《IMF-自动化与福利：任务与教育的作用（英）-2024.1.docx》由会员分享，可在线阅读，更多相关《IMF-自动化与福利：任务与教育的作用（英）-2024.1.docx（38页珍藏版）》请在三一办公上搜索。
1、INTERNATIONA1.MONETARYFUNDAutomationandWelfare:TheRoleofBequestsandEducationPreparedbyManukGhazanchyan,AlexeiGoumilevski,andAlexMourmourasWP/24/11IMFWorkingPapersdescriberesearchinprogressbytheauthor(三)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoft
2、heauthor(三)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.WoRK-NGPAPER2024JAN2023InternationalMonetaryFundWP/24/11AutomationandWelfare:TheRoleofBequestsandEducation*PreparedbyManukGhazanchyan5AlexeiGoumilevski,andAlexMourmourasAuthorizedfordistributionbyAlexMourmouras
3、January2024IMFWorkingPapersdescriberesearchinprogressbytheauthor(三)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(三)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.ABSTRACT:Thispaperexaminesthewelfareeffectsof
4、automationinneoclassicalgrowthmodelswithandwithoutIntergenerationaltransfers.Inastandardoverlappinggenerationsmodelwithoutsuchtransfers,improvementsinautomationtechnologiesthatwouldlowerwelfarecanbemitigatedbyshiftsinlaborsupplyrelatedtodemographicsorpandemics.WithperfectIntergenerationaltransfersba
5、sedonaltruism,automationcouldraisethewell-beingofallgenerations.Withimperfectaltruism,fiscaltransfers(universalbasicincome)andpublicpoliciestoexpandaccesstoeducationopportunitiescanalleviatemuchofthenegativeeffectofautomation.JE1.ClassificationNumbers:E13,E62,D641,1Automation;Aging;Altruism;FiscalPo
6、licy;Education;Keywords:,.OverlappingGenerationsmahazanchyanimf.orq：aqoumilevskiimf.orq;AuthorsE-MailAddresses:amourmourasimf.orqaManukGhazanchyanisanEconomistintheWesternHemisphereDepartment,AlexeiGoumilevskiisaSeniorScientificComputingEngineerintheInformationandTechnologyDepartment,andAlexMourmour
7、asisDivisionChiefintheAsiaPacificDepartment.Presentedatthe28thInternationalConferenceonComputinginEconomicsandFinanceinDallas,Texas,June17-19,h,2022.Wethankconferenceparticipantsfortheirusefulcomments.WearealsogratefultoPeterRangazas1MarinaMendesTavares,andJeremyCliftfortheinsightfulcommentsthatshap
8、edourpaper.ContentsI. Introduction1II. Relevant1.iterature3III. AutomationandWelfareinOverlappingGenerationModels5B. IntroducingaOne-Timetax9C. K&SModelwithaBequest11D. K&SModelwithBequestandEducation141. PrivateEducation152. PublicEducation17IV.ArtificialIntelligence22A.K&SModel22IV.Conclusions26Ap
9、pendices27Appendix A. K&SModelwithBequest27Appendix B. ReplicatingKotlikoffandSachs(2012)31Appendix C. RobustnessChecks31A.References3410000Decliningtrendsinworking-agepopulation(inmillions)900080007000202220242026202820302032203420362038204020422044204620482050Workingagepopulation(China)Workingagep
10、opulation(world)I. IntroductionAutomationhasacceleratedinrecentdecades,drivenbyongoingimprovementsincomputingandinformationtechnologiesandassociatedcostreductions.Machinesinawideningrangeofindustriesperformincreasinglycomplextasks,poweredbysophisticated,networkedsoftware.Theaccelerationinautomationa
11、nditseconomy-widediffusioninblue-andwhite-collaroccupationsalikeiscreatingnewemploymentcategoriesbutisalsocontributingtowideninginequalityandfuelingdemandforgovernmentpoliciestoreverselong-termincomelossesoflabor.Thislong-standingpromiseandconcernsarevividlyillustratedbythelatestbreakthroughinArtifi
12、cialIntelligenceinvolvinggenerative,pretrainedtransformers.1.ookingahead,whilethepaceofautomationislikelytocontinue,itseffectsmaybemitigatedbyoffsettingforces.Populationsareagingalmosteverywhere.Intheadvancedeconomies,theworkingagepopulationhasstartedshrinkingforthefirsttimesinceWorldWarIl(Spence,20
