-testonLeads0.190.351.140.83P-valueofJointF-test0.980.910.340.55Observations191,639175,246206,886208,105R-squared0.0090.0120.0620.136NumberofZipCodes23,74623,19524,22324,249Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.4)Leadsandlagsaretwo-monthaverages.92TableA.9:MeasuringCrowd-Out,2SLS,UsingLagsVARIABLESExchangeDirectLivingAnonymousDeceasedDiedonWL(FirstStage)TotalExchangesNearby0.0018***(One-MonthLag)(0.00030)[0.014]TotalExchangesNearby0.0015***(Two-MonthLag)(0.00031)ExchangeIndicator-0.68***0.21***0.48-1.02***(2SLSEstimates)(0.23)(0.061)(0.30)(0.21)[0.28][0.0052][0.49][0.21]Observations289,422283,036283,036283,036283,036NumberofZipCodes26,17819,79219,79219,79219,792Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.93TableA.10:QualityMeasures,2SLS,UsingLagsGraftSurvivalGraftSurvivalHLARegistrationVARIABLES>1year>2yearsMismatchesDuration(Days)ExchangeIndicator0.11-0.110.52-1,350***(2SLSEstimates)(0.13)(0.23)(0.76)(350)[0.93][0.89][3.64][598]Observations196,314179,925221,211222,557Numberofzipcodes17,62217,10318,36618,392FirstStageResults:DependentVariableisExchangeIndicatorTotalExchangesNearby0.0026***0.0027***0.0025***0.0024***(One-MonthLag)(0.00045)(0.00050)(0.00039)(0.00039)[0.015][0.012][0.018][0.018]TotalExchangesNearby0.0022***0.0015***0.0020***0.0020***(Two-MonthLag)(0.00048)(0.00052)(0.00040)(0.00040)Observations202,852186,459227,680229,030R-squared0.0220.0190.0220.022NumberofZipCodes24,16023,63724,83524,865Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.94TableA.11:MeasuringCrowd-Out,2SLS,30MileRadiusVARIABLESExchangeDirectLivingAnonymousDeceasedDiedonWL(FirstStage)TotalExchangesNearby0.0054***(ExcludingOwnifRelevant)(0.00042)[0.014]ExchangeIndicator-0.40***0.20***-0.23-0.57***(2SLSEstimates)(0.13)(0.043)(0.16)(0.12)[0.28][0.0052][0.49][0.21]Observations291,995285,601285,601285,601285,601NumberofZipCodes26,25019,85619,85619,85619,856Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.TableA.12:QualityMeasures,2SLS,30MileRadiusGraftSurvivalGraftSurvivalHLARegistrationVARIABLES>1year>2yearsMismatchesDuration(Days)ExchangeIndicator0.21***0.23**0.47-425**(2SLSEstimates)(0.075)(0.11)(0.46)(203)[0.93][0.88][3.64][597]Observations198,296181,901223,184224,536NumberofZipCodes17,68517,16518,42818,455FirstStageResults:DependentVariableisExchangeIndicatorTotalExchangesNearby0.0075***0.0079***0.0070***0.0071***(ExcludingOwnifRelevant)(0.00062)(0.00076)(0.00053)(0.00053)[0.015][0.012][0.017][0.018]Observations204,849188,456229,657231,015R-squared0.0250.0220.0250.025NumberofZipCodes24,23823,72024,90124,934Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.95TableA.13:MeasuringCrowd-Out,2SLS,75MileRadiusVARIABLESExchangeDirectLivingAnonymousDeceasedDiedonWL(FirstStage)TotalExchangesNearby0.0051***(ExcludingOwnifRelevant)(0.00029)[0.014]ExchangeIndicator-0.35***0.11***-0.13-0.63***(2SLSEstimates)(0.10)(0.027)(0.12)(0.092)[0.28][0.0052][0.49][0.21]Observations291,995285,601285,601285,601285,601NumberofZipCodes26,25019,85619,85619,85619,856Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.TableA.14:QualityMeasures,2SLS,75MileRadiusGraftSurvivalGraftSurvivalHLARegistrationVARIABLES>1year>2yearsMismatchesDuration(Days)ExchangeIndicator0.15**0.120.51-505***(2SLSEstimates)(0.065)(0.10)(0.36)(157)[0.93][0.88][3.64][597]Observations198,296181,901223,184224,536NumberofZipCodes17,68517,16518,42818,455FirstStageResults:DependentVariableisExchangeIndicatorTotalExchangesNearby0.0066***0.0064***0.0066***0.0067***(ExcludingOwnifRelevant)(0.00043)(0.00050)(0.00037)(0.00037)[0.015][0.012][0.017][0.018]Observations204,849188,456229,657231,015R-squared0.0260.0230.0270.027NumberofZipCodes24,23823,72024,90124,934Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.96TableA.15:MeasuringCrowd-Out,2SLS,Zip-LevelAggregatesExchangeDirectLivingAnonymousDeceasedDiedonWLVARIABLES(FirstStage)TotalExchangesNearby0.0068***(ExcludingOwnExchanges)(0.00040)[0.015]NumberofExchanges-0.52***0.13***-0.039-0.50***(2SLSEstimates)(0.11)(0.029)(0.13)(0.099)[0.31][0.0057][0.54][0.23]Observations264,217257,800257,800257,800257,800NumberofZipCodes26,25019,83319,83319,83319,833Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.TableA.16:MeasuringCrowd-Out,2SLS,IncludingWLRemovalsduetoDeteriorationofConditionandTransplantRefusalExchangeDirectAnonymousDeceasedDiedTooSickorVARIABLES(FirstStage)LivingonWLRefusedTxTotalExchangesNearby0.0049***(ExcludingOwnifRelevant)(0.00031)[0.013]ExchangeIndicator-0.53***0.13***-0.12-0.62***0.14*(2SLSEstimates)(0.11)(0.031)(0.13)(0.11)(0.081)[0.26][0.005][0.45][0.19][0.08]Observations317,689311,303311,303311,303311,303311,303NumberofZipCodes26,82420,43820,43820,43820,43820,438Clusteredstandarderrorsinparentheses(atzipcodelevel)***p<0.01,**p<0.05,*p<0.1Meansofdependentvariablesinbrackets1)Regressionsincludemonth-yearedeffects,zipcodeedeffects,aswellascontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.97FigureA.1:KidneyTransplantsandWaitingListSize,Per1,000,000U.S.ResidentsSource:OPTNDataasof5/6/2015,http://optn.transplant.hrsa.gov98FigureA.2:TransplantsbyDonorType,Per1,000,000U.S.PopulationSource:OPTNDataasof5/6/2015,http://optn.transplant.hrsa.govNote:DirectLivingandDeceaseddonortransplantsusethelefty-axis.Paired,List,Anonymoususetherighty-axis.FigureA.3:Two-wayandThree-wayExchange,andDonorChainDiagrams99FigureA.4:CentersPerformingAtLeastOnePairedorListExchange,orBothSource:OPTNSTARDataasof12/31/2014.100FigureA.5:NumberofExchangesWithin50Miles,ExcludingOwnSource:OPTNSTARDataasof12/31/2014.101APPENDIXBTABLESANDFIGURESFORCHAPTER2TableB.1:FrequencyofRegistrationOutcomesObserved2000-July201420072013OutcomeNPercentNPercentNPercentExchange4,1031.392020.967533.52Anonymous1,5280.52970.461760.82DirectLiving82,84427.995,72027.044,78622.37Deceased145,40849.1310,59150.0811,16452.19DiedonWL62,06420.974,54021.474,51221.09Total295,94721,15021,391Source:OPTNSTARDataasof12/31/2014.1)Includesalltransplantswheredonorrelationshipisobserved,anddeathsofthoseregisteredonthedeceaseddonorwaitinglist.2)Includestransplantswherezipcodeisunknown,despitethefactthatsuchobservationswillbedroppedinestimation.102TableB.2:Survival,MatchQuality,WaitingTime(2000-July2014)ObservedGraftSurvivalGraftSurvival#ofHLARegistrationOutcome>1year*>2years*MismatchesDuration(Days)**Exchange0.970.944.31467.68SD(0.18)(0.23)(1.25)(523.69)N2,9952,3004,0304,099Anonymous0.960.934.31654.10SD(0.20)(0.26)(1.26)(606.73)N1,2611,0971,4971,528DirectLiving0.960.933.16237.47SD(0.20)(0.26)(1.66)(346.70)N75,34870,61082,09782,698Deceased0.900.863.88813.43SD(0.29)(0.35)(1.75)(717.70)N127,580116,714144,482145,153Total0.930.883.64602.31SD(0.26)(0.32)(1.74)(667.25)N207,184190,721232,106233,478Source:OPTNSTARDataasof12/31/2014.1)Notethatsurvival,HLAmismatches,andregistrationdurationareonlyfortransplantrecipients.2)Includestransplantswherezipcodeisunknown,despitethefactthatsuchobservationswillbedroppedinestimation.3)Standarddeviationsinparentheses.*One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.