1SeeMendozaandTerrones2012andDell146Aricciaetal2012forabriefdiscussionontheformalde133nitionandempiricalidenti133cationofcreditboomsMendozaandTerrones2012analyzetheevolutionofprivatec ID: 230180
Download Pdf The PPT/PDF document "Creditmarketsplayanincreasinglycentralro..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Creditmarketsplayanincreasinglycentralroleinmoderneconomies.WithintheOECD,forinstance,domesticcredithasrisenfrom100%ofGDPin1970toapproximately160%ofGDPin2012(Figure1).Thisgrowthmaskslargevariationsacrosscountriesandovertime(Figure2).Andyet,thereisacommonfeaturetoallthesedi¤erentcountryexperiencesthatstandsout:credithasoftenalternatedperiodsofrapidgrowthorbooms,withperiodsofstagnationorsigni cantdeclineorbusts.Moreover,thereissomeevidencethatthesecreditboomsandbustshavebecomemorecommoninrecentyears.1ConsiderthecaseoftheUnitedStates,wherecredithasgrownmoreorlesscontinuouslythroughoutthesampleperiod.Eventhere,itgrewbyapproximately40percentofGDPbetween1990and2010,onlytocontractsharplyafterwards.InGreece,Ireland,SpainandPortugal,thedynamicsofcreditlooksimilarduringthelastdecades:stagnantordecliningcreditbetweenthemid1980sandthemid1990s,aspectacularsurgeincreditbetweenthemid-1990sand2010,andstagnationorasharpdeclinesincethen.Lookingahead,thesedropsincreditneednotbeshort-livedastheJapaneseandSwedishexperiencesshow.InJapan,forinstance,creditgrewrapidlyinthelate1990sandhasfallensteadilysinceits1999peak.InSweden,creditcollapsedduringthe nancialcrisisoftheearly1990sandittookoveradecadetoreturntoitspreviouspeak.Thesecreditboomsandbuststendtobeaccompaniedbychangesinkeyeconomicvariables.IthasbeenwelldocumentedthatcreditboomsareassociatedwithhighassetpricesandhighgrowthratesofrealGDP,consumptionandinvestment.Accordingtosomeestimatesthegrowthrateofinvestmentdoublesduringbooms.2Inspiteofthis,creditboomsarestillviewedwithconcernbypolicymakersandacademics.Thereasonisthattheyeventuallyend,andtheiraftermathsareoftencharacterizedby nancialcrisesandloweconomicgrowth.3Thishaspromptedcallsforpoliciesthatrestraincreditduringbooms,inthehopethatsmallerboomswillleadtosmallercrises. 1SeeMendozaandTerrones(2012)andDellAricciaetal.(2012)forabriefdiscussionontheformalde nitionandempiricalidenti cationofcreditbooms.MendozaandTerrones(2012)analyzetheevolutionofprivatecredittotheprivatesectorin61countriesbetween1960and2010andidentify70creditbooms:eachofthecountriesplottedinFigure2containsatleastoneofthese.DellAricciaetal.(2012)analyzedataonbankcredittothedomesticprivatesectorin170countriesbetween1960and2010.Theyidentify175creditbooms,whichtranslatesintoa14%probabilityofacountryexperiencingacreditboominanygivenyear.Theyalsodocumentthatcreditboomsappeartohavebecomemorecommonovertime,inthesensethatthefractionofcountriesexperiencingacreditboominanygivenyearhasbeenincreasingsincethe1980s.Claessensetal.(2011)useadi¤erentapproachandstudycreditcycles,buttheyalso ndthemtobecommon:inasampleof21advancedeconomiesbetween1960and2007,theyareabletoidentify114suchcycles.2SeeMendozaandTerrones(2012)andDellAricciaetal.(2012).3DellAricciaetal.(2012) ndthatonethirdofcreditboomsendupin nancialcrises.MendozaandTerrones(2012) ndslightlylowernumbers.Therecent nancialcrisisintheUnitedStatesalsoprovidesevidenceinthisregard:itwasthoseregionsthatexperiencedthelargestcreditboomsintherunuptothecrisisthatsu¤eredthegreaterincreaseincreditdelinquencyduringthecrisis(DellAricciaetal.2008,MianandSu 2009).1 Toevaluatethemeritofthesecallsforpolicy,onemusthaveaviewoftheforcesdrivingthesecreditcycles.Creditmayuctuateforvarietyofreasons,anddi¤erenttypesofuctuationsmaycallfordi¤erentpolicyresponses.Ataverygenerallevel,uctuationsincreditmayreectchangesindemandorinsupply.Andthese,inturn,mayreectchangesinavarietyoffactorslikepreferences,technologyorexpectations.Inthispaper,wefollowthelargebodyofrecentmacroeconomicresearchonthe nancialacceleratorandfocusoncredituctuationsthataredrivenbyuctuationsinborrowingconstraints.Thenarrativeissimple.Whenaborrowerobtainscredit,sheisexchanginggoodstodayforapromisetodelivergoodsinthefuture.Thesepromisesareonlyvaluedbysaversiftheyhavesomeguaranteeofbeingrepaid.Onecanthinkofthisguaranteeasthefutureincomeofentrepreneursthatcanbecrediblypledgedtosavers.Werefertothispledgeableincomeastheeconomysstockofcollateral.Ifentrepreneursareconstrained,itisthisstockofcollateralthatdeterminestheamountofpromisesthatcanbeissued.Ifentrepreneursareconstrained,understandingboom-bustepisodeslikethoseshownaboverequiresatheoryofcollateraluctuations.Thispaperprovidessuchatheory.Ourkeyinnovationistodistinguishbetweenfundamentalandbubblycollateral.Fundamentalcollateralisthepartofaborrowerspledgeableincomethatcorrespondstofutureoutput,i.e.itconsistsofaborrowersrightstofutureproduction.Bubblycollateralisinsteadthepartofaborrowerspledgeableincomethatcorrespondstofuturecredit,i.e.itconsistsofaborrowersrightstofutureborrowing.Wecallthistypeofcollateralbubblybecauseitconstitutesabubbleorpyramidschemeinwhichpresentcontributions(presentcredit)purchasefuturecontributions(futurecredit):aslongasthereturntothesebubblesorpyramidschemesisnolowerthantheinterestrate,lenderswillbewillingtoacceptthemascollateral.Themacroeconomicliteraturehasexclusivelyfocusedonfundamentalcollateral,studyingitsimplicationsforcredit,investmentandgrowth.Thisviewofcollateralisincomplete,though.Wheneverfundamentalcollateralisinsu¢cient(say,becauseofweakenforcementinstitutions),weshowthatthereisroomforinvestoroptimismtosustainbubblesthatexpandtheeconomysstockofcollateralandtotalcredit.Theintuitionbehindthisresultisthatthelackoffundamentalcollateralreducesthedemandforcreditandthustheequilibriuminterestrate,therebyloweringtherateatwhichbubblesmustgrowinorderforthemtobeattractivetosavers.Inthismanner,thelackoffundamentalcollateralrelaxestheconditionsforbubblycollateraltoexist.Whatarethemacroeconomice¤ectsofbubblycollateral?Byde nition,itenablesentrepreneurstoobtaincreditinexcessoftheirfundamentalcollateral:intuitively,currententrepreneurscan2 obtainexcesscredittodaybecauseitisexpectedthattherewillbeexcesscreditinthefutureaswell.Thisisthecrowding-ine¤ectofbubbles,whichceterisparibusincreasesinvestment.Butthisfutureexcesscreditwilldivertsomeoftheresourcesoffuturegenerationsawayfrominvestment.Thisisthecrowding-oute¤ectofbubbles,whichceterisparibusreducesinvestment.Themacroeconomicconsequencesofbubblesdependontherelativestrengthofthesetwoe¤ects.Inparticular,we ndthatthecrowding-ine¤ectdominateswhenbubblycollateralislow,andthecrowding-oute¤ectdominateswhenbubblycollateralishigh.Thisgivesrisetoanoptimalbubblewhichtrades-o¤thesetwoe¤ectsandprovidestheamountofbubblycollateralthatmaximizeslong-runoutputandconsumption.Anessentialfeatureofbubblycollateralisthatitsstockisdrivenbyinvestorsentimentormarketexpectations.Thecreditobtainedbyentrepreneurstodaydependsonmarketexpectationsaboutthecreditthatentrepreneurswillobtaintomorrow,whichinturndependontomorrowsmarketexpectationsaboutthecreditthatentrepreneurswillobtainonthedayafter,andsoon.Becauseofthis,marketsmaysometimesprovidetoomuchbubblycollateralandsometimestoolittleofit,whichcreatesanaturalroleforstabilizationpolicies.Weshowthatacreditmanagementagencywiththeauthoritytotaxandsubsidizecreditcaninfactreplicatetheoptimalbubbleallocation.Todoso,itmustadoptapolicyofleaning-against-the-wind,taxingcreditwhenbubblycollateralisexcessiveandsubsidizingitwhenbubblycollateralisscarce.Weexploretheimplicationsofthispolicyandshowthatitraisesoutputandconsumption,althoughitmaysometimesincreasemacroeconomicvolatility.Wearguethat,tosomeextent,thispolicycanbeinterpretedasaschemeinwhichthecreditmanagementagencypurchasesassetstomanagetheeconomysstockofcollateral.Finally,weshowthatthesameconditionsthatmakebubblycollateralpossibleguaranteethatacreditmanagementagencyalwayshasenoughresourcestoimplementthedesiredpolicy.Thepaperisorganizedasfollows.Section1developsourbasicmodelofcollateral,creditandinvestment.Section2exploresbubblyequilibriaandstudiestheimplicationsofbubblycollateralfordynamicsandwelfare.Section3introducesacreditmanagementagencyandshowshowcredit-marketinterventionscanbeusedtomaximizeoutputandconsumption.Section4explorestheissueofa scalbackstop,i.e.oftheresourcesrequiredtosustainthedesiredpolicy.Section5concludes.Beforemovingon,weo¤erashortreviewofrelatedliterature.Literaturereview:ThemodeldevelopedherebuildsuponpreviousworkbyMartinandVentura(2011,2012).InMartinandVentura(2012),weanalyzedtheexistenceandmacroeconomice¤ectsofbubblesin3 aneconomywithextreme nancialfrictionsthatimpedecredit.InMartinandVentura(2011),weextendedtheanalysistoaneconomywithcredit,andarguedthatbubblescouldbeasourceofcollateral.Thosepapersusedasimplegrowtheconomytoestablishtwokeyresultsregarding nancialfrictionsandbubbles.First, nancialfrictionsthatlimitborrowersabilitytopledgetheirfutureincome,andthustheeconomysstockoffundamentalcollateral,relaxtheconditionsfortheexistenceofbubbles:thereason,aswementionedabove,isthatsuchfrictionslowertheequilibriuminterestrateandthustherateatwhichbubblesmustgrowinorderforthemtobeattractivetosavers.Second, nancialfrictionsmakeitpossibleforbubblestobeexpansionary:thereasonisthat,byraisingtheeconomysstockofcollateral,bubblesmayintheoryraisecreditandinvestment.Relativetothosepapers,wenow:(i)fullycharacterizethee¤ectsofbubblycollateral,identifyingtheconditionsunderwhichitisexpansionaryorcontractionary,and;(ii)analyzethee¤ectsofmacroprudentialpolicies,andoftheresourcesrequiredtoimplementthem.Ourpaperisobviouslyrelatedtoalargebodyofresearchthatdealswithuctuationsincredit,bothempiricallyandtheoretically.Theformerhassoughttoidentifyempiricalregularitiesofcreditboomsandbusts:Gourinchasetal.