:FromStickyInflationtoStickyPricesTheNewKeynesianPhillipsCurve(NKPC),however,,-lookingbehaviorplaysasmallerroleduringthehighandvolatileinflationregimeto1981thaninthesubsequentperiodofmoderateinflation,:E31,E37,E52,E58Keywords:NewKeynesianPhillipsCurve,inflationsurveyforecasts,stickyprices,structuralbreaks,,,FEMES2006,andseminarparticipantsattheUniversitiesofAuckland,Otago,andManchester,forusefulcomments,,however,,ChinaFinancialPolicyResearchCen-tre,SchoolofFinance,RenminUniversityofChina(E-mail:zhangcs@).,CentreforGrowthandBusinessCycleResearch,Economics,SchoolofSocialSciences,UniversityofManchester(E-mail:@).DONGHEONKIMisAssistantProfessorofEconomics,Ko-reaUniversity(E-mail:dongkim@)andCentreforGrowthandBusinessCycleResearch,,2007;andacceptedinrevisedformOctober1,,CreditandBanking,,(June2008)©C2008TheOhioStateUniversity
668:MONEY,CREDITANDBANKINGThetheoriesusuallystopshort,however,ofspecifyingmodelsofaggregatesupplythatareintendedtoholdgenerally.—DavidRomerAdvancedMacroeconomics,2006,(NKPC),GaliandGertler(1999),Sborndone(2002,2005),andGali,Gertler,andLopez-Salido(2005)findapredominantroleforfutureexpectedinflation,whileFuhrerandMoore(1995),Fuhrer(1997),Rude-busch(2002),Linde(2005),Roberts(2005),andRuddandWhelan(2005),nonethe-less,crucialforunderstandingthedrivingprocessforinflation,namely,whetheritisdrivenonlybytheexpecteddiscountedsumofcurrentandfuturevaluesoftheoutputgap(asadvocatedbyGaliandGertler1999)orwhetherlaggedinflationalsopressuresinflation(asinFuhrer1997).Moreimportant,,Ball(1999)andSvensson(1999b)findthatnominalincomegrowthtargetingisdestabilizinginabackward-lookingmodel,whereasMcCallumandNelson(1999a,1999b),,overthelongspanofthepost-1960s,,(atanannu-alizedrate),theFederalFundsRate(FFR),over1960Q1–,,inaforward-lookingmonetarypolicyframework,Clarida,Gali,andGertler(2000)formallydocumentthattheresponseoftheFFRtoinflation(andtheoutputgap)(.,Clarida,Gali,andGertler1999),(2006),whoprovidesempiricalevidencethattheslopecoefficientofthereduced-form(backward-looking),agrowingliteratureintheunivariatecontext,includingAlogoskoufisandSmith(1991),Taylor(2000),CogleyandSargent(2001),Willis(2003),LevinandPiger(2004),andZhang(Forthcoming),’pricingbehavior(Taylor2000)andinturnaffecttheshort-run
CHENGSIZHANG,,ANDDONGHEONKIM:66920GDPIPDFFR1510501960 (%).NOTE:Samplespans1960Q1–-of-quarter.(Datasource:EconomicData-FRED,.),anddespitetheimportanceoftheNKPC,KimandKim(Forthcoming)(Forthcoming)haveasim-ilaraim,theyemployMarkovswitchingmodelsforthispurpose,whereaswestudythequestionthroughthetestsofAndrewsandPloberger(1994)(theSurveyofProfessionalForecasters(SPF),theMichi-ganSurvey,andtheGreenbookforecasts),weaimtocapturepeoples’responsestoeconomicperformanceandhencetomoreaccuratelymeasureinflationexpectationsthanapproachesexplicitlybasedonrationalexpectations;seealsoRoberts(1995).Ourapproachissupportedbyrecentstudiesthatpresentevidenceinfavorofusingobservedinflationforecastsasmeasuresofinflationexpectationsinmonetarypolicyanalysis(Croushore1993).Nevertheless,,,apredominantroleforinflationexpectations,withcorrespondingempiricalsupportforstickypricesmodels,-sizedbyourfinding(whensurveyinflationexpectationsareused),
670:MONEY,CREDITANDBANKINGstructuralchangesintheNKPC,employingobservedinflationforecastsandusingadynamicmodelthatisfreefromserialcorrelation,whileSection4checkstherobustnessoftheseresultstotheuseofthemorecommon“stylized”(1995),Roberts(1995),Fuhrer(1997),Yun(1996),GoodfriendandKing(1997),Clarida,Gali,andGertler(1999),GaliandGertler(1999),Jensen(2002),MankiwandReis(2002),Woodford(2003),andChristiano,Eichenbaum,andEvans(2005),amongothers,-foundedNKPCcanbeexpressedasepi=pi+αy,(1)tyttewherepiistherateofinflation,ydenotestheoutputgaporrealmarginalcost,andpitttistheinflationthatwouldprevailifoutputwereatitsnaturalrate(.,theoutputgapeiszero)andsupplyshocksareabsent;piissometimesknownascoreorunderlyingtinflation(Romer2006,).,GaliandGertlert(1999)elegantlyproposethatcoreinflationinthemicro-foundedNKPCisgivenbyepi=αEpi+αpi,(2)ftt+1bt−1twhereEpidenotesexpectedinflationforperiodt+1giveninformationavailablett+(1)with(2),the“New”Keynesianshort-runaggregatesupplycurveispi=αEpi+αpi+αy,0≤α≤1,0≤α≤1,andα>0,(3)tftt+1bt−1ytfbyinwhichtheconvexrestrictionα+α=,notablySvensson(1999a,1999b,2000),efurthersuggestsamoregeneralspecificationforcoreinflationpitakingaccountoftaspecificinflationtarget,∗∗pi=αEpi+αpi+αpi,(4)ftt+1bt−1t∗wherepidenotestheconstantinflationtargetofthecentralbank.
CHENGSIZHANG,,ANDDONGHEONKIM:671Whenα=0andα=1,(3),α=1andα=0fbgivesrisetothepurelyforward-lookingNKPCwithpricestickiness,,ifbothαandαarenonnegligible,tfbtheimpliedinflationprocessbecomes∞∞∑∑i−1i−1pi=ααpi+ααEy,(5)tbt−iytt+i−1ffi=1i=αandα,,Ball(1993,1994)andFuhrer(1997)suggestthattheNKPCwithalargevalueofα,Jensen(2002),Rudebusch(2002),Walsh(2003b),SvenssonandWoodford(2003,2004),McCallumandNelson(2004,2005),andSvensson(2005)(2005),byemployingamonetarypolicyanalysisframeworkincorporatingapurelyforward-lookingNKPC,RotembergandWoodford(1997),Levin,Wieland,andWilliams(1999)showthatinaPhillipscurvemodelwithadominantroleforinflationinertia,,therecentliterature,includingRudebusch(2002)andWalsh(2003a),,however,,GaliandGertler(1999)andGali,Gertler,andLopez-Salido(2005)employgeneralizedmethodofmoments(GMM)inconjunctionwithrationalexpectationstoestimate(3)withunitlaborcostasthemeasureofmarginalcostandconcludethatinflationinertiaismuchlessimpor-tantthansuggestedbyFuhrerandMoore(1995)andFuhrer(1997)(2002,2005)estimatesaclosed-formsolutionoftheNKPCandderivesresultsconsistentwithGaliandGertler(1999),(2005),however,(1999)isinducedbyanomittedvariableprobleminconjunctionwiththeuseofinstrumentalvariables(IV).Byestimatinganalternativeclosedforminflationequation,RuddandWhelan(2005)(1995),Fuhrer(1997),aswellasrecentempiricalstudiesofRudebusch(2002),EstrellaandFuhrer(2002,2003),Adamand
672:MONEY,CREDITANDBANKINGPadula(2003),Fuhrer(2006),Linde(2005),andZhang,Osborn,andKim(2006)providegenerallyconsistentevidencewithRuddandWhelan(2005).DespiteintensiveempiricalinvestigationofthehybridNKPCof(3),possiblechangesinbehavioroverdifferentregimeshasreceivedlittleattention,althoughtherecentstudiesofBarkbu,etal.(2005),JondeauandBihan(2005),andKimandKim1(Forthcoming).(2005)andJondeauandBihan(2005)imposearationalexpectationsassumptioninestimatingtheNKPCandthenapplyatestforasinglestructuralchangeasamodelspecificationcheck,,theMarkovswitchingmodelresultsofKimandKim,usingbothrealizedinflationandsurveyinflationdata(specificallytheSPFforecasts)tocapturetheforward-lookingcomponentin(3),-ditiontotheseconflictingconclusions,,be-causelaggedinflationappearsontheright-handsideofthedynamicmodelandhencethepresenceofserialcorrelationinthedisturbanceswouldrenderlaggedinflationaninvalidinstrument;seeZhang,Osborn,andKim(2006).Althoughthereisasyetnoclearevidenceonthestructural(in)stabilityoftheNKPC,(Willis2003)andmonetarypolicy(JuddandRudebusch1998,BrainardandPerry2000,Clarida,Gali,andGertler2000,Roberts2006),Willis(2003),includingthelabor,goods,andcapitalmarkets,arelikelytoaffectfirms’αandαarefunctionsof“deep”fbparametersrelevanttofirms’pricingbehavior,,monetarypolicyinfluencesinflationthroughapolicytransmissionmechanism,,Clarida,Gali,andGertler(2000)(2000),Taylor(2000)suggeststhatthereductionoffirms’pricingpoweroverthemostrecenttwodecadesispositivelycorrelatedwiththe(low)levelofinflation,(2000)(2003),however,anumberofrecentstudiesexaminingthepurelyforward-lookingNKPCmodel,includingEstrellaandFuhrer(2003)andCanova(2005),amongothers,thatfindnostructuralinstabilityintheforward-lookingparameter.
