PublicPublicDisclosureAuthoPolicyResearchWorkingPaper10735WORLDBANKGROUPPolicyResearchWorkingPaper10735InelasticDemandMeetsOptimalSupplyofRiskySovereignBonds*Mat′?asMorettiLorenzoPandolfiSergioL.SchmuklerGerm′anVillegasBauerWorldBankInternationalMonetaryFundTomsWilliamsKeywords:emergingmarketsbondindex,inelasticfinancialmarkets,institutionalinvestors,internationalcapitalmarkets,smallopeneconomies,JELCodes:F34,F41,G11,G15*WethankWalkerRayandtheparticipantsatapresentationheldattheASSAAnnualMeetingsforhelpfulfeedback.WearegratefultoPatricioYunisforresearchassistanceandAntonnParkforcopyediting.TheWorldBankChileResearchandDevelopmentCenterandKnowledgeforChangeProgram(KCP)andGeorgeWashingtonUniversityFacilitatingFund(GWUUFF)providedfinancialsupportforthispaper.LorenzoPandolfigratefullyacknowledgesfinancialsupportfromtheUnicreditFoundation.Thefindings,interpretations,andconclusionsexpressedinthispaperareentirelythoseoftheauthors.TheydonotnecessarilyrepresenttheviewsoftheIMFortheWorldBank,thoseoftheirExecutiveDirectors,orthegovernmentstheyrepresent.Moretti:matias.moretti@.Pandolfi:lorenzo.pandolfi@unina.it.Schmukler:sschmukler@.VillegasBauer:gvillegasbauer@.Williams:tomaswilliams@.11IntroductionGovernmentsinemergingeconomcapitalmarketsfortheiroverallfinancing.Thebehaviorofinvestorsinthesemarketsisthuscrucialtounderstandinggovernments’borrowingcosts,defaultrisk,andoptimaldebtelastic,implyingthatinvestorsarewillingtolendanyamountgovernmentsrequestattherisk-freerateplusadefaultriskpremium.Thisassumptiononinvestorbehaviorcontrastswithabodyofrecentworkforotherassetmarketsthatallowsforaricherinvestordemandstructure,typicallyinvolvinganinelasticordownward-slopingdemand(KoijenandYogo,etal.,2023).Inthispaper,wepresentnovelevidenceofdownward-slopingdemandcurvesinriskysovereignbondmarketsandanalyzetheirimpactongovernments’optimaldebtpolicies.Inthecontextofriskysovereignbonds,estimatingademandelasticityischallengingfortwomainreasons.Ideally,onewouldliketoidentifyshockstotheavailablebondsupplythatareunrelatedtocountryfundamentals,estimatepricereactionsaroundthem,andmapbyadjustingfutureissuancesortheirdefaultlikelihood.Thus,theestimatedpricereactionothercountriesintheindex.Wefindthatbondpricessignificantlyreacttotheseshocksinthehighfrequency,evenwhentheyareorthogonaltocountryfundamentals.Ourestimatesimplyaninversepricedemandelasticityof?0.33,whichwerefertoasareduced-formWeidentifythestructuralelasticitybyindirectinference.Weformulateaquantitativefunctionofobservables,andwedisciplineitbasedonourre2thestructuralelasticity.Ourfindingsshowthatoverone-thirdofourreduced-formelasticitycalibratedmodeltoanalyzetheimplicationsoffacinganinelasticbonddemand.Wefindthatdownward-slopingdemandcurvesactasacommitmentdevicethatlimitsgovernments’Westartouranalysiswithasimpleframeworktoguideouridentificationstrategy.Thispassivedemandastheportionofinvestors’holdingsaimedatreplicatingthecompositionof “effectivesupply,”namelythequantityavailablewhetherdemandcurvesforactiveinvestorsslopedownward(PandolfiandWilliams,2019;PavlovaandSikorskaya,2022).Nevertheless,ifexpectedpayoffsendogenouslyrespondtochangesintheeffectivesupply,anyobservedpricevariationresultingfromtheshiftcouldoverorunderestimatethedemandelasticity.Ontheempiricalfront,weidentifyexogenousshiftsinacountry’seffectivesupplyofsovereignbondsbyusingmonthlyrebIndexGlobalDiversified(EMBIGD),themostwidelytrackedindexbyinstitutionalinvestorsforU.S.dollar-denominatethecompositionofthisindexaffecttheeffectivebondsupplybecausetheyleadtosimilarrebalancingsintheportfoliosofpassiveinvestors.Duetopotentialtrackingerrorcosts,passiveinvestorstendnottodeviatefromtheindex.GiventheEMBIGD’spopularitWederiveameasureofflowsimpliedbyrebalancings(FIR)bycombiningtheassetspassivelytrackingtheEMBIGDwiththeindex’smonthlyrebalancings.QualifyingnewbondThesefrequentadjustmentsleadtochangesincountryweightpassivefundsflows.Toavoidendogeneityissues,weconstructaninstrumentthatexploitschangesintheFIRgeneratedbytheissuanceorretirementofbondsfromothercountriesintheindex.Assuch,thesechangesareorthogonaltoacountry’sownfundamentals.In3OuranalysisrevealsthatahigherFIRleadstohigherbondprices.Onaverage,a1prices.Theseestimatesimplyareduced-forminversedemandelasticityof?0.33.Wefindforcountrieswithhigherdefastatisticallynotsignificant.Onthequantitativefront,weformulateasovereigndebtmodelwherethegovernmenthaslimitedcommitmentandcanendogenouslydefaultonitsdebtobligations.Standardmodelsofthisnaturetypicallyassumeaperfectlyelasticdemandforsovereignbonds,withchangesinbondpricesdrivensolelybyvariationsindefaultrisk(Arellano,affectsthebondpricethroughtwointerconnectedmechanisms.First,becausetheactivedemandisdownwardsloping,theimpliedreductionintheefthepriceinvestorsarewillingtopayforit.Wedisciplinethemodelusingourempiricalestimates.Introducingpassiveandactiveestimatedreduced-formelasticity.Wethenusethemodeltobackoutastructuralelasticaccountfornearlyathirdofthereduced-formelasticity.Moreover,wefindthatthelargershockinstrumentisinherentlymoretemporarythanotherinstrumentsusedintheliterature,suchasindexadditionsordeletionsorindexmethodologicalrecompositions.Still,wefindthatthebiascanrepresentaboutone-thirdofthetotalpriceresponse.Ourmodelallowsustoexaminetheimpactofadownward-slopingdemandontheobservelowerdefaultriskandhigherbondpricescomparedtoascenariowithaperfectlyforholdingthedebt(whichcouldleadtoaconvenienceyield)butratherbytheinelasticdemandservingasacommitmentdeviceforthegovernment.Themechanismbehinditisasfollows:Withadownward-slopingdemand,issuinganadditionalunitofdebtdecreasesbondpricesevenifthedefaultriskremainsfixed.Asaresult,thegovernmentfindsissuinglargeamountsofdebttoocostlyandoptsnottodoso.Aninelasticdemandthuslimitsthemaximumamountofdebtthegovernmentiswillingtoissue.Wefindthatthislimitleadstoaquantitativereductionindefaultriskandanincreaseinbondprices.Ourfindingscontributetoseveralstrandsofliterature.First,wecontributetoalong-standingliteratureusingindexrebalancingstoestimateassetpricereactions,demandelasticities,andchangesininvestors’portfoliosacrossdiffereGurel,1986;Shleifer,1986;Greenwood,2005;Hauetal.,2010;Changetal.,2014;Raddetal.,2017;PandolfiandWilliams,2019;PavlovaandSikorskaya,2022).1Ourcontributiongovernmentfinancinginemergingeconomies:theinternationalU.S.dollarbondmarket.Anotherkeycontributionofourworkisshowingthat,eveninresponsetoexogenoussupply-shiftingshocks,partofthepricereactionisattributabletochangesintheasset’sbymovementsintheeffectivesupply.Assuch,itcanbeextendedtoavastliteraturethatTypicalexamplesaresovereignSecond,agrowingliteratureoninelasticfinancialmarketsemphasizestheroleoftheGabaixandKoijen,2021;Vayan1Beyondindexrebalancings,Drosteetal.