(2001),Claessensetal.(2011),MendozaandTerrones(2012),DellAricciaetal.(2012)andSchularickandTaylor(2012)fallwithinthiscategory.Onthetheoreticalfront,variouspapershavetriedtomodelcreditcyclesasanequilibriumoutcomeofcompetitionin nancialmarkets.SomeexamplesofthisworkareRuckes(2004),DellAricciaandMarquez(2006),Matsuyama(2007),GortonandHe(2008)andMartin(2008).Likeus,thesepapersmodeluctuationsincredit.Unlikeus,though,thesepapersemphasizetheroleofinstitutionalfeaturesof nancialmarketslikeregulationortheincentivesofcertainmarketparticipantsingeneratingandmagnifyinguctuationsincredit.Wetakeinsteadamacroeconomicperspectiveandarguethat,whenfundamentalcollateralisscarce,investorsentimentmaygiverisetocreditbubbles.Fromamethodologicalperspective,thus,thispaperisclosesttotheliteraturethathasstudiedthee¤ectsofbubblesinthepresenceof nancialfrictions:(i)unlikeus,CaballeroandKrishna-murthy(2006),FarhiandTirole(2011)andMiaoandWang(2011)focusontheroleofbubblesasausefulsourceofliquidity;(ii)likeus,Kocherlakota(2009)focusesontheroleofbubblesascollateralornetworth;and(iii)unlikeus,Ventura(2011)focusesonthee¤ectsofbubblesonthecostofcapital.Thereisalsoalongtraditionofpapersthatview atmoneyasabubble,likeSamuelson(1958)andKiyotakiandMoore(2008).Ourmodelisalsorelatedtothevastworkonmacroeconomicmodelswith nancialfrictions,goingbacktoBernankeandGertler(1989)and4 KiyotakiandMoore(1997),inwhichcollateralplaysakeyroleindeterminingthelevelof nancialintermediationandeconomicactivity.Ourcontributionrelativetothisworkisthatwedistinguishbetweenfundamentalandbubblycollateral.Ourstudyofpolicyisrelatedtotheliteratureonbailouts.Bailoutsarecommonlyrationalizedasawaytostimulateeconomicactivityintheeventofacrisis,bytransferringresourcestowardsdistressedagents.Inamacroeconomicmodelwith nancialfrictions,forinstance,GertlerandKiyotaki(2012)showhowpoliciesthatprovidecredittodistressedagentscanreducetheseverityofacrisis.Atthesametime,theprevailingviewisthatbailoutsgeneratecostsexantebydis-tortingincentives(e.g.FarhiandTirole2012).Thecreditsubsidiesthatwestudyarenotreallybailoutsbecausetheyare nancedthroughtaxesonentrepreneursthemselves,i.e.,theyentailnonettransferstoentrepreneursasawhole.Theyareakintobailouts,though,becausetheyaregiventoentrepreneurstorepaytheirdebts.Thee¤ectsofthesesubsidiesaredi¤erentfromthoseemphasizedintheliterature,however.Theyhavepositivee¤ectsoneconomicactivityexantebecause,asinRanciereetal.(2008),theexpectationoffuturebailoutsraisescollateral;buttheyarecostlytoexecuteexpostbecausetheydivertresourcesawayfrominvestment.Finally,thispaperisrelatedtotherecentworkthatstressestheroleofinformationinsensitivecollateralsuchasDangetal.(2011)andGortonandOrdoñez(2012).Thisresearchemphasizestheroleofthe nancialsystemingeneratingcollateralthatisacceptedbyeveryonewithoutfearofadverseselection,becauseitdoesnotpayforanyagenttoproduceprivateinformationregardingitsunderlyingcharacteristics.Ofcourse,theseincentivesmaychangeduringacrisis,inwhichcaseinformationisgeneratedandthevalueofcollateralmightexperiencelargeswings.Inourmodel,itisalsonaturaltothinkofbubblycollateralasinformationinsensitiverelativetofundamentalcollateral.Thevalueofbubblycollateraldependsoninvestorsentimentormarketexpectationswhicharepublicinformation.Thevalueoffundamentalcollateraldependsinsteadontheregularbusinessactivityoftheunderlying rmor nancialinstitution,whichislikelytobeinuencedbymanagersthroughavarietyofchannelsthataredi¢culttoobserveandgiverisetoprivateinformation.1ThebubblyeconomyThestartingpointofouranalysisistheclassicDiamond(1965)modelofcapitalaccumulationwithoverlappinggenerations.Theyoungsupplylabor,consumepartoftheirlaborincomeandsavethe5 rest.Thesesavingsareusedtopurchasecapitalfromtheoldandtoproducenewcapital.Theoldownthecapital,demandlaborandconsumetheircapitalincome.Tirole(1985)extendedtheDiamondmodelbyintroducinganadditionalmarketforbubblesorpyramidschemes.Participantsinapyramidschememakevoluntarycontributionsthatentitlethemtoreceivethenextcontribution.Thus,wesaythatanindividualpurchasesabubblewhenhe/shemakesacontributiontoapyramidscheme.Similarly,wesaythatanindividualsellsabubblewhenhe/shereceivesacontributionfromthepyramidscheme.Startingabubbleorpyramidschemeyieldsawindfalltothe rstparticipant,whichconsistsofthe rstcontributiontothescheme.Weassumethateachgenerationstartsnewbubbleswithrandominitialvalue.Thesenewbubblesaretradedinthemarketforbubblesalongsidetheoldbubblesstartedbypreviousgenerations.4ThemodelwepresenthereaddsacreditmarkettotheTirolemodel,assketchedinthelastsectionofMartinandVentura(2012).IntheDiamondandTirolemodelsthereisonlyintergen-erationaltrade.Herewecreategainsfromintragenerationaltradebyintroducingheterogeneitywithingenerations.Whilesomeindividualshavesavings(thesavers),otherindividualshavein-vestmentopportunities(theentrepreneurs).Byinvestmentopportunities,wemeanbothcapitalandbubbles.Youngsaversgivecredittoyoungentrepreneursand,inreturn,oldentrepreneurspaya(possiblycontingent)interestratetooldsavers.Ouremphasisispreciselyontheworkingsofthiscreditmarket.Ifthismarketworkedwell,theequilibriaofourmodelwouldnotdi¤erfromthoseoftheTirolemodel.Thecreditmarketwouldsimplyallowsaverstoe¤ectivelyowncapitalandbubbles,sincecompetitionamongentrepreneurswouldensurethatequilibriumcreditcontractsreplicatethepayo¤sofowningcapitalandbubbles.Butimperfectenforcementinstitutionslimitthecollateralofentrepreneursandimpairtheworkingsofthecreditmarket.5Bycollateral,wemeantheold-ageresourcesthatyoungentrepreneurscanpledgetoyoungsavers.Inthissection,wedescribeourmodelandshowhowthesetofequilibriadependsontheavailabilityofcollateral. 4ThearrivalofnewbubblesisanimportantdeparturefromtheoriginalTirolemodelthatplaysakeyroleinourtheory.5Imperfectenforcementinstitutionsdonotimpairtheworkingsofthemarketsforlabor,capitalorbubblessincetransactionsinthesemarketsdonotinvolveapromisebythesellertodeliverfuturepaymentstothebuyer.Thus,labor,capitalandbubblescanbetradedeveninsituationsinwhichcreditisnotpossibleorseverelyrestricted.6 1.1ThebasicsetupConsideraneconomyinhabitedbyasequenceofequal-sizedandoverlappinggenerationsofyoungandold.Timestartsinperiodt=0andthengoesonforever.Theobjectiveofindividualiofgenerationtistomaximizehis/herutilityUit,givenby:Uit=Cit;t+EtCit;t+1,(1)where0;andCit;tandCit;t+1aretheyoung-ageandold-ageconsumptionsofindividualigenerationt.Naturally,consumptionmustbenon-negative,i.e.Cit;t0andCit;t+10.Asusual,Etfgdenotesthemathematicalexpectationofthecorrespondingvariable.Eachgenerationcontainstwotypesofindividuals.Saversworkduringyouth,savepartoftheirlaborincomeandlenditto nanceold-ageconsumption.Entrepreneursborrowduringyouthto nancepurchasesofcapitalandbubbles,andproduceduringoldage.Saverssupplyoneunitoflaborduringyouthanddecidehowmuchoftheirlaborincometosave.Theyusetheirsavingstopurchasecreditcontractsinexchangeforagross,possiblycontingent,interestrateofRt+1unitsofoutputinperiodt+1foreachunitofoutputusedtopurchasecreditcontractsinperiodt.WerefertoRt+1asthecontingentorexpostinterestrate,andwerefertoEtRt+1astheinterestrate.Inoldage,saversreceivethereturntotheirsavingsandconsume.Thus,theirconsumptionsaregivenby:Cit;t=WtLt,(2)Cit;t+1=Rt+1Lt,(3)whereWtisthewageandLtissavingsorthevalueofcreditcontractspurchased.Equation(2)saysthatyoung-ageconsumptionequalslaborincomeminussavings,whileEquation(3)saysthatold-ageconsumptionequalsthereturntosavings.SaversmaximizeutilityinEquation(1)subjecttothebudgetconstraintsinEquation(2)and(3).Entrepreneursinvestandproduce.Theirproductiontechnologycanbedescribedbythefol-lowingCobb-Douglasfunction:F(Kt;Nt)=AtKt tNt1,(4)7 whereKtiscapital,2(0;1)and 1.Laborproductivitygrowsattheconstantrate 1.WerefertoAtasaproductivityshockandassumethatituctuatesrandomlyovertime.Theproductionofoneunitofcapitalrequiresoneunitoftheconsumptiongood.Weassumethatcapitalfullydepreciatesinproduction.Entrepreneursalsoinitiateandtradebubbles.WeuseBttodenotethevalueofthebubblesinperiodt.Someofthesebubblesareoldsincetheywereinitiatedbypreviousgenerationsofentrepreneursandwerepurchasedbyentrepreneursofgenerationt1.Someofthesebubblesarenewsincetheyareinitiatedbyentrepreneursofgenerationt1.Thus,theaggregatebubbleevolvesasfollows:Bt+1=RBt+1Bt+BNt+1,(5)whereRBt+1istheincreaseinpriceorreturntothebubblespurchasedfromentrepreneursofgenerationt1;andBNt+1thevalueofthebubblesinitiatedbytheentrepreneursofgenerationt.WerefertoBNt+1asbubblecreationandweassumethatitisnon-negative,i.e.BNt+10.To nancetheiractivities,entrepreneurssellcreditcontractstosavers.Thesecontractscanbefullystatecontingentbutmustbecollateralized,thatis,theymustbebackedbycrediblepromisesoffuturepayments.Thisbringsustothekeyfrictionthatunderliesalltheanalysisofthispaper:imperfectenforcementinstitutionslimittheamountofavailablecollateral.Inparticular,entrepreneurscanhideafraction1oftheircapitalincomefromenforcementinstitutions,sothat:Rt+1Lt[F(Kt+1;Nt+1)Wt+1Nt+1]+Bt+1.(6)Equation(6)statesthatpromisedpaymentscannotexceedafraction2[0;1]oftheircapitalincomeplustheirbubbles.Wethinkofthe rsttermasthefundamentalcollateralofentrepre-neurs,andofthesecondtermastheirbubblycollateral.6Throughoutthepaper,werefertoEquation(6)asthecreditorcollateralconstraint.As!1,theentirecapitalincomecanbeusedascollateralforcreditandthecreditconstraintcoincideswiththerestrictionthatold-ageconsumptionbenon-negative,i.e.Cit;t+10.Thislimitingcaseisthefrictionlesseconomy. 