Mexico's Balance-of-Payments Crisis: A Chronicle of Death Foretold

MEXICO’SBALANCE-OF-PAYMENTCSRISIS:ACHRONICLEOFADEATHFORETOLD . Note:InternationalFinanceDiscussionPapersarepreliminarymaterialscirculatedtostimulatediscussionand criticalcomment.ReferencesinpublicationstoInternationalFinanceDiscussionPapers(otherthanan acknowledgmentthatthewriterhadaccesstounpublishedmaterial)shouldbeclearedwiththeauthoror authors.

ThispaperclaimsthattherootsofMexico’sbalance-of-paymentscrisisarefoundintheprevailing highdegreeofcapitalmobilityandfinancialglobalization.Underthesecircumstances,shifisinforeigncapital flowsandanticipationofabanking-systembailoutmayproducelargeimbalancesbetweenstocksoffinancial assetsandforeignreserves,threateningthesustainabilityofcurrencypegs.Econometricanalysissuggeststhat 1/2ofMexico’sreservelossescouldbeaccountedforbythesephenomena.Largefinancialimbalancesarealso fertilegroundforself-fulfilling-prophesycriseswhichleaddevaluationstoproducedeeprecessions.These difficultiescanbepartlyremediedbyappropriatepolicies.

Mexico’sBalance-of-PaymentsCrisis:AChronicleofaDeathForetold GuillermoA.CalvoandEnriqueG.Mendoza’ 1.Introduction At first sight,the Mexicanfinancialcrashof December1994andthedeepeconomiccrisisthat followedresemblepreviousMexicancrises. Fivepreviousattemptsatfixingtheexchangeratesince1945 werefollowedbyeconomicexpansion,sharprealappreciationa,ndlargeexternaldeficits.Eventually,each attemptendedwithalargedevaluationandtheabruptreversaloftheinitialprocess.Inaremarkableaccount ofthesecycles,Gomez-Oliver(198)1showsthatexpansionarypoliciesandadverseexternalshockstriggered thedevaluations.Since1976thecriseshavebeenalmostperfectlytimedwiththepresidentialelections. In 1994,thecrashcoincidedagainwithpresidentialelections.Itfollowedagainfromaperiodof economicexpansion,realappreciationa,ndwideningexternalimbalancesassociatedwiththeexchange-ratebasedstabilizationinitiatedinMarch,1988.]Uponcloserscrutiny,however,severalelementsofthe 1994 crashdonotfitpreviouscrises.TheSalina.sadministrationimplementedafar-reachingprogramofstructural reform.Fiscalandmonetarydisciplinewererestored.Aggressivetradeliberalizationreplacedtheimport substitutionphilosophy.Publicenterpriseswereprivatized,andliberalizationandderegulationofseveral industrieswereundertaken.TheseeffortsculminatedintheNorthAmericanFreeTradeAgreement,which gave stability to the new outward growth strategy. Thus, by late 1993the familiar symptomsof expansionarypoliciesandfallingforeignreservestypicalof thenear-crisisstagesof the pastwerenot present.Theoverallpublicsectorbalancereportedasurplusof 1percentofGDP,comparedtoan11percent deficitin1988,inflationwasnearsingle-digitlevels,andgrossreserveswereatarecordlevelaboveU.S.$26 billion. lWithour apologiesto GarciaMarquez. Addresscorrespondenceto EnriqueG. Mendoza,Divisionof InternationaFl inance,Stop24,FederalReserveBoard,WashingtonD, C20551.Theauthorsarerespectively DistinguishedUniversityProfessor,DepartmentofEconomics,UniversityofMarylandandstaffeconomist intheDivisionofInternationaFl inanceoftheBoardofGovernorsoftheFederalReserveSystem.Theviews expressedherearetheauthors’onlyanddo notreflectthoseof theBoardofGovernorsoftheFederal ReserveSystemorothermembersofitsstaff. WearegratefultoDaveBackus,SaraCalve,FranciscoGil Diaz, SteveKamin,TimKehoe,JohnRogers,JulioSantaella,andJaimeSerraforhelpfulcomments. IThes~bili~tion planforma]]ystartedinDecember, 1987. TheexchangeratewasfixedFebruary29)1988.

2 Priortothecollapse,thelargeexternaldeficitandrealappreciationwerethesubjectofaheated policydebate. TraditionalargumentsoftheDombusch-Rodriguezovershootingmodelwereusedtopush foradevaluation(DornbuschandWerner(1994)),whilethelogicofequilibriumadjustmentsinducedby “ structuralrefom wasusedtodefendthecurrencypeg(Sachsetal.(1995)). However,therewasagreement onthecountry’ssound“fundamentals.”Intheworst-casescenario,adevaluationwouldre-alignthereal exchangerateandclosethecurrentaccountgap,andwouldbefollowedbyawidely-expectedtake-off. Why,then,didMexicoloseitsforeignexchangereserves?whydidthedevaluationtriggermassive runsagainstMexicaninvestments,causingtheworstrecessionofmodemtimes?whydidemergingmarkets world-widefeeltheimpactoftheMexicancrash?Thispaperarguesthatthesephenomenaarecharacteristic ofanewkindofbalance-of-payments(BOP)crisesintheeraoftheglobalcapitalmarket. Inthesenew crises,a country’sfixed-exchange-rateregimebecomesvulnerableas largeimbalancesemergebetween stocksof liquidfinancialassetsandgrossreserves. Bankingfragility,exogenousshifis“inworldcapital flows,andthepolicyresponseintheearlystagesofthecrisiscontributetotheseimbalances.Vulnerability leadsto largedevaluationsanda financialcrashbecauseof “panic”runsagainstfinancialassets. Thus, accordingtothisview,inthenewBOPcrisesflowimbalances(i.e.largecurrentaccountsandovervalued HI exchangerates),whilenotirrelevant,aremuchlesscriticalthan imbalances. Weproposeamechanismlinkingbankingfragilityandworldcapitalflowstocurrencyvulnerability thatproducessimilarpredictionsastheclassicmodelsof BOPcrisesbyKrugman(1979)andObstfeld (1986). Foreignreservesfallfirst graduallyandthen precipitouslyas the currencypeg is abandoned. However,thecrisisdoesnotoriginateinafiscaldeficit,asintheclassicmodels,butintheperceptionthat aIendingboomcausedbypoorly-managedcapitalinflowsleadstobankrunsandabanking-systembailout. TotheextentthattheMexicancrisisisconsistentwiththeaboveargumen~onecouldarguethatthe crisisor“death”wasforetold(orpredictable).2However,theviolenceofthecrisisthateruptedoncethe 2Weac~ow]~get however,thatMexico’sseverepoliticalcrisis,withassassinationsandsocialunrestunseen formorethan60years,alsoplaysaroleinanysensibleexplanationofthefinancialcrash. Ahighdegree ofpoliticaluncertaintyislikelytomagnifitheeffectsatworkinthemodelofBOPcriseswepropose.

3 exchangeratefloatedrequiresfurtherexplanation,sinceinourbanking-bailoutmodeldevaluationmarks theend,notthebeginning,ofthecrisis. Thusweofferasecondmechanismthatlinksthedevaluationto massiverunsagainstdomesticassets.Likeothermechanismsofferedtoexplainthisphenomenon(seeCole andKehoe(1996)),theoneweproposeincorporatesself-fulfillingorherd-behavior(i.e.multipleequilibria) elements.Ourapproachisperhapssimpler,however,becauseweshowthatherdingbyglobalinvestorscan beai~aturaloutcomeofmean-varianceportfoliooptimizationastheglobalmarketgrows. Thelargerthe marke~thesmallertheincentivesforgatheringinformationa,ndhencethemorelikelyinvestorsaretomove inherds.Weargue,however,thatthisherdbehaviormaynotbeenoughtojusti~ thedepthoftheMexican recession.Toexplainthelatter,thepolicyreactiontothefinancialcrashmustbehighlydistortionary. Thepaperisorganizedasfollows.Section2reviews economicdevelopmentsleadingto thedevaluatione,mphasizingfinancialvulnerability.Section3sketchesoutthetwokeycomponentsofour model.Section4provideseconometricevidenceshowingthatMexico’smonetarytransmissionmechanism reflectsco-movementsconsistentwithourview. Thelastsectiondrawsconclusionsandpolicylessons. 2.TheMexicanEconomy:1988-1995 2.1 InDecemberof1987,theMexicangovernmentintroducedanexchange-rate-basesdtabilizationplan. Akeycomponentofthisplanwasasocialagreement(the“pact”)bywhichworkers,firms,andgovernment convenedtomeetregularlytosetprice,wage,andexchange-ratepolicies.Asnotedearlier,theplanwas accompaniedbysweepingstructuralreforms.3Thecurrencywasnotfullypegged,astheexchangerate regimeshifiedfromafixedratetoanadjustablenarrowband.Still,theresponseoftheeconomywastypical ofexchange-rate-basedstabilizationsT.herewasaneconomicboom,monetaryaggregatesgrewfasterthan GDP,therealexchangerateappreciatedmarkedly,andtheexternalimbalanceswidened. Figure1illustratestherealeffectsoftheplan. Economicactivityrecoveredsharplyin GDPandprivateconsumptionroseby about30 percent,whileinvestmentgrewmorethan JSeeAspe(]993)foradetaileddescriptionofthePlan”