**Durationis0forlivingdonorkidneyrecipientswhodidnotregisteronthewaitinglist.103TableB.3:HeterogeneityinTransplantQuantityLDKTDDKTExchangeAnonDiedonWLNBloodTypeA0.3020.5070.0150.0060.172104,276(0.001)(0.002)(0.000)(0.000)(0.001)AB0.2410.5950.0160.0070.14112,813(0.004)(0.004)(0.001)(0.001)(0.003)B0.2640.4660.0180.0060.24739,691(0.002)(0.003)(0.001)(0.000)(0.002)O0.2720.4790.0120.0040.233136,223(0.001)(0.001)(0.000)(0.000)(0.001)All0.2800.4920.0140.0050.209293,003(0.001)(0.001)(0.000)(0.000)(0.001)PRARange0-400.3130.4840.0120.0050.185231,455(0.001)(0.001)(0.000)(0.000)(0.001)40-800.2030.5350.0240.0070.23224,511(0.003)(0.003)(0.001)(0.001)(0.003)80-1000.0930.5630.0180.0040.32132,643(0.002)(0.003)(0.001)(0.000)(0.003)All0.2790.4980.0140.0050.204288,609(0.001)(0.001)(0.000)(0.000)(0.001)AgeRange0-170.3890.5700.0090.0070.02612,797(0.004)(0.004)(0.001)(0.001)(0.001)18-350.3990.4640.0170.0060.11447,473(0.002)(0.002)(0.001)(0.000)(0.001)36-550.2780.4950.0150.0060.207129,588(0.001)(0.001)(0.000)(0.000)(0.001)Over550.2140.4920.0120.0040.278103,145(0.001)(0.002)(0.000)(0.000)(0.001)All0.2800.4920.0140.0050.209293,003(0.001)(0.001)(0.000)(0.000)(0.001)RaceWhite0.3560.4380.0180.0060.181155,132(0.001)(0.001)(0.000)(0.000)(0.001)Black0.1510.5800.0080.0030.25876,036(0.001)(0.002)(0.000)(0.000)(0.002)Hispanic0.2620.5050.0110.0030.21841,760(0.002)(0.002)(0.001)(0.000)(0.002)Other0.2160.5490.0130.0060.21520,071(0.003)(0.004)(0.001)(0.001)(0.003)All0.2800.4920.0140.0050.209292,999(0.001)(0.001)(0.000)(0.000)(0.001)EducationHS0.2350.5160.0110.0040.234122,753(0.001)(0.001)(0.000)(0.000)(0.001)>HS0.3310.4610.0210.0060.181114,620(0.001)(0.001)(0.000)(0.000)(0.001)All0.2810.4900.0160.0050.209237,373(0.001)(0.001)(0.000)(0.000)(0.001)Source:OPTNSTARDataasof12/31/2014.Tableincludesalltransplantswheredonorrelationshipisobserved,anddeathsofthoseregisteredonthedeceaseddonorwaitinglist.LDKT=LivingDirectKidneyTransplant,andDDKT=DeceasedDonorKidneyTransplant.Tableincludesonlyobservationswheretherelevantcharacteristic,aswellaspatientzipcode,areknown.Undereducation,Iexcluderoughly40,000patientswithunknowneducation,androughly12,000patientsundertheageof18yearsold.Standarderrorsareinparentheses.104TableB.4:HeterogeneityinTransplantQualityGraftSurvivalGraftSurvivalRegistrationHLA1year2yearsDuration(days)*Mismatches**BloodTypeA0.9250.885498.0513.576(0.001)(0.001)(1.921)(0.006)AB0.9270.886410.6723.888(0.003)(0.003)(4.505)(0.015)B0.9250.883705.3313.790(0.002)(0.002)(4.293)(0.010)O0.9260.884677.3643.616(0.001)(0.001)(2.218)(0.005)All0.9250.885601.4943.637(0.001)(0.001)(1.381)(0.004)PRARange0-400.9290.889554.7313.709(0.001)(0.001)(1.413)(0.004)40-800.9180.873785.1833.514(0.002)(0.003)(5.439)(0.014)80-1000.9040.851894.9793.159(0.002)(0.003)(5.902)(0.013)All0.9250.884606.4873.640(0.001)(0.001)(1.388)(0.004)AgeRange0-170.9430.901279.7903.734(0.002)(0.003)(4.140)(0.014)18-350.9400.897666.2303.431(0.001)(0.002)(3.753)(0.009)36-550.9310.896656.5063.656(0.001)(0.001)(2.166)(0.006)Over550.9050.857542.8023.709(0.001)(0.001)(2.047)(0.006)All0.9250.885601.4943.637(0.001)(0.001)(1.381)(0.004)RaceWhite0.9300.893465.2313.369(0.001)(0.001)(1.565)(0.005)Black0.9050.847815.1814.163(0.001)(0.002)(3.179)(0.006)Hispanic0.9380.907674.6053.584(0.001)(0.002)(3.939)(0.010)Other0.9380.905782.2034.010(0.002)(0.003)(5.883)(0.013)All0.9250.885601.5043.637(0.001)(0.001)(1.381)(0.004)EducationHS0.9210.877654.1103.647(0.001)(0.001)(2.200)(0.006)>HS0.9320.896556.5943.623(0.001)(0.001)(2.010)(0.006)All0.9260.886605.3863.635(0.001)(0.001)(1.494)(0.004)Source:OPTNSTARDataasof12/31/2014.Tableincludesalltransplantswheredonorrelationshipisobserved,anddeathsofthoseregisteredonthedeceaseddonorwaitinglist.LDKT=LivingDirectKidneyTransplant,andDDKT=DeceasedDonorKidneyTransplant.Tableincludesonlyobservationswheretherelevantcharacteristic,aswellaspatientzipcode,areknown.Undereducation,Iexcluderoughly40,000patientswithunknowneducation,androughly12,000patientsundertheageof18yearsold.Standarderrorsareinparentheses.*Duration=0forlivingdonorkidneyrecipientswhodidnotregisteronthewaitlist.**One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.105TableB.5:RegistrationOutcomeHeterogeneitybyLevelofKidneyExchangeActivityin2013AnyLivingDonorKidneyTransplantDeceasedDonorKidneyTransplantDiedWhileWaitingLeastActiveMostActiveDifferenceLeastActiveMostActiveDifferenceLeastActiveMostActiveDifferencePatientBloodTypeA0.2800.3038.4%0.5390.522-3.2%0.1810.175-3.3%(0.007)(0.008)(0.008)(0.008)(0.006)(0.006)AB0.2030.28339.1%0.6740.569-15.5%0.1230.14820.2%(0.019)(0.020)(0.022)(0.022)(0.016)(0.016)B0.2600.2714.0%0.5050.482-4.5%0.2350.2475.1%(0.012)(0.011)(0.014)(0.013)(0.012)(0.011)O0.2410.2628.8%0.5270.503-4.5%0.2330.2351.0%(0.006)(0.006)(0.007)(0.007)(0.006)(0.006)PRARange0-400.2880.3014.6%0.5100.493-3.3%0.2030.2061.8%(0.005)(0.005)(0.006)(0.005)(0.004)(0.004)40-800.2120.26022.7%0.6010.555-7.8%0.1870.185-0.7%(0.012)(0.013)(0.015)(0.015)(0.012)(0.012)80-1000.0980.13841.4%0.6340.585-7.7%0.2690.2773.1%(0.008)(0.009)(0.013)(0.013)(0.012)(0.012)AgeRange0-170.3410.3523.4%0.6430.638-0.9%0.0160.010-37.2%(0.023)(0.024)(0.023)(0.024)(0.006)(0.005)18-350.3830.44716.8%0.5030.460-8.6%0.1140.093-18.7%(0.013)(0.013)(0.013)(0.013)(0.008)(0.008)36-550.2580.2819.1%0.5420.520-4.0%0.2000.199-0.7%(0.007)(0.007)(0.008)(0.007)(0.006)(0.006)Over550.2040.2123.9%0.5270.504-4.4%0.2690.2855.8%(0.006)(0.006)(0.008)(0.008)(0.007)(0.007)RaceWhite0.3340.38214.4%0.4840.427-11.7%0.1820.1904.7%(0.006)(0.007)(0.006)(0.007)(0.005)(0.006)Black0.1140.14427.1%0.6370.606-4.9%0.2490.2500.3%(0.006)(0.006)(0.010)(0.009)(0.009)(0.008)Hispanic0.2070.26427.7%0.5380.531-1.3%0.2550.204-19.7%(0.010)(0.010)(0.013)(0.011)(0.011)(0.009)Other0.1900.23925.8%0.6080.546-10.1%0.2020.2156.2%(0.016)(0.013)(0.020)(0.015)(0.016)(0.013)EducationUptoHS0.2000.2189.0%0.5590.532-4.9%0.2410.2504.0%(0.006)(0.006)(0.007)(0.007)(0.006)(0.006)MorethanHS0.3130.3439.8%0.5040.473-6.1%0.1830.1840.1%(0.007)(0.007)(0.007)(0.007)(0.006)(0.005)Notes:Exchangeactivityismeasuredusingthemonthlyaverageofpairedandlistexchangesperformedwithin50milesofagivenpatient'szipcodeofresidence,includingthemonthofthegivenpatient'soutcomeandtheprevious11months.Themedianofthismeasureamongpatientsexperiencingatransplantordeathwhilewaitingin2013,theyearoffocus,is0.58.Standarderrorsinparentheses.106TableB.6:BloodType,2SLSResults,QuantityFirstStage2SLSEstimatesVARIABLESExchangeLivingDirectAnonymousDeceasedDiedonWLBloodTypeA0.0059***-0.26*0.090***-0.24-0.59***(0.00052)(0.14)(0.034)(0.16)(0.11)BloodTypeAB0.0081***-0.190.25**-0.40-0.66***(0.0014)(0.20)(0.11)(0.26)(0.19)BloodTypeB0.0069***-0.41***0.099**-0.32*-0.37***(0.00078)(0.15)(0.044)(0.18)(0.13)BloodTypeO0.0046***-0.81***0.15***0.23-0.56***(0.00042)(0.15)(0.040)(0.18)(0.13)Testing:AvsAB-0.0022-0.068-0.160.160.071(0.0014)(0.22)(0.10)(0.28)(0.20)AvsB-0.00100.15-0.00920.076-0.22(0.00090)(0.17)(0.046)(0.20)(0.15)AvsO0.0013**0.55***-0.055-0.47***-0.026(0.00058)(0.15)(0.038)(0.18)(0.12)ABvsB0.00120.220.16-0.086-0.29