IMF-美国真的摆脱了二战债务吗？（英）-2024.1.docx

INTERNATIONA1.MONETARYFUNDDidtheU.S.ReallyGrowOutofItsWorldWarIlDebt?JulienAcalinand1.aurenceBallWP/24/5IMFWorkingPapersdescriberesearchinprogressbytheauthor(三)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(三)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.WoRK-NGPAPER2024JANWP/24/5©2024InternationalMonetaryFundIMFWorkingPaperResearchDepartmentDidtheU.S.ReallyGrowOutofItsWorldWarIlDebt?PreparedbyJulienAcalinand1.aurenceBall*AuthorizedfordistributionbyPrachiMishraJanuary2024IMFWorkingPapersdescriberesearchinprogressbytheauthor(三)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheathor(三)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.ABSTRACT:ThefallintheU.S.publicdebt/GDPratiofrom106%in1946to23%in1974isoftenattributedtohighratesofeconomicgrowth.Thispaperexaminestherolesofthreeotherfactors:primarybudgetsurpluses,surpriseinflation,andpeggedinterestratesbeforetheFed-TreasuryAccordof1951.Ourcentralresultisasimulationofthepaththatthedebt/GDPratiowouldhavefollowedwithprimarybudgetbalanceandwithoutthedistortionsinrealinterestratescausedbysurpriseinflationandthepre-Accordpeg.Inthiscounterfactal,debt/GDPdeclinesonlyto74%in1974,not23%asinactualhistory.Moreover,theratiostartsrisingagainin1980andin2022itis84%.Thesefindingsimplythat,overthelast76years,onlyasmallamountofdebtreductionhasbeenachievedthroughgrowthratesthatexceedundistortedinterestrates.RECOMMENDEDCITATION:Acalin,Julienand1.aurenceBall(2024):'DidtheU.S.ReallyGrowOutofitsWorldWarIlDebt?',IMFWorkingPaper.JE1.ClassificationNumbers:E31,E43,E65,H60,H63Keywords:U.S.PublicDebt;FinancialRepression;SurpriseInflation;r-gAuthor,sE-MailAddress:iacalinimf.orq:lballihu.eduWearegratefultoGeorgeHallforhelpwiththedata.WethankFrancesBianchi,RuiEsteves,OlivierJeanne,N.GregoryMankiw,RicardoReis,andparticipantsintheJHUMacro-Financeseminar,theCEPRInternationalMacro-Historyonlineseminar,andtheDebtCon6Princetonconferenceforusefulcomments.WealsothankKyungWoongKohforoutstandingresearchassistance.ContentsIntroduction1FactorsInfluencingtheDebt/GDPRatio3ConstructingCounterfactualPathsoftheDebt/GDPRatio6DataandMeasurement13Results19ComparisontoHallandSargent(2011)22Conclusions25References28Tables29Figures30Appendix341 IntroductionDoesahighlevelofnationaldebtimposeaburdenonfuturegenerationswhomustpayitoff?Inrecentyears,economistssuchasBlanchard(2019)andFurmanandSummers(2020)havesuggestedthattheanswermaybeno,becauser<g:therealinterestrateondebtisusuallybelowthegrowthrateoftheeconomy.Underthatcondition,thegovernmentcanrolloverthedebtandaccumulatinginterestwithoutraisingtaxes,andthedebt/GDPratiowillfallovertime.Becauseofthispossibility,agrowingnumberofeconomistsagreeWithBlanchardthatupublicdebtmayhavenofiscalcost.”ThisideahasdecreasedconcernaboutthehighcurrentlevelofU.S.debt.Thinkingonthisissuehasbeeninfluencedbyasalienthistoricalexperience:thedeclineintheU.S.debt/GDPratioafterWorldWarI1.Payingforthewarincreasedthisratiofrom42%infiscalyear1941to106%in1946,butthenitstartedtofallandreachedatroughof23%infiscalyear1974.AsElmendorfandMankiw(1999)report,wanimportantfactorbehindthedramaticdropbetween1945and1975isthatthegrowthrateofGDPexceededtheinterestrateongovernmentdebtformostofthatperiod/5Krugman(2012)saysthatthe“debtfromWorldWarIIwasneverrepaidandjustbecameincreasinglyirrelevantastheU.S.economygrew.”Thisinterpretationofhistorylendscredencetotheideathatahighlevelofdebtshouldnotcausegreatconcern.However,otherresearchershavesuggestedreasonstoquestionthisinterpretation.First,asdiscussedbyauthorssuchasHallandSargent(2011)andEichengreenandEsteves(2022),theU.S.actuallypaidoffpartoftheWorldWarIIdebtbyrunningprimarysurplusesbylevyingtaxesinexcessofcurrentgovernmentspendingovermuchoftheperiodwhenthedebt/GDPratiowasfalling.Second,asdiscussedbyauthorssuchasReinhartandSbrancia(2015),interestrateswerehelddownrelativetoeconomicgrowththroughpoliciesthatarenotlikelytobefeasibleand/ordesirableinthefuture.Thesepoliciesincludedepisodesoffinancialrepression,mostclearlytheFedpeggingofinterestratesatlowlevelsfrom1942to1951,whichwasaimedatdecreasingthecostofthewar.Inaddition,ex-postrealinterestrateswerereducedbyunexpectedrisesininflationintheaftermathofthewarandlaterinthe1960sand1970s.Becauseofthesefactors,thepostwarexperiencedoesnotnecessarilysuggestthattheU.S.economynaturallygrowsoutofdebt.Thispaperseekstoexplainthepathofthedebt/GDPratiosinceits1946peakof106%.Weestimatetheeffectsonthispathofthegovernment'sprimarysurpluses,theinterestratepegbefore1951,andsurpriseinflation.WethenderiveaCounterfactualpaththatthedebt/GDPratiowouldhavefollowedintheabsenceofthesefactors.ThisCounterfactualshowshowmuchtheratiowasreducedbygrowthratesinexcessofundistortedrealinterestratesratesthatarenotreducedbyeitherapegorsurpriseinflation.WefindthatthisCounterfactualscenariodiffersgreatlyfromactualhistory.WithoutprimarySUrPlUSeSandinterestratedistortions,thedebt/GDPratiofallsonlyfrom106%in1946to74%in1974,ratherthanfallingto23%asinreality.OverthethreedecadesafterWorldWarII,thenaturalerosionofdebtfromeconomicgrowthwasconsiderablysmallerthanisoftensuggested.WealsoextendourCounterfactualtothepresent,Withevenmorenegativefindingsaboutgrowingoutofdebt.TheCounterfactualdebt/GDPratiostartsrisingagainin1980,andin2022itis84%:theearlierdeclineintheratioispartiallyreversed,leavingitonly22percentagepointsbelowits1946level.Theriseintheratioreflectsthefactthattheeconomy,sgrowthratehasaveragedlessthantheundistortedrealinterestrateondebtsince1980.OurmethodforconstructingCounterfactualdebtpathsbuildsonpreviouswork,butusesarichersetofinformationtomakeourquantitativeresultsasaccurateaspossible.Akeystepistomeasurethefractionsofoutstandingdebtinagivenyearthatwereissuedineachearlieryearthewreversematuritystructure,ofthedebtwhichwedousinggranulardataonTreasurysecuritiesproducedbyHall,Payne,andSargent(2018)beforei960andbyCRSPthereafter.Wealsoconstructatermstructureofinflationexpectationsfromsurveysofshort-andlong-termexpectations,whichallowsustoestimatetheeffectsofsurpriseinflationontherealreturnsondebtissuedindifferentyears.Finally,weestimatetheeffectsofthepre-1952interestratepegbycomparingthepeggedratesondebtofvariousmaturitiestomarketratesduringthepost-pegperiodof1952-1960.Ourvariouscalculationsrequiresomeassumptionsaboutunobservedvariables,butourresultsarenotgreatlychangedbyvaryingtheseassumptionsinreasonableways.Thenextsectionofthispaperprovidessomehistoricalbackgroundonthepost-WorldWarIIperiod.WethenpresentourmethodologyforconstructingCounterfactualpathsofthedebt/GDPratio,describeourdatasources,andpresentourresults.WealsocompareouranalysistoHallandSargenfs(2011)well-knownworkonthedebt/GDPratiosinceWorldWarI1.2 FactorsInfluencingtheDebt/GDPRatioFigure1showsthepathofthepublicdebt/GDPratiointheUnitedStatesfromfiscalyear1941tofiscalyear2022.WeseethatthisratiogrewrapidlyduringWorldWarII,risingfrom42%in1941to106%in1946.Itthenfellsteadilyuntilitreached23%in1974,anexperiencecommonlyattributedtoeconomicgrowth.Sincethetroughin1974,thedebt/GDPratiohasriseninmostyearsanditreached102%in2022.Thissectionprovidessomebackgroundonthreefactors,inadditiontoeconomicgrowth,thathaveinfluencedthedebt/GDPratio:primarysurplusesanddeficits,theFed,sinterestratepegfrom1942to1951,andunexpectedinflation.1.atersectionsquantifytheeffectsofthesefactors.2.1 