6Onewaytothinkaboutthisconstraintisthatcourtscangraballtheconsumptiongoodsthataretradedinthemarketbutonlyafractionofthosethatarenottraded.8 Withtheseassumptions,wecanwritetheconsumptionsofentrepreneursasfollows:Cit;t=LtKt+1Bt,(7)Cit;t+1=F(Kt+1;Nt+1)Wt+1Nt+1+Bt+1Rt+1Lt.(8)Equation(7)saysthatyoung-ageconsumptionequalscreditminusinvestmentandthepurchasesbubbles,whileEquation(8)saysthatold-ageconsumptionequalsproductionnetoflaborcostsplusthevalueofbubblesminuscreditpayments.EntrepreneursmaximizeutilityinEquation(1)subjecttothebudgetconstraintsinEquations(7),(8),thecreditconstraintinEquation(6)andthelawofmotionforbubblesinEquation(5).1.2MarketsandpricesWearereadytosolvethemodel.Inthelabormarket,oldentrepreneursdemandlaboruntilthemarginalproductequalsthewage.Sinceyoungsaverssupplyoneunitoflabor,Nt=1,thismeansthat:wt=(1)Atkt.(9)Asusual,weuselowercaseletterstorefertovariablesinunitsofe¢cientworkerssothat,forinstance,kt tKt.Equation(9)saysthatlaborincomeisaconstantfraction1ofproduction.Inthemarketforbubbles,oldentrepreneurssellbubblestoyoungones.Oldentrepreneurssupplyalltheirbubblesprovidedthepriceisnonnegative.Youngentrepreneurs,aswehaveseen, nancethepurchaseofbubblesbysellingcreditcontracts.Iftheexpectedreturntoholdingbubblesfallsshortoftheinterestrate,youngentrepreneursmakealosswhentheysellcreditcontractsto nancebubblepurchasesandthedemandforbubblesiszero.Iftheexpectedreturntoholdingbubblesexceedstheinterestrate,youngentrepreneursmakeapro twhentheyborrowtopurchasebubblesandthedemandforbubblesisunbounded.Thus,equilibriumrequiresthattheexpectedreturntoholdingbubblesequalstheinterestrate:EtRBt+1=EtRt+1.(10)Thisdoesnotmean,ofcourse,thattherealizedreturntoholdingbubblesequalstherealizedreturntocreditcontracts.9 Finally,wemovetothecreditmarketwhereyoungsaverssupplycredittoyoungentrepreneurs.Note rstthatEtAt+1k1t+1EtRt+11inequilibrium.7Usingthisobservation,thesupplyofcreditbyyoungsaverscanbedescribedasfollows:lt8:=wtifEtRt+1]TJ ; -1; .63; Td; [00;12[0;wt]ifEtRt+1=1.(11)Iftheinterestrateexceedsthediscountrate,youngsaverssavealltheirincome.Iftheinterestrateequalsthediscountrate,youngsaversareindi¤erentbetweensavingandconsuming.Thedemandforcreditbyyoungentrepreneurscanbewrittenasfollows:EtRt+1=8]TJ ; -1; .63; Td; [00;]TJ ; -1; .63; Td; [00;:EtAt+1k1t+1ifEtbNt+1(1)EtAt+1kt+1EtAt+1kt+1+Etbt+1 1ltifEtbNt+1(1)EtAt+1kt+1.(12)Equation(12)depictsthemaximuminterestratethatentrepreneursarewillingandabletopayperunitofcredit,respectively.Whetherthecreditconstraintisbindingornotdependsonwhetherexpectedbubblecreationexceedsthecapitalincomethatisnotpledgeable.Ifitdoes,thenthecreditconstraintisnotbindingbecauseentrepreneurshaveenoughcollateraltopledgetheentirereturntoinvestment:inthiscase,theinterestrateisequaltotheexpectedmarginalreturntoinvestment.Ifitdoesnot,thenthecreditconstraintisbindingbecauseentrepreneurialcollateralisinsu¢cienttopledgetheentirereturntoinvestment:inthiscase,themaximuminterestratethatentrepreneursareabletopayisdeterminedbytheratioofexpectedcollateral(fundamentalandbubbly)tocredit.8 7Thiscanbeshownbycontradiction.Assume rstthatEtRt+1EtAt+1k1t+1.Then,entrepreneurswouldnotwanttosellcreditcontractsandEtAt+1k1t+1=11;whilesaverswouldliketospendalltheirincomepurchasingcreditcontracts.Thus,itfollowsthatEtAt+1k1t+1EtRt+1.Assumenextthat1EtRt+1.Then,saverswouldnotwanttopurchasecreditcontractsandEtAt+1k1t+1=11;whileentrepreneurswouldwanttosellasmanycreditcontractsasallowedbytheirconstraint.Thus,itfollowsthatEtRt+11.8ThereisaninterestingwrinklethatariseswhenthecreditconstraintisbindingandEtbNt+1=0.Inthatcase,theonlypossibleequilibriumentailsEtRt+1=EtAt+1k1t+1:atthatinterestrate,though,individualentrepreneurs(whotakethereturntoinvestmentasgiven)areunconstrainedandtheydemandin nitecredit.Competitioncannotbiduptheinterestrate,moreover,becauseentrepreneursarealreadypledgingthemaximumshareoftheirincomethatisfeasiblegivenenforcementconstraints.Insuchasituation,clearingofthecreditmarketrequiresrationing.Weassumethatthisrationingworksasfollows:givenEtAt+1k1t+1entrepreneurso¤ertoborrowanamountLtataninterestrateEtRt+1;intheeventthattheinterestrateo¤eredbyallentrepreneursisthesame,thetotalsupplyofcreditisallocatedequallyacrossentrepreneurs.Therearetwopossibilities:(i)EtRt+11,inwhichcasesaversstrictlyprefertolendtheirresourcestoentrepreneursandthesupplyofcreditequalstheeconomyswagebill;(ii)EtRt+1=1,inwhichcasesaversareindi¤erentbetweenlendinginthecreditmarketorconsumingduringyouthandequilibriumlendingisexactlyequalto(EtAt+1)1 1.ThisissuedoesnotarisewhenEtbt+10,howeversmall,anditthereforeplaysaminorroleinouranalysis.10 Tocompletethesolution,weneedtodeterminehowentrepreneursusethecredittheyreceive,andthisisasfollows: kt+1+bt8:=ltifEtAt+1k1t+1]TJ ; -1; .63; Td; [00;12[0;lt]ifEtAt+1k1t+1=1.(13)Equation(13)saysthatentrepreneursconsumeinthe rstperiodonlyifthemarginalproductofcapitalequals1.Otherwise,entrepreneursuseallthecredittheyreceivetoinvestandpurchasebubbles.1.3EquilibriumdynamicsToobtainsomeintuitionsabouttheworkingsofthemodel,itisusefultocollapsethepreviousequationsasfollows:kt+18]TJ ; -1; .63; Td; [00;]TJ ; -1; .63; Td; [00;:=1 Atktbt ifEtRt+1120;1 Atktbt ifEtRt+1=1,(14)EtRt+1=minfEtAt+1;Etf(+nt+1)At+1ggk1t+1,(15)bt+1=EtRt+1+ut+1 bt+nt+1At+1kt+1,(16)whereut+1istheunexpectedcomponentofbubblereturnsandnt+1isthevalueofnewbubblesasashareofoutput,i.e.ut+1RBt+1EtRBt+1andnt+1bNt+1=(At+1kt+1).Werefertoutandntasbubble-returnandbubble-creationshocks,respectively.Equations(14)and(15)describethesupplyanddemandoffundsforinvestment,respectively.Equation(16)describesthedynamicsofthebubble.Jointly,Equations(14)-(16)provideafulldescriptionofthedynamicsofthestatevariables,i.e.ktandbt;foranyadmissiblesequenceofproductivityandbubbleshocks,i.e.At,utandnt.Thus,werefertothemasthelawofmotionofthesystem.Figure3showshowtheinterestrateandnextperiodscapitalstockaredeterminedbysolvingEquations(14)and(15).Ifthecapitalstockislow,theinterestrateisabovethediscountrateandyoungsaversusealltheirincometopurchasecreditcontracts.Inthiscase,collateralisabundantandthecapitalstockisdeterminedbythesupplyoffunds.Welabelthisrangeofcapitalstockstheregionoffullintermediation.Ifthecapitalstockishigh,theinterestrateequalsthediscountrateandyoungsaversconsumepartoftheirincome.Inthiscase,collateralisscarceandthecapital11 stockisdeterminedbythedemandforfunds.Welabelthisrangeofcapitalstockstheregionofpartialintermediation.Figure4showshowproductivityandbubblesa¤ectthesupplyanddemandoffundsforin-vestmentand,thus,thelawofmotion.Highrealizedproductivityraisesthesupplyoffunds,whilehighexpectedproductivityraisescollateralandthedemandforfunds.Bubblereturnshocksareembeddedinthesizeofthebubble.Thehigheristhesizeofthebubbletheloweristhesupplyoffunds.Highrealizedbubblecreationraisesthesizeofthebubbleandlowersthesupplyoffunds,buthighexpectedbubblecreationraisescollateralandthedemandforfunds.Theseintuitionsarecrucialtounderstandtheresultsthatfollow.Toconstructequilibriaforthiseconomy,wepropose rstajointstochasticprocessforbubblereturnandcreationshocks:fut;ntgforallt.Naturally,thisbubbleprocessmustbesuchthatEtut+1=0andnt0forallt.Weadopttheconventionthatb1=0andsetn0toensurethatb0takesthedesiredstartingvalue.Withthisstochasticprocessathand,wedetermineallpossiblesequencesforthestatevariablesfkt;btgusingEquations(14)-(16).Ifallthesesequencesaresuchthatkt0andbt0,wesaythattheproposedbubbleprocessisanequilibrium.Otherwise,wesaythattheproposedbubbleprocessisnotanequilibrium.Thebubblelessequilibriumisthatoneinwhichthebubbleprocesstakesthisform:fut;ntg=f0;0gforallt.Thebubblelessequilibriumalwaysexistsanditisausefulreference.Butthereareadditionalbubblyequilibriaandthesewe ndmoreinteresting.1.4DiscussionBeforemovingon,wewriteafewwordsaboutthenatureofbubblesforthosereadersthatremainskepticalaboutthismodelingstrategy.Convincedreaderscanskipthissection.Abubbleisnothingbutapyramidscheme.Participantsinapyramidschememakeavoluntarycontributiontotheschemethatgivesthemrighttothenextvoluntarycontribution.Eventhoughthismightseemquiteabstractorexoticat rstsight,itiseasyto ndreal-worldsituationsthatcorrespondfairlywelltothisconcept.Consider,forinstance,creditgiventoa rminexcessofthenetpresentvalueofthecash-owsthatthis rmwillgenerate.Thiscreditisoftenlabeledexcessive.Butthisexcessivecreditisrationalifcreditorsexpectthe rmtoobtainexcessivecreditalsointhefuture.Thus,excessivecreditcanbeinterpretedasavoluntarycontributiontothe rmthatgivestherighttothenextvoluntarycontribution.Thisleadsustothinkofreal-world rmsasportfoliosofcapitalandbubbles.Andthisisexactlywhattheentrepreneursinourmodelstand12 for.Naturally,thesaversinourmodelstandforthosethatrationallyprovidecredittoreal-world rms.Letustrytopursuethisviewabitfurtherandseehowwecouldre-interpretourmodelintermsof rmsandthestockmarket.Productionandinvestmentmusttakeplacewithin rmsthatareownedandmanagedbyentrepreneurs.Youngentrepreneurspurchasepre-existing rmsinthestockmarketorcreatenewonesatzerocost.LetVtdenotethevalueofall rms,i.e.thevalueofthestockmarket.To nancetheiractivities,entrepreneurssellcreditcontractsLt=It+Vt,whereItisinvestmentandthecapitalstockevolvesasfollows:Kt+1=It+(1)Kt,whereisthedepreciationrate,whichwehaveassumedtobeoneforsimplicity.Entrepreneurscanonlypledgetotheircreditorsafractionofpro tsandthere-salevalueoftheir rms.Asaresult,thefollowingcreditconstraintapplies:Rt+1Lt[F(Kt+1;Nt+1)Wt+1Nt+1]+Vt+1.Underthisre-interpretation,themodelshowsthattherearemanyequilibriainwhich rmpricestakethisform:Vt=(1)Kt+Bt,where(1)KtandBtarethefundamentalandbubblecomponentsof