CHENGSIZHANG,,ANDDONGHEONKIM:673ByinvestigatingthestructuralstabilityoftheNKPC,,weextendthesmallgroupofpapersthatestimatetheNKPCusingobservedinflationforecasts(includingRoberts1995,AdamandPadula2003,KimandKimForthcoming)(,JondeauandBihan2005,KimandKimForthcoming),(1983),inwhichfirmsareassumedtorevisetheirpricesinanygivenperiodwithafixedprobability(1−θ).FollowingGaliandGertler(1999),weassumebothforward-andbackward-lookingfirmscoexistinproportions(1−ω)andω(1999)furtherassumethatthebackward-lookingfirmsadjusttheirpriceusingB∗p=p+pi,(6)t−1tt−1B∗wherepdenotesthe(log)pricesetbybackward-lookingfirms,,quarterlyinflationisrelativelynoisy,sothatbackward-lookingagentsmayconsideraweightedaverageofpastinflationratherthanthestylizedsinglelagin(6),sothatB∗p=p+ρ(L)pi,(7)t−1tt−12q−1whereρ(L)=ρ+ρL+ρL+···+ρLisapolynomialinthelag123qoperatorwithρ(1)=,q=(7)inconjunctionwiththeusualassumptionsinCalvo’s(1983)model,itisshownintheAppendixthattheNKPCmodelhastheform3∑pi=c+αEpi+αpi+α pi+αy+η,(8)toftt+1bt−1 bit−iytti=1
674:MONEY,CREDITANDBANKINGwheretheconstanttermcreflectsthesteady-stateinflationrateandηcaptures0trandomfactorsthatalsoaffectinflation,-sentationin(8)summarizestheimpactofpastlevelsofinflationoncurrentinflationthroughthesinglecoefficientα,,Osborn,andKim(2006),(8)–,inwhatfollowswefocuson(8),asarobustnessassessment,wealsoconsiderthestylizedmodelwithasinglelagofinflation,namely,pi=c+αEpi+αpi+αy+η.(9)t0ftt+1bt−,,1979Q3isthestartoftheVolcker–Greenspaneraduringwhichmonetarypolicydifferedsignificantlyfromthepreviousregime;seeClarida,Gali,andGertler(2000).WhilethelinkbetweenmonetarypolicyandinflationmakesitplausiblethatsuchchangesmayleadtostructuralbreaksintheparametersoftheNKPC,,weperformbreaktestsusingthemethodologydevelopedbyAndrews(1993)andAn-drewsandPloberger(1994),takingintoaccountthepossibilityofmultiplebreaksthroughtherepartitionprocedureofBai(1997).Priortoexaminingthesetests,,,theforecastsfromtheSPFaregenerallyobtainedinthemiddleofthequarterandhencearelikelytobecorrelatedwiththesupplyshocksrepresentedbyηin(8).tInaddition,yisalsolikelytobecorrelatedwiththecontemporaneousnoise,,aswillbeevidentintheempiricalestimation,Durbin–Wu–Hausmanspecificationtestsindicatethatinmostcasesordinaryleastsquares(OLS)isnotconsistentbecausethenullhypothesisthatEpiandycanbetreatedasexogenousisrejectedatconventionallevelsoftt+,weuseIV,ormoregenerally,(8)consistsoftwolagseachofinflationexpectations,theoutputgap,unemploymentrate,growthrateofmoneyaggregate(M2),(8)isgenerallyfreeofsignificantserialcorrelationinempiricalestimations,laggedinflationvaluesontheright-handsideof(8),thebaselineestimationsareverifiedthroughtheIVserialcorrelationtest(DavidsonandMacK-innon1993,Godfrey1994),Hansen’s(1982)J-testforoveridentifyingrestrictions,andtheStockandYogo(2003)generalizedF-testforweakIV.