PrimarySurplusesandDeficitsFigure2showstheprimarybudgetsurplusasafractionofGDPfromfiscalyear1947,theyearafterdebt/GDPpeaked,tothepresent.Wecanseethatdeviationsfromprimarybalancecontributedtoboththefallinthedebt/GDPratiothrough1974andtherisesincethen.ThesharpfallingovernmentspendingafterthewarproducedPrimarysurplusesof3.6%ofGDPin1947and6.3%in1948.Afterthat,thesurplusremainedpositiveinmostyears,rangingbetween3.4%and-1.5%ofGDPthrough1974.ThisexperiencereflectedastrongpoliticalconsensusinfavorOfbudgetaryrestraintinthe1950sand1960s.Duringtheentireperiodfrom1947through1974,theprimarysurplusaveraged1.1%ofGDP,helpingtoreducethedebt/GDPratio.After1974,thepatternreversedandhabitualprimarydeficitscontributedtoarisingdebt/GDPratio.Theprimarybalancewasnegativeatalmostalltimesexceptinthelate1990s,andthedeficitwasespeciallyhigharoundtheGreatRecessionof2008andtherecentCOVIDpandemic.2.2 InterestRatesBeforetheFed-TreasuryAccordInApril1942,attherequestoftheTreasurydepartment,theFederalReserveadoptedapolicyofpegginginterestratesongovernmentbondsatlowlevels.Thispolicywasintendedtocontainthecostoffinancingthewar.TheFedcappedyieldsatlevelsrangingfrom0.375%forTreasurybillsto2.5%for30-yearbonds,maintainingthesecapsbystandingreadytobuyanyquantityofbonds.TheFed,spolicypresumablykeptinterestratesbelowtheneutrallevelandmadeitimpossibletoadjustratestocontrolinflation.DuringWorldWarII,inflationwascontainedthroughgovernmentpricentrols.PricecontrolswereeliminatedinJune1946,andinflationbecameunstable:infiscalyears1947through1951,CPIinflationratesrangedfrom-0.7percentto18.3percent,withanaverageof7.1percent.FedofficialsbecameincreasinglyunhappywiththeirinabilitytocontrolinflationandeventuallypersuadedtheTreasurythatthepegshouldbeabandoned,adecisionannouncedbythetwoagenciesintheirMarch1951“Accord.”】Theinterestratepeghadabigeffectondebtdynamicsbecauseitwasineffectduringthebuild-upandinitialrollingoverofthelargeWorldWarIIdebt.Inaddition,alargeshareofthedebtissuedwithlowinterestrateshadmaturitiesoftentothirtyyears,sotheinfluenceonthecostsofdebtservicewasfeltlongafterthepegendedin1951.Bothduringandafterthepeg,periodsofhighinflationproducedex-postrealinterestratesthatweredeeplynegativeformanygovernmentsecurities.Thepre-Accordpegisthemostclear-cutcaseoffinancialrepressionthatheldinterestratesdowninthepostwarU.S.ReinhartandSbrancia(2015)citeothertypesoffinancialrepression,suchascapsonbanks5interestratesunderRegulationQ.WeignoretheseotherpoliciesbecauseWedonotknowhowtoquantifytheireffectsontheinterestratesongovernmentdebt.Totheextenttheyareimportant,theeffectsofthepre-Accordpegthatwemeasurearealowerboundonthetotaleffectsoffinancialrepression.2.3 SurpriseInflationUnexpectedinflationreducesthedebt/GDPratiobypushingex-postrealinterestratesbelowex-anterealrates.Therelevantin11ationrateisthegrowthrateoftheGDPdeflator.Asdetailedbelow,theeffectonthedebt/GDPratioinagivenperioddependsonthecurrentlevelofinflationrelativetothelevelexpectedatvarioustimesinthepastwhenthecurrentlyoutstandingdebtwasissued.Figure3showsinflationsurprisesforfiscalyears1952through2022.Foreachyear,'SeeHetzeland1.each(2001)foradetailednarrativeofthisepisode.theFigurecomparestheactuallevelofinflationtothelevelexpectedoneyearearlierand(startingin1962)thelevelexpectedtenyearsearlier,basedonseveralsourcesofdataoninflationexpectations(seedetailsbelow).Weseethewell-knownriseininflationinthe1960sand1970sandthefactthatinflationpersistentlyexceededexpectedinflationduringthatperiodespeciallythelevelsexpectedtenyearsbefore,implyingabigerosionofrealinterestratesonlong-termsecurities.Theseinflationsurprisescontributedtothedecreaseinthedebt/GDPratiothrough1974andmoderatedthefirstpartofthesubsequentincreaseinthelate1970s.Startinginthe1980s,thestorywasreversed:asinflationfellfollowingtheVolckerregimeshift,actualinflationwasusuallylowerthanexpectedinflation.Thispatternpushedexpostrealinterestratesaboveex-anteratesandcontributedtotherisingdebt/GDPratio,althoughwewillseethatthiseffectwassmallerthanthatoftheearliersurpriseinflation,inpartbecausetheaveragematurityofthedebtwasshorter.Thesharpincreaseininflationin2021-2022meansthatunexpectedinflationhasagainbecomeafactorreducingthedebt/GDPratio.3 