rmprices.9Inthisre-interpretationofthetheory,thereisnomarketforbubbles,insteadthereisastockmarketinwhichportfoliosofcapitalandbubblesaretraded.Notethat,inthecaseoffulldepreciationthatwefocuson,wecandirectlyre-interpretthemarketforbubblesasthestockmarket.Thus,itisstockmarketbubbles(ormoregenerally,bubblesin rmvaluesevenifthesearenottradedinorganizedstockexchanges)thatfeedcreditboomsandbusts!Aninterestingissuethatarisesinthisre-interpretationofthemodelreferstothenatureofbubblecreation.DibaandGrossman(1988)arguedthat,ifabubbleexistsonastockor rmprice,thenitstartedonthe rstdateoftradingofthe rm.Toreachthisconclusion,theynotedthatbubblecreationatanydateafterthe rstdateoftradingwouldinvolveaninnovationinthe rm 9Thisassumesthatthepriceofcapitalisone.Thisisthecaseifoldandnewunitsofcapitalareindistinguishableandinvestmentisalwaysstrictlypositive.Ifdi¤erentcapitalvintagesweredistinguishable,forinstance,therecouldalsobeabubblecomponentinthepriceofcapital(s).Weignorethiscomplicationhere.13 price.Butthisinnovationmusthavehadazeroexpectedvalueifitwasnotpricedinthe rstdateoftrading.Uptohere,theirargumentindicatesthatexpectedbubblecreationafterthe rstdateoftradingofa rmmustbezero.DibaandGrossmandidnotstophereandnotedalsothatinsomepopularsetupsfreedisposalof rmsrulesoutnegativebubbles.Inthesesetups,thereisanevenstrongerrestrictiononbubblecreation.Ifnewbubblesmustbenonnegativeandtheirexpectedvaluemustbezeroafterthe rstdateoftrading,thenbubblecreationcanonlytakeplaceinthe rstdateoftrading.Sincewestudythroughoutequilibriainwhichexpectedbubblecreationispositive,itmightbeusefultoseehowwegoaroundtheDiba-Grossmanargument.Asthem,weassumefree-disposalandruleoutnegativebubblesastheydo,BNt0.Butonceweknowthatinourmodelthemarketforbubblesisnothingbutthemarketfor rms,Vt=Bt,itisstraightforwardtore-interpretEquation(5)(whichwereproducehereforconvenience)intermsof rmprices:Bt=RBtBt+BNt,Btisthepriceofallthe rmsintheeconomyanditincludesoldorpre-existing rms,RBtBt;andnewlycreated rms,BNt.Theargumentthatbubblecreationmustbezeroafterthe rstdateoftradingimpliesEtRBt+1=EtRt+1.Andindeed,thisiswhatEquation(10)says.ButtheDiba-Grossmanargumentdoesnotimposeanyrestrictiononthesizeofthebubbleinthe rsttradingdateof rms.ThisiswhyEtBNt+10ispossible.Positiveexpectedbubblecreationafterthe rstdateoftradingisalsopossibleoncesomeoftherestrictiveassumptionsusedbyDibaandGrossmanarerelaxed.Butprovingthiswouldtakeustoofara eldanditisnotneededhere.2BubblyequilibriaEntrepreneurswishtopurchasebubblesifthesegrowatleastasfastastheinterestrate.Entre-preneurscanraiseenoughfundstopurchasebubblesifthesedonotgrowfasterthantheeconomy.Thus,bubblyequilibriaarepossibleifandonlyiftheinterestratedoesnotexceedthegrowthrate.Itiswellknownthatthissituationarisesifthefrictionlesseconomyisdynamicallyine¢cient.Inthiscase,theinterestrateislowbecausethesupplyoffundsishighandtheeconomyoverinvests.Bubblesreduceunproductiveinvestmentandthisallowstheeconomytosustainahigherlevelof14 consumptionandwelfare.Wedonotwanttofocusonthiscasehowever.10Itissomewhatlessknownthat,evenifthefrictionlesseconomyisdynamicallye¢cient,theinterestratemightfallbelowthegrowthrateif nancialfrictionslimitthestockoffundamentalcollateral.Inthiscase,theinterestrateislowbecausethedemandforfundsislowand,ifanything,theremightbeunderinvestment.Thisisindeedthecasewefocusonhere.Throughout,weassumethat 1.Thisconditionensuresthatindividualsarepatientenoughtosaveeveniftheinterestrateisbelowthegrowthrate.Withoutthiscondition,theinterestratewouldneverfallshortofthegrowthrateandbubblyequilibriawouldnotbepossibleinourmodel.2.1BubblybusinesscyclesItseemsnaturaltostartwiththecaseinwhichtherearenoproductivityshocks,i.e.At=A.Wealsoadoptthefollowingassumption:1 21 1+.(17)The rstinequalityensuresthatthefrictionlesseconomyisdynamicallye¢cientandinvestmentisalwaysproductive.Thus,bubblyequilibriaexistonlyifthedemandforfundsisdepressed.Thesecondinequalityensuresthatthisisindeedthecase.Thatis,collateralisscarceandthislowersthedemandforfundsenoughtomakebubblyequilibriapossible.Theexistenceofbubblyequilibriaofthissortdoesnotimplythatthereisunderinvestmentinthebubblelessequilibrium.If1 1+ 1 1+,thesteadystateofthebubblelessequilibriumliesintheregionoffullintermediation.Theinterestrateislowenoughtomakebubblyequilibriapossible,butnotlowenoughtoleadtounderinvest-ment.If1 21 1+ ,thesteadystateofthebubblelessequilibriumliesintheregionofpartialintermediation.Inthiscase,theinterestrateislowenoughnotonlytomakebubblyequilibriapossible,butalsotoleadtounderinvestment. 10Onereasonisthatdynamicine¢ciencyisbelievedtoberareinpractice(seeAbeletal.1988),althoughthishasbeenrecentlychallengedbyGeerolf(2012).Anotherreasonisthat,inthiscasebubblyequilibriaarealwaysassociatedwithdeclinesininvestmentandoutput,whichseemscounterfactual.15 Letusconsider rstaquietbubbleprocess:fut;ntg=f0;ngforallt.Thisdeterministicbubbleisanequilibriumifnislowenough.Fromanyinitialcondition,theeconomyconvergestothefollowingsteadystate:k=min1 A(k)b ;[(+n)A]1 1,(18)b= nA(k) (+n)A(k)1.(19)Equation(18)and(19)jointlydeterminethesteadystatevaluesforthecapitalstockandbubble.Onecanthinkofnasanindexofthesizeofthebubble.Ahighervalueofnleadstoahighersteadystatebubble,foratleasttworeasons.The rstanddirectoneisthatahighervalueofnraisesthevalueofnewbubblesthatarecreatedineachperiod.Thesecondandmoreindirectreasonisthatahighervalueofnraisestheinterestrate,therebyraisingtherateatwhichthevalueofoldbubblesgrows.Figure5showsthee¤ectsofnonthesteadystatevaluesofcapital,thebubble,credit,theinterestrateandutility.Byutilitywemeanthesumofutilitiesofsaversandentrepreneurs,i.e.Ut=PiUit.11TheFigureshowsthecaseinwhichthesteadystateofthebubblelessequilibriumislocatedintheregionofpartialintermediationandthereisunderinvestment.Thee¤ectsofthesizeofthebubbleonthecapitalstockandutilityhaveaninverseU-shape.Smallbubblesraisecreditandinvestment.Thiscrowding-ine¤ectofbubblesraisesthecapitalstockandwelfare.Butifthebubblesaretoolarge,theeconomyenterstheregionoffullintermediationandthee¤ectsofbubbleschange.Largebubblesraisetheinterestrateanddisplaceproductiveinvestment.Thecrowding-oute¤ectreturns.Thelargerthebubblethelargeristhiscrowdingoute¤ect.Indeed,ifthebubbleislargeenough,utilityislowerthanifthereisunderinvestment.Thereisthereforeanoptimalbubblethatprovidestheamountofcollateralthatmaximizesthecapitalstockandutilityinthesteadystate.Thisbubbleistheonethatplacestheeconomyatthefrontierofthepartialandfullintermediationregions,eliminatingunderinvestmentwithoutraisingtheinterestrate.Asmallerbubblewouldprovidetoolittlecollateralandthefundsavailableforinvestmentwouldbetoolow.Alargerbubblewouldraisetheinterestrateandthefundsavailableforinvestmentwouldalsobetoolow.Thequietbubbleprovidesthebasicintuitionforwhatistocomelater.But,ifuctuations 11Fornow,thereadershouldfocusonlyonthesolidlinesinFigures5,6,and7.Thedashedlines,whichdepictthebehaviorofthesesamevariablesunderpolicyinterventions,willbeanalyzedinsection3.16 increditaredrivenbybubblesintherealworld,thesebubblesareeverythingbutquiet.Thus,weconsidernextaneconomythatexperiencesbubblyepisodes.Inparticular,weassumethattheeconomystartsinthefundamentalstatewherebt=nt=0.Theeconomycantransitiontooneofmanybubblystateswhichweindexbyj=1;:::;J.Let"jbetheprobabilityofreachingbubblystatejfromthefundamentalstate.Whentheeconomyabandonsabubblystate,itreturnstothefundamentalstate.Letjbetheprobabilitythattheeconomyreachesthefundamentalstatefrombubblystatej.Werefertotheintervalthatgoesfromthe rstperiodtheeconomyreachesbubblystatejtothe rstperiodinwhichitreturnstothefundamentalstateasabubblyepisodeoftypej.Letztbethestateoftheeconomy,i.e.zt2fF;B1;:::;BJg.Duringabubblyepisodeoftypej,wehavethatnt=njandut+1=8]TJ ; -1; .63; Td; [00;:[+(1j)nj]Ak1t+1j 1jifzt+1=Bj[+(1j)nj]Ak1t+1ifzt+1=F.(20)Theassumptionthatbubblecreationisconstantisjustforconvenience.Theassumptionaboutbubblereturnssaysthat,duringabubblyepisode,theexpectedreturntoholdingthebubbleequalsthatofcreditcontracts.Conditionalonthebubblyepisodenotending,thereturnonthebubblewillexceedthatofcreditcontracts.Butthisisonlycompensationfortheriskofthebubblyepisodeendingandthebubblebursting.12Withtheseassumptions,thedynamicsoftheeconomycanbewrittenasfollows:kt+18]TJ ; -1; .63; Td; [00;]TJ ; -1; .63; Td; [00;:min1 Akt;[(+"n)A]1 1ifzt=Fmin1 Aktbt ;[(+(1j)nj)A]1 1ifzt=Bj,(21)bt=8]TJ ; -1; .63; Td; [00;:0ifzt=F(+(1j)nj)Ak1t (1j)bt1+njAktifzt=Bj,(22)where"n=Pj"jnj.Equation(21)describesthebehaviorofthecapitalstock.Itsevolutionineachperioddependsonwhetherinvestmentisdeterminedbythesupplyordemandforfunds.Itisworthmentioningthattheprobabilityofabubblyepisodestartingraisesthedemandfor 12Naturally,theeconomywiththequietbubbleanalyzedinFigure3isthelimitingcaseoftheeconomywithbubblyepisodesinwhichthereisabubblystatewith"j!1,j!0andnj=n.Theeconomyentersthisstateatthebeginningoftimewithprobabilityoneandneverleavesit.17 fundsinthefundamentalstate.Thus,thebehavioroftheeconomyinthefundamentalstateisnotequivalenttothatofthebubblelessequilibrium.13Equation(22)describestheevolutionofthebubbleduringanepisodeoftypej.Notethatthegrowthofthebubbledependsonhowlargeandriskyitis,i.e.njandj.Thelargerandriskier,thefasteritgrows.Weshownextsomesimulations.Toproducethem,wechooseparametervaluesthatensurethatthesteadystateofthebubblelessequilibriumislocatedintheregionofpartialintermediation.Then,wecreatefourbubblystatesthatdi¤erintermsofthesizeofthebubbleanditsrisk: SizenRisk safe risky small (nS;S) (nS;R) large (nL;S) (nL;R) wherenSnLandSR.Wealsoassumethatallthesebubblesareequallylikely,i.e."j=".Finally,wegenerateasequenceof10,000valuesforzt,andstarttheeconomyinthesteadystatesothatwerecord10,000periodsofsteadystatebehavior.Figure6plotsthesimulatedtimeseriesofkt,EtRt+1,lt,Utandbtduringawindowof200periods.TheFigurealsoshowsthemeanandstandarddeviationofthesevariablescomputedwiththeentireseriesof10,000periods.Giventhelengthofthesimulation,thesesamplestatisticsareexcellentapproximationstotheircorrespondingtheoreticalcounterparts.The gureillustratesthat,dependingontheircharacteristics,somebubblescrowd-incapitalwhileotherscrowd-outcapital.Takethe rstbubbleinthesample:inthiscase,thebubbleissmallandsafeanditraisestheeconomyscapitalstock.Theinterestrateremainsconstantat1throughouttheepisode,whichcon rmsthatthereisnotenoughbubblycollateraltofullyintermediatewages.Contrastthesee¤ectswiththoseofthenexttwobubbles,whicharelarge.Initially,bothbubblesraisetheeconomyscapitalstock:afterawhile,though,theybothbecometoolargeandreducecapitalaccumulation.Theinterestrisesabove1duringbothepisodes,con rmingthatthereistoomuchbubblycollateral.The gurealsoillustratesthatriskierbubblesgrowfasterbecausetheymustcompensatepotentialbuyersfortheirriskofbursting.Toseethis,comparethetwoepisodesaroundperiods8030and8040.Bothbubblesareofthesamesizebutthe rstoneisriskier:thus, 13Wehaveassumedthattheprobabilityandtypeofbubblyepisodesareconstantovertimeandthereforeexpectedbubblecreationisconstantwhentheeconomyisinthefundamentalstate,i.e."n.Thisassumptionisnotnecessary.Wecouldhaveassumedinsteadthatthereareuctuationsinexpectedbubblecreationinthefundamentalstate.Inthiscase,eveninthefundamentalstatetheeconomywouldbesubjecttochangesininvestorsentiment.18 eventhoughthe rstepisodelastslessthanthesecondone,themaximumsizeofthebubbleislarger.Oneimplicationofthisisthat,conditionalonnotbursting,riskierbubblesbecomelargerandhavemoreseverecrowding-oute¤ects.Theseexamplesillustratetwokeyaspectsofcreditbubbles.The rstoneisthattheyhavemixede¤ectsontheeconomy.Bubblecreationprovidescollateraltothecurrentgenerationandthisallowsittoraisecreditandexpandinvestment.Butbubblecreationalsorequiresfuturegenerationstocancelthisadditionalcreditandthisdivertssomeoftheirresourcesawayfrominvestment.Inthisregard,wecanthinkofthecrowding-oute¤ectofbubblesasanoverhange¤ect,inthesensethatgrowthishurtbybubblesthatwerecreatedinthepast.Thesecondpointisthatcreditbubblescanfollowverydi¤erentpaths,potentiallymuchmorecomplicatedthantheonesexploredhere,andtheymayhaveverydi¤erente¤ectsonoutputandwelfare.Thisnaturallysuggestsaroleforpolicyinselectingorreplicatingadesirableequilibrium.Beforegoingthere,though,weneedtodealwithanotherlayerofcomplexity:real-worldeconomiesarenotonlysubjecttobubbleshocksbutalsotofundamentalones.2.2BubblyandrealbusinesscyclesLetusbringbackproductivityshocks.Inparticular,assumethatAtcantaketwopositivevalues:At2fAL;AHgwithALAH.Thetransitionprobabilityis0:5,i.e.Prob(At+16=At)=.Wede neEALEtfAt+1=At=ALgandEAHEtfAt+1=At=AHg,andnotethatALEALEAHAH.Then,weadoptthefollowingassumption:AH AH+EAHAL AL+EAL.(23)ThisassumptionconvergestothatinEquation(17)asproductivityshocksbecomesmall.The rstinequalityensuresthatthefrictionlesseconomyisdynamicallye¢cientand,asaresult,investmentisalwaysproductive.Thesecondinequalityensuresthat nancialfrictionsarestrongenoughtomakebubblyequilibriapossible.Consider rstthebubblelessequilibriumwithproductivityshocks.Theeconomygoesthroughperiodsofhighandlowproductivity.Periodsofhighproductivityarecharacterizedalargesupplyoffundsandbyahighexpectedproductivity,whichraisesthedemandforfunds.Theformerlowerstheinterestratewhilethelaterincreasesit(seeFigure4).Therelativestrengthofthesee¤ectsdependsonthepersistenceofproductivityasmeasuredby,andtheimportanceofdiminishing19 returnstoinvestmentasmeasuredby.If0:5,thee¤ectofproductivityonthesupplyoffundsdominates.Inthiscase,whenproductivityishightheinterestrateislowandthegapbetweentheactualandtheoptimallevelsofinvestmentishigh.Ifinstead0:5,thee¤ectofproductivityonthedemandforfundsdominatesandtheoppositeistrue.Toillustratehowproductivitya¤ectstheoptimalbubble,weextendtheeconomywithbubblyepisodestointroduceproductivityshocks.Tosimplify,weassumethattransitionprobabilitiesforproductivityandbubbleshocksareindependent.Withtheseassumptions,wehavethatthedynamicsoftheeconomyarenowgivenby:kt+1=8]TJ/;༕ ;.9; ;Tf 2;.35; 0 ;Td [;]TJ/;༕ ;.9; ;Tf 2;.35; 0 ;Td [;]TJ ; -1; .63; Td; [00;]TJ ; -1; .63; Td; [00;:min1 Atkt;[(+"n)EtAt+1]1 1ifzt=Fmin1 Atktbt ;[(+(1j)nj)EtAt+1]1 1ifzt=Bj,(24)bt=8]TJ ; -1; .63; Td; [00;:0ifzt=F(+(1j)nj)Et1Atk1t (1j)bt1+njAtktifzt=Bj.(25)Equations(24)and(25)aresimplegeneralizationsofEquations(21)and(22)toallowforproduc-tivityuctuations.Figure7showssimulationsforthiseconomy.Asbefore,thesimulationsstartinthesteadystateandrunfor10,000periods.Toproducethem,weusethesamemethodologyasintheprevioussection,exceptthatwenowhaveaprocessforproductivity.Inchoosingtheparametersforthisprocess,wehaveassumedthat0:5.Asmentioned,thisparameterconditionimpliesthatthebubblethatfullyintermediateswagesgrowswithproductivity.ThisiswhyinFigure7bubblesraisethecapitalstockwhenproductivityishighbutreduceitwhenproductivityislow.Thisisbestexempli edbythelongepisodethatstartsafterperiod6960andlastsuntilafterperiod7000,inwhichthebubbleislargeandsafe.Thecapitalstockincreasesduringtheepisodebutituctuateswithproductivity.Whenproductivityishigh,thecapitalstockgrowsbecausewagesarehighandthelargebubblechannelspartofthesewagestoinvestment.Whenproductivityfalls,however,thecapitalstockscontractsfortworeasons:notonlyarewageslower,butthelargebubbledivertspartofthemawayfrominvestment.Theevolutionoftheinterestratecon rmsthisnarrativesince,byrisingabove1wheneverproductivityislow,itindicatesthatthereistoomuchbubblycollateralandinvestmentisbeingcrowdedout.Inthisexample,theequilibriumbubbleistoosmallwhenproductivityishighbutitistoolargewhenproductivityissmall:anoptimalbubble20 wouldthusbeprocyclical,amplifyingtheresponseofoutputtoproductivityshocks.14Thegeneralinsightofthisexampleisthattherelativestrengthofcrowding-inandcrowding-oute¤ectschangeswiththestrengthoffundamentals.Thesameamountofbubblecreationtoday,forinstance,mayhavesmallorlargecrowding-oute¤ectsinthefuturedependingontheevolutionofproductivity.Iffutureproductivityishighandtheeconomyisintheregionofpartialintermedia-tion,thecrowding-oute¤ectofcurrentbubblecreationwillbesmall.Ifinsteadfutureproductivityislowandtheeconomyisintheregionoffullintermediation,thecrowding-oute¤ectofcurrentbubblecreationwillbehigh.Intheterminologyoftheprevioussection,bubblyepisodesmayhavelargeoverhange¤ectsshouldtheeconomysu¤erfromafallinproductivity.Takentogether,theexamplesofthelasttwosectionsshowthatbubblesmayexpandtheecon-omybyprovidingcollateral,buttheymayalsolowerinvestmentandoutputiftheygrowtoolarge.Canpublicpolicybeusedtoprovidetheeconomywiththedesiredamountofbubblycollateral?Weshowinthissectionthattheanswertothisquestionisa¢rmative.3AcreditmanagementagencyWeconsidernextanagencythatmanagesbubblycollateralthroughcredit-marketinterventions.Subsidizingoldentrepreneurs,thisagencyexpandstheircollateral.Taxingoldentrepreneurs,thisagencyreducestheircollateral.The scalimplicationsoftheseinterventionsarebornebyyoungentrepreneurs.Althoughtheseinterventionsdonotdirectlya¤ecttheequilibriumvalueofthebubble,theyenabletheagencytoselectanequilibriumallocation.Werefertothisagencyasacreditmanagementagency(CMA),sinceitguaranteesthecollateralofentrepreneursatthetargetedlevel.Ifex-postthebubbleturnsouttobetoosmallgiventhetarget,theCMAprovidesadditionalresourcestoentrepreneurs.Ifex-postthebubbleturnsouttobetoolargegiventhetarget,theCMAtakesawayresourcesfromthem.Beforeturningtotheanalysis,though,itisimportanttoponderbrieyonwhatitmeanstodesignpolicyinourenvironment.Thecompetitiveequilibriumisdrivenbyinvestorsentimentorexpectations,ascapturedbytheprocessfnt;utg.Thisposesaproblemfortheevaluationofpolicy,sincetheimplementationofagivenpolicymayinuenceexpectationsthemselvesandchangethe 14Moregenerally,thecyclicalityofthebubblethatattainsfullintermediationdependsontheshockthatiscon-sidered.Wehavealsoexplored,forinstance,aspeci cationofthemodelwith nancialshocks,i.e.withrandomuctuationsin.Theseshocksa¤ectthedemandforfundsbutnottheirsupplyand,asaresult,collateralispro-cyclical.Thismeansthat,tointermediatewagesfully,abubblemustbecounter-cyclicalwithrespectto,sothatitreducestheresponseofoutputto nancialshocks.21 underlyingprocessfnt;utg.Weavoidthisproblembyfocusingonpoliciesthatareexpectationallyrobust,inthesensethattheyimplementthesameallocationsregardlessofagentexpectationsorinvestorsentiment.3.1ThemodelwithacreditmanagementagencyWeintroducenowaCMAthatpromisesentrepreneursinperiodtatransferofStunitsoftheeconomysconsumptiongood.ThetransferStcanbenegativeandcontingentonthestateoftheeconomy.Ifthetransferispositive,itscostis nancedbyyoungentrepreneurs.Ifthetransferisnegative,itsbene tisdistributedtoyoungentrepreneurs.LettingXtdenotetheagencystaxrevenues,werequirethattheCMArunabalancedbudgetsothatSt=Xtineveryperiodt.De neSNt+1asthenetresourcesthatthispolicyprovidestotheentrepreneursofgenerationtintermsofgoodsinperiodt+1:SNt+1St+1EtRt+1 Xt.(26)Thatis,SNt+1isthedi¤erencebetweenthesubsidiesthatwillbeobtainedinoldageandthetaxespaidinyoungage,bothexpressedintermsofgoodsinperiodt+1.Insofarasexpectedsubsidiescanbepledgedbyentrepreneurstocancelcreditcontracts,thisisawealthtransferthata¤ectscollateral.IfSNt+10,thepolicycreatescollateralinperiodt.IfSNt+10,thepolicydestroyscollateralinperiodt.Thus,werefertoSNt+1ascollateralcreationbytheCMA.UsingthisnotationandthebudgetconstraintoftheCMA,wecanexpressthedynamicsofthepolicyasfollows:St+1=EtRt+1 St+SNt+1.