13、22).Globally,thepopulationofworkingageisexpectedtocontinuetogrowuntilabout2040,buttherat100ftheworkingagepopulationtothetotalisalreadydecliningglobally(Chart).InthecaseofChina,forexample,theworking-agepopulationisexpectedtoshrinkbyafifthoverthenext30years.AsGoodhartandPradhanstress,ourageisoneofdemo
14、graphicreversalinwhichtheuIongglutofinexpensivelaborthathadkeptpricesandwagesdownfordecades,isgivingwaytoaneraofworkershortages,andhencehigherprices.0.70.60.50.40.30.20.10Workingagepopulation/Totalpopulation(World)-rightWorkingagePOPUlatiOn/Totalpopulation(China)-rightSource:UnitedNations.WorldBank,
15、andstaffcalculationsRecurringglobalpandemicsalsoadverselyaffectlaborsupply,eitherbydepressinggrowthinthelaborforcedirectly(AIDSpandemic),orindirectlybyreducingtheparticipationofolderworkersandothersincontact-intensiveoccupations(pandemicrelatedtoCovid-19).Intheabsenceofmasssouth-northmigration,robot
16、smayturnouttobeessentialinmeetingmoreoftheneedsoftheelderlyandreversedeclinesinaggregateoutputandwelfarei.ThispaperexaminesthecombinedwelfareeffectsofautomationandlowerlaborsupplyusingneoclassicalgrowthmodelswithandwithoutIntergenerationaltransfers.Itbeginsbyreplicatingaversionofthewell-knownresulto
17、fKotlikoffandSachs(K&S,2012)thataone-timeimprovementinthetechnicalefficiencyofmachinesendsupimmiseratingallfuturegenerations.Thisstrikingresultreliesontwocrucialassumptions.First,machinesareverygoodsubstitutesforunskilledlaborthroughouttheeconomy,sothatimprovedautomationendsupdisplacingworkersandlow
18、eringwagesacrosstheboard.Second,therearenooperativeintergenerationaltransfersofanykind,sothattheownersofcapitalendupconsumingtheentirewindfallfromtheimprovementsinautomationduringtheirlifetime(ageneration,roughlythirtyyears).Thepositiveshocktotheefficiencyofmachinesdoesnotraisesaving,depressesinvest
19、mentinphysicalandhumancapital,andsetsinmotionanever-endingcycleofdecliningwelfare.Governmentpolicyisthereforeneededtospreadthiswindfallmoreequitablyacrossfuturegenerations.K&Sconsiderwealthtaxes,inparticularsocializingaportionoftheeconomyscapitalstockthatallowsthegovernmenttofinanceasustainableincom
20、estream(universalbasicincome)forallfuturegenerations.Resortingtocompulsionisessentialwhengenerationsareselfish,precludinganysortofvoluntaryintergenerationaltransfers.Infact,privateintergenerationaltransfersaresubstantial,withabouthalfofallhouseholdsplanningtoleaveestates(1.aitnerandJuster12017).Thef
21、irstobjectiveofthepaperistoreassessthewelfareeffectsofautomationinthepresenceofintergenerationaltransfers,bothbequestsandprivatelyandpubliclyfundedschooling.InaversionoftheK&Smodelwithoperativebequests,wefindthatintergenerationaltransfersarepositiveinequilibriumifthestrengthofaltruismexceedsacertain
22、threshold,mitigatingthenegativeeffectsofautomation.Butwhileitiscomfortabletoknowthatthegainsfromautomationmaybepassedtofuturegenerationswithouttheneedtonationalizecapital,thismodelofperfectaltruismisalsoextreme:manyfamiliesineachgenerationcannotmakeefficienttransferstotheirchildren.Whatisneededisamo
23、rebalancedmodel,onethatfeaturesheterogeneitybothwithinandacrossgenerations,withsomehouseholdsmakingefficientbequestsandothersstuckinacomersolution.WeassesswhetherautomationisimmiseratinginaversionofKSmodelthatincorporatespurealtruism,andinGlommandRavikumar,smodel(G&R,1992)inwhichparentsmakeinvestmen
24、tsintheschoolingoftheirchildren.Westudyhowfiscalandeducationalpoliciescanbestraisewelfare,byalleviatingfinancingconstraintsinthefinancingofhumancapitalinvestmentsandrestoringequalityofopportunity.SimilarresultsobtaininaversionofthemodelusedbyIvanyna,MourmourasandRangazas(2018)inwithtwotypesofhouseho
25、lds(thepoor,whoarebequest-constrained)andtherich(whoareunconstrained).Thepaperthenturnstoananalysisofacombinedshockinvolvingajumpinautomationandasimultaneousreductioninlaborsupplydrivenbydemographics.Asexpected,strengtheningaltruisticbondsraisebequestsandhumancapitalinvestmentsoftheyoung,providingan