4 However,growthduringthisperiodwasuneven,fallingin 1993to 0.6percent,wellbelowthe 1988-92 averageandthe 1994estimate(3.5percentinbothcases).4Thetradedeficitworsenedconsiderablyas . importsgrewmorethan300percentrelativetotheirlevelatthetimetheplanwasintroduced.Bylate1994, thetradebalanceshowedrecorddeficitsexceedingU.S.$1.5billionpermonth(seeFigure2).Figure2also showsthattherealeffectiveexchangerateappreciatedformostofthedurationoftheplan,withtwoepisodes of sharpappreciation--l988and 1991-1993.Atitspeak,therealexchangerateappreciatedbynearly60 percen~relativetotheMarch-1988level.Figure3showsthatM2rosesharplybothinrealconsumer-price termsandinU.S.dollarterms. RealM2doubled,whileM2inU.S.dollarswentupalmost3times. Explainingthesestylizedfactshasbeenthefmusofalargeliterature,recentlyreviewedbyRebelo andVegh(1995).Thispaperdoesnotaimtoaddtothatliterature.However,tomotivatetheideathatour approachtoBOPcriseshassomemerit wediscusshowthestylizedfactschallengeconventionaltheories. ConsiderfirstthehypothesisthatMexicoin1994wentthroughtheslow-growth,declining-inflation stageof theoutput-inflationspiralofanovershootingmodeldrivenbybackwardindexation.Themain weaknessofthisapproachistheevidencethatitskeypolicypredictionwaserroneous.Accordingtothis approach,aonce-and-for-aldlevaluationa,ccompaniedbyadismantlingofindexations,houldhaverealigned therealexchangeratewithoutcomplications.DombuschandWerner(1994)calledforarealdepreciation of 20 percentinMarch. Sincehalfofthat wasproducedpriorto thecrash--asa resultof thenominal depreciationallowedwithintheexistingexchange-ratebandandtheconvergenceof nontradablegoods inflation(seeFigure3)--amodest10-percentmanageddevaluationshouldhavecompletedtheprocess. However,theattemptatamanageddevaluationresultedina runagainstshort-term,dollar-denominated publicdebt(Tesobonos),andtherealcostsofthecreditcrunchthatfollowed in 1995wereenormous. It is also difficultto finddirectevidenceof backwardindexation.sInflationfolloweda highly persistentautoregressiveprocess,asevidentfromFigure3,butthisisnotsufficientevidencetosupportthe ‘Unlessotherwisenoted,datausedinthispaperarefromtheBankofMexico’s find econometricevidenceagainstbackwardindexation.

5 viewofthestickyinflationmodel. Moreover,therewerenoacross-the-boardincreasesinrealwages,as predictedbybackwardindexation.Infact,sinceanexplicitobjectiveofthepactwastheremovalofwage inertiaby settingminimumwageincreases(thebenchmarkforcontractualwages)belowpastinflation, sectorsinwhichsalariesweredirectlylinkedtothepactexperiencedlargerealwagedeclines.AsCalvoand Mendoza(1995)show,theaveragerealwageinmanufacturingrose24percentduring1989-1993,while averagewagesinconstructionandcontractualgovernmentwagesfellmorethan 18percentinrealterms. Equilibriumtheoriesview the trade deficitand the real appreciationas naturaloutcomesof transitionaldynamicsinducedbystructuralreforms.Thesephenomenaaretemporaryandreflectimproved long-rungrowthprospects.Largeinflowsofprivateforeigncapitalandgradualconvergenceofinflation tosingle-digitlevelsaregivenasevidenceinfavorofthisapproach.Inthissetting,thecrisisresultsfrom largeexogenousshocks. Onceagain,a keycriticismofthisviewisthattheDecembercrisisrefutesit. Despitetheresolutionofsomeofthepoliticaluncertaintyafierthepresidentialelections,capitalinflowsdid notreturn,andthe a economy. Thereisalsoquantitativeevidenceshowingthat equilibriumtheorycannoteasilyaccount a inexplainingactualreal-exchange-ratedynamics(seeObstfeld(1995)andMendoza(1995)). Athirdwell-knownexplanationforthedynamicsofexchange-rate-basedstabilizationsislackof policycredibility.EmpiricalstudiesbyCalvoetal.(1995),CalvoandMendoza(1994),MendozaandUribe (1996),andRebeloandVegh(1995)providepromisingresultsinthisarea. However,thesemodelsonly producetheinitialconditionsleadingtoreservelosses,anddonotexplainrunsagainstfinancialassets. 2.2 Inwhatfollows,wetakeasgiventheabove-reviewedfactsandfocusonkeyfinancialaspectsofthe crisis. First,wenotethatconventionaelxplanationsofthecrisisbasedonlargecurrentaccountdeficitsand realappreciationarenotveryconvincingbecausethedecisiontofloatthepesoresultedfromthecountry’s inabilityto meet its obligations.Before the crisis there were two indicatorsof financial

6 a a heldbytheprivatesector,alsoindollars,andgrossreserves. ThesegapsareplottedinFigure4. Beforethe devaluation,the stockof highly-liquidM2 in dollarsreachedU.S.$110billion,far exceedingmaximumresexves.Thedangerisnotsomuchthesizeofthisgap,buttheriskthatsuddenand largeshockstoM2canimplyalargedrainofforeignreserves,thuscompromisingthecumencypeg. Calvo (1995a)showsthatincountrieslikeAustriatheratioofM2toreservesisverylarge,butalsoverystable. ItistheinstabilityoftheratioincountrieslikeMexicoorArgentinathatisdangerous. The largeexpansionof M2 in Mexicois attributedto (1) the 1988-89fro-reachingfinancial Iiberalintion,whicheliminatedtightcreditcontrolsandreserverequirements,doublingtheM2 multiplier from 4.2in to8in1994,(2)thelargecapitalinflowsstartingin 1989,partofwhichtooktheformof bankdeposits,and(3)therecoveryofeconomicactivityandprivateexpenditures.Theimbalancebetween M2andreservesbecameacauseforwncem becauseofthesharpreversalofthelasttwoconditionsthat beganinlate1993.Tosubstantiatethisargument,weconjecturehereandprovelaterthatM2inMexicois influencedbyforeigncapitalflowsandprivateexpenditures.Theeffectofcapitalinflowsisconsistentwith theevidencedocumentedinCalvoetal.(1993),showingthatinflowsintohtin Americaarenegatively relatedtochangesinU.S.interestratesandoutput.Thisexplainspartofthesurgeofcapitalinflowsin1989- 1993and,likewise,predictsoutflowsafierU.S.interestratesroseearlyin 1994andclearsignsof U.S. recoverywerebeingdisplayed.Sincepartoftheinflowsaffected bankdeposits,theadverseeffectofthe outiowsonM2wastobeexpected.Theeffectofprivateexpenditurescanbeinterpretedastheoutcomeof an environment in which M2 is proportionalto expenditure(e.g., cash-in-advance). Thus, rising expendituresaccount in part for rapid M2 growthduring 1988-93,and thereforethe slow-downof consumptionandinvestmentgrowthin 1993shouldbeexpectedto reduceM2growth. Moreover,the effectsofcapitalflowsandexpendituresonM2mayberelatedif,asinMexico,capitalflowsresultinhigh currentaccountdeficits,sothataggregatedemandrises.