ConstructingCounterfactualPathsoftheDebt/GDPRatioThissectiondescribeshowweconstructCounterfactualpathsofthedebt/GDPratiowithzeroprimarysurplusesand/orwithoutdistortionsinrealinterestratesfromthepre-Accordpegandsurpriseinflation.Thedifferencesbetweenthesepathscapturethedifferentfactorsdrivingtheactualdebt/GDPpath.TheCounterfactualwithoutsurplusesorinterestratedistortionsshowshowmuchthedebt/GDPratiohasbeenreducedthrougheconomicgrowthinexcessofundistortedinterestrates.Weinterpretthiseffectasthenaturaltendencyoftheeconomytogrowoutofdebt,whichissometimescalledthe“negativesnowbalF,or“melting”effect(e.g.,Krugman2019；AizenmanandIto2020).3.1 OverviewWestartwithastandardequationdescribingtheevolutionofthenominalstockofdebt:Dt=(1+zt)Dt-Pt,(1)whereDtistheparvalueofthedebtinperiodtiitistheaverageinterestrateonthedebt,andPtistheprimarysurplus.Weconsidertheparvalueofthedebtbecauseitisafocusofpolicydiscussions.HallandSargent(2011)analyzeaversionofequation(1)thatappliestothemarketvalueofdebt.Takingthatapproachwouldcomplicateouranalysiswithoutmakingamaterialdifferenceforourresults,becausethepathsoftheparandmarketvaluesofthedebtareveryclosetoeachother(seeFigure2intheWebAppendixtoHallandSargent2011).Fromperiodr-1to/,thedebtgrowsatrateitasitisrolledover,andprimarysurplusesreducethedebt(whileprimarydeficitsincreaseit).Inourempiricalwork,aperiodisafiscalyear,anddebtismeasuredattheendoftheyear.TheinterestrateismeasuredastotalinterestpaymentsinfiscalyeartdividedbyDt-1.Inallsimulations,westartwiththedebtatitsactuallevelattheendoffiscalyear1946andderiveitspathafterthatforalternativeassumptionsaboutprimarysurplusesandinterestrates.Wethenderivethepathofthedebt/GDPratio,assuminginallcasesthatthepathofnominalGDPisthesameasinactualhistory.ItisstraightforwardtoconstructaCounterfactualwithnoprimarysurpluses:wesimplysetA=oforalltinequation(1).Adjustingforthepre-Accordpegandinflationsurprisesismorecomplex.Inourcoun-terfactualexercise,wereplacetheactualnominalinterestratesitwiththeratesthatwouldhaveprevailedundertwoconditions.Thefirst,whichisrelevantfordebtissuedbefore1952,isthattheex-anterealinterestrateistheonethatwouldhaveprevailedintheabsenceofthepeg,iftheFederalReservewereoperatingnormally(presumablysettingratestostabilizeoutputandinflation).Weestimatetheseundistortedinterestratesfromex-anterealratesobservedafterthepegperiod,asdetailedbelow.Thesecondconditionisthattheex-postrealinterestrateisalwaysequaltotheex-anterealrate.Weimplementthisassumptionbyaddingtheunexpectedcomponentofinflation(measuredusingsurveysofexpectations)toactualnominalinterestrates.Thisadjustmentyieldsthenominalratesthatwouldhaveprevailedgivenex-anterealratesiffinancialmarketparticipantshadknownthefuturepathofinflation.Wecanalsointe>retourCounterfactualinterestratesasthosethatwouldhavebeenobservedifalldebtwereindexedtoinflation.AndreolliandRey(2023)conductasimilarCounterfactualexerciseforEuroAreacountriesfortheperiodsince1999.Oursimulationsofdebt/GDPratiosassumethatbothundistortedrealinterestratesandrealGDParethesameinourCounterfactualscenariosasinactualhistory.ConventionalmacroeconomicsimpliesthatthehigherdebtlevelsintheCounterfactualswouldincreaserealratesandreducerealGDPbycrowdingoutcapital,andbothoftheseeffectswouldmagnifytheincreasesinthedebt/GDPratiorelat