(0.0015)(0.24)(0.11)(0.29)(0.21)ABvsO0.0036**0.62***0.11-0.63**-0.097(0.0014)(0.23)(0.11)(0.29)(0.20)BvsO0.0023***0.40**-0.046-0.54***0.19(0.00083)(0.17)(0.048)(0.21)(0.15)Observations291,995285,601285,601285,601285,601MeanofDep.Var.0.0140.280.00520.490.21NumberofZipCodes26,25019,85619,85619,85619,856Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.107TableB.7:BloodType,2SLSResults,QualityGraftSurvivalGraftSurvivalHLARegistrationVARIABLES1year2yearsMismatchesDuration(Days)BloodTypeA0.21**0.30**0.56-991***(0.086)(0.13)(0.51)(206)BloodTypeAB0.0970.30-1.45*-1,020***(0.20)(0.23)(0.76)(321)BloodTypeB0.20**0.19-0.48-221(0.088)(0.12)(0.53)(263)BloodTypeO0.20**0.0711.06**-351(0.087)(0.14)(0.52)(240)Testing:AvsAB0.110.00812.01**29.1(0.20)(0.25)(0.84)(343)AvsB0.0100.121.03-769***(0.11)(0.15)(0.62)(285)AvsO0.00850.23-0.51-639***(0.094)(0.15)(0.54)(235)ABvsB-0.100.11-0.97-799**(0.21)(0.25)(0.88)(384)ABvsO-0.100.22-2.52***-669*(0.20)(0.25)(0.83)(354)BvsO-0.00160.12-1.54**130(0.10)(0.15)(0.62)(305)Observations198,296181,901223,184224,536MeanofDep.Var.0.930.883.64597NumberofZipCodes17,68517,16518,42818,455Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.108TableB.8:PRACategory,2SLSResults,QuantityFirstStage2SLSEstimatesVARIABLESExchangeLivingDirectAnonymousDeceasedDiedonWLPRA0-400.0050***-0.61***0.13***0.14-0.65***(0.00036)(0.13)(0.035)(0.15)(0.11)PRA40-800.0093***-0.36**0.17***-0.88***0.067(0.0012)(0.16)(0.064)(0.21)(0.16)PRA80-1000.0062***-0.67***0.044-0.35-0.033(0.00091)(0.17)(0.047)(0.26)(0.23)Testing:0-40vs40-80-0.0042***-0.25-0.0471.02***-0.72***(0.0012)(0.19)(0.067)(0.24)(0.18)0-40vs80-100-0.00110.0520.0820.49*-0.62***(0.00095)(0.19)(0.053)(0.28)(0.24)40-80vs80-1000.0031**0.310.13*-0.53*0.099(0.0014)(0.22)(0.076)(0.32)(0.27)Observations287,608281,197281,197281,197281,197MeanofDep.Var.0.0140.280.00520.500.20NumberofZipCodes26,16119,75019,75019,75019,750Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.109TableB.9:PRACategory,2SLSResults,QualityGraftSurvivalGraftSurvivalHLARegistrationVARIABLES1year2yearsMismatchesDuration(Days)PRA0-400.16**0.140.068-885***(0.077)(0.12)(0.44)(198)PRA40-800.160.120.96-267(0.10)(0.13)(0.68)(288)PRA80-1000.39***0.54**2.00**652(0.15)(0.22)(0.86)(431)Testing:0-40vs40-800.00380.017-0.89-618*(0.12)(0.16)(0.76)(320)0-40vs80-100-0.23-0.41*-1.93**-1,537***(0.15)(0.23)(0.90)(443)40-80vs80-100-0.23-0.42*-1.04-919*(0.18)(0.25)(1.05)(490)Observations196,234179,944221,549222,511MeanofDep.Var.0.930.883.64602NumberofZipCodes17,61117,08918,36518,387Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.110TableB.10:AgeCategory,2SLSResults,QuantityFirstStage2SLSEstimatesVARIABLESExchangeLivingDirectAnonymousDeceasedDiedonWLAge0-170.0038***-3.32***0.20*3.05***-0.93***(0.0013)(0.65)(0.12)(0.78)(0.28)Age18-350.010***0.40***0.066**-0.99***-0.48***(0.00093)(0.14)(0.030)(0.13)(0.086)Age36-550.0064***-0.55***0.13***-0.026-0.55***(0.00047)(0.11)(0.033)(0.14)(0.097)AgeOver550.0034***-0.83***0.16***0.20-0.53***(0.00041)(0.17)(0.050)(0.22)(0.15)Testing:0-17vs18-35-0.0062***-3.72***0.134.04***-0.46*(0.0016)(0.63)(0.12)(0.76)(0.27)0-17vs36-55-0.0025*-2.77***0.0743.08***-0.38(0.0014)(0.63)(0.12)(0.76)(0.27)0-17vsOver550.00047-2.49***0.0392.85***-0.40(0.0014)(0.66)(0.13)(0.79)(0.29)18-35vs36-550.0036***0.95***-0.060*-0.96***0.076(0.00097)(0.13)(0.033)(0.14)(0.089)18-35vsOver550.0067***1.23***-0.096**-1.19***0.058(0.00097)(0.19)(0.049)(0.22)(0.15)36-55vsOver550.0030***0.28**-0.036-0.23-0.019(0.00055)(0.14)(0.040)(0.17)(0.12)Observations291,995285,601285,601285,601285,601MeanofDep.Var.0.0140.280.00520.490.21NumberofZipCodes26,25019,85619,85619,85619,856Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.111TableB.11:AgeCategory,2SLSResults,QualityGraftSurvivalGraftSurvivalHLARegistrationVARIABLES1year2yearsMismatchesDuration(Days)Age0-17-0.170.145.32***-9,085***(0.23)(0.40)(1.69)(1,695)Age18-350.120.060-2.56***-15.8(0.072)(0.11)(0.54)(215)Age36-550.35***0.48***0.97**608***(0.066)(0.099)(0.41)(201)AgeOver550.047-0.271.32**-1,997***(0.12)(0.18)(0.57)(287)Testing:0-17vs18-35-0.280.0757.88***-9,069***(0.22)(0.38)(1.64)(1661)0-17vs36-55-0.52**-0.354.36***-9,694***(0.22)(0.39)(1.63)(1666)0-17vsOver55-0.210.414.00**-7,089***(0.24)(0.42)(1.71)(1684)18-35vs36-55-0.23***-0.42***-3.53***-624***(0.070)(0.11)(0.50)(208)18-35vsOver550.0700.33*-3.88***1,981***(0.11)(0.18)(0.66)(287)36-55vsOver550.30***0.75***-0.352,605***(0.098)(0.16)(0.47)(244)Observations198,296181,901223,184224,536MeanofDep.Var.0.930.883.64597NumberofZipCodes17,68517,16518,42818,455Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.112TableB.12:Race,2SLSResults,QuantityFirstStage2SLSEstimatesVARIABLESExchangeLivingDirectAnonymousDeceasedDiedonWLWhite0.0091***-0.47***0.19***-0.39***-0.33***(0.00058)(0.095)(0.039)(0.11)(0.075)Black0.0028***-1.06***0.10**0.91***-0.96***(0.00042)(0.20)(0.047)(0.26)(0.19)Hispanic0.0039***0.260.0024-0.99***-0.27(0.00063)(0.25)(0.045)(0.28)(0.18)Other0.0046***-0.210.0170.096-0.90***(0.00090)(0.26)(0.071)(0.34)(0.26)Testing:WhitevsBlack0.0063***0.58***0.093*-1.30***0.63***(0.00068)(0.19)(0.049)(0.25)(0.17)WhitevsHisp.0.0052***-0.73***0.19***0.60**-0.060(0.00081)(0.24)(0.049)(0.27)(0.17)WhitevsOther0.0044***-0.260.18**-0.480.57**(0.0010)(0.26)(0.079)(0.33)(0.25)BlackvsHisp.-0.0011-1.31***0.100*1.90***-0.69***(0.00073)(0.28)(0.057)(0.33)(0.23)BlackvsOther-0.0018*-0.85***0.0850.82**-0.057(0.00096)(0.31)(0.080)(0.40)(0.29)Hisp.vsOther-0.000720.47-0.015-1.08***0.63**(0.0010)(0.33)(0.080)(0.41)(0.30)Observations291,991285,599285,599285,599285,599MeanofDep.Var.0.0140.280.00520.490.21NumberofZipCodes26,24819,85619,85619,85619,856Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.113TableB.13:Race,2SLSResults,QualityGraftSurvivalGraftSurvivalHLARegistrationVARIABLES1year2yearsMismatchesDuration(Days)White0.10*0.101.31***-719***(0.060)(0.092)(0.34)(131)Black0.45***0.61***-0.29-267(0.13)(0.19)(0.67)(352)Hispanic0.20*-0.0840.50-1,439***(0.11)(0.16)(0.84)(386)Other-0.16-0.18-2.86**625(0.17)(0.22)(1.13)(489)Testing:WhitevsBlack-0.35***-0.50***1.60**-452(0.12)(0.18)(0.63)(329)WhitevsHisp.