7 secondfinancialimbalancecriticalforunderstandingtheexcessivevulnerabilityofthepesois thegapbetweenshort-termpublicdebtandreserves. Theformerincludespeso-denominatedbills(Cetes, Pagafes,andBondes)andTesobonosbecauseunderanearly-fixedexchangeratethemonetaryauthoritymay becalledupontobackallshort-termpublicdebtwithforeignreserves. Threeconsiderationsareimportant inthisregard.First,dollardebthastheadditionaldisadvantagethatitcannotbedeflatedbyadevaluation. Second,whetherthedebtisheldbyforeignordomesticresidentsmayalsomakeadifferencebecausethe formerdonothavealiquiditymotiveforholdingit. Third,theshorterthematuritythelargertheliquidity crisisthatarefusaltorollitovermaygenerate.Asarguedbelow,themassiveswapofpeso-denominated bondsforTesobonosthattookplacein1994wasinappropriateinallthreeaccounts. From 1991to mid-1993short-termpublicdebtwas smallerthan grossreserves.A largedebtreservesimbalancedevelopedin1993-94a,ndendedwiththecollapseofthecurrency,whichoccurredwhen Tesobonosmaturedandcouldnotberolled-over.bWerefertothisphenomenonasa“bonds-ledspeculative attack.’”Whenthecrisiserupted,short-termpublicdebtwasnearly3timeslargerthanreserves.Tesobonos alone,includingcommercialbankholdings,exceededU.S.$22billioninDecemberof 1994,comparedto grossreservesoflessthanU.S.$13billionatthebeginningofthemonth. Byend-December,reservesfell tonearlyU.S.$6billion,wellbelowthecriticalU.S.$1ObillionsetbytheBankofMexico(seeAspe(1995)). Thepolicyresponsetolargepoliticalshocksandanincipientliquiditysqueezeinearly was akeyfactorthatcontributedtowidenthedebt-reservesimbalance.During1994,short-andlong-termpesodenominatedbondsweremassivelyconvertedintoTesobonos.Thefractionoftotalprivately-heldpublic debtallocatedto Tesobonosrosefrom4 to 75percentbetweenend-1993andend-1994(seeFigure5). Foreigninvestorsacquireda largershareofthedebt,asholdingsofMexicanresidentsfellfrom43to 34 blnterestinglyi,nmid1990Mexicoalsoexhibitedalargedebt-resewesimbalance.Inthisinstance, a heldmainlybyMexicanresidents. Moreover,in1990foreigncapitalinflowsweregainingstrength,whiletheoppositeoccurredin 1994. ‘ItisinterestingtonotethataslateasNovember20, 1994,thegovernmentwasnotconcernedbythehuge Tesobonos-Reservesimbalanceoritsabilitytoroll-overdebt(seethestatementbyAspe(1995)).

8 percent.Thus,priortomid-1993,arefusaltorolloverTesobonosandarunonCetescouldhavebeenmet byrunningdowninternationarleserves,andintheeventofadevaluationthemismatchdebt-reserveswould havebeendeflatedbythefallinthedollarvalueofCetes.Bylate1994theseoptionswerenolongerviable. Fromasimpleaccountingperspective,thelargeimbalancebetweendebtandreserveswastheresult ofanoverexpansionofcentralbankcredit. IntheafiermathoftheColosioassassinationtherewasarun againstCetesbybanks(whichenteredinlargerepurchaseagreementswiththecentralbank)andtheprivate sectorthatresultedinalossofaboutU.S.$10billionofforeignreserves.Inresponse,theauthoritiesdecided to(a)sterilizetheeffectonthemonetarybasebyexpandingdomesticcredit,and(b)beginthedollarization ofthedebtbyissuingTesobonostoptiially containtherunonCetes. Thesepolicieswerejustifiedonthe beliefthattheassassinationwasa largetransitoryshock. However,giventhe growingfragilityof the bankingsystem(asdescribedbelow),onecanalsointerpretthemasaimingtopropupcommercialbanks duringtheliquiditysqueezethatsurgedinApril,whenthedifferentialbetweentheinterbankinterestrate (TIIP)andthebenchmark28-dayCeteratewidenedsharply. Theaveragedifferentialclimbedto3.3and percentagepointsinthesecondandthirdquartersof 1994,comparedto 1.3percentinthefirstquarter. InNovembertherewasasecondlargerunonreservesandthecentralbankrespondedagainwith sterilizedinterventionand placementsof Tesobonos. The TIIP-Cetedifferentialrose to 5.6 and 9.5 percentagepointsinNovemberandDecemberrespectively.Thecentralbankfounditincreasinglydifficult toplacenewdebt,untilthecollapsearrivedDecember20.Theattemptatamanageddevaluationfailedin lessthanaday,andcausedafinalrunonreservesandTesobonosthatforcedtheBankofMexicotoallow theexchangeratetofloat. Figure6 showsclearlythesterilized-interventionresponsetotheattacksonreserves. Sterilized interventioneffectivelymeantsharpexpansionsofnetcredittocommercialbanks,withcreditflowsatend- Aprilandend-DecemberofmorethanN$10andN$96billionrespectivelyrelativetoend-December,1993. SinceMexicanbankliabilitiesare of muchshortermaturitythanassets(seeRojas-Suarezand Weisbrod(1995)),andnonperformingassetswererisingrapidly,theexpectationofslowerM2growthcould

9 helptoexplainthewideningoftheTIIP-Cetedifferential.8Ifthecentralbankhadnotenteredinasmany repurchasecontractsof Ceteswithbanksas itdid,a sharprise in sinceinprinciplereservelossescouldhavebeencontainedbya sufficientlylargemonetary contraction,sterilizedinterventionistheclearestproximatecauseforthelossofreserves.Moreover,the criticalsecondattackonreservesinNovembercannotbeeasilyattributedto largepoliticalshocks--infact, the “officialversion”thatthedevaluationwaspromptedbyhostilitiesinChiapasdoesnotmatchAspe’s (1995)accountoftheattack. Theabovediscussiontakesasgivenbankingfragility.Explainingitinfullisbeyondthescopeof thispaper,butwearguethatexternaleffectsandtheperverseincentivesprovidedtobanksbytheexchange ratepolicyandthefinancialreformplayedanimportantrole. Asdescribed,financialreformandforeign capitalinflowscontributedtoalargeincreaseinthe cumencyandmaturityrisks,asthesewere beingimplicitlytransferredtothecentralbank.Asthebanksweakened,thepegbecamelesscrediblesince agentsrealizedthattheintenttocommiteffectivelytoitwouldcompromisethecentralbank’sabilitytoact aslenderoflastresort,Underpressure,thecentralbankeitherwouldchoosetoprotectthebankingsystem usingsterilizedinterventiona,ndhenceweakenthecurrencytothepointthatitmaybedevalued,ortighten monetarypolicyasrequiredtomaintainthepeg,riskingtopushcommercialbanksintobankruptcy. ThesecondmaincomponentofMexico’sBOPcrisisistheabruptchangeofmindonthemeritsof Mexicaninvestmentsbyglobalinvestorsthathelpstorationalizethebonds-ledspeculativeattackinthedays andweeksafierthedevaluationa,ndtheTequilaEffect(i.e.thespill-overofMexico’scrisisintoemerging stockmarketsworldwide).Inearly1994,foreigninvestorsmaintainedtheirexpectationsofstrongMexican ‘By a

10 economicfundamentalsdespitepoliticaluncertaintyandfallingreserves. Thecurrencyinsuranceprovided byTesobonoswasenoughtogetthemtoholdonto Mexicanprivateandpublicdebt(theshareofpublicdebt heldbyforeignersrosefrom55to71percentbetweenJanuaryandOctoberof 1994,seeFigure5). Oneinterpretationofthesecondcomponentisthatinvestors’prospectsonMexico’sfundamentals changed,inpartbecauseoftheongoingpoliticalconflicts. However,mostoftheinformation availableuntiltheendof 1994,includingtheassessmentsofinternationalfinancialorganizations,praised Mexicoasacountrywithsoundpoliciesandsetforstronggrowthonthebasisofitsfar-reachingreforms. [n fac~onthedayofthedevaluationtherewaswideagreementthat itwastheright“medicine”forthe country’sproblems:StanleyFischerarguedthat“theexchangerateadjustments...willhelpreinforcethe economicrecoverythathasbeenevidentsinceearly 1994andsecuretheviabilityof Mexico’sexternal position”(IMFNewsBriefNo.94/18,12/22/94).Moreover,theTequilaEffmtisalsoseriouslyatoddswith theviewthatchangesinfundamentalsdetermineinvestors’behavior. Thereisampleevidence,althoughmostlyanecdotal,oftherunonMexicanfinancialassetsatir the devaluation. TheFinanceMinisterGuillermoOrtiznotedina statementto the AmericanChamberof Commercethat“afierthedevaluation,financialmarketsforM6xicopracticallydisappeared,andtherewas atruestampedeonthoseweeks,inwhich Mexicanpublicandprivatedebtdocumentswereliterally thrownaway”(b July26, 1995,translatedbytheauthors). Mexicanmarketshaveremained extremelyvulnerabletowildrumorsoriginatingathomeorabroad,asvividlyillustratedbythesharpfall inthepesoonNovember3, 1995onunfoundedrumorsofamilitarycoupandtheresignationofMr.Ortiz. AReuterscablequotedonetraderassaying“thedayhasbeenoneoftotalanguish,wedroppedaslowas 7.72pesosperdollarbutnowtherumorshavebeendenied,themarketbreathsagain.” Moreformally,the statisticalevidenceofstrongcontagioneffectsprovidedbyCalvoandReinhart(1995)isconsistentwiththe hypothesisofaherdingpanicinthedaysafiertheMexicancrashandwiththeTequilaEffect.

11 3. AModelofBalance-of-F’aymentCsrisesintheGlobalMarketplace 3.1 In governmentrunsadeficit(definedinrealtermsasd)thatisfullymonetized.Underperfectcapitalmobiliw andperfectforesight,thedomesticinterestrate(i)equalstheinternationalone(i*)duringthefixed-rate period,andthe internationalinterestrate forrealmoneybalancesisgivenbyL(i), L’(i)KO, a andthusnetdomesticcreditproduces anexcesssupplyofmoneythatinflationcannotabsorbbecauseofthecurrencypegandPPP. Theperiodofsmoothreservelossesendsat“switchtime”Twhenreservessuddenlyfalltotheir criticallevel.Thisoccursfortworeasons.First,afierreservesareexhaustedtheycannolongerbeusedto expanddomesticcredittofinancethepublicdeficit. Instead,sincetheexchangeratefloatsandpricesrise inlinewiththedevaluation,seignoragefinancesthedeficit.Defineinflation(equaltothedevaluationrate, duetoPPP)asn. Then,inthefloating-ratephasenL(i*+n~=d. Thus,thereisatTan abruptjumpiniand a sharpfallinmoneydemand. Second,withperfectforesighttheexchangeratecannotjumpatanytime because,ifit did,individualswouldreapunboundedarbitrageprofits.Thus,atTtheexchangeratedoesnot appreciateordepreciate,andthesuddendeclineinreservesisgivenbyAR=Lfi*J-Lfi*+nj. Tisanendogenousoutcomeofthemodeldeterminedby Giveninterestrates,determines the inflationrateneededtofinancethedeficit,andthusthesizeofAR.Thesmoothfallinreservesatrated lasts ifthecollapseispetiectlyanticipated.Themodelfailsdramatically,however,

12 becausein1993therewasafiscalsurplus.Incontrastbankswererapidlyaccumulatingbadloansandthere wereconcernsforthehealthofthefinancialsystem.Therefore,weborrowfromCalvo(1995b)amodelin whichtheanticipationofabanking-systembailouttriggersaKrugman-likeattack. Consideranepisodeofcapitalinflows,partofwhicharechanneledthroughbanks. Ifthecentral bankisalenderoflastresort,depositorslackincentivestomonitorthequalityandcharacteristicsofbank loans.Inparticular,theydonotreacttoagrowingmaturitymismatchbe~een long-termloansandshorttermdeposits.Ifbankers’expectationsofinflationinthelongrunarenotzero,theydonotofferattractive returnsonIong-temdepositsgiventheexpectedhighercost. Thus,theincentivestructuregivesrisetothe maturitymismatch,andasM2risesthecentralbank acquiresshort-ternobligations.Inthiscontext abankrunforcesthecentralbanktobailoutbanks,andthelossofreservesatTisaugmentedbythebailout. Thisresultisdemonstratedasfollows.Weintroducebankswhichliabilitiesarebondsdenominated in localcurrencyandgeneratingno Assumenooperationscostsandnoreserverequirements. Loananddepositinterestratesareidenticalinequilibrium,andbothareequaltoeither(a) beforeT,or (b)i“+n afier7’.Lettheinitialstockofdepositsbezero. Assumethatthediscoveryofanendowmentof naturalresources,ortheexpectedgainsofstructuralreforms,induceagentstoplaninvestmentexpenditures financedbyforeigncapital.Investmentgoodsareimporteduntiltheirprice-adjustedmarginalproductivity equalsi*. DenotetherealsumsinvolvedbyZ. Fundsareintermediatedbybanks,whichextendaninfinitematurityloanfinancedbyinstantmaturitydeposits. Abankrun isa situationinwhichdepositsarewithdrawntobuyforeignreserves. Tomakethis possible,thecentralbankissueshigh-poweredmoneyinexchangeforthebanks’portfolio.Afierabankrun: ThefiscaldeficitisreducedbytheyieldonthebankloanZi”.Furthermore,thelossofreservesatTis: AR = - - - Z >0. (2)

Z T Thelatierimplies,by(2),thatthefallinmoneydemandatTissmaller. Thus,thelossofreservesatTis largerorsmallerthanintheKrugmanmodeldependingontherelativestrengthoftwoopposingforces:(a) anegativeforcerepresentedbyahigherdemandformoneyafierT,and(b)apositiveforcerepresentedby thedirecteffectofa bankrunofsizeZ. Thisdirecteffectdominatesifi’=Obecause,by(2),nafierTwould thenbeindependenotf Z. Thus,bycontinuity,Zmagnifiesthefallinreservesat Sincethisdoesnotaltertheresultthatreservesinitiallydeclineattherate-d,theresuitthatatT aresuddenlydepletedremainsvalid. Inaddition,sinceARislarger,Tissmaller. Therefore,the perceptionofaninevitablebankingcrisisspeedsupthetimingoftheBOPcrisis, Theaboveanalysisisincompleteinmanyrespects.First,itdoesnotexplainwhattriggersthebank run. Banksoffercompetitiveinterestratesanddepositorsgainnothingbywithdrawingdeposits. The problemisthat a bankscouldnotmeettheirobligations.Thistriggersthecentral banktoactaslenderoflastresort,whichpromptsthelossofreserves.9 Second,wedefinedabankrunas completedepletionofbankdeposits,butwhatifonlyashare~ iswithdrawnat 7? Ifnofurtherrunsare anticipated,thisaffectsthetimingofthecrisis,butnotthecentralmessageoftheexperiment.However,if ~ isarbitraryand/orrunsoccurinstages,amultiple-equilibriumsituationcouldarise. Themodelis factisthatM2remainedhigh. Moreover,reservelossesreflectedliquidationofCetes,notwithdrawalsof bankdeposits.Reinterpretingthemodelintermsofa runonCetesisnotdifficult(seeCalvo(1995b)),in essenceone monetaryaggregatesarekepthighdespitestrongpressuresforacollapseismorecomplicated.Bynature, a BOPcrisisreflectsreduceddemandfordomesticfinancialassets. Calvo(1995a)arguesthatsterilized This analysisresemblesDiamondandDybvig(1983). However,theirstoryisnon-monetaryandrelieson technologicalconstraints. Intheircase,thelenderof lastresort(afiscalauthoritythatraiseslump-sum taxes)ispartofthesolution,whilehereitispartoftheproblem.

