Asymmetries and common cycles in Latin America evi

1AsymmetriesandcommoncyclesinLatinAmerica:evidencefromMarkov-switchingmodelsPabloMejía-Reyes*SchoolofEconomicStudiesTheUniversityofManchesterManchester,UnitedKingdomElColegioMexiquense,A.C.Ex-HaciendaSantaCruzdeLosPatosZinacantepec,MéxicoNovember1999AbstractMarkov-switchingmodelsareestimatedtoanalysedifferencesbetweenexpansionsandcontractionsperformances.Ingeneral,univariateanalysisresultsimplythatrecessionsaredeeperinabsolutemagnitude,lesspersistence,andmorevolatilethanexpansions.ThemagnitudeofourestimationsandthedirectionofsomeasymmetriessupporttheideathatbusinesscyclesinLatinAmericaaredifferenttobusinesscyclesintheUnitedStatesandotherdevelopedcountries.Onaninternationalperspective,itisfoundthatalthoughthereisnotacommonLatinAmericancycle,thereexistssomeevidenceaboutcommonregimeshiftsandcommoncyclesbetweenBrazil-PeruandChile-UnitedStates.However,itseemsthattheircausesareverydifferentandrelatedtocommonshocksandsimilarpolicies.Therefore,itisconcludedthatindividualbusinesscyclesarelargelyindependentinLatinAmerica.JELclassification:C22,C32,E32.Keywords:Markov-switchingmodels,Businesscycles,Internationalbusinesscycles.*TheauthorwouldliketoacknowledgethesupervisionofProfs.DeniseOsbornandKeithBlackburnaswellasthecommentsandhelpintheuseoftheprogrammesofMarianneSensier.Also,hethankstheConsejoNacionaldeCienciayTecnologíaofMéxicoforfinancialsupport.Theusualdisclaimerapplies.Correspondence:SchoolofEconomicStudies,TheUniversityofManchester,OxfordRoad,Manchester,M139PL,UK.E-mail:[e-mail]:00441612754934.Faxnumber:00441612754928.21.IntroductionPreviousstudiesofLatinAmericancyclicalfluctuationshavefoundthatrealGDPseriesexhibitsignificantpersistenceandthatcurrentshockshavepermanenteffects(Mejía-ReyesandHernández-Veleros,1998;Ruprah,1991;CuddingtonandUrzúa,1988).Also,theyhaveshownthatsupplyshocksandrealfactorstendtodominateeconomicfluctuationsevenintheshort-run(HoffmaisterandRoldós,1996,1997;KydlandandZarazaga,1997).Recently,evidenceofnonlinearityandasymmetriesovertheColombianbusinesscyclehasbeenfound(Mora,1997)andalsoasymmetriesoverthebusinesscyclesofanumberofLatinAmericancountries(Mejía-Reyes,1999).Ontheotherhand,thereareonlyfewstudiesthataddressinternationalbusinesscyclesforLatinAmericaandtheresultsarenotconclusive.Byusingdifferentmethodologies,ithasbeenfoundthatsignificantcorrelationsarelimitedtosmallgroupsofcountries(EngelandIssler,1993;ArnaudoandJacobo,1997),thateconomicfluctuationsarehighlyvariableandnottimeuniform(ArnaudoandJacobo,1997),andthatmostcorrelationsbecomenon-significantwhenthepost-debtcrisisperiodisincludedinthesample(IguíñizandAguilar,1998).Also,evidencepointstothesynchronisationofbusinesscyclesregimesforonlysomecountries(Mejía-Reyes,1999).ThesestudiesrepresentsignificantadvancesforthecomprehensionofLatinAmericanbusinesscycles.However,onlyafewofthemaddressissuesofnonlinearitiesandregimescharacteristics.Inthiscontext,weapplytheapproachproposedbyHamilton(1989)toanalysenonlinearitiesandasymmetriesinthebusinesscycleofeightLatinAmericancountriesandtheUnitedStates.Inparticular,weareinterestedinthedifferencesbetweenexpansionsandcontractionswithrespecttotheirmagnitude,persistence,durationandvolatility.Inaddition,weevaluatetheabilityoftheestimatedMarkov-switchingmodelstoidentifyperiodsofexpansionandcontractionintheprocessofeconomicgrowthbycomparingtheregimesimpliedbythiskindofmodelswiththoseobtainedbyMejía-Reyes(1999)(hereafter,MR)fromtheapplicationofaclassicalbusinesscyclesapproach.Also,weanalysetheinternationalnatureofbusinesscyclesbyapplyingmultivariateMarkov-switchingmodelswhichallowustomodel3economicgrowthacrosscountriesasajointstochasticprocesssothatwecanidentifycommonregimeshiftsandconsequentlycommoncycles.Thispaperisstructuredasfollows.InthenextsectionwepresentthemainfeaturesoftheMarkov-switchingmodelsappliedinthispaper.Thenwediscussthestrategyofspecificationandestimation,andreporttheresults.Finally,wesummariseourresultsandstatesomegeneralconclusions.2.Markov-SwitchingprocessesasstochasticbusinesscyclemodelsTheMarkov-switchingautoregressivemodelproposedbyHamilton(1989)hasbecomeincreasinglypopularfortheempiricalanalysisofmacroeconomicfluctuations(forexampleGoodwin,1993;KählerandMarnet,1992;Krolzig,1996and1997a;andClementsandKrolzig,1998).Incontrasttopreviousapproaches,Hamiltonspecifiesthefirstdifferenceoftheobservedseries(GNPinparticular)asanonlinearstationaryprocess,wherethenonlinearitiesariseiftheprocessissubjecttodiscreteshiftsinregimes-episodesacrosswhichthebehaviouroftheseriesismarkedlydifferent-andwherethestateoftheeconomyistreatedasanunobservedlatentvariable.Inthatsense,thisinnovativeapproachallowsresearcherstoovercometheshortcomingoflinearmodelstodealwiththeasymmetrybetweenexpansionsandcontractionsthathavebeendocumentedbyNeftci(1984),andrecentlybySichel(1993),fortheUSbusinesscycle,byArtis,Kontolemis,andOsborn(1997)foragroupofindustrialcountries,andbyMora(1997)andMejía-Reyes(1999)forsomeLatinAmericancountries.Inparticular,Markov-switchingautoregressiveprocessesinthegrowth