Tinbergen Institute has two locations Tinbergen Institute Amsterdam Gustav Mahlerplein 117 1082 MS Amsterdam The Netherlands Tel 31020 525 1600 Tinbergen Institute Rotterdam Burg Oudlaan 50 30 ID: 211395
Download Pdf The PPT/PDF document "Tinbergen Institute is the graduate scho..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Tinbergen Institute is the graduate school and research institute in economics of Erasmus University Rotterdam, the University of Amsterdam and VU University Amsterdam. More TI discussion papers can be downloaded at http://www.tinbergen.nl Tinbergen Institute has two locations: Tinbergen Institute Amsterdam Gustav Mahlerplein 117 1082 MS Amsterdam The Netherlands Tel.: +31(0)20 525 1600 Tinbergen Institute Rotterdam Burg. Oudlaan 50 3062 PA Rotterdam The Netherlands Tel.: +31(0)10 408 8900 Fax: +31(0)10 408 9031 Duisenberg school of finance is a collaboration of the Dutch financial sector and universities, with the ambition to support innovative research and offer top quality academic education in core areas of finance. DSF research papers can be downloaded at: http://www.dsf.nl/ Duisenberg school of finance Gustav Mahlerplein 117 1082 MS Amsterdam The Netherlands Tel.: +31(0)20 525 8579 TimeVaryingTransitionProbabilitiesforMarkovRegimeSwitchingModelsMarcoBazzi(a),FranciscoBlasques(b)SiemJanKoopman(b;c),AndreLucas(b)(a)UniversityofPadova,Italy(b)VUUniversityAmsterdamandTinbergenInstitute,TheNetherlands(c)CREATES,AarhusUniversity,DenmarkAbstractWeproposeanewMarkovswitchingmodelwithtimevaryingprobabilitiesforthetransitions.Thenoveltyofourmodelisthatthetransitionprobabilitiesevolveovertimebymeansofanobservationdrivenmodel.Theinnovationofthetimevaryingprobabilityisgeneratedbythescoreofthepredictivelikelihoodfunction.Weshowhowthemodeldynamicscanbereadilyinterpreted.WeinvestigatetheperformanceofthemodelinaMonteCarlostudyandshowthatthemodelissuccessfulinestimatingarangeofdierentdynamicpatternsforunobservedregimeswitchingprobabilities.WealsoillustratethenewmethodologyinanempiricalsettingbystudyingthedynamicmeanandvariancebehaviourofU.S.IndustrialProductiongrowth.Wendempiricalevidenceofchangesintheregimeswitchingprobabilities,withmorepersistenceforhighvolatilityregimesintheearlierpartofthesample,andmorepersistenceforlowvolatilityregimesinthelaterpartofthesample.Somekeywords:HiddenMarkovModels;observationdrivenmodels;generalizedautoregressivescoredynamics.JELclassication:C22,C32. Theauthorsthankparticipantsofthe\2014WorkshoponDynamicModelsdrivenbytheScoreofPredictiveLikelihoods",LaLaguna,andseminarparticipantsandVUUniversityAmsterdamforusefulcommentsanddiscussions.BlasquesandLucasthanktheDutchScienceFoundation(NWO,grantVICI453-09-005)fornancialsupport.KoopmanacknowledgessupportfromCREATES,CenterforResearchinEconometricAnalysisofTimeSeries(DNRF78),fundedbytheDanishNationalResearchFoundation.1 1IntroductionMarkovregimeswitchingmodelshavebeenwidelyappliedineconomicsandnance.SincetheseminalapplicationofHamilton(1989)toU.S.realGrossNationalProductgrowthandthewell-knownNBERbusinesscycleclassication,themodelhasbeenadoptedinnumerousotherapplications.Examplesareswitchesinthelevelofatimeseries,switchesinthe(autoregressive)dynamicsofvectortimeseries,switchesinvolatilities,andswitchesinthecorrelationordependencestructurebetweentimeseries;seeHamiltonandRaj(2002)forapartialsurvey.ThekeyattractivefeatureofMarkovswitchingmodelsisthattheconditionaldistributionofatimeseriesdependsonanunderlyinglatentstateorregime,whichcantakeonlyanitenumberofvalues.ThediscretestateevolvesthroughtimeasadiscreteMarkovchainandwecansummarizeitsstatisticalpropertiesbyatransitionprobabilitymatrix.Dieboldetal.(1994)andFilardo(1994)arguethattheassumptionofaconstanttransi-tionprobabilitymatrixforaMarkovswitchingmodelistoorestrictiveformanyempiricalsettings.TheyextendthebasicMarkovswitchingmodeltoallowthetransitionprobabili-tiestovaryovertimeusingobservablecovariates,includingstrictlyexogenousexplanatoryvariablesandlaggedvaluesofthedependentvariable.Althoughthisapproachcanbeusefulandeective,itisnotalwaysclearwhatvariablesorwhichfunctionalspecicationweshouldusefordescribingthedynamicsinthetransitionprobabilities.Ourmaincontributioninthispaperistoproposeanew,dynamicapproachtomodeltimevariationintransitionprobabilitiesinMarkovswitchingmodels.Weletthetransitionprobabilitiesvaryovertimeasspecictransformationsofthelaggedobservations.Henceweadoptanobservationdrivenapproachtotimevaryingparametermodels;seeCox(1981)foradetaileddiscussion.Observationdrivenmodelshavetheadvantagethatthelikeli-hoodistypicallyavailableinclosedformusingapredictionerrordecomposition.Ourmainchallengeistospecifyasuitablefunctionalformtolinkpastobservationstofuturetransi-tionprobabilities.Forthispurpose,weusethescoresofthepredictivelikelihoodfunction.SuchscoredrivendynamicshavebeenintroducedbyCrealetal.(2011,2013)andHarvey(2013).Scoredrivenmodelsencompassmanywell-knowntimeseriesmodelsineconomicsandnance,includingtheARCHmodelofEngle(1982),thegeneralizedARCH(GARCH)modelofBollerslev(1986),theexponentialGARCH(EGARCH)modelofNelson(1991),theautoregressiveconditionalduration(ACD)modelofEngleandRussell(1998),andmanymore.Inaddition,varioussuccessfulapplicationsofscoremodelshaveappearedinthere-centliterature.Forexample,Crealetal.(2011)andLucasetal.(2014)studydynamicvolatilitiesandcorrelationsunderfat-tailsandpossibleskewness;HarveyandLuati(2014)introducenewmodelsfordynamicchangesinlevelsunderfattails;Crealetal.(2014)inves-tigatescore-basedmixedmeasurementdynamicfactormodels;OhandPatton(2013)andDeLiraSalvatierraandPatton(2013)investigatefactorcopulasbasedonscoredynamics;andKoopmanetal.