资源描述：
《Balance Sheets, the Transfer Problem, and Financial Crises》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
InternationalTaxandPublicFinance,6,459–472(1999)°c1999KluwerAcademicPublishers,Boston.ManufacturedinTheNetherlands.BalanceSheets,theTransferProblem,andFinancialCrisesPAULKRUGMANkrugman@mit.eduMIT,DepartmentofEconomics,Cambridge,MA02139AbstractInaworldofhighcapitalmobility,thethreatofspeculativeattackbecomesacentralissueofmacroeconomicpolicy.While“first-generation”and“second-generation”modelsofspeculativeattacksbothhaveconsiderablerelevancetoparticularfinancialcrisesofthe1990s,a“third-generation”modelisneededtomakesenseofthenumberandnatureoftheemergingmarketcrisesof1997–98.Mostoftherecentattemptstoproducesuchamodelhavearguedthatthecoreoftheproblemliesinthebankingsystem.Thispapersketchesanothercandidateforthird-generationcrisismodeling—onethatemphasizestwofactsthathavebeenomittedfromformalmodelstodate:theroleofcompanies’balancesheetsindeterminingtheirabilitytoinvest,andthatofcapitalflowsinaffectingtherealexchangerate.Keywords:currencycrises,balancesheets,capitalflowsJELCode:F31,F32,F41Forthefoundingfathersofcurrency-crisistheory—afraternityamongwhomBobFloodholdsaplaceofhighhonor—theemergingmarketcrisesof1997–8inspirebothasenseofvindicationandasenseofhumility.Ononeside,thenumberandseverityofthesecriseshasdemonstratedinadevastatinglythoroughwaytheimportanceofthesubject;inaworldofhighcapitalmobility,itisnowclear,thethreatofspeculativeattackbecomesacentralissue—indeed,forsomecountriesthecentralissue—ofmacroeconomicpolicy.Ontheotherside,evenacasuallookatrecenteventsrevealstheinadequacyofexistingcrisismodels.True,theAsiancrisishassettledsomedisputes—asIwillarguebelow,itdecisivelyresolvestheargumentbetween“fundamentalist”and“self-fulfilling”crisisstories.(Iwaswrong;MauryObstfeldwasright).Butithasalsoraisednewquestions.Onewaytodescribetheproblemistothinkintermsofthecelebrated(Eichengreen,Rose,andWyplosz,1995)distinctionbetween“first-generation”and“second-generation”crisismodels.First-generationmodels,exemplifiedbyKrugman(1979)andthemuchcleanerpaperbyFloodandGarber(1984),ineffectexplaincrisesastheproductofbudgetdeficits:itistheultimatelyuncontrollableneedofthegovernmentforseignoragetocoveritsdeficitthatensurestheeventualcollapseofafixedexchangerate,andtheeffortsofinvestorstoavoidsufferingcapitallosses(ortoachievecapitalgains)whenthatcollapseoccursprovokeaspeculativeattackwhenforeignexchangereservesfallbelowacriticallevel.Second-generationmodels,exemplifiedbyObstfeld(1994),insteadexplaincrisesastheresultofaconflictbetweenafixedexchangerateandthedesiretopursueamoreexpansionary 460KRUGMANmonetarypolicy;wheninvestorsbegintosuspectthatthegovernmentwillchoosetolettheparitygo,theresultingpressureoninterestratescanitselfpushthegovernmentovertheedge.Bothfirst-andsecond-generationmodelshaveconsiderablerelevancetoparticularcrisesinthe1990s—forexample,theRussiancrisisof1998wasevidentlydriveninthefirstinstancebythe(correct)perceptionthattheweakgovernmentwasabouttobeforcedtofinanceitselfviatheprintingpress,whilethesterlingcrisisof1992wasequallyevidentlydrivenbytheperceptionthattheUKgovernmentwouldunderpressurechoosedomesticemploymentoverexchangestability.InthemajorcrisiscountriesofAsia,however,neitherofthesestoriesseemstohavemuchrelevance.Byconventionalfiscalmeasuresthegovernmentsoftheafflictedeconomieswereinquitegoodshapeatthebeginningof1997;whilegrowthhadslowedandsomesignsofexcesscapacityappearedin1996,noneofthemfacedthekindofcleartradeoffbetweenemploymentandexchangestabilitythatBritainhadfaced5yearsearlier(andifdepreciationwasintendedtoallowexpansionarypolicies,itratherconspicuouslyfailed!)Clearlysomethingelsewasatwork;webadlyneeda“third-generation”crisismodelbothtomakesenseoftherecentcrisesandtohelpwarnofcrisestocome.Butwhatshouldathird-generationmodellooklike?Mostoftherecentattemptstoproducesuchamodelhavearguedthatthecoreoftheproblemliesinthebankingsystem.McKinnonandPill(1996)andothers,myselfincluded(Krugman,1998),havesuggestedthatmoral-hazard-drivenlendingcouldhaveprovidedasortofhiddensubsidytoinvestment,whichcollapsedwhenvisiblelossesledgovernmentstowithdrawtheirimplicitguarantees;thislineofthoughthasbeentakentoconsiderablelengthsintheinfluentialpapersofCorsetti,Pesenti,andRoubini(1998).Meanwhile,analternativelineofwork,followedinparticularbyChangandVelasco(1998)attemptstoexplaincurrencycrisesasthebyproductofabankrun,modeledalaDiamondandDybvig(1983)asaself-fulfillinglossofconfidencethatforcesfinancialintermediariestoliquidatetheirinvestmentsprematurely.Butisabank-centeredviewofthecrisisreallyright?Certainlyinmostcasesthefinancialcrisisdidinvolvetroublesforbanksaswellasforcurrencies.Butitalsoinvolvedotherdifficulties,mostnotablyanepidemicoffinancialdistressthatcannotberesolvedsimplybyfixingthebanks.AsevidenceabouttheAsiancrisishasaccumulated,Ihavefoundmyselfincreasinglyskepticalaboutwhethereitheramoral-hazardoraDiamond–Dybvigstorycanreallygetattheessentialnatureofwhatwentwrong.Inanycase,thispapersketchesoutyetanothercandidateforthird-generationcrisismodeling,onethatemphasizestwofactorsthathavebeenomittedfromformalmodelstodate:theroleofcompanies’balancesheetsindeterminingtheirabilitytoinvest,andthatofcapitalflowsinaffectingtherealexchangerate.Themodelisatthispointquiteraw,withseverallooseendshanging.However,itseemstometotellastorywithamorerealistic“feel”thanearlierefforts,myownincluded.ItalsoshedssomelightonthepolicydilemmasfacedbytheIMFanditsclientsinthelasttwoyears.Theremainderofthispaperisinfiveparts.Thefirstdiscussesingeneraltermssomefeaturesofthefinancialcrisesof1997–8,andthefailure(inmyview)ofourmodelssofartoreproducesomekeystylizedfacts.ThesecondpartlaysoutaroughmodelintendedtocapturewhatInowbelievetobetwoessentialpiecesofthepuzzle:theroleofbalancesheetdifficultiesinconstraininginvestmentbyentrepreneurs,andtheimpactofthereal BALANCESHEETS461exchangerateonthosebalancesheets.Thethirdpartshowshowtheseeffectsproduceafeedbackloopthatcancauseapotentiallyhealthyeconomytoexperienceaself-fulfillingfinancialcrisis.Thefourthpartoffersacrudeinterpretationofthe“IMFstrategy”oflimitingcurrencydepreciationinordertoprotectagainstthisbalance-sheeteffect,andshowshowthisstrategymaysimplyreplaceonedestructivefeedbackloopwithanother.Afinalsectionofferssometentativepolicyconclusions.1.RecentCrises:StylizedFactsandModelsTheAsiancrisisarrivedwithlittlewarning.Bynormalcriteria,governmentbudgetswereingoodshape;currentaccountdeficitswerelargeinThailandandMalaysia,butrelativelymoderateinKoreaandIndonesia;despitesomeslowdowningrowthin1996,therewasnotastrongcasethatanyofthecountriesneededadevaluationforcompetitiveormacroeconomicreasons.Indeed,rightuptothesummerof1997manyobserversechoedtheconclusionofthenow-notoriousWorldBankreport,TheEastAsianMiracle(1993),thatgoodmacroeco-nomicandexchange-ratemanagementwasakeyingredientintheAsianrecipeforsuccess.AndasStiglitz(1998)hasemphasized,evenafterthefactitisverydifficulttocomeupwithanysetofconventionalindicatorsthatpicksouttheAsiancountriesasparticularlyatriskoffinancialcrisis,oridentifies1997–8asatimeofunusualrisk.Sowhatwentwrong?Asalreadysuggested,therearetwomajorviewsinthepost-crisistheoreticalliterature.Thefirstisthatunderneaththeapparentsoundnessofmacroeconomicpolicywasalarge,hiddensubsidytoinvestmentviaimplicitgovernmentguaranteestobanks,croniesofpoliticians,etc.The“over-borrowingsyndrome”wasmodeledinadvanceofthecrisisbyMcKinnonandPill(1996),andforatimebecamethereigningorthodoxyaftermyownbriefexposition(Krugman,1998);Corsetti,Pesenti,andRoubini(1998a,b)haveem-phasizedthattotheextentthatimplicitguaranteesledbankstoengageinmoral-hazardlending,itrepresentedahiddengovernmentbudgetdeficit,andtheunfundedliabilitiesofthesebanksrepresentedahiddengovernmentdebt.Accordingtothisview,then,theappar-entsoundnessofbudgetaryandmacroeconomicpolicywasanillusion:underthesurface,thegovernmentswereactuallyengagedinrecklessandunsustainablespending.Thealternativeview,stronglyexpressedbyRadeletandSachs(1998),isthatthecountrieswerenotdoinganythingwrong;theirinvestmentswerebasicallysound.Atmosttheycanbesaidtohavesufferedfromsomekindof“financialfragility”thatmadethemvulnerabletoself-fulfillingpessimismonthepartofinternationallenders.ChangandVelasco(1998a,b)havemadethemostthoroughlyworked-outattempttomodelthisfinancialfragility,relyingonaversionoftheDiamond–Dybvig(1983)modelofbankruns.Inthismodel,investorsfaceachoicebetweenshort-terminvestmentswithalowrateofreturnandlong-runinvest-mentswithahigherrateofreturn;unfortunately,thelong-runinvestmentsyieldrelativelylittleiftheymustbeliquidatedprematurely,andinvestorsareassumedtobeunsureexanteaboutwhentheywillwanttoconsume.Financialintermediariescanresolvethisdilemmabypoolingtheresourcesofmanyinvestorsandrelyingonthelawoflargenumberstoavoidholdingmoreshort-termassetsthannecessary.However,suchintermediariesthenbecomevulnerabletoself-fulfillingpanics,inwhichfearoflossesleadsdepositorstodemandimme- 