14 interventioncansustaintemporarilymonetaryaggregates,atthecostofenlargingthelossofreservesatT. 3.2 ViolentDeath:OptimalHerdingBehaviorbyGlobalInvestorsandSe~--l~l1ingProphesies Wenowpresentasimplemean-variance(p-u)modelofoptimalpotiolioallocationthatrationalizes ‘ thelargenegativereactionofglobalinvestorstotheDecemberdevaluation.Inthismodel,rationalinvestors becomeextremelysusceptibleto“small”newsasopportunitiesfordiversificationrise. Thisoccursbecause highly diversified investorshave lower incentivesto acquire informationthan investorswith fewer investmentopportunities.Thisinturnresultshornthefactthatasthenumberofcountriesinwhichtoinvest rises,themarginalgainfrominformation-gatherinegventuallydeclines.Outcomesinwhichtheequilibrium responsetonewsisaself-fulfillingpanicbecomeplausible,andthebehaviorofpolicy-makersbecomesas importantastheirpolicies(i.e.apoorly-handleddevaluationcanhavedisastrouseffects). Theseargumentsareformalizedasfollows.SupposethereareJcountrieswithdifferentinvestment projects,eachindexedbyj. Countryjhastherandomreturn/. Unlessinvestorsspendresourcesinlearning aboutaspecificcountry,r’sareperceivedasi.i.d.processeswithmeanpandvariance02.Therepresentative individualhasaVonNeumann-Morgenstemutilityfunction,U,thatisquadraticinthepotiolio’sreturn. Thus,arisk-averseinvestorinitiallyallocates amountsofwealthacrosscountries.Assuming,without lossofgenerali~,thathehasoneunitofwealth,expectedreturnandvariancearepanda2/Jrespectively. Theinvestorhearsnowa “market”rumorthatcountry1’sreturnhasa newmeanr, r~p,butits varianceisstillU2.Letobetheportfoliosharedevotedtocountries inthesecountries,itsallocationwillbeconstantacrossthem.Thus,theportfolio’smeanandvarianceare: ep - (1- e)r, (3) J - [ I Giventhequadraticutilityfunction.expectedutility,EU,is: i

Eu =ep +(1 - e)r - y 02 l - y > 0. i 1 Maximizing(5)withrespectto6yieldsthefollowingfirst-ordercondition: 1 “— . (6) J-1 - Thisframeworkallowsustoestablishtwokeyresults: Proposition1: of a Proof: ThechangeinportfoliocompositionisgivenbytheimplicitderivativeofOwithrespecttor in(6): 1 1 . . — m J - W. (7) J Y02 YU2 1 a infundsallocatedtocountry1asaproportionoftheoriginalinvestmentbecomesarbitrarylarge. Proposition2:~informationonassetreturnrcanbeacquiredatacost,thebene~tderivedfiomknowing r of Proof: Assumethatbyspendingafixedsum learningaboutcountry1,anindividuallearnstheactual Kin realizationofri beforechoosingaportfolio.Starttheanalysisatthepointinwhich isrevealed,anddenote itsrealizationbyr.Thevarianceonthereturnincountry1isbydefinitionzero. oisnowgivenby: e p -‘(J - 1). (8) y Rulingoutshortsales, 8=Oifr ~p,ando= 1ifrs where givenby: =p - —y . (9) Thus,interiorsolutionsrequire r < p. r~i~riseswith andconvergesto pasJgrows withoutbound.

16 By (9), in the !imitas~ reachesODin,formationgatheringpaysoff (ignoring onlyifr > p.]o K), However, for smallJ, information-gatheringpays off even thoughr s p (but as long as r z ~ Furthermore,inallcases,informationgatheringpaysoffif r> p.Thus,(a)ifexpostr > p,utilityisthe sameforhigh-andlow-diversifiedinvestorsand(b)ifexpost p,thenonlythelow-diversifiedinvestor hasachanceto gainfromhavingpaidforinformationoncountry1. Therefore,noticingthatexpectedutility increaseswithJ, themar~inalgainfrominformation-gatheringeventuallyfallsoffasJrises. Insum,Propositions1and2 showthat(1) u a Theabovemodelrationalizesmassiverunsagainstdomesticassetsasaresultofslightchangesin profitabilityexpectations.Thus,themodelmayexplainthesuddendumpingofMexico’sassetsbyglobal investorsinresponsetosmallnegativeshocks.Thedevaluationcouldqualifiasoneofsuchshocksbecause itmayhavesignaledthatthenewadministrationwaslessreliablethanthepreviousone,despiteitsintentions to maintainpreviouspolicyguidelines.The model,however,cannotexplainwhythe situationdid not stabilizeafieradropinstockmarketprices,whichwouldhaverestoredtheprofitabilityofdomesticassets. Inourview,theeffectoftheinitialrunwasmagnifiedbythelargestockofmaturingTesobonos. As the lattercameupfor redemptiontheonlyoptionIefiforgovernment,shortof default,wasto seek refinancing.However,thisrequiredinvestorstobepersuadedthatthegovernmentcouldoffer higherratesofretumonnewly-issuedTesobonos.Unfortunately,thelatterisnotaccomplishedbysimply offeringahigherrateofretum. Inparticular,ifsolvencyisatstake,adeclineinthepriceofTesobonosmay havetheoppositeeffectandaddtotheinvestors’misgivings. Howcouldsolvencybecomeanissuewhen,onthewhole,Mexicohadbeenpursuingprudent,ifnot tighqpolicies?Hereiswhereweenterintotherealmofself-fulfillingprophesies.Apossiblestoryisthat IOIf~ is 1argebutlessth~ infini~, information-gatheringisvaluabieforr in < r< P. However,sincermimconvergesto pasJgrows, theprobabilityofrmin<r c pgoestozeroasJdiverges to infinity.Hence,wecandisregardtheaboveopenintervalasJbecomes verylarge.

Mexicowasforcedtoadoptadraconianadjustmentprograminordertogetinternationalsupportand,thus, avoidoutrightdefaultonTesobonos. Theexpectationof socialunrestinreactiontotherequiredtough austeritymeasures,coupledwiththeseverepoliticalcrisisandthetensionsrelatedtotheChiapasuprising, may have a shock. Theirconsequentrefusalto rollover Tesobonosatareasonableinterestratemadethestrongadjustmentpolicyinevitable,causingtheeconomy tomovetoadeep-crisisequilibrium.Calvo(1995b)providesaformalmodelyieldingthisresult. 4.Empirical EvidenceontheDeath Foretoldand Monetary Transmission OurmodelofBOPcrisesreliesonamonetarytransmissionmechanisminwhichexpendituresand capitalflowsaffectM2. Thissectionprovidessomeeconometricevidencein favorof thisview,and quantifiestheextentof the reservelossesthat it predicted. Thegoalis notto assessthe relevanceof competingtheoriesofthecrash,butsimplytoprovidesomeempiricalevidenceconsistentwithourtheory. 4.1 KaminandRogers(1996)showthatanerror-correctionmodelofquarterlymoneydemandfitswell Mexicandataoverthelastdecade. regressions(afierthe 1988financialreform),actualrealM2exceededsignificantlypredictedM2in 1994. This thestablerelationshipfound intheda~ wasweakening.LineIinTable1reproducestheKamin-Rogersmodel.Themodelexplainsthe quarterlychangeinrealM2asafunctionofthe28-dayCeterate,theannualchangeintheCeterate,andthe first laginthe logarithmof velocityof circulation.1’Thespecificationassumesa strongco-integrating relationshipbetweenvelocityandtheCeteratethatcannotberejectedbythedata. WeidentifytwoadditionalfeaturesofthebehaviorofM2thatarecentralfromtheperspectiveof ourmodel:(1)privateexpendituresaresignificantdeterminantsofM2,and(2)variablesthatproxytheeffect offoreigncapitalinflowsweresignificantforpredictingM2uptothefirstquarterof 1994.Theseresults l’WeusequarterlynationalaccountsindicesprovidedinZndicadoresEcon6micos. Velocityisproxiedas the ratio of theGDPindexdividedoverreal M2at consumerprices.Thisincreasesthe coefficienton velocityfrom0.15intheKamin-Rogersmodelto0.19,withoutaffectingothercoefficientestimates.