26、additionalstimulustoeconomicgrowth.Inaddition,governmenttransfersoftaxrevenueleviedontherichcanimprovetheBusinessleadershavealsomadeaconnectionbetweenautomationandagingrecently.OneexamplewastherecentarticleinFortunebyIBMCEO,An/indKrishnaherehepointstodecliningpopulationstocalmfearsaboutA.I.takingjob
27、s.HefurtheraddedthatultimatelythereisgoingtobejobcreationwithA.l.,asjobswillalsobeaddedinareaswithmorevaluecreation.welfareofthepoorandreduceinequalitywithinandacrossgenerationswhenaltruisticlinksbetweengenerationsareweak.II. Relevant1.iteratureTheliteratureonautomationanditseconomicimpactisevolving
28、,withsomeearlierstudiesfromGordon(2012),Cowen(2011),AcemogluandRestrepo(2017,2018),SachsandKotlikoff(2012,2015),Ford,(2015);Freeman,(2015)amongstpessimists,andBrynjolfssonandMcAfee(2014),Autor(2014,2015)amongtheoptimists.Thekeyissueiswhetherautomationreplaceslaborshareandemploymentthroughreplacement
29、ofroutinetasksofever-increasingscopeandcomplexityorwhether,onnet,itincreaseslaborparticipationbycreatinghigh-payingjobsinemergingnewoccupations.SomegloomyscenariosforlaborresultingfromartificialintelligenceandsimultaneousautomationbreakthroughsaredescribedinBostrom(2014).Graetzet.al.(2018)examinedth
30、eeconomiccontributionofmodernindustrialrobotsin17countriesfortheperiod1993-2007.Contrarytothepessimisticview,theseauthorsfoundthattheincreasinguseofrobotsraisedtheannualgrowthofGDPandlaborproductivityby0.37and0.36percentagepoints,respectively.Authorsconcludethatthoserobotsdidnotsignificantlyreduceto
31、talemployment,althoughtheydidreduceIow-Sk川edworkers*employmentshare.Gaaitzenetal.(2020)studiedtheeffectsofadaptationofindustrialrobotsandoccupationalshiftsbytaskcontentinthethirty-sevencountriesfortheperiodfrom2005to2015.Theauthorsfoundthatincreaseduseofrobotsisassociatedwithpositivechangesintheempl
32、oymentshareofnon-routineanalyticjobsandnegativechangesintheshareofroutinemanualjobs.Ofcourse,enhancingpolicyincludingR&Dandtheregulatoryplatformsinbothprivateandpublicsectorstosupportdigitaltechnologiesiskeytoimproveproductivity.Whilethe2020-22pandemichelpedtoacceleratethedigitaltransformation,manys
33、ectors-includingthepublicsector-arelagging,andhenceconcernsabouttheeffectsofautomationonemploymentwillpersist(Spence,2022).OnlyafewstudiesexaminedtheeffectofautomationandpopulationagingonthelabormarketasidefromtheclassicalworkbyFreyandOsborne(2017)focusingontheprobabilityofautomationaffectingvarious
34、jobsandoccupations.OneoftheearlieststudiesonautomationandpopulationagingwasbyAcemogluandRestrepo(2017),wheretheauthorsexaminedtherelationshipbetweeneconomicgrowth,populationaging,andautomationatthecountrylevel.Phiromswadetal(2022)isamongstthemostrecentstudiestofocusonthoseeffectsbutalsoontheinteract
35、ioneffectsofautomationandpopulationagingonthelabormarket.ConsistentwithpreviousliteratureincludingwithGraetzandMichaels(2018)theauthorsfoundstrongevidencethatautomationnegativelyaffectsemploymentgrowthholdingotherfactorsconstant.Theyalsofoundstrongevidencethatthedisaggregatedmeasuresofage-relatedabi
36、litiesaffectemploymentgrowthbutnottheaggregatemeasure.Asexpectedandconsistentwithfindingsthatautomationisstillevolvinginaffectinghighvaluejobs,theauthorsfindthatwithoccupationswithlowscoreonboththeage-appreciatedcognitiveabilityaswellage-depreciatedphysicalability(suchasproductionoccupationsandfoodp
37、reparationandservingrelatedoccupations),thenegativeeffectofautomationonemploymenttendstobestrongest.However,foroccupationswithahighscoreinbothage-appreciatedcognitiveabilityaswellasage-depreciatedphysicalability(suchasprotectiveserviceoccupationsandhealthcarepractitionersandtechnicaloccupations),the