462KRUGMANdiatepayment,forcingdestructiveliquidationoflong-runassetsthatvalidatesthesefears.Inaclosedeconomythecentralbankcanprotectagainstsuchpanicsbyactingasalenderoflastresort;ChangandVelascoarguethatinanopeneconomywithafixedexchangerate,thelimitedsizeofthecentralbank’sreservesmaypreventitfromplayingthesamerole.ThereisnoquestionthatbothoftheseviewscapturesomeaspectsofwhathappenedtoAsia.Ononeside,“cronycapitalism”wascertainlyareality:theexcessesofThaifinancialcompanies,ofmembersoftheSuhartofamily,ofmegalomaniacchaebolareundeniable.Ontheotherside,bankrunsplayedanimportantroleintheunfoldingofthecrisis,particularlyinIndonesia,andafreezingupofthecreditsystemplayedatleastsomeroleindeepeningtherecessionafterthecrisishit.Yetasevidenceaboutthecrisishasaccumulated,bothexplanationshavecometoseeminadequatetothetaskofexplainingtheseverityoftheevent.Considerfirstthemoralhazardargument.Ifonereallytakesthatargumentseriously,itimpliesnotonlythatthereshouldbeover-investmentandexcessiverisk-takingbyen-trepreneurswithaccesstoguaranteedfinance,butalsothattheavailabilityofimplicitguar-anteesshouldtendtocrowdout“legitimate”investmentthatbearsthefullburdenofrisk.YetasRadeletandSachspointout,intherunuptothecrisisallformsofinvestmentintheemergingAsianeconomieswerebooming,includingdirectforeignpurchasesofequityandrealestate,investmentsthatclearlywerenotprotectedbyanyformofimplicitguarantee.Onemightpointtotheseverityoftheproblemofnon-performingloansafterthecrisisasevidencethatbadbankingwasakeyprobleminthecrisiseconomies.Butasmanyobservershavenoted,andasisdocumentedintherecentWorldBankreportTheRoadtoRecovery(1998),thebulkofthebadloanproblemisaconsequenceofthecrisis—oftheseverereces-sionsandcurrencydepreciationsthatfollowedthecollapseofcapitalinflows.Sincenobodyexpectedacrisisofanythinglikethisseverity,theprevalenceofbadloansweobserveexpostdoesnotmeanthatanythinglikethesameamountofbadlendingwastakingplaceexante.Whataboutthefinancialfragilitystory?HeremymainconcernisnotsomuchwithChangandVelascoaswithDiamond–Dybvig—specifically,withthewaythatfinancialfragilityanditsrealeffectsaremodeled.IntheDiamond–Dybvigmodelthecostsofprematureliquidationarephysical—abankrunliterallyleadstoinvestmentsbeingcannibalizedbeforecompletion,withtheoutputcosttotheeconomytheresultofaliteraldestructionofphysicalcapital.ThereareafewrealexamplesofthisprocessinAsia—half-completedstructureslefttodisintegrateforlackoffunding,ordismantledforscrapmetal.Therearealsosomemorecomplexstoriesthatcanbeviewedmetaphoricallyasexamplesofphysicalliquidation—forexample,potentiallyprofitableexportopportunitiesnottakenbecauseworkingcapitalhasbeensoldtopayoffbankloans.Butsurelythemainchannelsthroughwhichfinancialpanichasturnedgoodassetsintobadinvolvenotsomuchphysicalliquidationofunfinishedprojectsasmacroeconomiccrisis:companiesthatlookedsolventbeforethecrisishavegoneunderbecausecollapsinginvestmenthasproducedasevererecession,orbecausecapitalflighthasledtocurrencydepreciationthatmakestheirdollardebtsballoon.Ortoputitanotherway,DiamondandDybvigusedaphysicalmetaphorforthecostsofprematureliquidationasawaytofocusontheproblemofmultipleequilibriaonthepartofdepositors;fairenough.ButtomakesenseoftheAsiancrisisitisprobablyimportanttohaveabettermetaphor,onethatcomesclosertomatchingthestylizedfactsofactualexperience. BALANCESHEETS463Whatarethesestylizedfacts?Letmesuggestthreefactsthatamodelshouldprobablyaddress—andwhichsomeoralloftheexistingmodelsdonot,asfarasIcantell,seemtocapture.(i)Contagion:ThemoststunningaspectoftheglobalfinancialcrisishasbeenthewaythateventsinsmalleconomieslikeThailandorRussiahaveledmoreorlessdirectlytocrisesineconomiesthousandsofmilesdistant,withfewdirecttradeorfinanciallinks.Frommypointofviewthepowerofcontagioninthelasttwoyearssettlesalong-runningdisputeaboutcurrencycrisesingeneral:thedisputebetween“fundamentalists”and“self-fulfillers.”Intheoriginalfirst-generationmodels,thesuddennessofcurrencycrisesdidnotmeanthattheirtimingwasarbitrary;onthecontrary,suchcrisesemergedwhensomesetoffundamentalfactors(typicallythelevelofreserves)fellbelowacriticallevel.Obstfeld(1994)arguedthatinsecond-generationmodels,bycontrast,thetimingofcrisiswasindeedarbitrary;infact,acurrencycrisiscouldoccurtoacountrywhosefixedexchangeratemightotherwisehavesurvivedindefinitely.Iarguedinreply(Krugman,1996)thatthiswasamisleadingpoint:thereasonthatthetimingofcrisisseemeddeterminateinfirst-generationmodelswasnotbecauseofthedifferenceinthemechanismofcrisis,butbecauseinthosemodelstherewasassumedtobeaseculardeteriorationinthefundamentals—adeteriorationthatensured,throughbackwardinduction,thataspeculativeattackwouldalwaysoccurassoonasitcouldsucceed.Thispointwas,Istillthink,correct.However,Ithenwentontoarguethatweshouldviewapredictableseculardeteriorationinfundamentalsasthenormalcase,whateverthespecificsofthemodel,andthatspontaneousself-fulfillingcriseswouldthereforeberareevents.Iherebycapitulate.Icannotseeanywaytomakesenseofthecontagionof1997–8withoutsupposingtheexistenceofmultipleequilibria,withcountriesvulnerabletoself-validatingcollapsesinconfidence,collapsesthatcouldbesetoffbyeventsinfarawayeconomiesthatsomehowservedasatriggerforself-fulfillingpessimism.Itfollowsthatanyusefulmodelofthecrisismustinvolvesomemechanismthatproducesthesemultipleequilibria—acriterionmetbythefinancialfragilitymodels,butnotbythemoralhazardapproach.(ii)Thetransferproblem:IfthereisasinglestatisticthatcapturestheviolenceoftheshocktoAsiamostdramatically,itisthereversalinthecurrentaccount:inthecaseofThailand,forexample,thecountrywasforcedbythereversalofcapitalflowstogofromadeficitofsome10percentofGDPin1996toasurplusof8percentin1998.Theneedtoeffectsuchahugechangeinthecurrentaccountrepresentswhatmaybehistory’smostspectacularexampleoftheclassic“transferproblem”debatedbyKeynesandOhlininthe1920s.Inpracticethisswinghasbeenachievedpartlythroughmassiverealdepreciation,partlythoughsevererecessionthatproducesacompressionofimports.YetdespitetheevidentcentralityofthetransferproblemtowhatactuallyhappenedtoAsia,thisissuehasbeenremarkablyabsentfromformalmodels.Perhapsbecausethemodelershavebeenmainlyconcernedwiththebehaviorofinvestorsratherthanwiththerealeconomyperse,allofthemajormodelssofarhavebeenone-goodmodelsinwhichdomesticgoodscanbefreelyconvertedintoforeignandviceversawithoutanymovementinthetermsoftradeortherealexchangerate. 464KRUGMANIsthisanacceptablestrategicsimplification?Perhapsnot:inthemodelIdevelopbelow,thedifficultyofeffectingatransfer,theneedtoachievethecurrentaccountcounterpartofareversalofcapitalflowseitherviarealdepreciationorviarecession,turnsouttobetheheartofthestory.(iii)Balancesheetproblems:Finally,descriptiveaccountsbothoftheproblemsofthecrisiscountriesandofthepolicydiscussionsthatledthecrisistobehandledinthewayitwasplaceextensiveemphasisontheproblemsoffirms’balancesheets.Ononeside,thedeteriorationofthesebalancesheetsplayedakeyroleinthecrisisitself—notably,theexplosioninthedomesticcurrencyvalueofdollardebthadadisastrouseffectonIn-donesianfirms,andfearofcorrespondingbalancesheeteffectswasamainreasonwhytheIMFwasconcernedtoavoidmassivedepreciationofitsclients’currencies.Ontheotherside,theprospectsforrecoveryarenow,byallaccounts,especiallydifficultbe-causeoftheweakenedfinancialconditionoffirms,whosecapitalhasinmanycasesbeenwipedoutbythecombinationofdecliningsales,highinterestrates,andadepreciatedcur-rency.Noticethatwhilethesebalancesheetproblemsareinturnacauseoftheproblemofnon-performingloansatthebanks,theyarenotabankingproblemperse;evenarecap-italizationofthebankswouldstillleavetheproblemoffinanciallyweakenedcompaniesuntouched.Theroleofbalance-sheetproblemsinconstrainingfirmshasbeenthesubjectofsomerecentworkinthemacroeconomicsliterature,notablyKiyotakiandMoore(1997)andBernanke,Gertler,andGilchrist(forthcoming).Sofar,however,despitetheattentiongiventobalancesheetsinpracticaldiscussions,theissuehasbeenneglectedinthecurrencycrisisliterature.WhatIwilldointheremainderofthispaper,then,istotrytodevelopamodelinformedbytheseobservations.AsintheDiamond–Dybvigapproach,thisisamodelpotentiallycharacterizedbymultipleequilibria,inwhichalossofconfidencecanproduceafinancialcollapsethatvalidatesinvestorpessimism.However,themechanismofthatcollapseisdifferent:insteadofcreatinglossesviatheprematureliquidationofphysicalassets,alossofconfidenceleadstoatransferproblem.Thatis,inordertoachievetherequiredreversalofitscurrentaccount,thecountrymustexperiencealargerealdepreciation;thisdepreciation,inturn,worsensthebalancesheetsofdomesticfirms,validatingthelossofconfidence.Apolicythatattemptstolimittherealdepreciationimpliesadeclineinoutputinstead—andthis,too,canvalidatethecollapseofconfidence.Moreover,oncethecrisisoccursitcanhaveasustainedimpactontheeconomy,becauseofthatimpactonbalancesheets;asoneThaieconomistrecentlyputit,thecrisisleadstothe“decapitationoftheentrepreneurialclass,”andtheeconomycannotreturntonormaluntilitmanageseithertorepairthebalancesheetsofitsexistingentrepreneursorgrowsanewset.Itseemstomethatthisstory—inwhich,incidentally,banksdonotnecessarilyplayakeyrole,althoughtheycouldpresumablyalsobeintroduced—comescloserthananyofthepreviousmodelstohavingtheright“feel”formakingsenseofrecentevents.Butinanycase,letusnowproceedtothestatementandanalysisofthemodel. BALANCESHEETS4652.TheModelIconsideranopeneconomythatproducesasinglegoodeachperiodusingcapitalandlabor;forsimplicitytheproductionfunctionisassumedCobb–Douglas:yDG.K;L/DK®L1¡®(1)tttttCapitaliscreatedthroughinvestment;Iwillassume,againforsimplicity,thatcapitallastsonlyoneperiod,sothatthisperiod’scapitalisequaltolastperiod’sinvestment.