(27)TheformalsimilaritybetweenEquations(27)and(5),respectivelydescribingthedynamicsofpolicyandthebubble,hintsatsomeoftheresultsthatfollow.TheintroductionofaCMAdoesnota¤ecttheproblemofsavers,butitdoesa¤ecttheproblemofentrepreneurs.Sinceentrepreneurscanusesubsidiestocancelcreditcontracts,thecollateralconstraintinEquation(6)nowbecomesRt+1Lt[F(Kt+1;Nt+1)Wt+1Nt+1]+Bt+1+St+1,(28)22 whiletheirconsumptionsarenowgivenbyCit;t=LtKt+1BtXt,(29)Cit;t+1=F(Kt+1;Nt+1)Wt+1Nt+1+Bt+1+St+1Rt+1Lt.(30)Theonlydi¤erencesbetweenEquations(29)-(30)andEquations(7)-(8)areduetothetransfersthataremadebytheCMA.BorrowersmaximizeutilityinEquation(1)subjecttothebudgetconstraintsinEquations(29),(30),thecreditconstraintinEquation(28)andthelawsofmotionofbubblesandpolicyinEquations(5)and(27).TheintroductionofaCMAdoesnota¤ecttheequilibriuminthemarketsforlabororbubbles,whicharestillrespectivelydescribedbyEquations(9)and(10).Itdoesa¤ecttheequilibriuminthecreditmarket,though.WhilethesupplyofcreditisstilldescribedbyEquation(11),thedemandoffundsforinvestmentbyyoungentrepreneurscanbewrittenasfollows:EtRt+1=8]TJ ; -1; .63; Td; [00;:At+1k1t+1ifEtbNt+1+sNt+1(1)EtAt+1kt+1EtAt+1kt+1+Etbt+1+Etst+1 kt+1+ 1(bt+st)ifEtbNt+1+sNt+1(1)EtAt+1kt+1.(31)Asbefore,Equation(31)saysthattheinterestrateiseithertheexpectedmarginalproductofcapitalortheratioofentrepreneursexpectedcollateraltototalcredit.Theonlynoveltyisthattheconstraintisa¤ectedbytheCMAspolicies:expectedsubsidiesEtst+1providecollateraltoentrepreneursandtheyraisethedemandforcreditandthusinvestment,butcurrenttaxesstdiverttheresourcesofentrepreneursandtheyreducetheshareofcreditthatcanbeusedforinvestment.Tocompletethesolution,weneedtodeterminehowentrepreneursusethecredittheyreceive,andthisisasfollows: kt+1+bt+st8:=ltifEtAt+1k1t+1]TJ ; -1; .63; Td; [00;12[0;lt]ifEtAt+1k1t+1=1.(32)Equation(32)saysthatentrepreneursconsumeinthe rstperiodonlyifthemarginalproductofcapitalequals1.Otherwise,entrepreneursuseallthecredittheyreceivetoinvest,purchasebubbles,andpaytaxes.23 3.2EquilibriumdynamicsWecannowcollapsethemodelwithaCMAintothefollowingfourequations:kt+18]TJ ; -1; .63; Td; [00;]TJ ; -1; .63; Td; [00;:=(1)Atktbtst ifEtRt+1120;(1)Atktbtst ifEtRt+1=1,(33)EtRt+1=minfEtAt+1;Etf(+nt+1+mt+1)At+1ggk1t+1,(34)bt+1=EtRt+1+ut+1 bt+nt+1At+1kt+1,(35)st+1=EtRt+1 st+mt+1At+1kt+1.(36)wheremt+1measurescollateralcreationbytheCMAasashareofoutput,i.e.mt+1sNt+1=(At+1kt+1).Weshallthinkofmtasthekeypolicyinstrumentinthemodel,whichisdeterminedbytheCMA.Equations(33)-(36)provideafulldescriptionofthedynamicsofthestatevariables,i.e.kt,btandst;foranyadmissiblesequenceofproductivityandbubbleshocks,i.e.At,utandntandforanysequenceofpolicyinterventionsmt.Thus,werefertothesefourequationsasthelawofmotionofthesystemwithpolicy.Equations(33)-(36)showhowpolicyshocksa¤ectthesupplyanddemandoffundsforinvest-mentandthusthelawofmotion.Likebubbleshocks,pastandpresentpolicychoicesascapturedbym0;m1;:::;mtareembeddedinsubsidiesst.Throughtaxation,thesepolicychoicesreducethepresentsupplyoffundsavailableforinvestmentasdepictedinEquation(33).Simplyput,youngentrepreneursmustdevotepartoftheircredittopayingforthesubsidiespromisedtothepreviousgeneration.Futurepolicychoicesascapturedbymt+1insteadraisethecollateralofentrepreneurs,asshownbyEquation(34),enablingthemtoexpandtheirdemandforfundsandtheirinvestment.OnceaCMAisintroducedintothemodel,anequilibriumcannotbede nedwithoutspecifyingpolicy.Throughout,wetakepolicyasgivenanddescribeitbyastochasticprocessfmtg.Weadopttheconventionthats1=0sothats0isdeterminedbym0.Wethenproposeajointstochasticprocessforbubblereturnandbubblecreationshocks:fut;ntg,withEtut+1=0,bt0andnt0forallt.Withthisprocessathand,wesearchforasequenceforthestatevariables(kt;bt;st)thatsatis eskt0aswellasEquations(33)-(36).Ifsuchasequenceexists,wesaythattheproposedstochasticprocessisanequilibrium.Ifsuchasequencedoesnotexist,wesaythattheproposedstochasticprocessisnotanequilibrium.24 3.3ManagingcollateralduringbusinesscyclesWhatcanaCMAachievebytaxingandsubsidizingcredit?A rstimportantresult,whichfollowsdirectlyfromcomparingEquations(33)-(36)withEquations(14)-(16),isthataCMAcanusepolicytoreplicateanyoftheequilibriaoftheoriginaleconomy.Takeanyprocessfut;ntgforthebubblereturnandcreationshocksandacorrespondingbubblebt.Then,theCMAcanreplicatetheequilibriumallocationthatwouldariseunderanalternativeprocessf^ut;^ntg;andthecorrespondingbubble^bt,bychoosingapolicyfmtgthatsatis esmtAtkt=^ut^bt1utbt1 +(^ntnt)Atkt(37)forallt0:Thispolicyguaranteesthatst+bt=^btandthatEtf(mt+1+nt+1)At+1g=Etf^nt+1At+1ginallperiods.Thus,thetotalcrowding-outandcrowding-ine¤ectsofthepol-icysttogetherwithbubblebtareidenticaltothoseofbubble^btintheabsenceofpolicy.Notethatthispolicyisexpectationallyrobustinthesensethatitworksforanyprocessfut;ntg.Thisresultisimportantbecausewehaveseenthatmarketsmayprovidetoolittlebubblycollateral,whichisinsu¢cienttointermediateallwages,ortoomuchofit,whichcrowdsoutinvestmentandreducesoutputandconsumption.Itthereforeseemsnaturaltoconsiderpoliciesthatmanagetheeconomyscollateral.Inparticular,weconsiderpoliciesfmtgthatsimultaneouslysatisfyEtfmt+1At+1g=1k1t+1Etf(+nt+1)At+1g,(38)kt+1=(1)Atktbtst ,(39)inallperiodst,andstudytheire¤ectsonsteady-stateconsumptionandwelfare.Equation(38)guaranteesthattheproposedpolicystabilizesEtRt+1andsetsitequalto1atalltimes.Equa-tion(39)saysthatthisstabilizationmustbeachievedwhileguaranteeingthatwagesarefullyintermediatedinequilibrium.Throughthispolicy,theCMAe¤ectivelyleansagainstthewind,taxingcreditwhencollateralisexcessiveandEtRt+11andsubsidizingitwhencollateralisscarceandthereispartialintermediation.Notethatthepolicydependsonlyonobservables,sinceitneedonlybecontingentonaggregateproductivityAtandonthebubbleshocksfut;ntg.Wecanillustratethee¤ectsofsuchapolicybyrevisitingtheexampleofaquietbubblyprocess,inwhichAt=Aandfut;ntg=f0;ngforallt.Wehavealreadyarguedthatthereisanoptimallevelofbubblecreationinthisexample,whichmaximizesoutputandwelfareinthesteady25 state.Thisbubble,whichplacestheeconomyatthefrontierofthepartialandfullintermediationregions,correspondsexactlytotheoneprescribedbyourpolicyrule.Thus,ifweusentodenotesuchanoptimallevelofbubblecreationandfollowEquation(37),apolicythatsetsm=nninallperiodsattainsthemaximumlevelsofoutputandwelfareinsteadystate.ThedashedlinesinFigure5depictthesteadystatevaluesofcapital,credit,theinterestrate,utility,thebubbleandthesubsidysasafunctionofnundertheproposedpolicy.Inthiscase,allvariablesareindependentofn.Onceagain,thereasonisthattheCMAintervenesinthecreditmarkettomaximizeintermediationwhileminimizingcrowdingoute¤ects.Intheregionofpartialintermediationitdoessobysubsidizingcreditpaymentsbyoldentrepreneurs,i.e.bysettings0,whereasintheregionoffullintermediationitdoessobytaxingcreditpaymentsbyoldentrepreneurs,i.e.bysettings0untilEtRt+1=1.TheseinsightsareeasilyextendedtotheeconomywithbubblyepisodesthatwasanalyzedinSection2.1.Thisisthesameeconomyasbefore,withtheonlydi¤erencethatittransitionsbetweenfundamentalandbubblystatesaccordingtoasunspotvariablezt2fF;Bg.Thus,theoptimalbubbleisalsothesameasbefore,andthecorrespondingallocationcanbeimplementedbytheCMAbysettingmtAtkt=(nnt)Atktutbt1 ,inallperiods.ThedashedlinesinFigure6showthee¤ectsofthispolicyonthemainvariablesoftheeconomy.TheCMAsintervention,whichisnowstatecontingent,raisesthecapitalstockinallperiodsandfullystabilizesit.Wheneverbubblycollateralisscarceandtheeconomyisinthepartialintermediationregion,creditissubsidizeduntilitequalstotalwages.Whenevercollateralisabundantandtheinterestrateexceeds1,creditistaxeduntiltheEtRt+1=1.ThisexampleillustratesthattheCMAsinterventiononlya¤ectsthemarketforbubblesthroughitse¤ectontheinterestrate.Theeconomystillexperiencesbubblyepisodesandthepolicyrulehasnodirecte¤ectoverthem.Instead,theCMAmonitorstheseuctuationsandintervenesinthemarketforcreditaccordingly.Whenbubblesburst,theCMAstepsinandhelpsentrepreneurscanceltheircreditcontracts.Whenbubblesbecometoolarge,asinthesecondandthirdepisodesoftheexample,theCMAstepsinandtaxesentrepreneurs.Bydoingso,theCMAreducesthedemandforcredit,eliminatesthecrowdingoute¤ectandboostscapitalaccumulation.Ofcourse,noteveryoneishappywiththispolicy:generationswithalotofbubblecreationlosebecausepart26 oftheirwealthistaxedaway.Yet,asthetableinFigure6con rms,averagewelfareincreasesasaresultoftheintervention.Figures5and6depictthee¤ectofourpolicyruleineconomieswithnofundamentaluncertainty.Inthiscase,theruleisoptimalinthesenseofmaximizingaveragewelfareandoutputinthesteadystate.Theanalysisbecomesmorecomplicatedineconomieswithfundamentaluncertainty,though.Aswehavediscussed,alevelofbubblecreationthatboostsgrowthwhenproductivityishighmightleadtolargecrowding-oute¤ectswhenproductivityfalls.Thiscomplicatesthecharacterizationofanoptimalbubble.Nonetheless,weapplynextoursamepolicyruletothiseconomyandshowthatitalsoraisestheaveragelevelsofoutputandwelfareinthesteadystate.Figure7buildsontheexampleofSection2.3anddepictstheimpactofthepolicyruleinaneconomywithboth,productivityandbubbleshocks.Asbefore,thesolidlinesrepresentthevaluesofthedi¤erentvariablesintheabsenceofpolicywhilethedashedlinesrepresenttheirvaluesunderthepolicyrule.The gureshowsthat,oncemore,thepolicyrulestabilizestheinterestrateanditalsoraisestheaveragecapitalstockandwelfare.Thenewinsightisthatthepolicyhasanambiguouse¤ectoneconomicvolatility.Fluctuationsthataredrivenbybubbleshocksarestabilizedbythepolicy,whileuctuationsthataredrivenbyproductivityshocksareampli