(2023)usehigh-frequencyU.S.TreasuryauctionstoestimatetheeffectofdemandshocksonTreasuryyields.45 etal.,2021b)andinternationalfinancialassets(KoijenandYogo,2020;Gourinchasetal.,wefocusontheinterplaybetweenadownward-slopingdemandcurve,defaultrisk,andtheprovisionofriskybonds.Weshowthatthedemandelasticityinteractswithdefaultriskandinfluencesagovernment’ssupplyofriskybaseofgovernmentdebtimpactbondyields(WarnockandWarnock,2009;Dell’Erbaetal.,2013;Peiris,2013;ArslanalpandPoghosyan,2016;AhmedandRebucci,2022).Acloselybondyieldsininternationalmarkets.Inthispaper,weexploitexogenouschangesintheframeworkextendsstandardmodels(AguiarandGopinath,2006;Arellano,andEyigungor,2012)byincorporatingtwodifferentinvestortypes(activeandpassive)andintroducingadownward-slopingdemand.Thisricherstructureallowsustodisciplinetheriskbehindthoseestimatesandtobackoutthestructuraldemandelasticity.Inouranalysis,weareagnosticaboutthemechanismsbehindthedownward-slopingdemand.PreviousworkbyBorriandVerdelhan(2010),Lizarazo(2013),Pouzoa(2016),andArellanoetal.(2017)analyzesovereigndebtmodelswithrisk-averseinvestors.Intheirmodels,investors’downward-slopingdemandisaby-productoftheirriskaversion.Inotherwords,investorsareinelasticonlybecausetheymustbecompensatedforeachadditionalunitofriskydebttheyhold.However,severaldifferentmechanismscaadownward-slopingdemand.Forexample,itcanbeexplainedbyregulatorylimitations,suchasaValue-at-Risk(VaR)constraint(asinMiranda-AgrippinoandRey,2020),byinvestors’buy-and-holdstrategies(whichcanberationalizedbyatasteforsimplicityor2ArelatedliteraturefocusesonU.S.andinternationalcorporatebondmarkets(DathanandDavydenko,2020;Bretscheretal.,2022;Calomirisetal.,2022;Kubitza,2023).3Inthisregard,ourpaperconnectswithrecentworkbyCostainetal.(2022),whointroduceendogenousdefaultriskintoaVayanos-VilapreferredhabitatmodeltoanalyzethetermstructureofinterestratesintheEuropeanMonetaryUnion.6agencyfrictions),orbyfixed-sharemandatesspecifyinghowinvestorsshouldallocatetheirstructurethatcanaccommodateanyofthesepotentialdrivers.Ouraimisnottouncoverthecausesofinvestors’inelasticbehaviorbutrathertoexamineitsaggregateimplications.Therestofthepaperisstructuredasfollows.Section2introducesasimpleframeworktoguideouranalysis.Section3presentstheempiricalanalysis,includingdetailsontheinstitutionalsetupofEMBIindexes,datasources,theidentificationstrategy,andresults.2DemandElasticityforRiskyBondsToguideouranalysis,weintroduceasimpleframeworkfeaturingheterogeneousinvestorswhodifferinhowtheyallocatetheirfundsacrossriskyassets.Thisstylizedframeworkcanbeusedtoestimateareduced-formpricedemandelasticity.Indoingthat,weshowAlthoughwefocusonthecaseofriskybonds,ourframeworkcanbeappliedtoanyrisky2.1ModelInvestorsareheterogeneousinhowtheyallocatetheirfundsacrossriskyassets.Letj={1,...,J}denotetheinvestor.AsinGabaixandKoijen(2021),weassumethateachinvestorjhasamandateorrulethatspecifieshowtheyshouldallocatetheirfundsacrossi={1,...,N}riskybonds.Totightlylinkthemodelwithourempiricalanalysis,wefurtherassumethatinvestorstrackthecompositionofabenchmarkindexIanddifferinhowactivelyorpassivelytheydoso.Letwt={w,...,w}denotethevectoroftime-varyingindexweightsforeachconstituentbondofindexI.Marketsarecompetitive,andinvestorstakeWedefinext=astheshareofwealththatinvestorjinvestsinbondiattimet.Thetermqdenotestheunitpriceofbondi,B,tdenotestheend-of-periodholdingsofinvestorjinbondi,andWj,tdenotestheirwealth.Thesharextisgivenbythefollowingxt=θj(ξeΛji,t(r+1))+(1?θj)w.