(ThisassumptionalsoputstoonesideDiamond–Dybvig-typeconcernsovermaturitymismatch).Theresidentsofthiseconomyaredividedintotwodistinctclasses.Workersplayapassiverole—theylackaccesstothecapitalmarket,andthereforemustspendalltheirincomewithineachperiod.Capitalisbothcreatedandownedbyaclassofentrepreneurs,whoareassumedtobesingle-mindedlyengagedinaccumulationatthispoint,savingandinvesting(eitherathomeorabroad)alltheirincome.Onlytheseentrepreneurshavetheabilitytoundertakedomesticinvestment,whichaswewillsoonseeplaysacrucialroleinthestory.Thegoodproducedbythiscountryisnotaperfectsubstitutefortradedgoodsproducedelsewhere.Indeed,Iwillassume(yetanothersimplification)thatthereisaunitaryelasticityofsubstitution,withashare¹ofbothconsumptionandinvestmentspendingonimports,1¡¹ondomesticgoods.Therestoftheworldisassumedtobemuchlargerthanthedomesticeconomy,andtospendanegligiblefractionofitsincomeondomesticgoods.(Thedisparitybetweenthedomesticandforeignmarginalpropensitiestospendondomesticgoods—1¡¹inthecaseofdomesticspending,0forforeignspending—givesrisetothetransferproblemthatiscrucialtothisapproach).Iftheforeignelasticityofsubstitutionisalso1,thevalueofdomesticexportsintermsofforeigngoodsisfixed,sayatX,andthevalueintermsofdomesticgoodsisthereforepX,wherepistherelativepriceofforeigngoods(a.k.arealexchangerate).Bearinginmindthatashare1¡®ofdomesticincomeaccruestoworkerswhomustspendit,anddefiningIandCasinvestmentandconsumptionexpendituresintermsofdomesticgoods,wecandeterminetherealexchangerateasfollows.MarketclearingfordomesticgoodsrequiresthatyD.1¡¹/IC.1¡¹/CCpXD.1¡¹/IC.1¡®/.1¡¹/yCpX(2)whichimpliesyt[1¡.1¡®/.1¡¹/]¡.1¡¹/ItptD(3)XWecanimmediatelynoticethatthehigherisinvestment,thelowertherealexchangerate.Thenextstepistodescribethedeterminationofinvestment.Thecentralideahereisthattheabilityofentrepreneurstoinvestmaybelimitedbytheirwealth.Specifically,followingBernankeetal(forthcoming)Iassumethatlendersimposealimitonleverage:entrepreneurscanborrowatmost¸timestheirinitialwealth.It·.1C¸/Wt(4) 466KRUGMANUnderlyingthislimitationonborrowing,presumably,aresomekindofmicroeconomicmotives,probablyinvolvingasymmetricinformation.Forthepurposesofthispaper,how-ever,Isimplyassumetheexistenceoftheconstraintandtake¸asagiven.Thisconstraintneednotbebinding;althoughentrepreneursareassumedtosavealloftheirincome,theymaychoosenottoborrowuptothelimit.Inparticular,theywillnotborrowbeyondthepointatwhichtherealreturnondomesticinvestmentequalsthatonforeigninvestment.Onewaytodeterminethislimitistocomparetheforeignrealinterestrate,r¤,withthereturnachievedbyconvertingforeigngoodsintodomestic,thenconvertingthenext-periodreturnbackintoforeigngoods.Becauseashare¹ofinvestmentfallsonforeigngoods,thepriceindexforinvestmentrelativetothatofdomesticoutputisp¡¹;thereturnoninvestmentintermsofdomesticgoodsistherefore1CrDG.Ip¡¹;L/(5)tkt¡1Butaunitofforeigngoodscanbeconvertedintoptunitsofdomesticgoodsthisperiod,thereturnconvertedinto1=ptC1unitsnextperiod;sothestatementthatthereturnondomesticinvestmentmustbeatleastaslargeasthatonforeignbondsmaybewritten.1Cr/.p=p/¸1Cr¤(6)tttC1Finally,investmentcannotbenegative:It¸0(7)Aswewillsee,dependingoncircumstances(4),(6),or(7)maybethebindingconstraint.Thelastelementinthestatementofthemodelisthedefinitionofentrepreneurs’wealth.Domesticentrepreneursownalldomesticcapital;theymayalsoownotherclaimsonforeigners,and=orhavedebtstoforeigners.Iassumethatsomeclaimsaredenominatedintermsofthedomesticgood,othersintermsoftheforeigngood;meanwhile,sincecapitallastsonlyoneperiod,thevalueofdomesticcapitalissimplytheincomeaccruingtocapitalwithinthecurrentperiod.LetD;Fbethenetdebtsofdomesticentrepreneursindexedtodomesticandforeigngoodsrespectively;Iwillsloppilyrefertotheseas“domesticcurrency”and“foreigncurrency”debtrespectively,althoughtheyarereallydenominatedingoodsratherthanmoneys.ThenthewealthofentrepreneursinperiodtisWtD®y¡D¡pF(8)Obviouslyafullmodelshouldtrytoendogenizethe“currencycomposition”(again,actuallygoodscomposition,sincethemodelisnotexplicitlymonetary)ofdebt;again,however,Isimplytakeitasagiven.Wenowhavearoughbutworkablemodelthatcanbeusedtoexamineonewayinwhichafinancialcrisiscanoccurinanopeneconomy.3.TheTransferProblemandFinancialCrisisAccordingtoourmodel,theamountthatdomesticentrepreneurscanborrowfromforeignerstofinanceinvestmentdependsontheirwealth.Atthesametime,however,thewealthof