arereportedinlinesII-VIofTable1.Consumptionandinvestmentprovideusefuladditionalinformation forexplainingM2notincludedinthevariablesusedinlineI. Examinationofresidualplotsshowsthatthis new informationaccountsforpaxtof the 1994residualsofthefirstregression. Note,however,that in reversedregressionsrealM2isasignificantdeterminantofconsumptionandinvestment.Weexploreissues ofsimultaneityanddynamicinteractionbetweenvariableslaterinthesection. LinesIV-VIpresentthesecondresult.During1 theperiodoffinancialLiberalization andlargecapitalinflows,twowidely-usedindicatorsofcapitalinflows(theinterestrateof3-monthU.S. Tbillsorthestockofgrossforeignreserves)weresignificantdeterminantsofM2. Theshort-termelasticity of M2withrespectto theTbillrateisnearly4timesiargerthanthatwithrespecttotheCeterate. This resultsinpartfromthelongermaturityofTbills(3months)comparedtoCetes(28days). Moreover,the Tbillrateismeanttocapturenotonlyanopportunity-cosfteature,butisalsoaproxyfortheeffectsofcapital inflowschanneledthroughbanks.ThisfollowsfromtheempiricalevidenceinCalvoet.al(1993)showing thattheTbil!rateisamajordeterminantofcapitalinflows.Alternatively,lineVreportsresultsthatdetect asignificanteffectofcapitalinflowsonM2usinggoss reservesasamoredirectmeasureofthoseinflows. Our theoreticalargumentstendto viewthelinkbetweenM2andcapitalinflowsasa featureof moneydemand,butintheempiricalteststheymayalsocaptureaneffectthroughmoneysupply.Theintent ofthetestsistoshowthatpriorto 1994therewasastablelinkbetweenM2andcapitalinflows,andthaton thatbasisasharpcontractionofM2waspredictablein1994.Thekeyissueofidenti~ingwhetherthiswas a supplyordemandeffectisleftforfuturework. Note,however,thatunderapurecurrencypegHume’s specie-flowmechanismimpliesthatmonetaryaggregatescannotrisesystematicallyinresponsetocapital inflows moneydemandrises. Inpractice,theBankofMexicointervenedattimestolessentheeffect ofcapitalinflowsondomesticprices,andthustheinflowsmayhaveaffectedmoneysupply. Afierthefirstquarterof 1994,capitalflowsarenolongersignificantforexplainingM2. Thereis suchalargestructuralbreakintheaftermathofthefirstattackonresemesthatestimatingtheregressions inlinesIV-VIaddingonlyanextraqumer ofdatafailstoproducestatisticallysignificantcoefficientson

i 19 theTbillrateorforeignreserves.EstimationofrecursivecoefficientsshowsthattheTbillelasticityinline IVjumpshornaround-2toOin 1994:2.Thisevidenceiscriticaltosupporttheargumentthatthepolicies ofsterilizedinterventionandTesobonoplacementsplayedakeyroleinthecrisis. Figure7plotsthegrowth andlevelforecastofM2basedontheobservedvaluesoftheright-hand-sidevariablesintheregressionof lineIV,togetherwithactualrealM2. In a squeezethatwas wouldhavelefionlyU.S.$3billionofreserves inexcessofthenoted“critical”levelofU.S.$10billion.Therefore,thereisstrongevidencethatMexicowas treadingonverydangerousfinancialgroundsandthataBOPcrisiswasbynomeansunlikely.Onthisbasis, weconcludethatthecrisiswashighlypredictable.Deathwasforetold. Additionalimportantevidenceontheeffectsofsterilizationisobtainedbycomparingthestaticand dynamicrelationshipbetweennetdomesticcreditofthecentralbank,reservesandbasemoneydemand,all valuedinMexicanpesos.Understerilizedinterventionc,reditexpansionisdeterminedbyexogenousreserve movementsandsystemicbasemoneydemandchanges.Accordinglyw, efoundstrongevidenceofa one-toonestaticlinkbetweencurrentnetdomesticcreditandforeignreservesinaregressioninwhichchangesin foreignreserves,the Cete rate, and seasonaldummiesexplainchangesin centralbank credit--using instrumentalvariablesandmonthlydatafromJanuary,1992toApril,1995.12Incontrast,whendynamic elementsareconsidered,usingmultivariateGranger-causalityregressionswithtwolagsandcontrollingfor changesintheCeterate,pastchangesinnetdomesticcreditpredictcurrentreservechanges,buttheopposite ‘2Theanalysisislimitedtothissamplebecause,duetorecentmethodologicalchanges,consistentdataare availableasof 1992.Theinstrumentsare3lagsofreservesandtheCeterate,thecurrentandlaggedTbill rate,andtheseasonaldummies.Thecoefficientonthechangeinreservesis-0.96withat-statisticof-23.54.

isnottrue.13Thus,thecontemporaneouslinkbetweenreservesandcentralbankcreditreflectssterilized interventionb,utthedynamicrelationshipisinlinewiththeearlierstagesofaBOPcrisis,inwhichreserves fallgraduallyinresponsetoexcessivenetdomesticcreditexpansion. Thesingle-equationanalysisofM2featuredalong-runcointegratingrelationshipbetweenM2and itsdeterminantsa,ndahigh-frequencyrelationshipbetweenfirstdifferencesofthevariables.However,the interactionsemphasizedbyourtheorymaybefeaturesofmacroeconomicdynamicsatthe“businesscycle” frequency,whichliesbetweenthehighfrequencyoffirst-differenceddataandthelong-runtrendsofcointegratingvectors.Moreover,single-equationanalysisignorestheinteractionamongseveralvariablesthat arelikelytobeendogenousT. oaddresstheseissues,weproceedtoexaminewhetherthelinkagesbetween M2,capitalflows,andexpendituresfoundearlierarefeaturesofMexicanbusinesscycles,andwhetherthe directionofstatisticalcausalityinthedataisconsistentwithourinterpretationoncedynamicinteractions areallowed.Thisanalysisalsoprovidesthemeansforstudyingthemonetarytransmissionmechanismof themaininstrumentofmonetarypolicy--theCeterate--inaVARsettingsimilartothosetypicaloftherecent literatureonmonetarytransmission(seeSims(1992)). Awareofthedebatesonmethodsforisolatingcyclicalcomponentsoftimeseries,weexaminedthe datausingbothaquadratictimetrend(QT)andtheapproximateBaxter-Kingband-pass(BP)filter(see CalvoandMendoza(1995)fordetails).QTregressionscovertheperiod1983:1-1994:4andincludeseasonal dummies.Augmented-Dickey-FulleUrnitRoottestsconfirmedthatQTcyclicalcomponentsarestationary. TheBP between5and24quarters,or 1and6years,using6 leadsandlagsin centered,weightedmovingaverages. ThistakesoutseasonalcomponentsandreflectsMexico’ssix-year businesscycle.TheBPfilterwascalculatedextendingthesampleto 1980:1sothatbothfiltersproduce48 usableobservations.Sincethetwofiltersproducedsimilarco-movementsa,lthoughcycleswiththeBPfilter ‘3TheChi-squarestatisticforcreditinthereservesregressionis5.46,withaprobabilityvalueof615percent. Thesamestatisticforreservesinthecreditregressionis3.28,withaprobabilityvalueof 19.4percent.