(1)7Thetermθjparameterizesthedegreeofactivenessorpassivenessofinvestorj.Purelypassiveinvestorscanbecharacterizedbyθj=0,indicatingthattheirportfoliosimplyθje(0,1],whichcapturesthefractionoftheirportfoliothatisnotlinkedtoindex裝.Withintheiractiveallocation,investorsapportionafixedfraction,ξ,oftheirwealthtobondiandavaryingcomponentdeterminedbyΛji,t（r+1）,whereΛj>0parameteelasticityandi,tisanarbitraryfunctionofthenext-periodexcessreturnofbondi,r+1.Forinstance,ifi,t（r+1）=Et（r+1）,investorsallocateahighershareoftheirwealthtoAsweshownext,thereduced-formmandateinEquation(1)allowsustointroduceanaggregatedemandelasticitythatcanbeparameterizedbyΛ三{Λ1,..,ΛJ}.Whilethismandatecanhavedifferentmicrofoundations(asshowninAppendix),wetakeitasgivenbondsbutrathertoexamineitsimplications.4Afteraddingupalltheindividualdemands,wecanwritethemarket-clearingconditionqB=+ti(w),(2)whereBisthebondsupply,and三ΣjWj,tθj(ξeΛji,t)andti(w)三ΣjWj,t(1-θj)wdenotethemarket-valueactiveandpassivedemands,respectivelyti(w),istheportionofinvestors’holdingsaimedatreplicatingtheindexcompositiontheyfollow.Itcapturestheholdingsofbothsemi-andfullypassiveinvestors.Weexplicitlywriteti(w)asafunctionofwtoemphasizeitsdependenceontheindexweights.Fortherestoftheanalysis,itisusefultorewritethismarket-clearingconditionasB=第+lti（w）,Foranybondiinfixedsupply,aninthesupplyofbondsavailabletoactiveinvestors(i.e.,aleftwardshiftintheeffectiveorthedemandcurvesforactiveinvestorsslopedownward.Figure1illustratesthispoint.Iftheactivedemandisfullyelastic,thenanexogenousshiftinlti(w)shouldnotaffectbondprices(Panel(a)).Conversely,bondpricesshouldreacttothisshiftiftheactivedemand4AsarguedbyGabaixandKoijen(2021):“Whileidentifyingtheexactreasonsforlowmarketelasticityisinteresting,thisquestionhasalargenumberofplausibleanswers.Fortunately,itispossibletowriteaframeworkinawaythatisrelativelyindependenttotheexactsourceoflowelasticity[...].”5InSection2.2,weimposeadditionalstructuretoobtainclosed-formexpressionsforAandTti(w,.8Figure1Indexrebalancingandthedemandelasticity(a)PerfectlyelasticdemandPriceEffectivePriceSupplyq=qAiBi?TiQuantity10BBi?TiQuantity10(b)InelasticdemandPriceEffectivePriceSupplyAiAiBi?TiQuantity10BBi?TiQuantity10Note:ThefiguredepictsadecreaseintheeffectivesupplydrivenbyanincreaseinTi.Panel(a)considersthecasewhentheresidualdemandisfullyelasticandPanel(b)whenitispricesensitive.Basedonthisgraphicalintuition,onecouldexploitexogenow(w+1)?Tti(w),canthenestimatethefollowingreduced-forminversedemandelasticity:i=(?)?qB?Ttitt(3)isastandardpracticeintheliteraturetoestimatedemandelasticitiesinequitymarkets.for?Ttitoanalyzewhetherdemandcurvesforequitiesslopedownward.Morerecently,onRussellequityindexes.Forthecaseofbonds(sovereignorcorporate),exploitingindexcoincidewithlargebondissuancesthattendtoreflectinformationabouttheissuer’sown2.2EndogenousIssuerResponsesandBondPayoffsTodirectlymapEquation(3)intoastructuralelasticity,ηi,weneedtoassumethattheintrinsicvalueoftheassetisunaffectedby?Tti.However,exogenousshiftsintheeffective9ofriskybonds,apositive?Ttithatleadstoahigherbondpriceimpliesalowerborrowingcost,potentiallyaffectingtheissuer’sdefaultlikelihoodorits(future)bondissuances.EvenifthecurrentsupplyBrepayofffromholdingthebondand,therefore,itsprice.Priceresponsestoexogenousshiftsintheeffectivesupplycanthuscapturebothanformsolutionfortheprice.Weassumethati,t(r+1)=sothattheactivedemandisafunctionofthebond’sexpectedexcessreturnanditsvariance(theSharperatio).6WedefineR+1asthenext-periodrepaymentperunitofthebondsothatr+1≡?rf,whererfdenotestherisk-freerate.7Basedonthesedefinitionsandthemarket-clearingconditioninEquation(2),wecanwritetheequilibriumbondpriceasThetermcapturesthepriceunderperfectlyelasticinvestors,whichisafunctionoftheexpectednext-periodrepayment.Ontheotherhand,theΨfunctioncapturestheΨ≡1?κ(Λ)(B?Tti?,.