BALANCESHEETS467eachindividualentrepreneuritselfdependsonthelevelofsuchborrowingintheeconomyasawhole,becausethevolumeofcapitalinflowaffectsthetermsoftradeandhencethevaluationofforeign-currency-denominateddebt.Wecanthereforeimmediatelyseetheoutlinesofastoryaboutfinancialcrisis:adeclineincapitalinflowscanadverselyaffectthebalancesheetsofdomesticentrepreneurs,reducingtheirabilitytoborrowandhencefurtherreducingcapitalinflows.Butweneedtobeabitmoreprecise.Imagineagameinwhichlendersdecide,inrandomorder,howmuchcredittooffertosuccessivedomesticentrepreneurs.Theofferofcreditdependsonwhatthelendersthinkwillbethevalueoftheborrower’scollateral.Butbecausesomedebtisdenominatedinforeigngoods,thisvaluedependsontherealexchangerate,andhenceontheactuallevelofborrowingthattakesplace.Arational-expectationsequilibriumofthisgamewillbeasetofself-confirmingguesses—thatis,theexpectedlevelofinvestmentimplicitinthecreditoffersmustmatchtheactuallevelofinvestmentthattakesplacegiventhoseoffers.Asafirststep,letusderivetherelationshipbetweeninvestmentandthewealthofen-trepreneurs.From(8),weknowthatwealthdepends,otherthingsbeingthesame,ontherealexchangeratep;from(3)weknowthatpdependsonI.WethereforehavethatdW.1¡¹/FD(9)dIXLetusdefineIfasthe“financeable”levelofinvestment—thatis,thelevelofinvest-mentthatwouldoccuriftheleverageconstraint(4)werebinding.Sincetheabilityofentrepreneurstoborrowdependsontheirwealth,wehavedIf.1C¸/.1¡¹/FD(10)dIXIfdIf=dIislessthan1,thebehaviorofthismodelisrelativelyuninteresting:aneconomywithahighrateofreturnoninvestmentmayfindthatadjustmentinitscapitalstockisdelayedbyfinancingconstraints,buttherewillbenothingresemblinganAsian-stylefinancialcrisis.ButsupposethatdIf=dI>1.Thentherecanindeedbemultipleequilibria,withthepossibilitythatalossoflenderconfidencewillbevalidatedbyfinancialcollapse.ThepicturewouldlooklikeFigure1.1Onthehorizontalaxisistheexpectedlevelofinvestment,whichdeterminesviaitseffectontherealexchangerate,andhenceonbalancesheets,howmuchcreditisextendedtodomesticfirms.Ontheverticalaxisistheactuallevelofinvestmentthatresults.(Thepicturecouldalternativelybedrawnintermsoftheexpectedandactuallevelsofp).AthighlevelsofexpectedIthefinancingconstraint(4)isnotbinding;instead,investmentisdeterminedbytherate-of-returnconstraint(6).AtlowlevelsofexpectedIfirmsarebankrupt,andcannotinvestatall—thatis,theyarehardagainstthenon-negativityconstraint(7).InanintermediaterangeIisconstrainedbyfinancing,andthescheduleisthereforesteeperthanthe45-degreeline.Thereareclearlythreeequilibriainthispicture.Theintermediate,internalequilibriummaybedismissedaslikelytobeunstableunderanyplausiblemechanismofexpectationformation.Thisleavesuswithtwopossibleoutcomes:ahigh-leveloutcomeHinwhichinvestmenttakesplaceuptothepointwheredomesticandforeignratesofreturnareequal;andalow-leveloutcomeLinwhichlendersdonotbelievethatdomesticentrepreneurshave 468KRUGMANFigure1.anycollateral,theirfailuretoprovidefundsmeansadepreciatedrealexchangerate,andthatunfavorablerealexchangeratemeansthatentrepreneursareinfactbankrupt,validatinglenders’pooropinion.AndwethereforenowhaveourextremelystylizedversionoftheAsianfinancialcrisis:something—itdoesnotmatterwhat—causedlenderstobecomesuddenlypessimistic,andtheresultwasacollapsefromHtoL.Thecollapsedoesnotindicatethatthepreviousinvestmentswereunsound;theproblemisinsteadoneoffinancialfragility.ThedifferencebetweenthisstoryoffinancialfragilityandthattoldbyChangandVelascocanbehighlightedbyconsideringtheconditionsunderwhichthisfragilitycanoccur—namely,whendIf=dI>1.Byconstructionhere,thiscriterionhasnothingtodowiththemismatchbetweenshort-termdebtandlong-terminvestments;nordoesitappeartodependonforeignexchangereserves.Instead,asweseefrom(10),thefactorsthatcanmakefinancialcollapsepossibleare:(i)Highleverage(ii)Lowmarginalpropensitytoimport(iii)Largeforeign-currencydebtrelativetoexportsThesefactorsmatter,ofcourse,becausetheymakethecircularloopfrominvestmenttorealexchangeratetobalancesheetstoinvestmentmorepowerful.Wecannowalsoofferapossibleanswertothegreatmystery:WhyAsia?Whynow?IfweaskwhatwasspecialaboutAsianeconomies,somethingthatmayhavemadethempeculiarlyvulnerabletofinancialcrisis,theanswerishighleverage:allofthenowafflictedcountrieshadunusuallyhighlevelsof¸.Ifweaskwhynow—giventhathighleverage,“cronycapitalism,”etc.havebeencharacteristicofAsianeconomiesfordecades—theansweristhatonlyafter1990didtheseeconomiesbeginextensiveborrowingdenominated