edbyit.Thereasonforthislastresultisthat,underourassumptions,thepolicyruleisprocyclicalrelativetoproductivityshocks.Whenproductivityishigh,investmentisconstrainedbythedemandoffunds.Wagesarehighintheseperiodsandcollateralisinsu¢cienttointermediatethemall.Byfollowingthepolicy,theCMAincreasesexpectedsubsidies,raisescreditandexacerbatestheboom.Whenproductivityislow,however,investmentisconstrainedbythesupplyoffunds.Inthesetimes,thereisnotmuchthatthepolicycandotoexpandoutputbesidestaxingborrowingtoreducecrowding-oute¤ects.Hence,thepolicyruleincreasesthestrengthofproductivity-drivenboomsrelativetoproductivity-drivenrecessions.Itsoveralle¤ectoneconomicvolatility,however,dependsontherelativevarianceofbubblevis-à -visproductivityshocks.Inourparticularexample,asthetableinFigure7shows,itincreasesthevolatilityofboththecapitalstockandofwelfare.3.4DiscussionTheexamplesofthissectionshowthatthebubblyeconomyprovidesanewrationaleformacropru-dentialpolicy.Whenbubblycollateralislowerthandesired,theCMAsupplementsitbyprovidingfundstoentrepreneurs.Whenbubblycollateralisinsteadhigherthandesired,theCMAreducesit27 bysubstractingfundsfromentrepreneurs.Thisleaning-against-the-windpolicycanreplicatetheoptimalbubbleandmaximizesteady-stateoutputandconsumption.Naturally,thispolicygeneratesresourceswhenbubblycollateralishighanditrequiresresourceswhenbubblycollateralislow.Underourassumptions,theselossesandgainsarebornefullybyyoungentrepreneurs.Thepolicycanthereforebeinterpretedasanentrepreneurialinsuranceorcounter-cyclicalfund,towhichentrepreneurscontributeduringyouthintheexpectationofreceivingatransferduringoldageiftheircollateralturnsouttobelow.Notehoweverthatthisinsuranceisnotactuariallyfair.IfEtSNt+10,itprovidesnetresourcestotheentrepreneursofgenerationt.IfEtSNt+10,itsubstractsnetresourcestotheentrepreneursofgenerationt.Couldthispolicyalsobeinterpretedasconventionalmonetarypolicy?Afterall,itfullysta-bilizestheinterestrateandsetsitequalto1.Butthisisonlyonepartofthepolicy,though.Therearemanypoliciesthatstabilizetheinterestrateatthatlevel(i.e.,theysatisfyEquation(38))withoutfullyintermediatingwages(i.e.,theydonotsatisfyEquation(39)).Thecrucialpartofthepolicyisthatitmanagescollateral,providingfundstoentrepreneurswhencollateralislowandtakingfundsfromthemwhencollateralishigh.Butperhapsourleaning-against-the-windpolicycouldbeinterpretedasunconventionalmon-etarypolicy.Inparticular,itcouldberecastasanassetpurchasescheme,notunliketheonesadoptedbyvariousgovernmentssincetheonsetoftherecent nancialcrisis.Underthisinterpre-tation,theCMApurchasescreditcontractsfromsaversatapriceofAtkt+bt+st.ItthencollectsAtkt+btinpaymentsfromoldentrepreneurs.Whenst]TJ/;༕ ;.9; ;Tf 1;.51; 0 ;Td [;0,thisinterpretationisverynatural:itmeansthattheCMApaystoomuchforassetsandsu¤ersalossequaltost.Bydoingso,though,itrelaxesentrepreneurialconstraintsandraisesborrowingexante.Whenst0,instead,thisinterpretationofthepolicyisabitlessnatural:itmeansthattheCMApaystoolittleforassetsandmakesapro tofst.Butwhywouldasaversellhercontractsatsuchalowprice?Indeed,thisinterpretationonlymakessenseiftheCMAhasthelegalabilitytoforcesuchsales.Thereisanotherinterestingaspectofinterpretingourpolicyasanassetpurchasescheme.Uptonow,wehaveassumedthatalllossesgeneratedbythepolicy(i.e.,st]TJ/;༕ ;.9; ;Tf 1;.51; 0 ;Td [;0)are nancedthroughtaxation.Inreality,though,theseschemesaremostly nancedbyissuinggovernmentsecurities.Aswenowargue,though,nothingchangesifweallowtheCMAto nanceitslossesbyissuingdebt.28 4The scalbackstopWehaveassumedsofarthattheCMAcanguaranteeanyamountofcollateralprovidedthatitdoesnotexceedsavings,i.e.theupperboundonXtisgivenbytheeconomysentiresavings.Thisisobviouslyanunrealisticassumption.WhathappensiftheCMAsabilitytotaxyoungentrepreneursislimited?Weshownextthatthesameconditionsthatmakeapolicydesirablealsoguaranteethatacreditmanagementagencyhasenoughresourcestoimplementit.4.1ThemodelwithdebtWeaddressthisquestionbyassumingthattheCMAhasataxationcapacity0,i.e.amaximumamountthatitcantaxpere¤ectiveworker.Clearly,thiscapacityconstraintdoesnotbindforpoliciesthatsatisfystinallperiods.Tosustainpoliciesthatrequiretransfersst-285;insomeperiods,though,theCMAneedstoissuecreditcontracts.Todistinguishthesecontractsfromthoseissuedbyentrepreneurswerefertothemaspubliccreditcontractsorpublicdebt.Wethenrefertocreditcontractsissuedbyentrepreneursasprivatecreditcontractsorprivatedebt.Publiccreditcontractsissuedinperiodtpayagross,possiblycontingent,amountDt+1unitsofoutputinperiodt+1.Weuseqttodenotetheperiod-tpriceofapubliccontractthatpromisestodeliver,inexpectation,oneunitofoutputinperiodt+1.Withthisnotation,wecanwritethebudgetconstraintoftheCMAasfollows:St+Dt=Xt+qtEtDt+1.(40)whereqtEtDt+1isthevalueofpubliccreditcontractsissuedinperiodt.Equation(40)saysthattotalspendingbytheCMAintransfersanddebtpaymentscannotexceedtotalrevenuesfromtaxationandthesalesofpubliccreditcontracts.Themainchangerelativetoourpreviousanalysisisthatnowwehaveanadditionalmarketforpubliccreditcontracts.Inthismarket,theCMAsuppliesdebtinelastically.Lendermaximizationimpliesthat,inequilibrium,thereturntopubliccreditcontractsmustequalthereturntoprivatelending,sothatqt=(EtRt+1)1.15Thisdoesnotimply,however,thattheex-postinterestrateonpublicandprivatecreditcontractscoincide.De neDNt+1measuresthenetresourcesthattheevolutionofpubliccreditprovidestothe 15ToseethisnotethattheconsumptionsoflendersarenowgivenbyCSt;t=WtLtqtEtDt+1andCSt;t+1=Rt+1Lt+Dt+1.Clearly,thedemandforpubliccreditcontractswillbeeitherzeroorin nityifqt6=(EtRt+1)1:29 entrepreneursofgenerationtintermsofgoodsinperiodt+1:DNt+1Dt+1EtRt+1 Dt.(41)Thatis,DNt+1isthedi¤erencebetweenthepublicdebtthatgenerationtleavesinoldageandthepublicdebtthatthisgenerationreceivesinyoungage,bothexpressedintermsofgoodsinperiodt+1.Itrepresentsawealthtransfertogenerationtanda¤ectsitscollateral.IfDNt+10,thepolicycreatescollateralinperiodt.IfDNt+10,thepolicydestroyscollateralinperiodt.Thus,werefertoDNt+1asadditionalcollateralcreationbytheCMA.WhileSNt+1isthecollateralcreationthatisdeterminedbytheevolutionofsubsidies,DNt+1isthecollateralcreationthatisdeterminedbytheevolutionofdebtandultimatelyoftaxes.4.2EquilibriumdynamicsFollowingstepsthatarenowfamiliar,wecanrewritethemodelwithaCMAasfollows:kt+18]TJ/;༕ ;.9; ;Tf 1;.98; 0 ;Td [;]TJ/;༕ ;.9; ;Tf 1;.98; 0 ;Td [;]TJ ; -1; .63; Td; [00;]TJ ; -1; .63; Td; [00;:=(1)Atktbtstdt ifEtRt+1120;(1)Atktbtstdt ifEtRt+1=1(42)EtRt+1=minEtAt+1;Et+nt+1+mSt+1+mGt+1At+1 k1t+1(43)bt+1=EtRt+1+ut+1 bt+nt+1At+1kt+1(44)st+1=EtRt+1 xt+mSt+1At+1kt+1andxt(45)dt+1=EtRt+1 dt+mGt+1At+1kt+1(46)wheremSt+1andmGt+1reectnowcollateralcreationthroughbothsubsidies/taxesandpublicdebt,i.e.mSt+1sNt+1=At+1kt+1andmGt+1dNt+1=At+1kt+1.AcomparisonbetweenEquations(42)-(43)and(33)-(34)showsthattheintroductionofdebtdoesnotsubstantiallya¤ectthesupplyanddemandoffundsforinvestment.Aslongass0t=st+dtinallperiods,anypolicywithsubsidiess0tthatisfully nancedthroughtaxationcanbeperfectlyreplicatedwithanalternativepolicywithsubsidiessts0tthatis nancedpartlythroughtaxationandpartlythroughdebt.Equations(42)and(43)alsoshowthat,likebubblesandsubsidies,theintroductionofdebt nancinghasconictinge¤ectsonintermediationandcapitalaccumulation.30 Allelseequals,debtissuedinthepastrequirestaxationanditreducesthesupplyoffundsavailableforinvestment.ThisiswhydtenterswithanegativesigninEquation(42).Debtissuedinthepresent,though,makesitpossibletoreducetaxesonthecurrentgenerationofentrepreneurstherebyenablingthemtoexpandtheirdemandforcredit.ThisiswhymGt+1enterswithapositivesigninEquation(43).Nowpolicyconsistsofasequenceforthetripletxt;mSt;mGt .Weadopttheconventionthats1=d1=0sothats0andd0isdeterminedbyx0;mS0andmD0.Wethenproposeajointstochasticprocessforbubblereturnandbubblecreationshocks:fut;ntg,withEtut+1=0,bt0andnt0forallt.Withthisprocessathand,wesearchforasequenceforthestatevariables(kt;bt;st;dt)thatsatis eskt0aswellasEquations(42)-(46).Ifsuchasequenceexists,wesaythattheproposedstochasticprocessisanequilibrium.Ifsuchasequencedoesnotexist,wesaythattheproposedstochasticprocessisnotanequilibrium.4.3Aretherelimitstodebt nancing?AquickanalysisofEquations(42)-(46)suggeststhatdebt nancingcanhelpachievepoliciesthatwouldnotbefeasibleotherwiseduetothe scalconstraintthatxt.Whatensures,though,thatthegovernmentcanissuepublicdebtcontracts?Thatis,whatarethelimitstodebt nancing?Thisquestionisrelevantbecauseanimportantdi¤erencebetweentaxesanddebtisthatpayingtaxesismandatorywhilepurchasingdebtisnot.Inanyperiodt,thedebtissuedbytheCMAisbackedbytaxesbutalsobytheexpectationofnewdebtthatwillbeissuedinperiodt+1.Ifthisexpectationchanges,somightthewillingnessofsaverstoholddebt.Thus,debt nancingseemssubjecttochangesininvestorsentimentthatmightlimittheCMAsabilitytomanagetheeconomyscollateral.Inshort,publiccreditmightbejustasbubblyasprivatecredit.Toseewhetherthisconcernisjusti ed,wefocusonthecaseinwhichtheCMAusesthepolicyrulederivedinSection3.3tointermediateallwageswhilesettingEtRt+1=1forallt.Weassumethatissmall,sothatdebt nancingisisrequiredtofundthedesiredsubsidies.Inequilibrium,themostthattheCMAcanpromisetorepayattimet+1isgivenbyEtdt+1+ EtfEt+1dt+2g,i.e.expecteddebtpaymentscannotexceedtheexpectedresourcesobtainedthroughtaxationandthroughtheissueofnewdebt.ButtheexpecteddebtpaymentsthattheCMAcanmakeattime31 t+2arethemselveslimitedbyexpectedtaxationandthesaleofnewdebtatt+3,andsoon.Iteratingthisprocessforwardweobtainanupperboundfordebtpayments:Etdt+11Xj=0( )j+limT!1h( )TEtdt+Ti.