-0.100.190.81720**(0.11)(0.16)(0.80)(366)WhitevsOther0.270.284.17***-1,344***(0.17)(0.22)(1.11)(484)BlackvsHisp.0.25*0.69***-0.791,172**(0.15)(0.22)(0.95)(472)BlackvsOther0.62***0.78***2.57**-892(0.20)(0.27)(1.21)(568)Hisp.vsOther0.37*0.0943.35***-2,064***(0.19)(0.25)(1.30)(583)Observations198,294181,899223,182224,534MeanofDep.Var.0.930.883.64597NumberofZipCodes17,68517,16518,42818,455Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.114TableB.14:Education,2SLSResults,QuantityFirstStage2SLSEstimatesVARIABLESExchangeLivingDirectAnonymousDeceasedDiedonWLUptoHSEducation0.0037***-0.57***0.096**0.24-0.77***(0.00043)(0.18)(0.046)(0.23)(0.17)MorethanHS0.0080***-0.44***0.099***-0.20*-0.45***(0.00053)(0.10)(0.030)(0.12)(0.084)Testing:UptoHSvs-0.0043***-0.13-0.00280.44**-0.31**MorethanHS(0.00059)(0.16)(0.042)(0.20)(0.15)Observations237,373230,981230,981230,981230,981MeanofDep.Var.0.0160.280.00520.490.21NumberofZipCodes24,85418,46218,46218,46218,462Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,race,education.TableB.15:Education,2SLSResults,QualityGraftSurvivalGraftSurvivalHLARegistrationVARIABLES1year2yearsMismatchesDuration(Days)UptoHSEducation0.30**0.121.95***-480(0.12)(0.16)(0.69)(306)MorethanHS0.13**0.140.32-870***(0.062)(0.093)(0.36)(161)Testing:UptoHSvsMorethanHS0.17-0.0201.63***390(0.11)(0.14)(0.62)(264)Observations157,433142,931179,991181,139MeanofDep.Var.0.930.893.63605NumberofZipCodes16,23515,67617,03417,061Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffects,zipcodeedeffects,andcontrolsforageatlisting,PRA,previoustransplantstatus,bloodtype,gender,ethnicity,education.2)Thenon-death-censoredgraftsurvivalvariablesalsoassumetransplantsurvivalforthosewhoselastknownstatusisalivewithafunctioningkidneytransplant.3)One-yeargraftsurvivalexcludes2013-14data,twoyearsexcludes2012-14.115TableB.16:HeterogeneousEffectsonLivingDonationbyBloodTypeTypeOTypeATypeBTypeABVARIABLESDonorsDonorsDonorsDonorsTotalExchangesNearby0.0047***0.0028***0.0013***0.00025ClusteredStd.Errors(0.0012)(0.00071)(0.00049)(0.00017)MeanofDep.Var.[0.21][0.087][0.026][0.0032]PercentageChange2.2%3.2%5.0%7.9%Observations251,157251,157251,157251,157NumberofZipCodes25,86025,86025,86025,860Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffectsandzipcodeedeffects.TableB.17:HeterogeneousEffectsonAnonymousDonationbyBloodTypeTypeOTypeATypeBTypeABVARIABLESAnon.DonorsAnon.DonorsAnon.DonorsAnon.DonorsTotalExchangesNearby0.00089***0.00039***0.00026**0.000081*ClusteredStd.Errors(0.00021)(0.00015)(0.00012)(0.000047)MeanofDep.Var.[0.0027][0.0022][0.0007][0.0002]PercentageChange32.6%17.4%34.9%33.3%Observations251,157251,157251,157251,157NumberofZipCodes25,86025,86025,86025,860Clusteredstandarderrorsinparentheses(atzipcodelevel).***p<0.01,**p<0.05,*p<0.11)Regressionsincludemonth-yearedeffectsandzipcodeedeffects.116FigureB.1:KidneyTransplants/WaitingListSize,Per1,000,000U.S.ResidentsSource:OPTNDataasof5/6/2015,http://optn.transplant.hrsa.gov117APPENDIXCTABLESANDFIGURESFORCHAPTER3TableC.1:WaitingListAdditions,byOrganandYearListingYearHeartIntestineKidneyLiverLungPancreasAll19883,05412,1942,18234026518,03519893,15912,7742,95043555019,86819903,78413,4183,68369475222,33119913,983113,8714,1761,10587124,00719924,12615,9234,8071,3571,09827,31119933,9975917,1355,5221,5201,26829,50119943,8848517,7476,2291,7281,41731,09019954,3829119,2717,3291,8901,61634,57919964,0308819,7058,0541,9901,66035,52719973,89713420,4608,6202,0791,73936,92919984,07515221,7059,5372,2211,93039,62019993,65014922,80310,5202,0992,33041,55120003,56517024,29010,8802,0902,79443,78920013,50621924,12211,1262,1382,73743,84820023,31820325,2269,6451,9752,64143,00820033,00820526,05310,3242,0222,56144,17320042,96025028,85210,8562,0782,71347,70920052,89528430,92510,9871,6232,67549,38920063,11431733,17111,0371,8522,55152,04220073,16228134,47911,0832,0092,39853,41220083,43726734,65411,1752,0582,35053,94120093,57926035,65811,2622,3442,23155,33420103,58424136,44412,0102,5262,14456,94920113,50418435,59511,9252,5201,87655,60420123,70615937,05811,6112,3921,93556,86120134,03118038,62512,0202,5761,75059,182Note:Thesenumbersarecalculatedatasinglepointintimeineachyear.Source:authors'calculationsfromSRTRdata.118TableC.2:WaitingListExitsbyYearandReasonforLeavingDeceasedTransferredImproved,TransplantLivingDonorMedicallytoanotherTransplantDeteriorated,atAnotherDonorTotalYearTransplantUnsuitablecenterDiedOtherNotNeededTooSickCenterTransplantAllOthersWaitlistExits19871,9461047035763178551023,34419889,9464712821,5801,90443031427815,217198910,6265472861,7842,14363541024716,692199012,4736754142,0562,28070852829819,445199113,0337984822,5322,19968363440320,775199213,3281,0196842,7692,13758379830021,622199314,5871,1899573,1402,37669393728524,184199415,0981,3597623,3432,6256821,21832825,435199515,8343921,0473,7082,1214847039141,49735527,055199615,8689714,2881,9237961,0619911,72336027,981199716,1701,1504,8321,9335701,1491,0571,97833629,176199816,9041,2545,5372,0616021,2281,0652,25926731,178199916,9191,3486,8352,0946801,3691,3142,69247433,728200017,2401,8626,4551,6806741,4731,4233,45577035,039200117,5541,9297,0651,4136341,5841,4564,00292536,574200218,1881,7477,2022,1431,3091,8621,5564,159131239,498200318,5611,7247,1381,8361,0531,6461,5684,350112539,019200419,9491,7297,3731,9031,1311,6321,9734,765101241,491200521,1172,6947,3732,0671,0611,9052,0734,952113444,378200622,1352,2937,3702,7259662,1192,0955,063120745,974200721,9992,1727,1353,1211,5902,4632,2904,911144447,129200821,7031,9987,1703,9511,4982,9472,3595,121155748,305200921,8151,8587,1583,5211,3153,4272,4325,633124248,403201022,0581,9567,0493,7971,2763,8802,5385,766124349,564201122,4571,9847,3014,1381,1464,4412,7325,425133150,955201222,1412,3766,9863,6671,2574,7392,6705,361186651,063201322,9352,4326,7333,7611,1835,2422,7415,534191952,480Notes:fiAllothersflincludesfiremovedinerrorfl,fichangedtokidney/pancreasfl,fideceaseddonoremergencytransplantfl,fideceaseddonormulti-organtransplantfl,fiinactiveprogramfl,fidiedduringtransplantfl,andfiunabletocontracttransplantandrefusedtransplantfl.Notethatmedicallyunsuitablebecamefiimproved,transplantnotneededflandfideteriorated,toosickflin1995.Blankcellsindicatethatthecountisbelow25.119TableC.3:NumberofPersonsonWaitlists,byOrganandYearYearHeartIntestineKidneyLiverLungPancreas19851,1521986343,70885198761710,1314312835198896912,4465538714219891,26614,97569912128219901,67916,7051,02034139419912,13818,4491,44365535119922,62521,5192,11292913819932,7774324,2262,8051,20120519942,8327126,7613,7911,57025119953,3367830,0835,2881,84831519963,5197833,3716,9302,20135819973,6648736,6658,8312,53337919983,8829339,98910,9362,97745419993,72810042,70313,1133,22751720003,71313546,09515,0743,38074620013,64016048,95316,6153,5161,04320023,46817351,46915,5053,5191,14320033,20816254,34815,5763,5861,31520042,93317958,11115,4053,5711,38720052,68917660,99415,2072,9001,37220062,51619864,30614,6372,5991,44120072,36017667,30114,1062,0771,32220082,49016872,08713,7771,8881,25920092,76318177,29613,8231,7601,17620102,98022082,41314,2621,7341,09420112,95823185,81914,3911,6311,00320123,20321890,82814,2081,56091920133,51222496,52014,3011,562906Notes:Theseincludebothactiveandinactivepatientsonwaitlistsforsingleorgans,i.e.,theyexcludethoseregisteredoneithertheheart-lungorthekidney-pancreasmulti-organwaitlists.Theyarecalculatedatasinglepointintimeineachyear.Source:authors'calculationsfromSRTRdata.Blankcellsindicatethatthecountisbelow25.120TableC.4:ChangesinStateMotorcycleHelmetLaws,1988-2012YearUniversaltoPartialPartialtoUniversal1988OR(6)1989NE(1),TX(9)1990WA(6)19911992CA(1),MD(10)...1997AR(8),TX(9)1998KY(7)1999LA(8)2000FL(7)200120022003PA(9)2004LA(8)...2012MI(4)Note:Themonthalawchangedislistedinparentheses,wherefi1fldenotesJanuary,fi2fldenotesFebruary,etc.Source:InsuranceInstituteforHighwaySafety,http://www.iihs.org/laws/default.aspx121TableC.5:NumberofDonorsandNumberofOrgansTransplantedfromMVAandNon-