(2012)showthatscoredriventimeseriesmodelshaveasimilarfore-2 castingperformanceascorrectlyspeciednonlinearnon-Gaussianstatespacemodelsoverarangeofmodelspecications.WeshowthatthescorefunctioninourMarkovswitchingmodelhasahighlyintuitiveform.Thescorecombinesallrelevantinnovativeinformationfromtheseparatemodelsassociatedwiththelatentstates.Theupdatesofthetimevaryingparametersarethereforebasedontheprobabilitiesofthestates,givenallinformationuptotimet1.Inoursimulationexperiments,thenewmodelperformswellandsucceedsincapturingarangeoftimevaryingpatternsfortheunobservedtransitionprobabilities.WeapplyourmodeltostudythemonthlyevolutionofU.S.IndustrialProductiongrowthfromJanuary1919toOctober2013.Weuncoverthreeregimesforthemeanandtworegimesforthevarianceoverthesampleperiodconsidered.Thecorrespondingtransitionprobabilitiesaretimevarying.Inparticular,thehighvolatilityregimeappearstobemuchmorepersistentintheearlierpartofthesamplecomparedtothelaterpart.Theconverseholdsforthelowvolatilityregime.Suchchangesinthedynamicsofthetimeseriesarecapturedinastraightforwardwaywithinourmodel.Moreover,thetofthenewmodeloutperformsthetofseveralcompetingmodels.Asanalcontribution,itisworthwhilementioningthatourmodelalsopresentsaninterestingmixofparameterdriven(Markovswitching)dynamicswithobservationdrivenscoredynamicsforthecorresponding(transitionprobability)parameters.Inparticular,itisinterestingtoseethatscoredrivenmodelscanstillbeadoptedwhenanadditionallteringstep(fortheunobserveddiscretestates)isrequiredtocomputethescoreoftheresultingconditionalobservationdensity.Thisfeatureofthenewdynamicswitchingmodelisinterestinginitsownright.SimilardevelopmentsforalinearGaussianstatespacemodelhavebeenreportedbyCrealetal.(2008)andDelleMonacheandPetrella(2014)Theremainderofthepaperisorganizedasfollows.InSection2webrie ydiscussthemainset-upoftheMarkovswitchingmodelanditsresidualdiagnostics.InSection3weintroducethenewMarkovswitchingmodelwithtimevaryingtransitionprobabilitiesbasedonthescoreofthepredictivelikelihoodfunction.InSection4wediscusssomeofthestatisticalpropertiesofthemodel.InSection5wereporttheresultsofaMonteCarlostudy.InSection6wepresenttheresultsofourempiricalstudyintothedynamicsalientfeaturesofU.S.IndustrialProductiongrowth.Section7concludes.2MarkovswitchingmodelsMarkovswitchingmodelsarewell-knownandwidelyusedinappliedeconometricstudies.WerefertothetextbookofFruhwirth-Schnatter(2006)foranextensiveintroductionanddiscussion.ThetreatmentbelowestablishesthenotationanddiscussessomebasicnotionsofMarkovswitchingmodels.3 Letfyt;t=1;:::;TgdenoteatimeseriesofTunivariateobservations.Weconsiderthetimeseriesfyt;t=1;:::;Tgasasubsetofastochasticprocessfytg.Theprobabilitydistributionofthestochasticprocessytdependsontherealizationsofahiddendiscretestochasticprocesszt.Thestochasticprocessytisdirectlyobservable,whereasztisalatentrandomvariablethatisobservableonlyindirectlythroughitseectontherealizationsofyt.ThehiddenprocessfztgisassumedtobeanirreducibleandaperiodicMarkovchainwithnitestatespacef0;:::;K1g.ItsstochasticpropertiesaresucientlydescribedbytheKKtransitionmatrix,,whereijisthe(i+1;j+1)elementofandisequaltothetransitionprobabilityfromstateitostatej.Allelementsofarenonnegativeandtheelementsofeachrowsumto1,thatisij=P[zt=jjzt1=i];K1Xj=0ij=1;ij0;8i;j2f0;:::;K1g:(1)Letp(ji; )beaparametricconditionaldensityindexedbyparametersi2and 2 ,whereiisaregimedependentparameterand isnotregime-specic.Weassumethattherandomvariablesy1;:::;yTareconditionallyindependentgivenz1;:::;zT,withdensitiesytj(zt=i)p(ji; ):(2)Forthejointstochasticprocessfzt;ytg,theconditionaldensityofytisp(ytj ;It1)=K1Xi=0p(ytji; )P(zt=ij ;It1);(3)whereIt1=fyt1;yt2;:::gistheobservedinformationavailableattimet1.Allparam-eters and0;:::;K1areunknownandneedtobeestimated.TheconditionalmeanofytgivenztandIt1maycontainlagsofytitself.FrancqandRoussignol(1998)andFrancqandZakoan(2001)derivetheconditionsfortheexistenceofanergodicandstationarysolutionforthegeneralclassofMarkovswitchingARMAmodels.Inparticular,theyshowthatglobalstationarityofytdoesnotrequirethestationarityconditionswithineachregimeseparately.Asanexample,considerthecaseK=2foracontinuousvariableytwithconditionaldensityp(jzt)=N(1zt)0+zt1;2;(4)where0and1arestaticregime-dependentmeans,and2isthecommonvariance.Thelatenttwo-stateprocessfztgisdrivenbythetransitionprobabilitymatrix= 0010011111!;(5)4 wherethetransitionprobabilitiessatisfy000;111.Wehavei=ifori=0;1,and =(2;00;11)0.Toevaluateequation(3),werequirethequantitiesP(zt=ij ;It1)forallt.WecancomputetheseecientlyusingtherecursivelteringapproachofHamilton(1989).AssumingwehaveanexpressionforthelteredprobabilityP(zt1=ij ;It1),wecanobtainthepredictiveprobabilitiesP(zt=ij ;It1)asP(zt=ij ;It1)=K1Xk=0kiP(zt1=kj ;It1):(6)Hence,theconditionaldensityofytgivenIt1isgivenbyp(ytj ;It1)=K1Xi=0K1Xk=0p(ytji; )kiP(zt1=kj ;It1):(7)Wecanrewritethisexpressionmorecompactlyinmatrixnotation.Denet1astheKdimensionalvectorcontainingthelteredprobabilitiesP(zt1=ij ;It1)attimet1andlettbetheKdimensionalvectorcollectingthedensitiesp(ytji; )attimetfori=0;:::;K1.Itfollowsthat(7)reducestop(ytj ;It1)=0t1t:(8)ThelteredprobabilitiestcanbeupdatedbytheHamiltonrecursiont=0t1t 0t1t;(9)wheredenotestheHadamardelementbyelementproduct.ThelterneedstobestartedfromanappropriatesetofinitialprobabilitiesP(z0=ij ;I0).ThesmoothedestimatesoftheregimeprobabilitiesP(zt=ij ;IT)canbeobtainedfromthealgorithmofKim(1994).TheHamiltonlterin(9)isimplementedfortheevaluationofthethelog-likelihoodfunctionwhichisnumericallymaximizedwithrespecttotheparametervector(00;:::;0K1; 0)0usingaquasi-Newtonoptimizationalgorithm.Toavoidlocalmaxima,weconsiderdierentstartingvaluesforthenumericaloptimization.DiagnosticcheckinginMarkovregimeswitchingmodelsissomewhatmorecomplicatedwhencomparedtoothertimeseriesmodelsbecausethetrueresidualsdependonthelatentvariablezt.Hencetheresidualsareunobserved.AstandardsolutionistheuseofgeneralizedresidualswhichhavebeenintroducedbyGourierouxetal.(1987)inthecontextoflatentvariablemodels.TheyhavebeenusedinthecontextofMarkovregimeswitchingmodelsbyTurneretal.(1989),Gray(1996),MaheuandMcCurdy(2000),andKimetal.(2004).GiventhelteredregimeprobabilitiesP(zt=ij ;It1),fori=0;:::;K1,letiand2i5 denotetheconditionalmeanandtheconditionalvarianceofytinregimei.Thestandardizedgeneralizedresidualetisdenedaset=K1Xi=0yti iP(zt=ij ;It1);t=1;:::;T:(10)Alsointhecontextofswitchingmodels,Smith(2008)adoptsthetransformationproposedbyRosenblatt(1952)anddenestheRosenblattresidual~etas~et=1 K1Xi=0P(zt=ij ;It1)1i(yti)!;(11)wheredenotesthecumulativedistributionfunctionofastandardnormalwiththecorre-spondinginversefunction1.IfytisgeneratedbythedistributionimpliedbytheMarkovswitchingmodel,thentheRosenblattresidual~etisstandardnormallydistributed.Further-more,Smith(2008)showsinanextensiveMonteCarlostudythatLjung-BoxtestsbasedontheRosenblatttransformationhavegoodnite-samplepropertiesforthediagnosticcheckingofserialcorrelationinthecontextofMarkovregimeswitchingmodels.3TimevaryingtransitionprobabilitiesIntheprevioussectionweconsideredthetransitionprobabilitymatrixtobeconstantovertime.Dieboldetal.