(47)Equation(47)providesanexpressionfortheCMAs scalbackstop.Itsaysthatexpecteddebtpaymentsarebackedbyfundamentalresources,whichcorrespondtothenetpresentvalueoftaxes,andbybubblyresources,whichcorrespondtothepublicdebtthatisexpectedtoberolledoverforever.Thekeyaspectofthisexpressionisthat,since 1,thefundamentalresourcesoftheCMAarein niteforanypositivevalueof,regardlessofhowsmallitis.Thus,eveniftheCMAmustissuedebtto nanceitspolicies,itcanalwaysdosowithoutbeingexposedtoroll-overcrises:thereasonisthat,nomatterhowlargedebtpaymentsare,theyarealwaysfullybackedbyfuturetaxes.Inthebubblyeconomy,the scalbackstopoftheCMAisunlimited!Theintuitionbehindthisresultisquitenatural.TheCMAs scalbackstopincludesallfuturetaxes,includingthosethatwillbepaidbygenerationsthatareyettobeborn.Thesetaxrevenuesgrowwiththeeconomy.Whencollateralisscarceandoutputisbelowpotential,though,thisgrowthrateishigherthanthemarketinterestrateandthenetpresentvalueoftaxrevenuesisunbounded.Thus,thesameconditionthatmakesbubblyepisodespossibleguaranteethat,eveniftheCMAmustissuedebtto nanceitspolicies,thisdebtisnotpronetoarollovercrisisbecauseitisfullybackedbytaxrevenues.5Whathavewelearned?Sowhatdowetakehomefromthispaper?Creditboomsandbustsareafactoflifeinmoderneconomies.Thereisawidespreadviewamongmacroeconomiststhatuctuationsincollateralareanimportantdriveroftheseuctuationsincredit.Thispaperbuildsonthisviewandmakesthefollowingobservations:1.Economieswithbindingborrowingconstraintshavetwotypesofcollateral,fundamentalandbubbly.Fluctuationsinbothtypesofcollateralgenerateboom-bustcyclesincredit,investmentandgrowth.Theseuctuationsmightbedrivenbytraditionalorfundamentalshocks,butalsobychangesininvestorsentimentormarketexpectations.Bothtypesofshocka¤ecttheamountsoffundamentalandbubblycollateral.32 2.Bubblycollateralraisesequilibriumcredit(crowding-in)butdivertspartofthiscreditawayfrominvestment(crowding-out).Whenbubblesaresmall,thecrowding-ine¤ectdominatesandinvestmentandoutputincrease.Whenbubblesarelarge,thecrowdingoute¤ectdominatesandinvestmentandoutputarelow.Thereisthereforeanoptimalbubblesizethattradeso¤thesetwoe¤ectstomaximizelong-termoutputandconsumption.3.Marketsaregenericallyunabletoprovidetheoptimalamountofbubblycollateral,butacreditmanagementagencycanreplicatetheoptimalbubbleallocationbytaxingcreditwhenbubblycollateralisexcessiveandsubsidizingitwhenitisinsu¢cient.Suchapolicycanbecharacterizedasleaning-against-the-wind,butnotasapolicyofpreventingbubbles.Itcannotbecharacterizedasabailoutpolicyeither,sincethecredit-marketinterventionsthatarerequiredtoimplementitpayforthemselves.Tosomeextent,wearguethatthispolicycanbeinterpretedasaschemeinwhichthecreditmanagementagencypurchasesassetstomanagetheeconomysstockofcollateral.Theseresultsprovideacoherentandrichviewofcreditboomsandbusts,inwhichbothfunda-mentalandbubblycollateralplayakeyrole.Theyalsoprovideausefulblueprinttoguidepolicyindealingwithcreditbubbles.Butthetheoryhaslimitationsandthereismuchworktobedone.A rstlimitationofthemodelisthatitdepictssituationsinwhichonlythetotalamountofbubblycollateralmatters.Buttherearesituationsinwhichitisnotjusttheamount,butalsothetypeofbubblycollateralthatisimporttant.Aneconomywith nancialintermediariesthatcollectdepositsandlendtoentrepreneurs,forinstance,willrequiredi¤erenttypesofbubblycollateraltobackdepositandloancontracts.Aneconomyinwhichsaversdi¤eraccordingtotheirriskaversion,inturn,mightrequiredi¤erenttypesofcollateraltobacksafeandriskycreditcontracts.Inbothcases,theremaybetoomuchcollateralofonetypebuttoolittlecollateraloftheothertype.Thus,toimproveonmarketoutcomes,theCMAmustsimultaneouslysubsidizesometypesofcreditwhiletaxingothers.16Asecondlimitationofthemodelisthatitconsidersaclosedeconomy.Butoneofthecharac-teristicsofrecentcreditboomsandbustsisthattheyhavebeenassociatedwithsurgesandsuddenstopsincapitalowstogetherwithlargeuctuationsinrealexchangerates.Itwouldbeinter-estingtodevelopanopen-economyversionofthemodelthatallowsustoinvestigatetheroleofcapitalowsandexchangeratesinshapingthesecreditboomsandbusts.Moreover,thisextension 16Wereferthereadertoanonlineappendixforamoreformaltreatmentoftheseissues.33 wouldallowsusalsotostudytheinternationaltransmissionofcreditbubblesandtheinternationalcoordinationofcreditmanagementpoliciesliketheoneswehaveanalyzedhere.17Athirdlimitationofthemodelisthateveryoneobservesthegapbetweentheoptimalbubbleandtheexistingone.Thus,theroleofpolicyissimplytobridgethisgap.Realityismorecomplicatedbecausemarketparticipantsandpolicymakersmaybeuncertainastowhetheructuationsaredrivenbyfundamentalorbubblycollateral.Introducingthistypeofuncertaintyisanimportantnextstepinthisresearchagenda.Afourthlimitationofthemodelisthatitessentiallytakesfundamentalcollateralasexogenous.Onemaythink,though,thatfundamentalcollateralisproducedbythe nancialsystemwhenitinvestsinthescreeningandmonitoringofentrepreneurs.Bubblycollateralprovidesaninexpensivealternativetofundamentalcollateralbecauseitissustainedbymarketexpectations.Thismayentailcosts,however:bydiscouraginginvestmentinfundamentalcollateral,bubblycollateralmaybeexpansionaryintheshortrunattheexpenseoflargerdownturnswhenbubblyepisodesend.Takingintoaccountthesenegativeincentivee¤ectsofbubbles,andanalyzinghowtheyimpactthedesignofpolicy,isanotherexcitingnextstepinthisresearchagenda.References[1]Aoki,K.andK.Nikolov(2011),Bubbles,BanksandFinancialStability,workingpaperCARF-F-253,TheUniversityofTokio.[2]Bernanke,B.andM.Gertler(1989),AgencyCosts,NetWorthandBusinessFluctuations,AmericanEconomicReview79,14-31.[3]Caballero,R.andA.Krishnamurthy(2006),BubblesandCapitalFlowVolatility:CausesandRiskManagement,JournalofMonetaryEconomics53(1),33-53.[4]Claessens,S.,A.KoseandM.Terrones(2011),FinancialCycles:What?How?When?,NBERInternationalSeminaronMacroeconomics,UniversityofChicagoPress,7(1),303-344.[5]Dang,T.V.,G.GortonandB.Holmstrom(2012),Ignorance,DebtandFinancialCrises,mimeo,Yale. 17Forsuchanextension,seeMartinandVentura(2015).34 [6]DellAriccia,G.,D.Igan,L.LaevenandH.Tong,withB.BakkerandJ.Vandenbussche(2012),PoliciesforMacro nancialStability:HowtoDealwithCreditBooms,IMFSta¤DiscussionNote,SDN12/06.[7]DellAriccia,G.,D.IganandL.Laeven(2012),CreditBoomsandLendingStandards:Evi-dencefromtheSubprimeMortgageMarket,JournalofMoney,Credit,andBanking44(2-3),367-384..[8]DellAriccia,G.andR.Marquez(2006),LendongBoomsandLendingStandards,TheJour-nalofFinance51(5),25112546.[9]Diba,B.andH.Grossman(1988),"TheTheoryofRationalBubblesinStockPrices,"Eco-nomicJournal,98(3),746-754.[10]Farhi,E.andJ.Tirole(2011),BubblyLiquidity,ReviewofEconomicStudies79,678-706.[11]Farhi,E.andJ.Tirole(2012),CollectiveMoralHazard,MaturityMismatchandSystemicBailouts,AmericanEconomicReview102,60-93.[12]Geerolf,F.(2013),ReassessingDynamicE¢ciency,mimeo,ToulouseSchoolofEconomics.[13]Gertler,M.,andN.Kiyotaki(2010),FinancialIntermediationandCreditPolicyinBusinessCycleAnalysis,inFriedmanandWoodford,HandbookofMonetaryEconomics,Vol.3A,Elsevier.[14]Gorton,G.andP.He(2008),BankCreditCycles,ReviewofEconomicStudies75(4),1181-1214.[15]Gorton,G.andG.Ordoñez(2014),CollateralCrises,AmericanEconomicReview104(2),343-378.[16]Gourinchas,P.O.,R.ValdesandO.Landerretche(2001),LendingBooms:LatinAmericaandtheWorld,Economia1(2),47-99.[17]Kiyotaki,N.,andJ.Moore(1997),CreditCycles,JournalofPoliticalEconomy105,211-248.[18]Kiyotaki,N.andJ.Moore(2008),Liquidity,BusinessCyclesandMonetaryPolicy,mimeo,PrincetonUniversity.35 [19]Kocherlakota,N.(2009),BurstingBubbles:ConsequencesandCures,mimeo,FederalRe-serveBankofMinneapolis.[20]Martin,A.(2008),EndogenousCreditCycles,mimeo,CREI.[21]Martin,A.andJ.Ventura(2011),TheoreticalNotesonBubblesandtheCurrentCrisis,IMFEconomicReview59,6-40.[22]Martin,A.andJ.Ventura(2012),EconomicGrowthwithBubbles,AmericanEconomicReview102,3033-3058.[23]Martin,A.andJ.Ventura(2015),InternationalTransmissionofCreditBubbles:TheoryandPolicy,mimeo,CREI.[24]Matsuyama,K.(2007),CreditTrapsandCreditCycles,AmericanEconomicReview97(1),503-516.[25]Mendoza,E.andM.Terrones(2012),AnAnatomyofCreditBoomsandtheirDemise,WP18379,NBER.[26]Mian,A.andA.Su (2009),TheConsequencesofMortgageCreditExpansion:EvidencefromtheU.S.MortgageDefaultCrisis,TheQuarterlyJournalofEconomics124(4),1449-1496.[27]Miao,J.andP.Wang(2011),BubblesandCreditConstraints,mimeo,BostonUniversity.[28]Rancière,R.,A.TornellandF.Westermann(2008),SystemicCrisesandGrowth,TheQuarterlyJournalofEconomics123(1),359-406.[29]Ruckes,M.(2004),BankCompetitionandCreditStandards,TheReviewofFinancialStud-ies17(4),1073-1102.[30]Samuelson,P.(1958),AnExactConsumption-loanModelofInterestwithorwithouttheSocialContrivanceofMoney,JournalofPoliticalEconomy66,467-482.[31]Schularick,M.andA.Taylor(2012),CreditBoomsGoneBust:MonetaryPolicy,LeverageCyclesandFinancialCrises,1870-2008,AmericanEconomicReview102(2),10291061.[32]Tirole,J.(1985),AssetBubblesandOverlappingGenerations,Econometrica53(6),1499-1528.36 [33]Ventura,J.(2011),BubblesandCapitalFlows,JournalofEconomicTheory147(2),738-758.37 79207940796079808000802080408060808081000.70.80.91InterestRate,EtRt+17900792079407960798080008020804080608080810000.020.04BubbleCreation,nt79007920794079607980800080208040806080808100!0.100.10.2Bubble,bt79007920794079607980800080208040806080808100!0.100.10.2PolicyIntervention,st7900792079407960798080008020804080608080810011.21.4Welfare,EtUt+1790079207940796079808000802080408060808081000.350.40.450.5Credit,lt7900792079407960798080008020804080608080810000.20.4BurstingProbability,!t790079207940796079808000802080408060808081001.522.5Productivity,AtWithoutPolicyWithPolicyWithoutPolicyWithPolicykt+1µ0.2970.337s.d.0.0270EtUt+1µ1.1361.202 InterestRate,EtRt+168606880690069206940696069807000702000.020.040.06BubbleCreation,nt68606880690069206940696069807000702000.51Bubble,bt686068806900692069406960698070007020456Productivity,At68606880690069206940696069807000702051015Welfare,