21 weresmootherandsmaller,weconductmostoftheanalysisbelowusingthesimplerquadratictrend. Figure8plotscyclicalcomponentsofconsumption,investmen~andrealM2. Thescatterdiagrams inthetoppartoftheFigureshowastrongpositiverelationshipbetweenM2andinvestmentorconsumption. Thecha~ belowplotsthethreeseriesandshowssizableboomsinexpendituresandM2in 1984-85and 1990-93,followingtheimplementationofstabilizationprograms. Notethatbymid-1993expendituresand M2weremovingbelowtrend.Contemporaneoucsorrelationcoefficientsconfirmthestrongco-movements relatingM2andexpenditures.ThecorrelationsofM2withconsumptionandinvestmentare0.48and0.44 respectively. Thereisalsoa strongnegativecorrelationbetweenM2andtheTbillrate(at -0.5)which providesfurtherevidencesuggestingthatamajordrivingforceofcapitalinflowsiscloselyrelatedtoM2. We conductedmultivariateGrangercausalityteststo determinewhethertheorderingin which variablescontaininformationrelevanttopredicteachotherisconsistentwithourviewofthetransmission mechanism,oncetheinformationofotherrelevantvariablesisconsidered.Thetestsexaminethepredictive poweroflaggedGDP,realM2(M2/P),interestrates(CETEandUSTBILL),expenditures(consumption, C,orinvestment,I),thetermsoftrade(TOT),andtherealexchangerate(RER),toexplaincurrentvalues of M2/P,C,I,GDP,TB,andCETE. Theregressionswereestimatedusing2-and4-lagstructures,and replacingGDPwithTOTtoavoidpossiblesimultaneitybiasbetweenC,I,and C 2 2 Theseresultsshowthattherapidcyclicalexpansionofrealbalancescannotbe explainedbythetraditionalscaleeffectattributedto GDPgrowth,andinsteadwaspredictedbycapital inflows(proxiedby a 14CalvoandMendoza(1995)providecompleteresultsofthetests. 15GDpand~R arealmostsignificant,atthe 10percentlevelwith4lags,topredictCETE. T’husC, ETE mayreflectinpartthepolicy-makers’reactiontothepaceofeconomicactivityandrealappreciation.

22 withthemodelsofTab!e1.butthereisalsoanindirecteffectthatdependsontheeffectofUSTBILLon CETE.Inspectionofcoeff~cienetstimatesshowsthatdeclinesintheTbillratepredictincreasesintheCete rate, in accordancewiththeevidenceof limitedsterilizationof capitalinflowsbefore1994. Therapid growthofM2contributessignificantlytopredicttheboomsinGDP,C,andI,andthewideningtradedeficit. TheCeterateandM2alsohelppredictthetradedeficit,addingto conventionaleffectsviafasterGDP growthandrealappreciationthatarestatisticallysignificant. These causalitylinkssupporta cash-in-advance(CIA) frameworkas a first approximationto modelingmoneydemand.ThisframeworkisconsistentwithM2/PGranger-causingexpendituresbecause initmoneycarriedoverfromthepastpaysforcurrentpurchases.Thelimitedevidenceoftwo-waycausality betweenCandM2/P,andthesignificanteffectsofCandI intheM2models,arenotinconsistentwiththis view, since the relevanttiming of transactionsvaries and, if the CIA constraintis binding,current expendituresshouldstillbe relatedtomoneyholdings. Causalityfrommoneytoexpendituresmayalsoreflectcreditconstraints.Consider,forexample, thatdespitefinancialreformequilibriumcreditrationingcontinuesbecauseofincompleteinformationinan adverse-selectionsetting, with low-income“good”borrowers unable to offer enough collateralto differentiatethemselvesfrom“bad”borrowers.Anexpansionofthedepositbasewouldeasecreditrationing andrelaxcreditconstraints.Thus,ariseinM2/PshouldGranger-cause(i.e.pre-date)expenditures. The causalitytests are also usefulfor determiningthe degreeof statisticalerogeneityof the variables,which inturn iskeyinthedesignoftheVARmodelto beexaminednext.TheVARsystem includesfiveendogenousvariables(CETE,M2/P,RER,GDP,and TB)and two exogenousvariables (USTBILLandTOT). TBisusedasaproxyforexpenditurestokeepthesystemsmall,butessentiallythe sameresultsareproducedusingI. Similarly,thenearly-fixedexchangerateandstableU.S.inflationimply thatRecaptures boththedynamicsofrealappreciationandCPIinflation.AICstatisticsofthecausality tests suggestthata2-quarterlagstructureisaparsimoniousrepresentationfortheVAR. Moreover,the evidencethat CETE is domesticvariables,and that M2/P is exogenousto GDP and

23 expenditures,favors a specificationin which CETE and M2/P are first in the ordering inlinewithresultsofvariancedecompositions.CETEandM2/Pexplain70percent oftheirownforecasterrorsafier24quarters--althoughCETEexplains17percentoftheM2/Perror. Figure9plotstheVAR’Simpulseresponsefunctionstol-percentstandard-deviationshockstoCETE andM2/P.CETEimpulseresponsesreflectthetransmissionmechanismofmone~ policy,sincetheCete rate isexogenousto anyotherMexicanvariable. Moreover,sinceanelementofthatpolicywaspartial sterilizationofcapitalinflows,theCETEshockalsoreflectsinparttheeffectsofashocktocapitalinflows. AnincreaseinCETEhassignificantnegativeeffectsonM2/PandGDP,andcausesanimprovementinTB. SinceTBimprovesasGDPfalls,absorption,andhenceprivateexpenditures,arefalling. Thus,incontrast totheM2models,theVARaccountsforthesteadyexpansionofM2andexpendituresin 1994asresulting fromtheinterest-rateeffectofsterilizedintervention--keepingCETEbelowtrendinducesM2/PandTBto remainabovetrend. Interestingly,theVAR’SresponsetoaCETEshockalsosuggeststhatasharprisein theCeterate,asanalternativetosterilizedintervention,wouldhavebeenaneffectivepolicytoaddressthe vulnerabilityexpressedinthelargeM2imbalance,aswellastherealappreciationandthelargetradedeficit. Thedirecteffectofshockstocapitalinflowschanneledthroughbanks,ortheliquidity a 9 a a 2 ConclusionsandPolicyLessons This a

24 rateregimeshavecollapsedinMexico,butthelastepisodewasuniqueinthatitoccurreddespitethetight policiespursuedin1988-1993.ItwasassociatedwithamassiverunagainstTesobonosbyglobalinvestors, anditspilledoverintoemergingmarketsworld-wide.Theseverecreditcontractionthatfollowedin1995 causedtheworserecessioninMexicanhistoryandIefithebankingsystematthebrinkofcollapse.These eventsshockedadvocatesofthedevaluationofthepesoasameansforcorrectingtheovervaluationofthe realexchangerateandthelargecurrentaccountdeficit. OurdiscussionshowsthatatheoryofMexico-typecrisescouldbebasedontwokeycomponents. First,amodelexplaininghowbankingfragilityandglobalcapitalflowscontributetocauselargeimbalances betweenstocksofliquid a a Second,amodelofhowdevaluationinsuchavulnerable situationleadstomassiverunsagainstfinancialassets,seeminglyinconsistenwt ithacountry’sfundamentals. Weproposetwomodelsthat partofthenewtheory.Amodelinwhichanticipationofabankingsystembailoutleadstoanattackonforeignreserves,andamodelinwhichstandardpotiolio diversification leadstoherdingbehaviorastheglobalmarketgrowsandgainsfrominformationgatheringdiminish. Econometricanalysisprovidesevidenceconsistentwithourinterpretationoftheprocessleadingto currencyvulnerabili~inMexico.Thereisstrongevidencethatglobalcapitalflowshavesignificanteffects onM2andthatthereisastronglinkbetweenmoneyandprivateexpenditures.AmodelofM2thatcaptures theseeffectspredictsasharpcontractionofthequantityofmoneyin1994thatexplainsmorethan1/2ofthe observedloss in foreignreserves. However,the centralbank’spoliciesof sterilizedinterventionand TesobonoplacementscausedaseverestructuralchangeinthebehaviorofM2andmanagedtosustainits growthuntilthedevaluation.Theeffectsofglobalcapitalflowsandinterestratepolicyarefirtherexamined usingaVAR. TheevidenceoftheinfluenceofworldcapitalmarketsonMexico’sM2andthecloselink betweenM2andexpendituresisrobusttotheadditionoftheVAR’Scomplexdynamicinteractions. Therearesomecriticalpolicylessonstobelearned.Firs~closeattentionmustbepaidtoindicators of financialvulnerability. Second,it mustbe recognizedthatthe possibilityof herdingbehaviorisan

25 inescapablefeatureoftheglobalcapitalmarketthatcomesalongwithitswell-knownbenefits.Thecosts ofthisherdingbehavioraremagnifiedifthestockofshort-termpublicdebtislargerelativetointernational reserves. Third,early-warningsystemsaimedatidentifyingandcontainingthesourcesofthedestabilizing dynamicsofexchange-rate-basedstabilizationsneedtobedevelopedandimplemented.