(1994)andFilardo(1994)argueforhavingtimevaryingtransitionprobabilitiest.Theyproposetolettheelementsoftbefunctionsofpastvaluesofthedependentvariableytandofexogenousvariables.TheHamiltonlterandKimsmoothercaneasilybegeneralizedtohandlesuchcasesoftimevaryingt.Akeychallengeistospecifyanappropriateandparsimoniousfunctionthatlinksthelaggeddependentvariablestofuturetransitionprobabilities.Forthespecicationofthedynamicsoft,weadoptthegeneralizedautoregressivescoredynamicsofCrealetal.(2013);similardynamicscoremodelshavebeenproposedbyCrealetal.(2011)andHarvey(2013).WeprovidethedetailsofthescoredrivenmodelfortimevaryingtransitionprobabilitiesintheMarkovregimeswitchingmodel.Thenewdynamicmodelisparsimoniousandtheupdatingmechanismishighlyintuitive.Eachprobabilityupdateisbasedontheweightingofthelikelihoodinformationp(ji; )in(2)foreachseparateregimei.3.1DynamicsdrivenbythescoreofpredictivelikelihoodTheparametervector containsboththetransitionprobabilitiesaswellasotherparameterscapturingtheshapeoftheconditionaldistributionsp(ytj ;It1).Withaslightabuseofnotation,wesplit intoadynamicparameterftthatweusetocapturethedynamic6 transitionprobabilities,andanewstaticparameter thatgathersallremainingstaticparametersinthemodel,aswellassomenewstaticparametersthatgovernthetransitiondynamicsofft.Forexample,inthetwo-stateexampleofSection2wemaychooseft=(f00;t;f11;t)0withf00;t=logit(00;t)andf11;t=logit(11;t),wherelogit(00;t)=log(00;t)log(100;t),andlog()referstothenaturallogarithm.Atthesametime,weset =(2;!;A;B),where!,A,andBaredenedbelowinequation(12).Fortheremainderofthispaper,wedenotetheconditionalobservationdensitybyp(ytjft; ;It1).IntheframeworkofCrealetal.(2013),thedynamicprocessesfortheparametersaredrivenbyinformationcontainedinthescoreoftheconditionalobservationdensityp(ytjft; ;It1)withrespecttoft.ThemainchallengeinthecontextofMarkovswitchingmodelsisthattheconditionalobservationdensityisitselfamixtureofdensitiesusingthelatentmixingvariablezt.Therefore,theshapeofourconditionalobservationdensityasgivenbyequation(3)issomewhatinvolved.Theupdatingequationforthetimevaryingparameterftbasedonthescoreofthepredictivedensityisgivenbyft+1=!+Ast+Bft;st=Strt;rt=@ @ftlogp(ytjft; ;It1);(12)where!isavectorofconstants,AandBarecoecientmatrices,andstisthescaledscoreofthepredictiveobservationdensitywithrespecttoftusingthescalingmatrixSt.Theupdatingequation(12)canbeviewedasasteepestascentorNewtonstepforftusingthelogconditionaldensityattimetasitscriterionfunction.AninterestingchoiceforSt,asrecognizedbyCrealetal.(2013),isthesquarerootmatrixoftheinverseFisherinformationmatrix.ThisparticularchoiceofStaccountsforthecurvatureofrtasafunctionofft.Also,forthischoiceofStandundercorrectmodelspecication,thescaledscorefunctionsthasaunitvariance;seealsoSection4.3.2Timevaryingtransitionprobabilities:thecaseof2statesWerstconsiderthetwo-stateMarkovregimeswitchingmodel,K=2.Weletthetransitionprobabilities00;tand11;tvaryovertimewhilethetworemainingprobabilitiesaresetto01;t=100;tand10;t=111;tasin(5).Wespecifythetransitionprobabilitiesasii;t=ii+(12ii)exp(fii;t)=(1+exp(fii;t));i=0;1;wheref00;tandf11;taretheonlytwoelementsinthetimevaryingparametervectorft,andwherethetwoparameters0ii0:5,fori=0;1,canbesetbytheeconometriciantolimittherangeoverwhichii;tcanvary.IntheapplicationinSection6,wesetwesetii=0,fori=0;1,suchthatii;tcantakeanyvalueintheinterval(0;1).WeprefertoworkwithaparsimoniousmodelspecicationandthereforewetypicallyhavediagonalmatricesforAandBin(12).Theupdatingequationsforthetimevarying7 parameterftisgivenbyequation(12)wherethescalingissettoSt=I0:5t1whereIt1isthe22Fisherinformationmatrixcorrespondingtothe21scorevectorrtdenedin(12).Thescorevectorfortheconditionaldensityin(7)takestheformrt=p(ytj0; )p(ytj1; ) p(ytj ;It1)gft; ;It1;(13)gft; ;It1= P[zt1=0j ;It1](1200)00;t(100;t)P[zt1=1j ;It1](1211)11;t(111;t)!:(14)Thisexpressionhasahighlyintuitiveform.Therstfactorin(13)isthedierenceinthelikelihoodofytgivenzt=0versuszt=1.Thedierenceisscaledbythetotallikelihoodoftheobservationgivenallthestaticparameters.Ifthelikelihoodofytgivenzt=0isrelativelylargecomparedtothatforzt=1,weexpectf00;ttoriseandf11;ttodecrease.Thisispreciselywhathappensinequations(13)and(14).Themagnitudesofthestepsaredeterminedbytheconditionalprobabilitiesofbeinginregimezt1=0orzt1=1,respectively,attimet1.Theremainingfactors(12ii)ii;t(1ii;t),fori=0;1,areduetothelogitparameterization.Inparticular,ifwearealmostcertainofbeinginregimezt1=0attimet1,thatisP[zt1=0j ;It1]1,thenwetakealargestepwithf00;tbutwedonotmovef11;tbymuch.Obviously,ifwearealmostcertainofbeinginregimezt1=0,ytcanonlylearnussomethingabout00;t.Wedonotlearnmuchabout11;tinthiscase.Theconverseholdsifwearealmostcertainofbeinginregimezt1=1attimet1,inwhichcasewecanonlylearnaboutf11;t=logit(11;t).Theweightsforthelteredprobabilitiesinthevectorg(ft; ;It1)in(13)takesaccountofthis.TheconditionalFisherinformationmatrixbasedon(13)issingularbydesign.Thevectorg(ft; ;It1)ontheright-handsideof(13)isIt1-measurableandhencetheexpectationofitsouterproductremainsofrank1.Therefore,wescalethescorebyasquarerootMoore-Penrosepseudo-inverse1oftheconditionalFisherinformationmatrix.Wehavest=Gt[p(ytj0; )p(ytj1; )]=p(ytj ;It1) q R11[p(ytj0; )p(ytj1; )]2=p(ytj ;It1)dyt;(15)withGt=gft; ;It1= gft; ;It1 ,andwheretheintegralhasnoclosedformingeneralandiscomputednumerically,forexampleusingGauss-Hermitequadraturemethods.AnalternativetotheanalyticMoore-Penrosepseudo-inverseisanumericalpseudoinverse;forexample,wecouldusetheTikhonovregularizedmatrixinverseasgivenbyI1=2t1=I+(1)It11=2,withunitmatrixIandxedscalar01.For!0theTikhonovinversecollapsestotheMoore-Penrosepseudo-inverse. 