MVADonorsDonorsOrgansTransplantedperMVADonorOrgansTransplantedpernon-MVADonorYearNon-MVAMVALungKidneyHeartIntestineLiverPancreasTotalLungKidneyHeartIntestineLiverPancreasTotal1988274914040.041.850.4900.440.062.890.041.70.3800.390.062.571989288412120.051.810.4900.580.123.060.061.70.400.50.092.751990334612550.081.810.5700.620.143.220.071.670.4200.560.12.831991340511650.121.830.5700.680.153.360.131.670.4400.620.112.97199235579990.171.830.580.010.710.153.450.171.670.4600.650.113.061993380011050.191.830.620.010.750.23.60.181.650.440.010.670.143.091994387412600.181.820.5900.750.23.540.21.60.4300.680.153.071995397514100.281.80.580.010.780.263.720.231.570.40.010.70.163.071996409313490.251.80.590.010.790.273.70.211.550.380.010.720.163.031997410813850.291.810.580.010.840.293.820.251.540.380.010.710.163.051998440614000.251.810.580.010.840.313.810.211.530.360.010.740.183.021999448513450.241.830.550.010.850.353.830.221.50.340.010.750.1832000454114490.291.80.510.020.840.343.810.231.470.330.010.740.192.972001470613770.281.810.550.020.860.393.920.241.470.320.020.740.182.962002473214640.311.830.510.020.880.383.940.251.480.310.020.780.193.022003503714250.311.820.510.020.90.393.940.251.430.270.020.810.162.932004563115210.321.80.470.020.880.373.860.251.380.240.020.80.172.852005607615190.361.80.470.030.890.373.920.291.380.240.020.790.152.862006637916440.341.820.470.030.880.313.840.281.40.230.020.770.142.842007651115830.341.780.460.020.860.323.770.291.370.240.020.750.132.82008671712760.351.820.480.020.850.33.820.311.390.240.020.740.132.832009678812350.391.810.470.030.820.283.80.341.380.250.020.750.132.872010669012560.431.80.480.030.830.283.860.371.420.270.020.740.122.942011684712830.441.820.450.020.810.263.80.371.430.260.010.740.112.932012688112670.421.770.470.020.810.253.740.351.420.270.010.720.112.882013697413000.471.810.50.020.810.243.850.381.420.270.010.740.12.93Note:Theseincludebothactiveandinactivepatientsonwaitlists.Theyarecalculatedatasinglepointintimeineachyear.122TableC.6:EstimatesoftheEffectoftheRepealofHelmetLawsonPerCapitaOrganDonors,OrganDonations,andOrganTransplants,byOrganMVAOrganMVAOrganNon-MVAOrganNon-MVAOrganDonorsTransplantsDonorsTransplants(1)(2)(3)(4)Overall0.9063.4290.336-0.79(0.234)(0.785)(0.989)(1.914)[4.886][17.092][16.949][49.974]ByOrganLung0.4340.253(0.126)(0.400)[1.299][4.452]Kidney1.577-0.933(0.392)(1.073)[8.386][25.295]Heart0.421-0.516(0.104)(0.262)[2.409][5.334]Intestine0.0060.105(0.017)(0.050)[0.072][0.235]Liver0.7740.361(0.206)(0.609)[3.665][12.279]Pancreas0.216-0.06(0.155)(0.229)[1.261][2.380]Notes:1)Allestimationsamplesconsistof57DSAsfrom1987to2013.TheunitofobservationisaDSA-year.AllmodelsincludeindicatorsforyearsandDSAs.2)Standarderrors,inparentheses,arerobusttoclusteringwithinDSAovertime.3)Samplemeansforrelevantdependentvariablesarelistedinbrackets.123TableC.7:EstimatesoftheEffectoftheRepealofHelmetLawsonWaitingListAdditionsbyIn-VersusOut-of-AreaAllAdditionsIn-DSAOut-of-DSA(1)(2)(3)Overall18.66510.3228.343(8.733)(6.213)(4.906)[147.717][114.885][32.832]ByOrganLung2.11.190.91(1.341)(0.576)(0.923)[7.911][4.790][3.121]Kidney8.4813.0875.395(6.887)(5.477)(2.736)[87.803][71.861][15.942]Heart0.0910.355-0.264(1.354)(1.041)(0.498)[12.339][9.472][2.867]Intestine0.7960.0940.703(0.427)(0.223)(0.305)[1.668][0.568][1.100]Liver5.2263.9771.249(3.696)(1.700)(2.479)[35.257][25.719][9.538]Pancreas-0.014-0.3310.317(0.816)(0.454)(0.435)[2.423][1.619][0.803]1)Allestimationsamplesconsistof57DSAsfrom1987to2013.TheunitofobservationisaDSA-year.AllmodelsincludeindicatorsforyearsandDSAs.2)Standarderrors,inparentheses,arerobusttoclusteringwithinDSAovertime.3)Samplemeansforrelevantdependentvariablesarelistedinbrackets.124TableC.8:EstimatesoftheEffectofHelmetLawRepealsonWaitingListAdditionsbyIn-VersusOut-of-AreaandbyMultilistingStatusNoMultilistingsMultilistingsAllAdditionsIn-DSAOut-of-DSAAllAdditionsIn-DSAOut-of-DSA(1)(2)(3)(4)(5)(6)Overall6.3784.232.14911.9955.7446.252(6.918)(5.445)(3.034)(6.244)(3.231)(3.787)[118.406][97.010][21.396][41.081][26.967][14.114]ByOrganLung2.3481.2781.070.3060.180.126(1.234)(0.578)(0.765)(0.304)(0.098)(0.227)[5.597][3.522][2.076][1.076][0.509][0.566]Kidney-0.283-0.9790.6967.7393.4674.272(5.426)(4.810)(1.399)(3.729)(1.883)(2.410)[67.597][59.197][8.400][26.144][17.483][8.660]Heart0.2720.408-0.137-0.0440-0.043(1.104)(0.871)(0.402)(0.251)(0.164)(0.108)[10.718][8.329][2.390][1.200][0.822][0.378]Intestine0.144-0.050.193-0.089-0.003-0.086(0.289)(0.091)(0.227)(0.041)(0.016)(0.042)[0.479][0.172][0.307][0.126][0.035][0.092]Liver2.6362.706-0.0712.7291.41.329(2.572)(1.504)(1.361)(1.870)(0.744)(1.241)[27.791][21.043][6.749][5.735][3.415][2.319]Pancreas0.052-0.2020.254-0.175-0.1790.004(0.463)(0.249)(0.255)(0.159)(0.110)(0.078)[1.284][0.876][0.408][0.755][0.486][0.269]1)Allestimationsamplesconsistof57DSAsfrom1987to2013.TheunitofobservationisaDSA-year.AllmodelsincludeindicatorsforyearsandDSAs.2)Standarderrors,inparentheses,arerobusttoclusteringwithinDSAovertime.3)Samplemeansforrelevantdependentvariablesarelistedinbrackets.125TableC.9:EstimatesoftheEffectofHelmetLawRepealsonPerCapitaLivingOrganDonors,byRelationtotheRecipientOverall-3.445(1.403)[15.200]ByDonor'sRelationshiptoIntendedRecipientParent-0.366(0.18)[2.162]Child-0.481(0.248)[2.415]Sibling-0.683(0.347)[4.725]OtherRelative-0.233(0.101)[1.022]Spouse-0.44(0.193)[1.544]AllOtherDirectedDonations-1.202(0.36)[2.373]Anonymous(Undirected)-0.125(0.041)[0.152]1)Allestimationsamplesconsistof57DSAsfrom1987to2013.TheunitofobservationisaDSA-year.AllmodelsincludeindicatorsforyearsandDSAs.2)Standarderrors,inparentheses,arerobusttoclusteringwithinDSAovertime.3)Samplemeansforrelevantdependentvariablesarelistedinbrackets.126TableC.10:EstimatesoftheEffectofHelmetLawRepealsonGraftSurvivalamongTransplantRecipients,byOrganGraftSurvivalAfterGraftSurvivalAfterGraftSurvivalAfter1year3years5years(1)(2)(3)Overall0.0070.0050.002(0.004)(0.004)(0.005)[0.887][0.796][0.708]ByOrganKidney-0.002-0.004-0.004(0.006)(0.008)(0.008)[0.919][0.832][0.737]Liver0.0240.0210.015(0.011)(0.012)(0.013)[0.834][0.747][0.681]Heart0.0330.0290.011(0.010)(0.015)(0.014)[0.865][0.788][0.715]Lung0.0490.0230.001(0.022)(0.019)(0.023)[0.780][0.594][0.459]Pancreas0.1350.0920.087(0.051)(0.050)(0.055)[0.773][0.634][0.539]Intestine0.0540.02-0.03(0.115)(0.124)(0.136)[0.678][0.510][0.410]1)Allestimationsamplesconsistof57OPOsfrom1987to2013.TheunitofobservationisaDSA-year.AllmodelsincludeindicatorsforyearsandDSAs.2)Standarderrors,inparentheses,arerobusttoclusteringwithinDSAovertime.3)Samplemeansforrelevantdependentvariablesarelistedinbrackets.127FigureC.1:DonationServiceAreaMapoftheUnitedStatesSource:Wedd,HarperandBiggins(2013).128FigureC.2:EventStudyEstimatesofHelmetLawRepealsonPerCapitaOrganDonorsandTransplantsNotes:Authors'calculationsfromSRTRdata.129FigureC.3:EventStudyEstimatesofHelmetLawRepealsonPerCapitaWaitlistAdditionsNotes:Authors'calculationsfromSRTRdata.130FigureC.4:EstimatesofHelmetLawRepealsonFlowofPerCapitaWaitlistAdditionstoNeighboringOPOswithaLawChange(Treatment)relativetotheFlowtoNeighboringOPOswithoutaLawChange(Control):AllOrgansNotes:Authors'calculationsfromSRTRdata.Meansofpercapitaws.131FigureC.5:EstimatesofHelmetLawRepealsonFlowofPerCapitaWaitlistAdditionstoNeighboringOPOswithaLawChange(Treatment)relativetotheFlowtoNeighboringOPOswithoutaLawChange(Control):KidneysandNon-KidneysNotes:Authors'calculationsfromSRTRdata.Meansofpercapitaws.132FigureC.6:EstimatesofHelmetLawRepealsonFlowofPerCapitaWaitlistAdditionstoNeighboringOPOswithaLawChange(Treatment)relativetotheFlowtoNeighboringOPOswithoutaLawChange(Control):KidneyMultilistersandKidneyNon-MultilistersNotes:Authors'calculationsfromSRTRdata.Meansofpercapitaws.133FigureC.7:EventStudyEstimatesofHelmetLawRepealsonOne-yearGraftSurvivalNotes:Authors'calculationsfromSRTRdata.134APPENDIXDPROOFSFORCHAPTER1D.1ProofforProposition1Proposition1.Whenexchangeisintroduced,thequalitythreshholdQwhereiisindifferentbetweendonatingdirectlyandnotdonatingincreases.Assuch,Prob[Ui(Direct)Ui(None)]willdecreasemeaningthatfewerindividualswilldonatedirectly.Proof.PotentialdonoriwilldonatedirectlywhenUi(Direct)Ui(None),orWiC(Direct;Z)+Qj(Direct;Z)+aiE[Qj(Direct;Z)]Wi+E[Qj(None;Z)]+aiE[Qj(None;Z)](D.1)IfUi(None)increasesmorethanUi(Direct)whenexchangeisintroduced,thenProb[Ui(Direct)Ui(None)]willdecrease.Thisimpliesthatthequalitythreshholdforindifferencebetweendonatingdirectlyandnotatall,Q,willincrease.Iassumethatexchangeintroductionhasnoeffectonthecostofdonatingdirectly,northequalityofi'sdirectdonationtopatientj.Ialsoassumethattheeffectofexchangeontheoutcomesofotherpatientsisthesamewhetheridonatesdirectlyordoesnotdonateatall.Theeffectofexchangeintroductionontheutilityderivedfromeachchoiceisgivenbythefollowing:Ui(DirectjZ=1)Ui(DirectjZ=0)=aihE[Qj(Direct;Z=1)]E[Qj(Direct;Z=0)]i=aihE[Qj(None;Z=1)]E[Qj(None;Z=0)]i(D.2)135andUi(NonejZ=1)Ui(NonejZ=0)=E[Qj(None;Z=1)]E[Qj(None;Z=0)]+aihE[Qj(None;Z=1)]E[Qj(None;Z=0)]i(D.3)Therefore,thegaintoUi(None)islargerthanthegaintoUi(Direct),providedthatj'sexpectedoutcomeimproveswiththeintroductionofexchangewhenidoesnotdonate.E[Qj(None;Z=1)]E[Qj(None;Z=0)]+aihE[Qj(None;Z=1)]E[Qj(None;Z=0)]i>aihE[Qj(None;Z=1)]E[Qj(None;Z=0)]i(D.4)orE[Qj(None;Z=1)]E[Qj(None;Z=0)]>0(D.5)Sincetheeffectofexchangeintroductionontheoutcomesofothersisthesamewhetheridonatesdirectlyornotatall,thisresultholdsbecausepatientj'soutcomedoesnotimprovewithexchangeintroductionunderdirectdonationwhileherexpectedoutcomedoesimprovewhenidoesnotdonate.136D.2ProofforProposition2Proposition2.Whenexchangeisintroduced,thequalitythreshholdQwhereiisindifferentbetweendonatinganonymouslyandnotdonatingchangesambiguously.Individualswillbemorelikelytodonateanonymouslyifitistostartadonorchain,andlesslikelytodonateanonymouslytoasingleindividual.Proof.PotentialdonoriwilldonateanonymouslywhenUi(Anon)Ui(None),orWiC(Anon;Z)+aiE[Qk(Anon;Z)]Wi+aiE[Qk(None;Z)](D.6)IfUi(Anon)increasesmorethanUi(None)whenexchangeisintroduced,thenProb[Ui(Anon)Ui(None)]willincrease.Iassumethatthecostofdonatinganonymouslyisthesamewhetherornotexchangehasbeenintroduced.Theeffectofexchangeintroductionontheutilityderivedfromeachchoiceisgivenbythefollowing:Ui(AnonjZ=1)Ui(AnonjZ=0)=aihE[Qk(Anon;Z=1)]E[Qk(Anon;Z=0)]i(D.7)andUi(NonejZ=1)Ui(NonejZ=0)=aihE[Qk(None;Z=1)]E[Qk(None;Z=0)]i(D.8)SoiwillbeatleastaslikelytodonateanonymouslyifE[Qk(Anon;Z=1)]E[Qk(Anon;Z=0)]E[Qk(None;Z=1)]E[Qk(None;Z=0)](D.9)Whendonatinganonymously,atleastonepersonreceivesatransplant.Then,letusdecomposeQkintoQaandQa.Qamayincludetheoutcomesofpatientstransplantedviadonorchain,137whereQaistheinitialrecipient.TheinequalitythenbecomesQa(Anon;Z=1)+E[Qa(Anon;Z=1)]Qa(Anon;Z=0)E[Qa(Anon;Z=0)]E[Qa(None;Z=1)]+E[Qa(None;Z=1)]E[Qa(None;Z=0)]E[Qa(None;Z=0)](D.10)Now,supposetheoutcomesofotherswouldnotimprovewithi'sdonationwhenexchangedoesnotexist,suchthatE[Qa(Anon;Z=0)]=E[Qa(None;Z=0)].Theinequalitythenreducesto:Qa(Anon;Z=1)+E[Qa(Anon;Z=1)]Qa(Anon;Z=0)E[Qa(None;Z=1)]+E[Qa(None;Z=1)]E[Qa(None;Z=0)](D.11)orQa(Anon;Z=1)E[Qa(None;Z=1)]+E[Qa(Anon;Z=1)]E[Qa(None;Z=1)]Qa(Anon;Z=0)E[Qa(None;Z=0)](D.12)ThismeansthatProb[Ui(Anon)Ui(None)]willincreaseif,withi'sdonationwhenexchangeexists,theoutcomesofaand-aimprovebyatleastasmuchastheimprovementina'soutcomewhenidonatesintheabsenceofexchange.Itislikelythattheinitialrecipient'soutcomeimprovesbyasmalleramountwithi'sdona-tionwhenexchangeexistscomparedtowhenitisabsent.Thefollowingshouldholdiftheinitialrecipientisthesame:Qa(Anon;Z=1)=Qa(Anon;Z=0),whileE[Qa(None;Z=1)]>E[Qa(None;Z=0)].Iftheinitialrecipientisdifferentwhenexchangeexistscomparedtowhenitisabsent,thenwewouldexpectQa(Anon;Z=1)Qa(Anon;Z=0).Tobeconservative,assumethatQa(Anon;Z=1)=Qa(Anon;Z=0).ThentheinequalityfurtherreducestoE[Qa(Anon;Z=1)]E[Qa(None;Z=1)]E[Qa(None;Z=1)]E[Qa(None;Z=0)](D.13)ThismeansthatProb[Ui(Anon)Ui(None)]willincreaseiftheimprovementinotherpatients'outcomeswhenidonateswhileexchangeexistsoutweighstheimprovementina'sexpectedout-comeduetotheintroductionofexchangeinthecasewhereidoesnotdonate.138Ifidoesnotdonatetostartadonorchain,thenitisreasonabletoassumethatE[Qa(Anon;Z=1)]E[Qa(None;Z=1)]=0,andthereforethatiwouldbelesslikelytodonateanonymouslytojustonepersonwhenexchangeexists.However,ifi'sdonationhelpsevenoneadditionalperson,b,whowouldreceiveanexchangetransplantaspartofadonorchain,thentheinequalityismuchmorelikelytohold.TheimprovementinQb,Qb(Exch;Z=1)E[Qb(None;Z=1)],isverylikelytooutweighthesecond-orderimprovementinQa,E[Qa(None;Z=1)]E[Qa(None;Z=0)].139APPENDIXESUPPLEMENTARYINFORMATIONFORCHAPTER3E.1IdentifyingTransplantCandidatesWhoMultilistThereisnosinglecodeinthedatathatwhetheratransplantcandidateismultilisted.Usingauniquepatientvariable,wecancreateamultilisting.Foreachpatient,weidentifyallregistrationsthatbelongtothesamespellforasingle-organtransplantbyworkingbackwardsfromwhenaspellforends.Aspellendswhentheindividualreceivesatransplant,leavesthewaitlistwhentherearenoopenregistrations,dieswithoutreceivingatransplant,orisstillonawaitlistwhenthedatawereextractedin2014.Foreachindividual,allregistrationsinwhichthelistingdateisthesamedateorearlierthantheobservedtransplantordeatharecodedaspartoftheindividual'sspell.Asubsequentspellbeginswhenaregistrationoccursfollowingtheendofapreviousspellfromtransplantandendswhenweobserveatransplantordeath.Allregistrationsthatbeginafterthedateofthe2ndspellandendbeforeoratthesametimeasthe2ndspellarecodedinthe2ndspellandsoon.Registrationsthatoccurafterthemostrecenttransplantarecountedasthespell.Ifapatienthasneverhadatransplant,alloftheirregistrationsforthatgivenorganarecategorizedasasinglespell.Wecodeallregistrationsaspartofamultiplelistingornot.Wealsocodethechronologicalorderofthemultilistingregistrationswithinspells.Aspecialcaseinvolvedcandidateswhoarelistedformulti-organregistrations(kidney-pancreasandheart-lung).Wesplittheseintotwosingle-organobservations.Forexample,kidney-pancreasregistrationsaresplitintoonekidneyandonepancreaslisting.Therefore,wefollowthekidneyspellandthepancreasspell.Wedothisbecausesomeofthesemultiorganwaitlistregistrationsendwhenthecandidatesreceivesatransplantforoneofthetwoorgans.Observationsoftransplantswithoutacorrespondingwaitinglistregistrationareassumedtonotbeonthewaitinglist.140E.2DesignatingCountiestoDonationServiceAreasDonationServiceAreas(DSAs)aretheunitofobservationthroughoutmostofouranalysis.AsFigure1illustrates,theDSAshavetheirownboundaries.In2015thereare58DSAs;weuseonly57,droppingPuertoRico.AccordingtopersonalcorrespondenceswithPeggyeWilkersonattheCentersforMedicareandMedicaidServices(CMS),thecurrentcountydesignationswereeffectiveMay31,2006.TheSRTRdataprovideuswiththismappingforthecurrentcountiesandDSAs.WedesignateeachcountytoitscurrentDSAbasedonthedataprovidedbySRTR.SRTRfurtherwhencountiesaresplitbetweentwoDSAs.Inparticular,thereisavariableinthecountyleveldatacalledfiCountyfraction:Fractionofcounty'sreferralstothisOPOinthisyear,"whichistheshareofdeathsinacountythatwerehandledbyaOPO.Only2percentofthecountiesaresplitbetweentwoDSAs,andonly1.4percentofthecountiesaresplitbetweentwoDSAsformorethanoneyear.Whenthecountyissplitinonlyoneyear,itisbecausetheDSA'sboundarieswerechanginginthatyear(seebelow).Thesharesdonotchangemuchfromyeartoyear.Toaccountforthis,weassign,forallyears,theentirecountytotheDSAwherethelargershareofthecountyreferralswasmadein2013.ManythankstoBrynThompsonatSRTRforhelpingussorttheseissuesout.AsecondissueforassigningthecountiestoDSAsisthattheDSAschangedovertimeand,insomecases,thenamesoftheOPOsthatadministertheDSAschangedovertime.Therefore,inourindividual-leveldata,wehavetransplantcandidateslistedinOPOs/DSAsthatnolongerexistandarenotavailableinthecurrentmappingbetweencountiesandDSAs.MarkPasterattheAssociationofOrganProcurementOrganizationandChasMacKenzieattheLifeChoiceDonorServicesprovidedvaluableinformationonthehistoryofaWisconsinOPOnamechangeandaConnecticutnamediscrepancyinourdata.AmoresubstantiveissueisthatmanyoftheoriginalDSAseventuallymergedintothecurrentsetofDSAs.Thatis,30OPOs/DSAsintheSRTRdatasetwereinexistenceatonepointbutnolongerexist.13ofthosewereonlyinexistence141between1987and1988.WedonothavedataonwhichcountieswereintheDSAsintheearlyyears;weonlyknowthattheDSAsexisted.PeggyeWilkersonofCMSsuggestedthatthecounty-DSAconcordancefromthoseyearsisnotreadilyavailable.Themoststraightforwardsolution,webelieve,istoassumethatthecurrentcountytoDSAdesignationwasalwaysinplace.ItseemsunlikelythatthiswouldsubstantivelyaffectourresultssincewearesimplytreatingtwoDSAsasiftheywerealwaysoneandtheDSAsarelikelytobeaffectedbythesamestatelaws,exceptinfewcaseswhereDSAscrossstatelines.Mostofthe30DSAsthatnolongerexistareinstateswheretheDSAsarewhollycontainedinasinglestate.142BIBLIOGRAPHY143BIBLIOGRAPHYAbadie,Alberto,andSebastienGay.2006.fiTheimpactofpresumedconsentlegislationoncadavericorgandonation:across-countrystudy.flJournalofHealthEconomics,25(4):599Œ620.Anderson,DrewM.2016.fiDirectandIndirectEffectsofPoliciestoIncreaseKidneyDonations.flWorkingPaper.UniversityofWisconsin-Madison.Andreoni,James.1990.fiImpureAltruismandDonationstoPublicGoods:ATheoryofWarm-GlowGiving?flEconomicJournal,100(401):464Œ77.Ashlagi,I.,D.S.Gilchrist,A.E.Roth,andM.A.Rees.2011.fiNonsimultaneousChainsandDominosinKidney-PairedDonationŠRevisited.flAmericanJournalofTransplantation,11(5):984Œ994.Ausubel,LawrenceM,andThayerMorrill.2014.fiSequentialKidneyExchange.flAmericanEconomicJournal:Microeconomics,6(3):265Œ285.Becker,GaryS,andJulioJorgeElias.2007.fiIntroducingincentivesinthemarketforliveandcadavericorgandonations.flTheJournalofEconomicPerspectives,21(3):3Œ24.Bilgel,2012.fiTheimpactofpresumedconsentlawsandinstitutionsondeceasedorgandonation.flTheEuropeanJournalofHealthEconomics,13(1):29Œ38.Bingaman,A.W.,F.H.WrightJr.,M.Kapturczak,L.Shen,S.Vick,andC.L.Murphey.2012.fiSingle-CenterKidneyPairedDonation:TheMethodistSanAntonioExperience.flAmer-icanJournalofTransplantation,12(8):2125Œ2132.Bray,RA,JDLNolen,CLarsen,TPearson,KANewell,KKokko,AGuasch,PTso,JBMendel,andHMGebel.2006.fiTransplantingtheHighlySensitizedPatient:TheEmoryAl-gorithm.flAmericanjournaloftransplantation,6(10):2307Œ2315.Byrne,MargaretM,andPeterThompson.2001.fiApositiveanalysisofincentivesforcadavericorgandonation.flJournalofHealthEconomics,20(1):69Œ83.Cameron,AndrewM,ABMassie,CEAlexander,BStewart,RobertAMontgomery,NRBenavides,GDFleming,andDLSegev.2013.fiSocialmediaandorgandonorregistration:theFacebookeffect.flAmericanJournalofTransplantation,13(8):2059Œ2065.Cecka,JM.2010.fiCalculatedPRA(CPRA):thenewmeasureofsensitizationfortransplantcan-didates.flAmericanJournalofTransplantation,10(1):26Œ29.Clarke,RobertaN.2008.fiOrgandonation:amarketingchallenge.flHealthMarketingQuarterly,24(3-4):189Œ200.144Cypel,Marcelo,JonathanCYeung,MingyaoLiu,MasakiAnraku,FengshiChen,WojtekKarolak,MasaakiSato,JaneLaratta,SassanAzad,MindyMadonik,etal.2011.fiNor-mothermicExVivoLungPerfusioninClinicalLungTransplantation.flNewEnglandJournalofMedicine,364(15):1431Œ1440.Davis,ConnieL,andFrancisLDelmonico.2005.fiLiving-donorkidneytransplantation:are-viewofthecurrentpracticesforthelivedonor.flJournaloftheAmericanSocietyofNephrology,16(7):2098Œ2110.Deffains,Bruno,andJeanMercierYthier.2010.fiOptimalproductionoftransplantcareser-vices.flJournalofPublicEconomics,94(9):638Œ653.Delmonico,FrancisL.2004.fiExchangingkidneys-advancesinliving-donortransplantation.flNewEnglandJournalofMedicine,350(18):1812Œ1813.Delmonico,FrancisL,PaulEMorrissey,GeorgeSLipkowitz,JeffreySStoff,JonathanHim-melfarb,WilliamHarmon,MarthaPavlakis,HelenMah,JaneGoguen,RichardLuskin,etal.2004.fiDonorkidneyexchanges.flAmericanJournalofTransplantation,4(10):1628Œ1634.Dickert-Conlin,Stacy,ToddElder,andBrianMoore.2011.fiDonorcycles:MotorcycleHelmetLawsandtheSupplyofOrganDonors.flJournalofLawandEconomics,54(4):907Œ935.Faravardeh,Arman,MieEickhoff,ScottJackson,RichardSpong,AleksandraKukla,NaimIssa,ArthurJMatas,andHassanNIbrahim.2013.fiPredictorsofGraftFailureandDeathinElderlyKidneyTransplantRecipients.flTransplantation,96(12):1089Œ1096.Fernandez,JoseM,DavidHHoward,andLisaStohrKroese.2013.fiTheEffectofCadavericKidneyDonationsonLivingKidneyDonations:AnInstrumentalVariablesApproach.flEco-nomicInquiry,51(3):1696Œ1714.Fuggle,SusanV,JoanneEAllen,RachelJJohnson,DaveCollett,PhilipDMason,Christo-pherDudley,ChristopherJRudge,JAndrewBradley,ChristopherJEWatson,etal.2010.fiFactorsAffectingGraftandPatientSurvivalAfterLiveDonorKidneyTransplantationintheUK.flTransplantation,89(6):694Œ701.Gentry,SE,DLSegev,MarySimmerling,andRAMontgomery.2007.fiExpandingKidneyPairedDonationthroughParticipationbyCompatiblePairs.flAmericanJournalofTransplanta-tion,7(10):2361Œ2370.Gentry,SommerE,RobertAMontgomery,andDorryLSegev.2011.fiKidneypaireddonation:fundamentals,limitations,andexpansions.flAmericanJournalofKidneyDiseases,57(1):144Œ151.Gjertson,DavidW,andJMichaelCecka.2000.fiLivingunrelateddonorkidneytransplantation.flKidneyInternational,58(2):491Œ499.Gordon,ElisaJ,DanielaPLadner,JuanCarlosCaicedo,andJohnFranklin.2010.fiDispari-tiesinKidneyTransplantOutcomes:AReview.flSeminarsinNephrology,30(1):81Œ89.145Gore,JohnL,GMDanovitch,MSLitwin