BALANCESHEETS469inforeigncurrencies,borrowingthatplacedthematriskoffinancialcollapseiftherealexchangeratedepreciated.4.TheDilemmaofStabilizationAlthoughstandardmodelsofcurrencycrisishavenottodatetakenaccountoftheproblemsposedbyforeign-currencydebt,practitionershavebeenawareofthisissuefordecades.AndtherisksoffinancialtraumabecauseofthatdebtwereamajorreasonwhytheIMFadviseditsAsianclientstofollowthemuch-criticized“IMFstrategy”ofdefendingtheircurrencieswithhighinterestratesratherthansimplylettingthemdecline.Thismodeldoesnotallowadirectanalysisofmonetarypolicy.Wecan,however,takeaveryroughcutatthenatureandconsequencesoftheIMFstrategybyimaginingthattheeffectofthatstrategyistoholdtherealexchangeratepconstantevenwhenthewillingnessofforeignlenderstofinanceinvestmentdeclines.Inthatcase,ofcourse,somethingelsemustgive;andthenaturalassumptionisthatoutputdeclinesinstead.Indeed,ifweholdpconstant,outputwillbedeterminedbyasortofquasi-Keynesianmultiplierprocess;rearranging(2)wehavepXC.1¡¹/IyD(11)1¡.1¡®/.1¡¹/Butgiventhatashare®ofoutputgoestoprofits,adeclineininvestmentwillreduceentrepreneurs’wealth:dW®.1¡¹/D(12)dI1¡.1¡®/.1¡¹/andhenceonceagaintherewillbeafeedbackfromactualtofinanceableinvestment:dIf.1C¸/®.1¡¹/D(13)dI1¡.1¡®/.1¡¹/Itisimmediatelyclearthatstabilizingtherealexchangerate,whileclosingonechannelforpotentialfinancialcollapse,opensanother:ifleverageishigh,theeconomymaystabilizeitsrealexchangerateonlyattheexpenseofaself-reinforcingdeclineinoutputthatproducesanequivalentdecapitationoftheentrepreneurialclass.5.PolicyImplicationsOnewouldordinarilybesomewhatdiffidentaboutdrawingpolicyimplicationsfromsoroughaframework.However,policymustbeandisbeingmade,byandlargewithoutanyexplicitanalyticalframeworkatall;soherearesomeconclusionsinspiredfromthemodel.Theypertaintothreeratherdifferentquestions.Firstisthequestionofprophylacticmeasures:whatcanwedotopreventsuchcrisesinthefuture?Secondisthequestionofpolicyinthecrisis:howcanthecrisisbehaltedoratleastlimited?Finallythereisthequestionofwhattodooncethecrisishasoccurred:howdoesonerebuildtheeconomy? 470KRUGMANProphylacticmeasures:IntheaftermathoftheAsiancrisis,abroadconsensushasemergedamongresponsiblepeoplethatcountriesneedtotakemuchgreatercarewiththeirbankingsystems—thattheyneed“transparency,”bettercapitalstandards,morecarefulregulationofrisk-taking,anendtocronyism,etc.Underlyingsuchrecommendationsisthebeliefthatthecrisiswaslargelydueeithertomoralhazard,Diamond–Dybvig-typeproblems,orboth.Anditishardtodisagreethatsuchmeasuresareagoodthing.IfIamrightaboutthemechanismofcrisis,however,evenaverycleanandprudentbankingsystemmaynotbeenoughtoprotectopeneconomiesfromtheriskofself-reinforcingfinancialcollapse.AmorecontroversialproposalisforthewidespreadimpositionofChilean-typerestric-tionsonshort-termborrowingdenominatedinforeigncurrencies.Theideahereisthatbyreducingshort-termforeign-currencyexposure,countriescanreducetherisksofbeingforcedintocrisisbyalossofconfidence.Ihavebeenskepticalaboutthisargumentonthegeneralgroundsthataslongasacountryhasfreeconvertibilityofcapital,short-termforeignloansareonlyoneofmanydifferentpossiblesourcesofcapitalflight.Wecannotdealwiththeissueofmaturitystructureinthismodel,sincesuchissueshavebeenruledoutbyassumption.Butinthespiritofthemodel,considerthefollowingsituation:domesticfirmsarefinancedbyamixtureofshort-termdebtdenominatedindomesticcurrency,andlong-termdebtdenominatedinforeigncurrency.Doesthefactthattheforeigncurrencydebtislong-termprotectthecountryfromfinancialcrisis?Surelynot:ifpeopleexpectafinancialcrisis,theholdersofdomesticshort-termdebtwillrefusetorollitover,generatinganexchangeratedepreciationthatbankruptsthefirmseventhoughtheforeign-currencydebtitselfislongterm.Sowhatistheappropriateprophylacticpolicy?Theanswerfromthismodel,atanyrate,seemstobetodiscouragefirmsfromtakingonforeign-currency-denominateddebtofanymaturity.Looselyspeaking,thereappearstobeasortofexternaldiseconomytoborrowinginforeigncurrencies:becausesuchborrowingmagnifiesthereal-exchange-rateimpactofadverseshocks,andbecauserealdepreciationinteractswithcapital-marketimperfectionstocauseeconomicdistress,thedecisionbyanindividualfirmtoborrowindollarsimposescostsontherestoftheeconomy.Dealingwithcrisis:Muchofthevituperativepublicdebateoverhowtodealwithcrisishasinvolvedthequestionofwhethertolettheexchangerategoorstabilizeit.Theanswersuggestedbythismodelisthatthisisarealchoice,butthatbothanswersmaybeequallybad.Isthereathirdway?Onepossibilitywouldbetheprovisionofemergencylinesofcredit.However,inthecontextofthismodelitappearsthatthesecreditlineswouldhavetodomorethanprovidebalance-of-paymentsfinancing,orevenprovidelender-of-lastresortfacilitiestobanks:theywouldhavetomakeupthecreditbeinglostbyfirms,soastoallowinvestmenttocontinue.Thusthecreditlineswouldhavetobeverylargeindeed,andalsobeaccompaniedbyamechanismthatfunnelsthefundstotroubledentrepreneurs.