1Ifx2Rnisavector,thentheMoore-Penrosepseudo-inverseofxx0isgivenbykxk4xx0,anditssquarerootbykxk3xx0,askxk3xx0kxk3xx0=kxk4xx0.Asg(ft; ;It1)isIt1-measurable,scalingthescorebythesquarerootMoore-Penrosepseudo-inverseoftheconditionalFisherinformationmatrixyeldsanexpressionproportionaltokg(ft; ;It1)k3g(ft; ;It1)g(ft; ;It1)0g(ft; ;It1)=kg(ft; ;It1)k1g(ft; ;It1).8 3.3Timevaryingtransitionprobabilities:thecaseofKstatesWecaneasilygeneralizethetwo-regimemodeltoKregimes.Toenforcethatalltransitionprobabilitiesarenon-negativeandsumtoone(row-wise),weusethemultinomiallogitspecication.Givenasetofvaluesfor0ij0:5,wesetij;t=ij+(12ij)exp(fij;t)"1+K1Xj=1exp(fij;t)#1;i;K1;t=1K1Xj=1ij;t(ij);(16)fori=0;:::;K1andj=0;:::;K2.Thetimevaryingparametersfij;t,correspondingtothetimevaryingtransitionprobabilitiesij;t,arecollectedintheK(K1)1vectorft.Thevectorftissubjecttotheupdatingequation(12).Theingredientsforthescaledscorevectorintheupdatingequation(12)aregivenbyrt=J0trt;It1=E[J0rtrt0Jt];rt=@logp(ytj ;It1) @vec()0=t t1 p(ytj ;It1);where istheKroneckerproductandtheelementsofJt=@vec(t)=@f0taregivenby@ij;t @fi0j0;t=8]TJ ; -2;.51; Td; [00;:(12ij)ij;t(1ij;t);fori=i0^j=j0;(12ij)ij;tij0;t;fori=i0^j6=j0;0;otherwise:;fori;i0=0;:::;K1andj;j0=0;:::;K2.4StatisticalpropertiesInthissectionwestudythestochasticpropertiesoftheestimateddynamictransitionprob-abilitiesinourscoredrivenMarkovswitchingmodel.Inparticular,weanalyzethebehavioroftheestimatedtimevaryingparameterasafunctionofpastobservationsy1;:::;yt1,pa-rametervector ,andinitialpointf1.Wewritetheprocessasf~ftgwith~ft:=~ft( ;f1),fort=1;:::;T.WefollowBlasquesetal.(2012),whousethestationarityandergodicity(SE)conditionsformulatedbyBougerol(1993)andStraumannandMikosch(2006)forgeneralstochasticrecurrenceequations.Dene~Xt=(~f0t;~0t)0asthestackedvectoroflteredtimevaryingparameters~ftandlteredprobabilities~tasdenedin(9).Wedene~t=~t( ;1)forsomeinitialpoint1.Ourstochasticrecurrenceequationforthelteredprocessf~Xtgnowtakestheform~Xt+1=H(~Xt;yt; ),where~Xt+1="~t+1~ft+1#=H(~Xt;yt; ):="(~t;~ft;yt; )!+As(~t;~ft;yt; )+B~ft#;wheres(~t;~ft;yt; )isthescaledscoredenedin(15)and(~t;~ft;yt; )isthefractiondenedin(9)fortherecursion~t+1=(~ft;~t;yt; ):=0~t1t=~0t1t.The9 followingpropositionstatessucientconditionsforthelteredprocessf~Xt( ;X1)gwithinitializationatX1:=(01;f01)0toconvergealmostsurelyandexponentiallyfast(e.a.s)toauniquelimitSEprocessf~Xt( )g.2Proposition1.LetfytgbeSE,withijin(16)satisfyingij0forallpairs(i;j),andassumethatforevery 2 (i)Elog+ H1;f1;y1; 1;(ii)Elnsup(f;) _Hf;;y1; 0;where_H(X;y1; )=@H(X;y1; )=@XdenotestheJacobianfunctionofHw.r.t.X.Thenf~Xt( ;X1)gconvergese.a.s.toauniqueSEprocessf~Xt( )g,forevery 2 ,thatisk~Xt( ;X1)~X( )ke:a:s:!0ast!18 2 .Proof.TheassumptionthatfytgisSEimpliesthatftgin(9)isSE.TogetherwiththecontinuityofH(andtheresultingmeasurabilityw.r.t.theBorel-algebra),itfollowsthatfH(;;yt; )gisSEforevery byKrengel(1985,Proposition4.3).ConditionC1inBougerol(1993,Theorem3.1)isimmediatelyimpliedbyassumption(i)forevery 2 .ConditionC2inBougerol(1993,Theorem3.1)isimplied,forevery 2 ,bycondition(ii).Asaresult,forevery 2 ,f~Xt( ;f1)gconvergesalmostsurelytoanSEprocessf~X( )g.Uniquenessande.a.s.convergenceisobtainedbyStraumannandMikosch(2006,Theorem2.8). Proposition1eectivelydenesthosecombinationsof,AandBforwhichwecanensurethatthelteredsequencef~Xt( ;f1)gconvergese.a.s.toanSElimitforagivenSEdatasequencefytg.Weemphasizethatanite!isrequiredforcondition(i)toholdsince!enterstheHfunction.However,!playsnoroleinthecontractioncondition(ii)asitdoesnotin uencetheJacobian_H.Numericalexperiments(notreportedhere)suggestthatstabilityisachievedbyawiderangeofcombinationsoftheparameters,AandB,whereisavectorcontainingallij.Inparticular,thesetofstablecombinations(A,B)becomeslargerforhighervaluesof.Thismechanismisintuitivebecausetheentriesofthevectorboundtheelementsofthevectorandawayfromzeroandone.Asaresult,0ensuresthatthedenominatorin(9)isboundedawayfromzeroandhencethesequenceftgbecomesmorestable.Proposition1isessentialincharacterizingthestochasticpropertiesofthelteredtimevaryingparameters.ItdoesnotonlyallowustohavefurtherinsightsintothenatureofthelteredestimatesftintheMonteCarlostudyofSection5,butitalsoenablesustointerprettheparameterestimatesoftheGASmodel.TheSEnatureofthelteredsequenceisalsoanimportantingredientinobtainingproofsofconsistencyandasymptoticnormalityofthemaximumlikelihoodestimatorthatrelyonanapplicationoflawsoflargenumbersandcentrallimittheorems;seeBlasquesetal.(2014)formoredetails. 2Wesaythatarandomsequencef~Xtgconvergese.a.s.toanotherrandomsequencef~Xtgifthereisaconstantc1suchthatctk~Xt~Xtka:s:!0;seeStraumannandMikosch(2006)forfurtherdetails.10 Table1:Simulationpatternsfor00;tand11;t Model00;t11;t 1.Constant0:950:852.SlowSine0:5+0:45cos(4t=T)0:50:45cos(4t=T)3.Sine0:5+0:45cos(8t=T)0:50:45cos(8t=T)4.FastSine0:5+0:45cos(20t=T)0:50:45cos(20t=T)5.Break0:21fT=2g+0:81ftT=2g0:81fT=2g+0:21ftT=2g 5MonteCarlostudy5.1DesignofthesimulationstudyToinvestigatetheperformanceofourestimationprocedurefortheMarkovregimeswitchingmodelwithtimevaryingtransitionprobabilities,weconsideraMonteCarlostudyforthetworegimemodel(4).Thetworegimesconsistoftwonormaldistributionswithcommonvariance2=0:5andmeans0=1and1=1.Weset00=11=0andconsider5dierentformsoftimevariationforthetransitionprobabilities00;tand11;t.ThepatternsaresummarizedinTable1andrangefromaconstantsetoftransitionprobabilities,viaslowandfastcontinuouslychangingtransitionprobabilitiestoanincidentalstructuralbreakinthemiddleofthesample.Weinvestigatetherobustnessofourestimationprocedureinthesedierentsettings.Forallofthedatagenerationprocessesconsidered,ourregimeswitchingmodelwithtimevaryingparametersisclearlymisspecied.InourMonteCarlostudyweconsiderthreedierentsamplesizes:T=250;500;1000.ThenumberofMonteCarloreplicationsissetto100.WeadoptthemodelwithK=2statesasdescribedindetailinSections3.1and3.2.Foreachdatageneratingprocessandsamplesize,weestimatethestaticparameters usingthemethodofmaximumlikelihood.Giventheestimatedvaluesforthestaticparameters,wecomputethelteredparameters^00;tand^11;tusingtheupdatingequationsin(12).WecomparetheresultstothosefortheMarkovswitchingmodelwithtimeinvarianttransitionprobabilities.5.2ThesimulationresultsInTable2,wepresentMonteCarloaveragesofthemaximizedlog-likelihoodvalue,agoodness-of-tstatisticandaforecastprecisionmeasure.OurstatisticformodeltisthecorrectedAkaikeInformationCriterion(AICc).TheAICcistheoriginalAICofAkaike(1973)butwithastrongernitesamplepenalty;seeHurvichandTsai(1991).Ourmeasureofforecastprecisionisthemeansquaredone-stepaheadforecasterror(MSFE).TheMSFEusesthepastobservationsy1;:::;yt1tomakeaforecastofyt,usingthestaticparameter11 Table2:SimulationresultsIWehavesimulated100timeseriesfromeachdatagenerationprocess(DGP)listedinTable1andforsamplesizesT=250;500;1000.Thestaticparametersareestimatedbythemethodofmaximumlikelihood,bothfortheMarkovregimeswitchingmodelwithstatic(Static )andwithtimevaryingtransitionprobabilities(TV ;ft).Inthelattercase,theunderlyingtimevaryingparametersareupdatedusingequation(12).Wereportthesampleaveragesforthe100simulatedseriesofthemaximizedlog-likelihoodvalue(LogLik),thecorrectedAkaikeInformationCriterion(AICc)andthemeansquarederroroftheone-stepaheadforecastofyt(MSFE).Theone-stepaheadforecasterrorsarecomputedwithinthesamplethatisusedforparameterestimation. LogLikAICcMSFEDGPTStaticTV ;ftStatic TV ;ftStatic TV ;ft Constant250-300.299-293.037610.859604.8711.6951.583500-617.392-611.9471244.9101242.2711.6701.6271000-1255.957-1252.0462521.9752522.2761.6831.668 SlowSine250-355.189-344.200720.639707.1962.3732.240500-730.993-704.0581472.1101426.4932.2122.1001000-1483.296-1412.0452976.6522842.2742.2192.148 Sine250-354.946-346.154720.153711.1052.3892.237500-732.544-711.5641475.2131441.5072.2432.1741000-1485.593-1426.2652981.2482870.7142.2372.152 FastSine250-356.440-347.479723.141713.7542.6022.323500-733.108-720.6501476.3411459.6782.3582.2771000-1487.584-1449.2172985.2292916.6192.2162.212 Break250-356.643-343.721723.547706.2392.4382.197500-735.293-706.5441480.7111431.4672.3842.1371000-1493.767-1429.8122997.5962877.8092.3722.104 estimatesobtainedfromtheentiresampley1;:::;yT.Thecriterionthereforemeasurestheforecastprecisionofthemodelforthemaximumlikelihoodestimateof inthestaticmodelandforthemaximumlikelihoodestimateof inthetimevaryingparametermodel.WelearnfromTable2thattheaveragemaximizedlog-likelihoodvaluesareuniformlyhigherforthemodelwithtimevaryingtransitionprobabilitiescomparedtothemodelwithconstanttransitionprobabilities.Thisresultisnotverysurprisingsincethetimevaryingmodelisadynamicgeneralizationofthestaticmodelandhasmoreparameters.AmoreconvincingresultisthatthetimevaryingmodelproducesoverallsubstantiallysmallerAICcvaluesthantheconstantmodel.Asexpected,theonlyexceptionisthedatageneratingprocesswithconstanttransitionprobabilities.Fortimeseriessimulatedwithtimevaryingtransitionprobabilitieschangegraduallyorareofastructuralbreaktype,themodelwithtimevaryingtransitionprobabilitiesperformssubstantiallybetterthanthestaticmodel.Finally,whenwefocusontheforecastprecisionmeasureMSFE,wealsoconcludethatthe12 Table3:SimulationresultsIIWehavesimulated100timeseriesfromeachdatagenerationprocess(DGP)listedinTable1andforsamplesizesT=250;500;1000.Thestaticparametersareestimatedbythemethodofmaximumlikelihood,bothfortheMarkovregimeswitchingmodelwithstatic(Static)andwithtimevaryingtransitionprobabilities(TV).Inthelattercase,theunderlyingtimevaryingparametersareupdatedusingequation(12).Wereportthesampleaveragesforthe100simulatedseriesofthemeansquarederror(MSE)andthemeanabsoluteerror(MAE)oftheone-stepaheadforecastoftwotransitionprobabilities00and11.Theone-stepaheadforecasterrorsarecomputedwithinthesamplethatisusedforparameterestimation. MSEMAEMSEMAE0011 DGPTStaticTVStaticTVStaticTVStaticTV Constant2500.0000.0100.0150.0460.0060.0370.0490.1365000.0000.0050.0120.0320.0020.0150.0320.08910000.0000.0020.0070.0190.0010.0060.0210.057 SlowSine2500.1890.0870.3560.2230.1970.0970.3620.2385000.2040.0730.3690.1940.2080.0850.3720.21210000.2020.0550.3670.1600.2050.0550.3690.161 Sine2500.1840.1080.3500.2550.1900.1220.3550.2755000.2000.0840.3650.2190.2040.0940.3670.23410000.1990.0650.3640.1820.2010.0680.3650.190 FastSine2500.1590.1640.3290.3260.1550.1670.3260.3285000.1820.1140.3480.2650.1860.1190.3500.27410000.1990.0790.3630.2160.2010.0860.3650.229 Break2500.1770.1010.3410.2200.1700.0940.3420.2215000.1750.0820.3390.1810.1720.0790.3400.18810000.1740.0760.3370.1620.1720.0770.3380.171 Markovswitchingmodelwithtimevaryingtransitionprobabilitiesconvincinglyoutperformsitsstaticcounterpart.Nextweverifytheprecisionofthelteredtransitionprobabilityestimatesforthestaticmodelwith00and11,andforthetimevaryingmodelwith00;tand11;t.InaMonteCarlostudy,thetransitionprobabilitiesaresimulatedaspartofthedatagenerationprocess.Henceweareabletocomparetruetransitionprobabilitieswiththeirlteredestimatesandcomputethemeansquarederror(MSE)andthemeanabsoluteerror(MAE)statistics.InTable3wepresenttheMonteCarloaveragesofthesetwostatistics,forthedierentdatagenerationprocesses,thedierentsamplessizesandthetwomodels:thestaticmodel(static)andthetimevaryingmodel(TV).TheresultsinTable3providestrongevidencethattheMarkovregimeswitchingmodelwithtimevaryingtransitionprobabilitiesissuccessfulinproducingaccuratelteredesti-13 matesoftheprobabilitiesforalldierenttimevaryingpatterns.ItisonlyfortheseriesthataresimulatedfromamodelwithconstanttransitionprobabilitiesthattheMSEandMAEstatisticsforthestaticmodelaresmallerthanthoseforthetimevaryingmodel.Forthiscase,however,theabsolutevalueofallstatisticsareofanorderofmagnitudesmallerthanforalltheotherdatageneratingprocesses.Forconstanttransitionprobabilities,thedierencesbetweenthestaticandtimevaryingmodelcanthusbequaliedassmall.WealsoobservethatforincreasingsamplesizesT,theMSEandMAEstatisticsmostlydecreaseforthetimevaryingmodel,whilethisdoesnotoccurforthestaticmodel.Toputtheseobservationsinsomeperspective,wenoticethatthesinusoidpatternshavethesamenumberofswingsovertheentiresamplefordierentsamplesizes.Therefore,thechangeinthetransitionproba-bilitiesgetssmallerperunitoftimeasTincreases.Itfollowsthattheupdatingequation(12)forthetimevaryingmodelcantrackthetruetransitionprobabilitymoreaccuratelyasTincreases.Thismechanismdoesnotaecttheinaccuracyoftheestimatesobtainedfromthestaticmodel.6AnempiricalstudyofU.S.IndustrialProductionMarkovregimeswitchingmodelsarepopularinempiricalstudiesofmacroeconomictimeseries.Wethereforeillustrateournewmethodologyfortimevaryingtransitionprobabilitiesinanempiricalstudyconcerningakeyvariableformacroeconomicpolicy,U.S.IndustrialProduction(IP).ThetimeseriesforIPisobtainedfromtheFederalReserveBankofSt.Louiseconomicdatabase(FRED);wehavemonthlyseasonallyadjustedobservationsfromJanuary1919toOctober2013,T=1137.WeanalyzethepercentagegrowthofIP(log-dierences100)andconsidertheresultingseriesasourytvariable.Figure1presentsboththeIPindexandtheIPpercentagelogdierencesyt.6.1ThreemodelspecicationsThetimeseriesofIPgrowthisrelativelylong.Inspectingthetimeseriesplotofyt,weanticipatethattheremaybepossiblechangesinboththemeanandvarianceoftheseriesovertime.Toremainwithinaparsimoniousmodellingframework,weadoptthemodelofDoornik(2013).Doornikanalyzesaquarterlytimeseriesofpost-warU.S.grossdomesticproductgrowthbymeansofaMarkovswitchingmean-variancecomponentmodel.Inourspecicsetting,weconsideramodelwiththreeregimesforthemean(m=0;1;2)andtworegimesforthevariance(v=0;1).Thethreeregimesforthemeanmayrepresentrecession,stableandgrowthperiodsinU.S.production.Eachregimeforthemeanconsistsofaconstantandpmlaggeddependentvariablesforyt,withm=0;1;2.Theconstantandthepmautoregressivecoecientsarecollectivelysubjecttotheregimetowhichtheybelong.14 