38、negativeeffectofautomationonemploymenttendstobeweakest.Aghion-Jones-Jones(2017)studytheimplicationsofartificialintelligenceforeconomicgrowthinlightofreconcilingevolvingautomationwiththeobservedstabilityinthecapitalshareandpercapitaGDPgrowthoverthepastcentury.Theauthorscreatesufficientconditionstogen
39、erateoverallbalancedgrowthwithaconstantcapitalsharethatstayswellbelow100percent,evenwithnearlycompleteautomation.Inotherwords,whileBaumoscostdiseaseleadstoadeclineintheshareofGDPassociatedwithmanufacturingoragriculture(oncetheyareautomated),thisisbalancedbytheincreasingfractionoftheeconomythatisauto
40、matedovertime.TheauthorsalsostudytheeffectsofintroducingA.l.intheproductiontechnologyfornewideasandthepossibilitythatA.l.couldgeneratesomeformofasingularity,wheretheauthorsneverthelessclaimthattheBaumolthemeherealsoremainsrelevant:evenifmanytasksareautomated,growthmayremainlimitedduetoareasthatremai
41、nessentialyetarehardtoimprove.Pizzinelliandothers(forthcoming)examinetheimpactofArtificialIntelligence(AI)onlabormarketsinbothAdvancedEconomies(AEs)andEmergingMarkets(EMs).TheauthorsproposeanextensiontoastandardmeasureofAlexposure,accountingforAspotentialaseitheracomplementorasubstituteforlabor,wher
42、ecomplementarityreflectslowerrisksofjobdisplacement.ThentheyanalyzeworkerlevelmicrodatafromtwoAEs(USandUK)andfourEMs(Brazil,Colombia,India,andSouthAfrica),revealingsubstantialvariationsinunadjustedAlexposureacrosscountries.TheauthorsfoundthatwhileAlposesrisksoflabordisplacementduetotaskautomation,it
43、alsoholdspromiseinitscapacitytoenhanceproductivityandcomplementhumanlabor,especiallyinoccupationsthatrequireahighlevelofcognitiveengagementandadvancedskills.TheauthorsalsofindthatAEsmayexpectamorepolarizedimpactofAlonthelabormarketandarethuspoisedtofacegreaterriskoflaborsubstitutionbutalsogreaterben
44、efitsforproductivity.Theextentandformofvoluntaryintergenerationaltransfersisdictatedbythestrengthofintergenerationalaltruismandisanimportantconsiderationinmacroeconomicsthatisrelevantforourpaper.Kotlikoff(2001)providesanexcellentsurveyofkeyworksontheroleofintergenerationalaltruism,includingempirical
45、findingsforexample,theresultsofAltonjiandHayashi(1994)whichareconsistentwiththepurealtruismtheory.Acloselyrelatedareaofresearchconcernstheformofhumancapitalinvestments,specificallytherationalebehindeducationorotherbequestsinkind.RazinandRosenthal(1990)showthatfamilytaxationasaresponsetoinformationas
46、ymmetrybetweenaparentandachildcouldreducetheneedforgovernmentinterventionandtaxation.HoodandJoyce(2017)provideanexcellentupdatetotheempiricalrelevanceofaltruism.OurpaperismostcloselyrelatedtoMichel,Thibault,andVidal(2004),whostudytheeffectsofaltruismandfiscalpoliciesongrowthinanoverlappinggeneration
47、smodelinthetraditionofDiamond,andtoGlommandRavikumar(1992)whostudybequestsintheformofhumancapitalinvestmentsinchildren.Westudyprivatelyfundedschoolingforfamilieswithoperativebequestsandpubliclyfundededucation.Wefindthatgovernmentspendingoneducationpromoteeconomicgrowth.Theseconclusionsaresupportedby
48、avastvolumeofresearchthatlinkindividualseducationattainmenttoeconomy-wideprosperity.FabrizioCarmignani(2016)studiedeffectsofgovernmentexpendituresoneducationtoeconomy.AuthorusedtheWorldBank,sWorldDevelopmentIndicatordatabasedataon151countriesfor2000-2010years.Heconcludedthat*4increaseineducationexpe
49、nditureby1pointofGDPincreasesGDPgrowthby0.9percentagepoints.Gheraia,Zouheyretal.(2021)investigatedrelationshipbetweenthecostofeducationandeconomicgrowthintheKingdomofSaudiArabiafortheperiod1990-2017.Authorsfoundthatinthelongruntheriseineducationalexpenditureby1%wouldleadtoanincreaseineconomicgrowthby0.89%.SimilarresultswereobtainedbyYahya,Mohdetal.(2012).Authorsanalyzedthelong-runrelationship