26 References Aspe,Pedro,(1993),EconomicTransformation:TheMexicanW~, MITPress,Cambridge:MA. ,(1995),“Mdxicoen 1994:LasRazonesdelaPoliticaCambiaria,” Friday,July14,p.2A.,No.587. BancodeMexico, BancodeMexico,MexicoCity,Mexico. Calve,GuillermoA.(1995a),“CapitalFlowsandMacroeconomicManagement:TequilaLessons,”mimeo, CenterforInternationalEconomics(CIE).UniversityofMaryland. ,(1995b),“VarietiesofCapital-MarketCrises,”mimeo,CIE,UniversityofMaryland. ,andEnriqueG.Mendoza(1995),“ReflectionsonMexico’sBalanceofPaymentsCrisis:AChronicle ofaDeathForetold,”mimeo,CIE,UniversityofMaryland. andEnriqueG.Mendoza(1994),“TradeReformsofUncertainDurationandRealUncertainty:A FirstApproximation,” LeonardoLeiderman,andCarmenM.Reinhart(1993),“CapitalInflowsandRealExchangeRate AppreciationinLatinAmerica,” vol. Calve,SaraandCarmenM.Reinhart“CapitalFlowsto LatinAmerica:IsthereEvidenceofContagion Effects.”Mimeo,WorldBank,WashingtonD.C.,1995. Cole,HaroldL.,andTimothyJ.Kehoe,(1996)“Self-FulfillingDebtCrises,”mimeo,FederalReserveBank ofMinneapolis. Diamond,DouglasW.andPhilipH.Dybvig(1983),“BankRuns,DepositInsurance,andLiquidity,” Dombusch,Rudigerand AlejandroWerner,(1994),“Mexico:Stabilization,Reform,andNo Growth,” 1 BrookingsInstitution,WashingtonD.C. I Edwards,Sebastian,(1993),“ExchangeRatesasNominalAnchors,” KielInstituteofWorldEconomics,pp.1-32.

y Posguerra,1946-1976,FondodeCulturaEconomic%MexicoCi~, Mexico. Krugman,Paul,(1979),“AModelofBalance-of-PaymentCsrises,”JournalofMoney,Credit,andBanking, VO].11,pp.311-325. InternationalEconomicReview, 36,No. 1,pp. , “TheSyndromeofExchange-Rate-BasedStabilizations:AnAssessment oftheCredibilityHypothesisunderUncertainty,”mimeo,FederalReserveBoard. Obstfeld,Maurice(1986),“RationalandSelf-FulfillingBalance-of-paymentsCrises,” pp.72-81. ,(1995),“InternationaCl urrencyExperience:NewLessonsandLessonsRelearned,”unpublished manuscript,DepartmentofEconomics,UniversityofCaliforniaBerkeley. Rebelo,SergioT. andCarlosA. Vegh,(1995)“RealEffectsof Exchange-Rate-BasedStabilization:An AnalysisofCompetingTheories,”mimeo,DepartmentofEconomics,UniversityofRochester. Rojas-Suarez,Lilianaand StevenR. Weisbrod, (1995)“BankingCrisesand America:The1980sand 1990s,”mimeo,Inter-AmericanDevelopment Sachs,Jeffrey,AaronTornell,and AndresVelasco(1995),“Lessonsfrom Mexico,”mimeo,Harvard University. Santaella.JulioA. andAbrahamE.Vela,(1994)“The1987MexicanInflation-StabilizationM: atching StylizedFactsandTheory,”unpublished,InternationalMonetaryFund,WashingtonD.C. Sims,ChristopherA.“InterpretingtheMacroeconomicTimeSeriesFacts:theEffectsofMonetaryPolicy.” EuroDeanEconomicReview,1992,(36),pp.975-1OOO.

Table1. Quarterly I 0.762 0.040 83:1-94:4 II 0.800 0.037 83:1-94:4 (4.190)’ 0.778 0.039 83:1-94:4 Iv -1.989 0.861 0.032 88:2-94:1 (3.789)” f-2,061)** v 0.053 0.868 0.031 88:2-94:1 (2.197)** -0.988 0.520 0.026 88:06-94:02 (2.607)” (3.026)’ (-1.761)*** change 1 5 ‘% .) .

ofthe 1988-1994 Exchan~e-Rate-BaSed Stabilizatic)n Plan (1988:1=100) 180 160 /h\ /\ /; /. 140 . - I \ 120 1O(I 8C ,/’ 6C — GDP ---- fixedinvestment -----= privateconsumption—-—ogovernmentconsumption . . i ., . r . . / la .. . \ \ t \ 50 0 80- 82- 84 86 88 92 94 exports ---- imports —

50C o -1500 -2000 I I I I E I 1 v Q t 85 86 87 88 89 90 91 92 93 94 95 — tradebalance(monthlyflow) t 160 6 140 . 3 4 9 120 2 0 80 85 86 87 88 89 90 91 92 93 94 95 nominalexchangerate ---- realeffectiveexchangerate

i 160000 c 85 86 87 88 89 90 91 92 93 94 95 — M2indollars ---- realM2 (1988 prices) 50 c 86 87 88 89 90 91 92 93 94 95 CPI total ----- CPI durables(tradables) ––- CPI setvices(nontradables)

i . 85 86 87 88 89 90 91 92 93 94 ’95 — short-termpublicdebt ---- grossresemes 0 85 86 87 88 89 90 91 92 93 94 95 M2 ---- grossreserves

I I 33 -%4- -\\ / \ I \ \ \ \ % t f ‘- --- /- \\ \ 20 \ -------- -$... --—- 0 1 Bondes —-– -- / # 0 -- - --- ,~ 50 94:01 94:03 94:05 94:07 94:09 94:11 —

i 34 - I L e 8 l o 4 uE 8 \ ; : l ’ 0 1 0 0 0 0 m i sosad soAanu suo!l[!q . L

Actual 0.3 0.2 , 0,1 0.0 -0.1 -0.2 -0.3 -0.4 I B I 1 I I I 1 1 1 I I # I 1 I 1 I I F I I I c s I actual growthrate ---- predictedgrowth rate 80000 60000., 1988 ‘1989 ‘1990 .1991 .1992 ‘1993 ‘1994 ‘1995 actual end-of-period stock ---- predictedend-of-period stock

i . of and Investment 0.1 + + $ + + + + + + + + + ++ + + + + + + + + -0.1 0.0 0.1 I I I I I 0.1 0.0 -0.1 + 13 , 1 I 8 , 4 a1 8 m 5 , , I 8 , 6 , , I 8 , 7 , , I 8 1 8 1 a1 89 , , I9 1 0 , , 19 1 1 1 , I92 , 1 1 9 v 3 , 1 19 , 4 , —

9 I Shocks Real M2 Cete rate 0,04 :~ 0.03 \ \ s-. t 0.02 \ \ \ \ \ \ \ \ \l 4- \lll- -. -------- I \ \ * 0- I , q , 1 , , I 1 a , 1 , , I , 1 , 1 1 I , 6 8 10 12 14 16 18 20 22 24 8 10 12 14 16 18 20 22 24 GDP Real Exchange Rate o.oo2-: :“’’’’*--- II ‘** .\ l* % . *\ - -0.002- -0.003- -0.03! , , , 1 , , 1 1 , , v 1 1 , , n 1 v m * 1 6 8 10 12 14 16 18 20 22 : 8 10 12 14 16 18 20 22 24 Trade Balance -0.02 ! J 1 r , I , T r , , , , , , n T , , 8 10 12 14 16 18 20 22 24 — Cete

International Finance Discussion Papers * U uthor@ A Chronicle GuillermoA. Calvo of DeathForetold EnriqueG. Mendoza TheTwinCrises: GracielaL.Kaminsky CarmenM.Reinhart GracielaL. Kaminsky a LeonardoLeiderman a CarolC. Bertaut MichaelHaliassos William in a

International Finance Discussion Papers [FDP Number ModellinglnflationinAustralia A Excess 521 The