(5)Thetermκ(Λ)≡characterizesthedegreeofinelasticityinthemarketforbondi.Whenκ(Λ)=0,thedemandisperfectlyelasticandthepriceonlydependsontheexpectedrepayment.TheB?Tticomponentiswhatwehavereferredtoastheresidualsupply,andcapturestheinelasticportionoftheactivedemand,whichdependsonthefixedcomponentofinvestors’mandates,ξ.SeeAppendixforthedetailsandderivations.thestructuralelasticity,ηi,andendogenouschangesinexpectedrepayment,αi.Inpart6ThisisasimilarspecificationtotheoneinGabaixandKoijen(2021),whichisafunctionofexpectedexcessreturnsandashocktotastesorperceptionsofrisk.Asshowninourquantitativeanalysis,thisspecificationallowsustocaptureademandelasticitythatdiffersacrosscountrieswithdifferentlevelsofdefaultrisk,whichisconsistentwithourempiricalfindings.7Therepaymentfunctiondependsontheexpecteddefaultandnext-periodissuances.Forashort-term(one-period)riskybond,itisgivenbyR+1=1?d+1,whered=1denotesadefault.Forlong-termbonds,italsodependsonnext-periodissuancesastheyaffectthenext-periodbondprice.InSection4,weanalyzethisfunctionindetail.Figure2PriceEffectivePriceSupplyAABi?Bi?TiBi?Ti01PriceEffectivePriceSupplyABi?TiQuantity10BBi?TiQuantity10Note:ThefiguredepictsareductionintheeffectivesupplydrivenbyanincreaseinTi.Panel(a)considersacaseinwhichtheexpectedassetpayoffsincreaseaftertheeffectivesupplydecreases,whilePanel(b)showstheoppositecase.?)=?)+?)?E).(6)Figure2providesagraphicalillustration.Ifapositive?Ttiraisesthenext-periodexpectedrepayment(i.e.,αi<0),investorswillbewillingtopayahigherpriceforanygivenB,mightleadtotheconclusionthatthedemandcurveissteeper(moreinelastic)thanittrulyis.Conversely,ifapositive?Ttilowersthenext-periodexpectedrepayment(i.e.,αi>0),itwouldcausetheactivedemandtoshiftdownward(Panel(b)).Thisshiftmightleadtothedemandcurvebeingestimatedasflatter(moreelastic)thanittrulyis.Sincebondpricesandpayoffsarejointlydetermined,itischallengingtodisentangletheeffectsonbondpricesduetothedownward-slopingdemandfromthoseresultingfromchangesinexpectedpayoffs.Toformallyquantifyeachmechanismseparately,onewouldneedastructuralmodelinwhichbondprices,thesupplyofthebond,anditspayoffsareendogenousoutcomes.Putdifferently,toquantifyαi,wemustfirstunderstandhow?Ttiaffectsissuers’policies(debtissuancesanddefault).OnecouldarguethatmorchangesinTtiwouldlikelyhavealargerimpactontheissuers’policies,potentiallymakingtheabsolutevalueofαilarger.Conversely,amoretransitoryshockcouldleadtoasmallerInthenextsection,weconstructanovelinstrumentfor?Tti,basedonmonthlyindexrebalancingsforamajorsovereignbondindexforemergingeconomies.WeestimatebondSection4,weformulateastructuralmodeltobackoutthestructuralelasticity.3EmpiricalAnalysis3.1IndexRebalancingsasPassiveDemandShocksInthissection,weexploitmonthlyrebalancingsintheEMBIGDtoidentifyexogenousshiftsintheavailablebondsupplyforactiveinvestors(theeffectivesupply).TheEMBIGDtrackstheperformanceofemergingmarketsovereignandquasi-sovereignbondsinU.S.dollarsissuedininternationalmarkets.8Unlikeotherindexesthatuseatraditionalmarketcapitalization-basedweightingscheme,theEMBIGDrestrictstheweightsofcountrieswithabove-averagedebtoutstandingbyincludingonlyafractionoftheirfaceamountofdebtbillionin2018(AppendixFigure2).10RebalancingsintheEMBIGDindex,triggeredbybondinclusionsandexclusions,occurareportdetailingtheupdatedindexcomposition.Consequently,passiveinvestorstrackingtheindexadjusttheirportfoliosbybuyingorsellingbondstomatchthenewindexweights. FollowingPandolfiandWilliams(2019),weconstructtheflowsi(FIR)measureforeachcountryateachrebalancingdate.TheFIRquantitativelymeasurestherelativechangeinpassivedemandforacountry’ssovereignbondsresultin8Theindexincludesbondswithamaturityofatleast2.5yearsandafaceamountoutstandingofatleastUS$500million.Tobeclassifiedasanemergingeconomy,acountry’sgrossnationalincome(GNI)percapitamustbebelowanIndexIncomeCeiling(IIC)forthreeconsecutiveyears.TheIICisdefinedbyJ.P.MorganandadjustedeveryyearbythegrowthrateoftheWorldGNIpercapita,Atlasmethod(currentUS$),providedbytheWorldBank.Bondsintheindexmustsettleinternationallyandhaveaccessibleandverifiablebidandaskprices.Onceincluded,theycanremainintheindexuntil12monthsbeforematurity.Locallawinstrumentsarenoteligible.9TheJ.P.MorganEmergingMarketsBondIndexGlobal(EMBIG)hasthesamebondinclusioncriteriaastheEMBIGD.TheonlydifferencebetweenthemisthatwhiletheEMBIGusesamarketcapitalizationweightingscheme,theEMBIGDmodifiesthisschemetolimittheweightsofcountrieswithabove-averagedebtoutstanding.AppendixFigure1plotstheEMBIGcountryweightsofboththeEMBIG(amoretraditionalmarket-basedindex)andEMBIGDversionsforDecember2018.AppendixdescribestherulesthattheEMBIGDusestocomputetheweightsoftheinstrumentsincludedintheindex.10AppendixFigures3and4showthehighpreponderanceofU.S.dollar-denominatedsovereigndebtissuedbyemergingeconomiesininternationalmarkets.FIRc,t=?c,t.qc,t?1Bc,t?1-wc,t?1At?1.(7)Thenumeratorcapturesthechangeinpassivedemandimpliedbytheindexrebalancing,andisdefinedas?c,t=(wc,t-w)At.Itmeasurestheamountoffundsthat,onagivenrebalancingdate,enterorleaveacountryduetotherebalancingintheportfolioofpassiveinvestorstrackingindexbondsavailabletoactiveinvestors,qc,t?1Bc,t?1-wc,t?1At?1.Thefirsttermintheparentheses,wc,t,isthebenchmarkweightforcountryc,attimet,inindex裝.Itisdefinedaswc,t=qc,tfc,t,whereqc,tBc,tdenotesthemarketvalueofbondsfromcountrycattimet.qc,tdenotesthepriceandBc,tdenotesthefaceamountthecountry’samountofoutstandingdebtratherthanonmarketvalues.11ThedenominatoroftheindexandItisthenumberofavailableindexunits.Thatis,itcapturestherelativemarketcapitalizationofcountryc’ssovereignbondsincludedin裝.Thesecondtermintheparentheses,w,isdefinedasw=wc,t?1q1.12Itcapturesthe“buy-and-holdweight,”definedastheweightcountrycwouldhavehadattimetiftheindexcompositionhadremainedunchanged.13Atin?c,trepresentstheAUMofinvestorspassivelytrackingtheEMBIGD.country’sfundamentals,fortworeasons.First,theFIRisaffectedbycountries’sovereignconstant,theFIRcanbemechanicallycorrelatedtopresentorpastbondpricechanges.theformerderivativewillnotbeequaltozerogiventheeffectofcurrentpricesthroughthenumerator.Inturn,thelatterderivativemightnotbeequaltozeroduetotheeffectso11Topreservediversification,J.P.Morganappliesaschemethatentailsacaptotheweightofcountrieswithgreater-than-averagesovereignbondmarkets,forwhomthediversificationcoefficientisthereforesmallerthanone,fc,t<1.IncontrasttotheEMBIGD,forpurelymarketcapitalization-weightedindexes(suchastheEMBIGlobal),fc,t=1,Ac,t.12Thisbuy-and-holdweightiscomputedasifnobondshadenteredorexitedtheindexattimet.13Notethatw=.Absentanychangeintheindexcomposition(i.e.,inclusionsorexclusionsofnewbondsorcountries),ifthepriceofacountry’ssovereignbondsincreasesmorethanthatofothercountriesintheindex,theweightofthatcountryintheindexincreases.Nevertheless,investorsdonotneedtorebalancetheirportfoliosasthe“buy-and-holdweight”coincideswiththenewweightintheindex,wc,t.pricesthroughthedenominator.Giventhatweaimtoshocksonbondprices,thepotentialendogeneityoftheFIRcouldbiasoWeaddressthepotentialchallengeoftheFIRendogeneityintwoways.First,foreachnotchangerelativetothepreviousperiod.Inotherwords,wefocusonlyoncountriesthatexperiencenonewissuances,bondrepurcduetomaturityont