(Thiswouldbeespeciallydifficultpolitically,sinceinthemidstofcrisisthereiswidespreadandoftenjustifiedvilificationofthosesameentrepreneurs,onthegroundsthattheirexcessesbroughtonthecrisisinthefirstplace).Ofcourseifonetakesthemodelseriously,asufficientlylargecreditlinewouldneveractuallyhavetobeused,sinceitsveryexistencewouldpreventthe BALANCESHEETS471crisisfromevergettingunderway(butonehastobecrediblywillingtouseitinordernottohaveto).Anotherpossibilityistoruleoutthepossibilityofadownwardfinancialspiralbybeingreadytoimposeacurfewoncapitalflight.Again,thereissubstantialsympathyevenamongrespectableopinionforstandstillagreementsonforeign-currencydebt;butthismaywellnotbesufficient,ifcapital-accountconvertibilitymeansthatotherformsofcapitalflightarestillpossible.Allofwhichraisesthepossibilitythatitmightbenecessary,andevenintheinterestsofinvestorsthemselves,toimposeemergencycapitalcontrols:::enoughsaid.Afterthefall:Finally,whatwehopeisthecurrentquestion:oncethecrisishashappened,howdoesonegettheeconomygoingagain?Todatemostactualeffortshavefocusedonbankrestructuringandrecapitalization;butifthismodelisontherighttrack,thiswillnotbesufficient.Themainproblematthispoint,themodel(likemanypractitioners)suggests,isthatthefirmsandentrepreneurswhodroveinvestmentandgrowthbeforethecrisisarenoweffectivelybankruptandunabletoraisecapital.Ifthisisright,thekeytoresuminggrowthiseithertorescuethoseentrepreneurs,throughsomekindof“privatesectorBradyPlan,”ortogrowanewsetofentrepreneurs—orboth.Alikelysourceofnewentrepreneursis,ofcourse,fromabroad:awelcomematforforeigndirectinvestmentmightbejustwhatthedoctorordered.Again,allofthisisbasedonaliberalinterpretationofaveryroughmodel.Itseemstome,however,thatthismodeldoesprovideatleastadifferentperspectiveonhowtothinkabouttheseissues.AsIsaidatthebeginningofthispaper,theAsiancrisisinspiresmixedemotionsinthoseofuswho,likeBobFlood,havesharedadecades-longfascinationwiththeissueofcurrencycrises.Ourobsessionhasbeenspectacularlyandtragicallyvindicated;buttheworldseemstokeepfindingnewwaystogeneratecrises.Letushopethatthelessonsofthis“third-generation”crisisarelearned,andthatnofuturecrisesariseinthesameway;butevenifthathopeisfulfilled,onecanbesurethattherearemanymoregenerationstocome.Notes1.Strictlyspeaking,therearetwootherpossibilitiesevenifdIf=dI>1.Ifdomestic-currencydebtisveryhigh,entrepreneursmaybeunabletoborrowevenwithafavorableexchangerate;ifDislow,evenaveryunfavorableratewillnotcausefinancialcollapse.Ineglectthesecasesforthesakeofthemainstory.ReferencesBernanke,Ben,MarkGertler,andSimonGilchrist.(forthcoming).“TheFinancialAcceleratorinaQuantitativeBusinessCycleFramework.”InJ.TaylorandM.Woodford(eds.),HandbookofMacroeconomics.Chang,Roberto,andAndresVelasco.(1998).“FinancialCrisesinEmergingMarkets:ACanonicalModel.”NBERWorkingPaperNo.6606.Corsetti,Giancarlo,PaoloPesenti,andNourielRoubini.(1998).“PaperTigers?AModeloftheAsianCrisis.”Mimeo. 472KRUGMANDiamond,Douglas,andPhilipDybvig.(1983).“BankRuns,DepositInsurance,andLiquidity.”JournalofPoliticalEconomy91401–419.Eichengreen,Barry,AndrewRose,andCharlesWyplosz.(1995).“ExchangeMarketMayhem:TheAntecedentsandAftermathofSpeculativeAttacks.”EconomicPolicy,251–312.Flood,Robert,andPeterGarber.(1984).“CollapsingExchangeRateRegimes:SomeLinearExamples.”JournalofInternationalEconomics17,1–13.Kiyotaki,Nobuhiro,andJohnMoore.(1997).“CreditCycles.”JournalofPoliticalEconomy105,211–248.Krugman,Paul.(1979).“AModelofBalance-of-PaymentsCrises.”JournalofMoney,Credit,andBanking11,311–325.Krugman,Paul.(1996).“AreCurrencyCrisesSelf-Fulfilling?”InB.BernankeandJ.Rotemberg(eds.),NBERMacroeconomicsAnnual1996.Cambridge:MITPress.Krugman,Paul.(1998).“WhatHappenedtoAsia?”Mimeo,MIT.McKinnon,Ronald,andHuwPill.(1996).“CredibleLiberalizationsandInternationalCapitalFlows:The‘OverborrowingSyndrome.’”InT.ItoandA.Krueger(eds.),FinancialDeregulationandIntegrationinEastAsia.Chicago:UniversityofChicagoPress.Obstfeld,Maurice.(1994).“TheLogicofCurrencyCrises.”Cahierseconomiqueetmonetaires43,189–212.Radelet,Steven,andJeffreySachs.(1998).“TheOnsetoftheEastAsianFinancialCrisis.”Mimeo.