Figure1:U.S.IndustrialProduction(monthly,seasonallyadjusted)andIPpercentagegrowth(log-dierences100)Hence,incaseofthreeregimesforthemean,wehave3+P3m=1pmcoecients.Thetworegimesforthevariancemaysimplydistinguishperiodsoflowandhighvolatility.Hencethenumberofcoecientsforthevariancepartequals2.ModelI:staticspecicationLetfztgandfztgdenotethehiddenprocessesthatdeterminethemeanandthevarianceforthedensityofyt,respectively.Wehaveytj(zt=m;zt=v;It1)Nm;t;2v;m=0;1;2;v=0;1;(17)withthethreemeanequationsm;t=0;m+1;myt1+:::+pm;mytpm;m=0;1;2;(18)where0;misanintercept,1;m;:::;pm;mareautoregressivecoecients,forpm2N+,and21and22arethetwovariances.Thetransitionprobabilitiesforthemeanandvariancearecollectedinthematricesand,respectively,whicharegivenby=0B@0001100011011110112021120211CA;= 0010011111!:(19)15 WefollowDoornik(2013)inspecifyingthetransitionprobabilitymatrixforthe32=6regimesas= ;wherethe36probabilitiesinareafunctionof6meanand2varianceprobabilities.TheconditionaldensityofytgivenIt1canbeexpressedintermsofthelteredprobabilitiesasin(8).Wehavep(ytj ;It1)=266664P[zt1=0;zt1=0j ;It1]P[zt1=1;zt1=0j ;It1]...P[zt1=2;zt1=1j ;It1]3777750( )266664p(yt;0;t;20;It1)p(yt;1;t;20;It1)...p(yt;2;t;21;It1)377775=0t1( )t:(20)Nextweconsidertwotimevaryingextensionsofthevarianceregimesinthestaticmodel.Afterexperimentationwithtimevaryingparametersforthemeanregimes,wehaveconcludedthatinourempiricalsettingonlytheintroductionoftimevaryingtransitionprobabilitiesforthevarianceregimescanimprovethetofthestaticmodel.Wenoticethatoncetheprobabilitiesinaretimevarying,allprobabilitiesin= aretimevarying.ModelII:timevaryingvarianceprobabilitiesasafunctionofjyt1jInmodelII,weconstructabenchmarkinthespiritofDieboldetal.(1994)andFilardo(1994).Wespecifythetimevaryingtransitionprobabilitiesforthevarianceregimesintasalogistictransformationofthelaggeddependentvariable,thatisvv;t=exp(gvv;t) 1+exp(gvv;t);gvv;t=c0;v+c1;vjyt1j;v=0;1;(21)wherec0;visaninterceptandc1;visaxedcoecient,forv=0;1.Thefourccoecientsareestimatedbythemethodofmaximumlikelihood,jointlywiththeothercoecients.ModelIII:timevaryingvarianceprobabilitiesasafunctionofthescoreInthisspecicationoftimevaryingtransitionprobabilitiesforthevarianceregimes,weadopttheframeworkofSection3.2.Empirically,itturnsoutthereisnoneedtoshrinktherangeofii;texante.Wethereforesetii=0andspecifythetimevaryingmatrixtasvv;t=exp(fvv;t) 1+exp(fvv;t);v=0;1;ft=(f00;t;f11;t)0;whereftisupdatedovertimeasin(12).Theresultingconditionaldensityforytisgivenbyp(ytj ;It1)=0t1(t )t.TheregimeprobabilitystructureofDoornik(2013)is16 morerestrictedthantheoneforthegeneralK-regimeMarkovswitchingmodelinSection3.3.Wethereforehavedierentexpressionsforthescorevectorandscalingmatrix.Thescorevectorisgivenbyrt=(r00;t;r11;t)0wherervv;t=vv;t(1vv;t) p(ytj ;It1)0t1@t @vv;t t;v=0;1;witht1andtdenedimplicitlyin(20).TocomputetheconditionalFisherInformationE[rtr0t],whereEiswithrespecttop(ytj ;It1),weevaluate4numericalintegrals,foreverytimetandforeachvaluefortheparametervector ,byaGauss-Hermitemethod.6.2Parameterestimates,modeltandresidualdiagnosticsTable4presentstheparameterestimatesforthethreemodelspecications.Forallmodelswehavetakenpm=3in(18),withm=0;1;2.Wehaveexperimentedwithothervaluesofpmbutthischoiceprovidesanadequatet.Theparameterestimatesareobtainedbynumericallymaximizingthelog-likelihoodfunctionwithrespecttothestaticparametervector or .TheassociatedstandarderrorsareobtainedbynumericallyinvertingtheHessianmatrixatthemaximizedlog-likelihoodvalue.Thesetsofestimatedcoecientsforthethreemeanregimesareverysimilaracrossthethreemodelspecications.Theintroductionoftimevaryingvariancetransitionprobabilitiesdoesnotappeartoaectthemeanspecicationmuch.Thecoecient0;mdeterminestheinterpretationofaregime.Wecanlearnfromourestimationresultsthatregimem=0correspondstolowIPgrowth,m=1representsrecessionandm=2identieshighgrowthinIP.Theautoregressivecoecients1;m;:::;3;mshowthatin"normalyears"IPgrowthispersistent,whileduringrecessionsIPgrowthissubjecttopersistentcyclicaldynamics.PeriodsofhighIPgrowthareveryshortlivedgiventhestrongnegativelyestimatedautoregressivecoecientsform=2.Theestimatedtransitionprobabilitiesrevealthetypicalsituationinregimeswitchingmodelsthatonceweareinarecessionorlowgrowthregime,itismostlikelythatweremaininthisstate.Itisonlyforthehighgrowthregimethatitismorelikelytomovetoalowgrowthregimewhiletheprobabilitytostay,22=12021,isestimatedaround0:35.Forthevarianceregimes,themodelsclearlydistinguishbetweenalow(approximately0:3)andahigh(approximately5:5to6:0)varianceregime.Themagnitudesofthesevariancesareagaincomparableacrossthedierentmodels.ForModelI,bothvarianceregimesarehighlypersistentwithprobabilities00and11bothestimatedcloseto1.WealsolearnfromTable4thatthemodelthasimprovedafterintroducingtimevaryingvariancetransitionprobabilities,bothforModelIIandModelIII.Themaximizedlog-likelihoodvalueshaveincreasedby8and17basispointsatthecostofanadditional2and4parametersfortherespectiveModelsIIandIII.ThecorrectedAkaikeinformationcriterionclearlypointstoModelIIIasthebestcomprimisetomodeltandaparsimoniousmodelspecication.17 