,P-TTPham,andJenniferSSinger.2009.fiDispar-itiesintheUtilizationofLiveDonorRenalTransplantation.flAmericanJournalofTransplanta-tion,9(5):1124Œ1133.Hanto,RuthanneL,SSaidman,AlvinERoth,andFDelmonico.2010.fiTheEvolutionofaSuccessfulKidneyPairedDonationProgram.flTransplantation,90:940.Harrison,TylerR,SusanEMorgan,andLisaVChewning.2008.fiThechallengesofsocialmarketingoforgandonation:Newsandentertainmentcoverageofdonationandtransplanta-tion.flHealthMarketingQuarterly,25(1-2):33Œ65.Howard,DavidH.2007.fiProducingorgandonors.flTheJournalofEconomicPerspectives,21(3):25Œ36.Howard,DavidH.2011.fiWaitingtimeasapricefordeceaseddonorkidneys.flContemporaryEconomicPolicy,29(3):295Œ303.Keith,Douglas,ValarieBAshby,FriedrichKPort,andAlanBLeichtman.2008.fiInsuranceTypeandMinorityStatusAssociatedwithLargeDisparitiesinPrelistingDialysisAmongCan-didatesforKidneyTransplantation.flClinicalJournaloftheAmericanSocietyofNephrology,3(2):463Œ470.Kessler,JuddB,andAlvinERoth.2012.fiOrganallocationpolicyandthedecisiontodonate.flTheAmericanEconomicReview,102(5):2018Œ2047.Kessler,JuddB,andAlvinERoth.2014.fiLoopholesunderminedonation:Anexperimentmo-tivatedbyanorgandonationpriorityloopholeinIsrael.flJournalofPublicEconomics,114:19Œ28.Lacetera,Nicola,MarioMacis,andSarahSStith.2014.fiRemovingbarrierstoorganandbonemarrowdonation:TheeffectofleaveandtaxlegislationintheUS.flJournalofHealthEconomics,33:43Œ56.Leppke,Susan,TabithaLeighton,DavidZaun,Shu-ChengChen,MelissaSkeans,AjayKIsrani,JonJSnyder,andBertramLKasiske.2013.RegistryofTransplantRecip-ients:collecting,analyzing,andreportingdataontransplantationintheUnitedStates.flTrans-plantationReviews,27(2):50Œ56.Li,Danyang,ZackaryHawley,andKurtSchnier.2013.fiIncreasingOrganDonationviaChangesintheDefaultChoiceorAllocationRule.flJournalofHealthEconomics,32(6):1117Œ1129.Lindsay,CottonM,andBernardFeigenbaum.1984.fiRationingbywaitinglists.flTheAmericanEconomicReview,74(3):404Œ417.Massie,A.B.,S.E.Gentry,R.A.Montgomery,A.A.Bingaman,andD.L.Segev.2013.fiCenter-LevelUtilizationofKidneyPairedDonation.flAmericanJournalofTransplantation,13(5):1317Œ1322.146Meier-Kriesche,Herwig-Ulf,andBruceKaplan.2002.fiWaitingtimeondialysisasthestrongestriskfactorforrenaltransplantoutcomes:APairedDonorKidneyAnal-ysis.flTransplantation,74(10):1377Œ1381.Melcher,ML,DBLeeser,HAGritsch,JMilner,SKapur,SBusque,JPRoberts,SKatznel-son,WBry,HYang,etal.2012.fiChaintransplantation:initialexperienceofalargemulticen-terprogram.flAmericanJournalofTransplantation,12(9):2429Œ2436.Mierzejewska,Beata,MagdalenaDurlik,WojciechLisik,CaitlinBaum,PaulSchroder,JonathanKopke,MichaelRees,andStanislawStepkowski.2013.fiCurrentApproachesinNationalKidneyPairedDonationPrograms.flAnnTransplant,18:112Œ124.Nassir,BisharaA,CarlEDean,SuyingLi,NicholasSalkowski,CraigASolid,MarkASchnitzler,JonJSnyder,SJosephKim,BertramLKasiske,MarkLinzer,etal.2015.fiVariationinCostandQualityinKidneyTransplantation.flTransplantation,99(10):2150Œ2157.NationalCancerInstitute.2015(ReleasedJanuary2015,Accessed8/22/15).fiU.S.PopulationData1969-2013.flhttp://seer.cancer.gov/popdata/.Obialo,ChamberlainI,ElizabethOAlexanderQuarshie,andPhyllisCMartin.2005.fiUltralateReferralandPresentationforRenalReplacementTherapy:SocioeconomicImplica-tions.flAmericanJournalofKidneyDiseases,46(5):881Œ886.O'Brien,Miles,andAnnKellan.2012(accessedMay11,2015).fiNewSoftwareMatchesMoreKidneyDonations,Faster.flhttp://www.nsf.gov/news/special_reports/science_nation/kidneyexchange.jsp.Opelz,Gerhard.1997.fiImpactofHLAcompatibilityonsurvivalofkidneytransplantsfromunrelatedlivedonors.flTransplantation,64(10):1473Œ1475.OPTN.2009.fi2009OPTN/SRTRAnnualReport:TransplantData1999-2008.flhttp://www.ustransplant.org/annual_reports/current/.OPTN.2015.fiOrganProcurementandTransplantationNetworkPolicies.flhttps://optn.transplant.hrsa.gov/ContentDocuments/OPTN_Policies.pdf.OPTN.2016.fiOrganProcurementandTransplantationNetworkPolicies.flhttps://optn.transplant.hrsa.gov/governance/policies/.Patzer,RE,JPPerryman,JDSchrager,SPastan,SAmaral,JAGazmararian,MKlein,NKutner,andWMMcClellan.2012.fiTheRoleofRaceandPovertyonStepstoKidneyTrans-plantationintheSoutheasternUnitedStates.flAmericanJournalofTransplantation,12(2):358Œ368.Rees,MichaelA,JonathanEKopke,RonaldPPelletier,DorryLSegev,MatthewERut-ter,AlfredoJFabrega,JeffreyRogers,OlehGPankewycz,JanetHiller,AlvinERoth,etal.2009.fiANonsimultaneous,Extended,Altruistic-DonorChain.flNewEnglandJournalofMedicine,360(11):1096Œ1101.147Rodrigue,JR,DLCornell,JKLin,BruceKaplan,andRJHoward.2007.fiIncreasingLiveDonorKidneyTransplantation:ARandomizedControlledTrialofaHome-BasedEducationalIntervention.flAmericanJournalofTransplantation,7(2):394Œ401.Roth,A.E.,T.Sönmez,andM.U.Ünver.2004.fiKidneyExchange.flTheQuarterlyJournalofEconomics,119(2):457Œ488.Roth,A.E.,T.Sönmez,andM.U.Ünver.2007.fiEfkidneyexchange:Coincidenceofwantsinmarketswithcompatibility-basedpreferences.flAmericanEconomicReview,97(3):828Œ851.Roth,AlvinE,TayfunSönmez,andMUtkuÜnver.2005.fiPairwiseKidneyExchange.flJournalofEconomicTheory,125:151Œ188.Schnier,K,NTurgeon,RMerion,andDHoward.2014.fiSubsidizingAltruisminLivingOrganDonation.flWorkingPaper.Segev,DorryL,LaurenMKucirka,PoojaCOberai,RulanSParekh,LEbonyBoulware,NeilRPowe,andRobertAMontgomery.2009.fiAgeandComorbiditiesareEffectofGenderDisparitiesinRenalTransplantation.flJournaloftheAmericanSocietyofNephrology,20(3):621Œ628.Siminoff,LauraA,HeatherMMarshall,LeventDumenci,GordonBowen,ArunaSwami-nathan,andNahidaGordon.2009.fiCommunicatingeffectivelyaboutdonation:aneduca-tionalinterventiontoincreaseconsenttodonation.flProgressinTransplantation,19(1):35Œ43.Sönmez,Tayfun,andMUtkuÜnver.2015.fiEnhancingtheEfyofandEquityinTrans-plantOrganAllocationviaIncentivizedExchange.flSSRNWorkingPaper2551344.SRTR.2013.fi2012OPTN/SRTRAnnualReport.flhttp://srtr.transplant.hrsa.gov/annual_reports/2012/pdf/07_dod_13.pdf.Sweeney,Matthew.2010.fiDoesthePublicProvisionofKidneysCrowdOutLivingKidneyDo-nation?ARegressionDiscontinuityAnalysis.flCornellUniversity,mimeo.UnitedStatesDepartmentofHealthandHumanServices.2015(Accessed9/30/15).fiMedicareCoverageofKidneyDialysisandKidneyTransplantServices.flhttps://www.medicare.gov/Pubs/pdf/10128.pdf.UNOS.2014(Accessed8/15/15).fiQuestionsandAnswersforTransplantCandi-datesaboutMultipleListingandWaitingTimeTransfer.flhttps://www.unos.org/wp-content/uploads/unos/Multiple_Listing.pdf.Wallis,CBradley,KannanPSamy,AlvinERoth,andMichaelARees.2011.fiKidneypaireddonation.flNephrologyDialysisTransplantation,26(7):2091Œ2099.Waterman,AmyD,JohnDPeipert,ShelleySHyland,MelanieSMcCabe,EmilyASchenk,andJingxiaLiu.2013.PatientCharacteristicsandRacialDisparitiesinEvalua-tionCompletionandLivingDonorTransplant.flClinicalJournaloftheAmericanSocietyofNephrology,8(6):995Œ1002.148Wedd,JoelP,AnnMHarper,andScottWBiggins.2013.fiMELDscore,allocation,anddistri-butionintheUnitedStates.flClinicalLiverDisease,2(4):148Œ151.Wellington,AlisonJ,andEdwardASayre.2011.fiAnevaluationofincentivepoliciesfororgandonationsintheUnitedStates.flContemporaryEconomicPolicy,29(1):1Œ13.149