Thetimevaryingtransitionprobabilitiesforthelowvolatility(v=0)andhighvolatility(v=1)regimesinModelIIIarehighlypersistentprocessesovertime:theestimatesforthediagonalelementsofBareveryclosetounity.FromModelIIwemayconcludethatonlythetransitionprobabilityofalowvolatilityregimeistimevarying:theestimateofc1;1isnotsignicant.ForModelIIIwendsomewhatstrongerevidencethatbothtransitionprobabilitiesaretimevarying:theestimatesofbothdiagonalelementsofAaresignicantattheusuallevelof5%.Finally,Table4presentsdiagnosticteststatisticsforthegeneralizedandRosenblattresidualswhichwehavediscussedinSection2.Thep-valuesforthewell-knownJarque-Bera2normalitytestandtheLjung-Box2serialcorrelationtest,fortheresidualsandthesquaredresiduals,indicatethatallmodelsarecapableofdescribingthesalientfeaturesinIPgrowth.TherearesomedierencesbetweenthestatisticsforthegeneralizedandRosenblattresiduals,buttheyaresmallandhavenobearingsonthemainconclusions.TheJarque-BeratestmayindicatethattheIPgrowthtimeseriesissubjecttoafewoutlyingobservations.6.3Signalextraction:regimeandtransitionprobabilitiesInFigure2wepresentthesmoothedestimatesofprobabilitiesformeanandvarianceregimesandthelteredestimatesof(timevarying)transitionprobabilitiesforModelsI,IIandIII.ModelIIappearsunabletocapturethedynamicsinthetransitionprobabilities.WehavelearnedfromTable4thattheestimateofcoecientc1;1isnotsignicant;itisalsore ectedinFigure2withthetimeseriesplotofthelteredprobabilityestimatesforthehighvarianceregimethatisalmostconstantovertime.Ontheotherhand,thelteredprobabilityestimatesforthelowvolatilityregimearehighlyerratic.ThelteredprobabilitiesforModelIIIshowanentirelydierentpattern.Boththelowandhighvolatilitytransitionprobabilitiesevolvegraduallyovertime.Inparticular,thepersistenceofthelowvolatilityregimeappearstohavegoneupovertime,withvaluesaround0.7intheearlypartofthesample,andvaluescloseto1inthesecondhalfofthesample.Theconverseholdsforthehighvolatilityregime.Thepersistenceprobability11iscloseto1uptothe1940s.Afterthat,theprobabilitydecreasessubstantiallytovaluesaround0.5,andslowlyrisestowardstheendofthesampleagain.ThepatternforthelteredprobabilitiesisconsistentwiththeempiricalpatterninthedatainFigure1.Intheearlierpartofthesample,highvolatilitylevelsarepredominant.Towardsthemiddleofthesample,largevolatilitiesareincidentalandshort-lived,whereastowardstheendofthesampleduringtheyearsofthenancialcrisis,U.S.debtceilingcrisis,andtheEuropeansovereigndebtcrisis,highervolatilitylevelsappeartoclusteragainmore.TheempiricalpatternsarealsocorroboratedbytheparameterestimatesinTable4.Inparticular,theparameterestimatesforthediagonalelementsofBarebothcloseto1;itsuggeststhatthedynamictransitionprobabilitiesevolvegraduallyovertime.Theestimates18 Table4:Parameterestimates,modeltandresidualdiagnosticsInthersttwopanelswereportthemaximumlikelihoodestimateswithstandarderrorsinparanthesesbelow,forModelsI,IIandIII.Intherstpaneltheparameterestimatesforthemeanm;tin(17)arereportedforeachregimem=0;1;2:theintercept0;m,theautoregressivecoecients1;m;:::;3;m,andthetransitionprobabilitiesmj,forj=0;1,inof(19).Inthesecondpanelthetworegimevarianceestimatesfor2varereported.ThevariancetransitionprobabilityforModelIvvisestimateddirectlywhilewehavevv=logit1(xv)forModelII(xv=c0;v)andforModelIII(xv=!v=(1Bvv)),forv=0;1,where!vandBvvarethe(v+1)thelementsofvector!anddiagonalmatrixBin(12),respectively.ThetimevaryingvarianceprobabilitiesaredeterminedinModelIIbyc1;v,andinModelIIIbyAvvandBvvwhicharethe(v+1)thdiagonalelementsofAandBin(12),respectively,forv=0;1.Inthethirdpanelwereportmodeltstatistics:Fit(1)isnumberofstaticparameters;Fit(2)ismaximizedlog-likelihoodvalue;Fit(3)isAICc,seeSection5.2.Wefurtherreportthep-valuesoftheresidualdiagnostic(RD)teststatisticsforthegeneralized(et)andRosenblatt'sresiduals(~et):RD(1)isJarque-Beranormality2(2)test;RD(2)isLjung-Boxserialcorrelation2(6)test;RD(3)isasRD(2)forsquaredresiduals. ModelIModelIIModelIII m=0m=1m=2m=0m=1m=2m=0m=1m=20;m0.076-0.2120.8460.068-0.2770.8370.043-0.1280.737(0.038)(0.182)(0.168)(0.037)(0.140)(0.172)(0.033)(0.095)(0.128)1;m0.3161.121-0.6090.3271.126-0.6210.3511.087-0.479(0.050)(0.096)(0.135)(0.050)(0.088)(0.130)(0.040)(0.079)(0.107)2;m0.212-0.569-0.3950.220-0.526-0.4840.234-0.537-0.221(0.050)(0.146)(0.123)(0.040)(0.139)(0.086)(0.037)(0.108)(0.086)3;m0.1050.0760.0390.1130.0110.1330.1120.0360.103(0.037)(0.107)(0.124)(0.035)(0.115)(0.101)(0.030)(0.065)(0.092)m00.9090.1110.5770.9090.1130.5920.8640.1450.576(0.041)(0.103)(0.140)(0.032)(0.072)(0.142)(0.036)(0.070)(0.175)m10.0160.8580.0550.0140.8580.0730.0210.8420.048(0.019)(0.092)(0.052)(0.012)(0.064)(0.062)(0.015)(0.059)(0.055) v=0v=1v=0v=1v=0v=12v0.3365.5790.3515.8660.3175.920(0.025)(0.629)(0.026)(0.691)(0.023)(0.541)vv0.9800.9470.9960.8830.8860.702(0.007)(0.018)(0.003)(0.053)(0.080)(0.200)c1;v-1.8990.108(0.419)(0.209)Avv0.1320.148(0.058)(0.074)Bvv0.9980.989(0.003)(0.011) i=1i=2i=3i=1i=2i=3i=1i=2i=3Fit(i)22-1642333024-1634331726-16253302RD(i)et0.0650.7720.5560.0320.9680.6590.0190.7920.924RD(i)~et0.0110.4090.6480.0120.8300.7380.0070.6320.676 19 Figure2:Smoothedprobabilityestimatesfortherecessionregimeinthemeanandforthehighvarianceregime.Filteredtransitionprobabilityestimatesforthelowandhighvarianceregimes.Intherstgraph,theverticalgrayareasindicaterecessionsaccordingtotheNBERbusinesscycleclassications.20 ofbothdiagonalelementsofAhavethecorrectsignandleadtoparameterchangesthatincreasethelocaltofthemodelintermsoflog-likelihood.Finally,wepresentthesmoothedestimatesofztinthetoppanelsofFigure2,togetherwiththeNBERbusinesscycleclassications.WemayconcludethatallmodelsresultinhighersmoothedrecessionprobabilitiesintheNBERclassiedperiods.Themodeltforamodelwithtimevaryingtransitionprobabilitiesforthevarianceregimes(ModelIIorIII)istypicallyhigherthanthestaticModelI.Fromthesmoothedprobabilitiesforthehighvarianceregime,mostofthehighvarianceregimeislocatedinthersthalfofthesample.Thesecondepisodeofhighvarianceisduringthenancialcrisis,withtheintermediateperiodhavingpredominantlyalowlevelofvolatility.Wenoticethatsome,butnotall,NBERrecessionscorrespondtoperiodsofhighvolatility.Thissupportstheuseofourcurrentframeworkwithseparateregimesforthe(conditional)meansandforthevariances.7ConclusionWehaveintroducedanewmethodologyfortimevaryingtransitionprobabilitiesinMarkovswitchingmodels.Wehaveshownthattheuseofthescoreofthepredictivelikelihoodandthegeneralizedautoregressivescore(GAS)modellingframeworkofCrealetal.(2013)candrivethedynamicsofthetransitionprobabilitieseectivelyovertime.Thecorrespondingdynamicscaneasilybeinterpretedwhiletheinformationembeddedintheconditionalob-servationdensitiesarefullyincorporated.Wehaveformulatedconditionsfortheestimatedtimevaryingprobabilitiesfromourscoredrivenmodeltoconvergetostationaryandergodicstochasticprocesses.BymeansofanextensiveMonteCarlostudy,wehaveshownthattheourproposedobservationdrivenmodelisabletoadequatelytrackthedynamicpatternsintransitionprobabilities,eveniftheunderlyingdynamicsthemselvesarepossiblymisspecied.Bothfordeterministicstructuralbreaksanddeterministicsinusoidpatterns,ourmodelyieldsalargeimprovementinmodeltcomparedtoamodelwithconstanttransitionprobabilitiesonly.InourempiricalstudyforIndustrialProductiongrowth,wehaveshownthatwecaneectivelyusethemodeldynamicfeaturesinthemeanandvariancesimultaneously.WehavefoundthatourproposedmodeloutperformsboththeMarkovswitchingmodelwithconstantprobabilitiesandwithtransitionprobabilitiesdependingonalaggeddependentvariable.Inparticular,thepatternslteredbyourmodelcanbeeasilyinterpreted,withhigher(lower)persistenceforhigh(low)volatilityregimesinthebeginningofthesamplecomparedtothelaterpartofthesample.Highervolatilitiesappeartore-occuragainattheveryendofthesample,duringthenancialandsovereigndebtcrises.Weconcludethatthemodelcanprovideausefulbenchmarkinsettingswheretransitionprobabilitiesinaregimeswitchingmodelmayvaryovertime.21 ReferencesAkaike,H.(1973).MaximumlikelihoodidenticationofGaussianautoregressivemovingaveragemodels.Biometrika60(2),255{265.Blasques,F.,S.J.Koopman,andA.Lucas(2012).Stationarityandergodicityofunivariategeneralizedautoregressivescoreprocesses.DiscussionPaperTinbergenInstituteTI12-059/4.Blasques,F.,S.J.Koopman,andA.Lucas(2014).Maximumlikelihoodestimationforgeneralizedautoregressivescoremodels.DiscussionPaperTinbergenInstituteTI14-029/III.Bollerslev,T.(1986).Generalizedautoregressiveconditionalheteroskedasticity.Journalofeconometrics31(3),307{327.Bougerol,P.(1993).Kalmanlteringwithrandomcoecientsandcontractions.SIAMJournalonControlandOptimization31(4),942{959.Cox,D.R.(1981).Statisticalanalysisoftimeseries:somerecentdevelopments.Scandina-vianJournalofStatistics8,93{115.Creal,D.,S.J.Koopman,andA.Lucas(2008).Ageneralframeworkforobservationdriventime-varyingparametermodels.DiscussionPaperTinbergenInstituteTI08-108/4.Creal,D.,S.J.Koopman,andA.Lucas(2011).ADynamicMultivariateHeavy-TailedModelforTime-VaryingVolatilitiesandCorrelations.JournalofBusiness&EconomicStatistics29(4),552{563.Creal,D.,S.J.Koopman,andA.Lucas(2013).Generalizedautoregressivescoremodelswithapplications.JournalofAppliedEconometrics28(5),777{795.Creal,D.,B.Schwaab,S.J.Koopman,andA.Lucas(2014).Observationdrivenmixed-measurementdynamicfactormodels.ReviewofEconomicsandStatistics,forthcoming.DeLiraSalvatierra,I.andA.J.Patton(2013).Dynamiccopulamodelsandhighfrequencydata.DukeUniversityDiscussionPaper.DelleMonache,D.andI.Petrella(2014).Ascoredrivenapproachforgaussianstate-spacemodelswithtime-varyingparameter.WorkingPaper,ImperialCollegeLondon.Diebold,F.,J.Lee,andG.Weinbach(1994).RegimeSwitchingwithTime-VaryingTran-sitionProbabilities.InC.Hargreaves(Ed.),NonstationaryTimeSeriesAnalysisandCointegration,pp.283{302.OxfordUniversityPress.22 Doornik,J.(2013).AMarkov-switchingmodelwithcomponentstructureforUSGNP.EconomicsLetters118(2),265{268.Engle,R.F.(1982).AutoregressiveconditionalheteroscedasticitywithestimatesofthevarianceofUnitedKingdomin ation.Econometrica50(4),987{1007.Engle,R.F.andJ.R.Russell(1998).AutoregressiveConditionalDuration:ANewModelforirregularlySpacedTransactionData.Econometrica66(5),1127{1162.Filardo,A.J.(1994).Business-cyclephasesandtheirtransitionaldynamics.JournalofBusiness&EconomicStatistics12(3),299{308.Francq,C.andM.Roussignol(1998).ErgodicityofautoregressiveprocesseswithMarkov-switchingandconsistencyofthemaximum-likelihoodestimator.Statistics:AJournalofTheoreticalandAppliedStatistics32(2),151{173.Francq,C.andJ.-M.Zakoan(2001).Stationarityofmultivariatemarkov{switchingARMAmodels.JournalofEconometrics102(2),339{364.Fruhwirth-Schnatter,S.(2006).FiniteMixtureandMarkovSwitchingModels.Springer.Gourieroux,C.,A.Monfort,E.Renault,andA.Trognon(1987).Generalisedresiduals.JournalofEconometrics34(1),5{32.Gray,S.F.(1996).Modelingtheconditionaldistributionofinterestratesasaregime-switchingprocess.JournalofFinancialEconomics42(1),27{62.Hamilton,J.(1989).ANewApproachtotheEconomicAnalysisofNonstationaryTimeSeriesandtheBusinessCycle.Econometrica57(2),357{384.Hamilton,J.D.andB.Raj(2002).Newdirectionsinbusinesscycleresearchandnancialanalysis.EmpiricalEconomics27(2),149{162.Harvey,A.C.(2013).DynamicModelsforVolatilityandHeavyTails:WithApplicationstoFinancialandEconomicTimeSeries.EconometricSeriesMonographs.CambridgeUniversityPress.Harvey,A.C.andA.Luati(2014).Filteringwithheavytails.JournaloftheAmericanStatisticalAssociation,forthcoming.Hurvich,C.M.andC.-L.Tsai(1991).BiasofthecorrectedAICcriterionforunderttedregressionandtimeseriesmodels.Biometrika78(3),499{509.Kim,C.(1994).DynamiclinearmodelswithMarkov-switching.JournalofEconomet-rics60(1),1{22.23 Kim,C.-J.,J.C.Morley,andC.R.Nelson(2004).Isthereapositiverelationshipbe-tweenstockmarketvolatilityandtheequitypremium?JournalofMoney,CreditandBanking36,339{360.Koopman,S.J.,A.Lucas,andM.Scharth(2012).Predictingtime-varyingparameterswithparameter-drivenandobservation-drivenmodels.TinbergenInstituteDiscussionPapers12-020/4.Krengel,U.(1985).Ergodictheorems.Berlin:DeGruyterstudiesinMathematics.Lucas,A.,B.Schwaab,andX.Zhang(2014).Measuringcreditriskinalargebankingsystem:econometricmodelingandempirics.JournalofBusinessandEconomicStatistics,forthcoming.Maheu,J.M.andT.H.McCurdy(2000).Identifyingbullandbearmarketsinstockreturns.JournalofBusiness&EconomicStatistics18(1),100{112.Nelson,D.B.(1991).ConditionalHeteroskedasticityinAssetReturns:ANewApproach.Econometrica59(2),347{370.Oh,D.H.andA.J.Patton(2013).Time-varyingsystemicrisk:EvidencefromadynamiccopulamodelofCDSspreads.DukeUniversityDiscussionPaper.Rosenblatt,M.(1952).Remarksonamultivariatetransformation.TheAnnalsofMathe-maticalStatistics23(3),470{472.Smith,D.R.(2008).EvaluatingSpecicationTestsforMarkov-SwitchingTime-SeriesMod-els.JournalofTimeSeriesAnalysis29(4),629{652.Straumann,D.andT.Mikosch(2006).Quasi-maximum-likelihoodestimationincondition-allyheteroscedastictimeseries:astochasticrecurrenceequationsapproach.TheAnnalsofStatistics34(5),2449{2495.Turner,C.M.,R.Startz,andC.R.Nelson(1989).AMarkovmodelofheteroskedasticity,risk,andlearninginthestockmarket.JournalofFinancialEconomics25(1),3{22.24 TI 2014-072/III Tinbergen Institute Discussion Paper z University of Padova, Italy; Faculty of Economics and Business Administration, VU University Amsterdam, the Netherlands.