CHENGSIZHANG,,ANDDONGHEONKIM:675Basedontheprecedingdesign,,writetheNKPCmodelinvectornotationasY=Xβ+Xβ+η,(10)1122whereYconsistsofobservationsoninflation,pi,andthematrixofexplanatorytvariablesin(8)isX=[XX].Ourfocusofinterestisthecoefficientsα,αthat12fbrepresentforward-andbackward-lookingbehaviorin(8),togetherwiththeinterceptwhosechangesmaycaptureshiftsintheinflationtarget,oritsperceptionbyagents,through(4).Inthiscasethestabilitytestrelatestothecoefficientsubvectorβ=2′(c,α,α),τ∈[0,1],whichsplitstheentiresampleintotwosubsamples,say,s1=τTands2=(1−τ)T,,theIVestimateforβovers1isgivenby2[]−1s1s1′s1s1s1′s1s1˜β=WMWWMy,(11)22W22W11s1s1s1s1s1whereM=I−P,W=PX(i=1,2),Zdenotesthematrixofobserva-s1WWiZi11tionsfortheinstrumentalvariablesandPistheprojectionmatrixfortheobservations′−1′indicated,forexample,P=X(XX)(11).Inaddition,theheteroskedasticity-consistentcovari-ancematrixestimates(HCCMEs)forβovertheseparatesubsamplesarecomputed2by()sisi′si−1sisisisi′si−1˜˜˜Vβ=(WW)W W(WW),i=1,2,(12)2′sisisisi˜where =diag(η˜η˜)andη˜τ,theWaldstatisticfortestingthenullhypothesisβ=βis22[][()()][]′−1s1s2s1s2s1s2˜˜˜˜˜˜˜˜Wald(τ)=β−βVβ+Vββ−β.(13)T222222TheAndrews–PlobergerSup-WaldstatisticfortestingabreakatanunknownpointisthencomputedasthemaximumvalueofWaldstatisticin(13)overallpossiblebreakpoints(.,τ∈[,]).Theassociatedexponential-Waldstatistics,givenby∫τmaxExpWald=lnexp[(τ)]dτ,(14)Tτminarealsoreported,asHansen(2000)(9),thesestatisticsareimplementedwiththeNewey–West(fixedbandwidth)HACmatrixusingtheBartlettkernel,-WaldandExpWald,withasymptoticp-valuescomputedusingthemethodofHansen(1997).
676:MONEY,CREDITANDBANKINGTocapturepossiblemultiplebreaks,wesequentiallyapplythestructuralbreaksteststosubsamplesandperformtherefined(repartition)proceduresuggestedbyBai(1997)andBaiandPerron(1998).Thatis,ifthenullhypothesisofstabilityisrejected,,theirdatesarerefined(reestimated).Forexample,supposetwo˜˜˜˜breakdates(k,i=1,2)areidentifiedandk<˜˜˜subsample[1,k]and,likewise,kisrefinedusingthesubsample[k,T].,,pi;inflationexpecta-ttions,Epi;andameasureoftheoutputgap,+1tthecontextofastandardmeasureofinflationandtofacilitatecomparisonswiththeliterature,wemeasureinflationbytheannualizedquarterlygrowthrateoftheGDPdeflator(.,400timesthefirstdifferenceofthelogGDPdeflator).TheoutputgapisobtainedfromtheestimatesofrealpotentialGDPpublishedbytheCongressionalBudgetOffice(CBO).Weemployfourdifferentinflationforecastsintheanalysisofthissection,namely,the1-quarter-and1-year-aheadmedianforecastsfromtheSPF(denotedSPF1QandSPF1Y,respectively),theGreenbook1-quarter-aheadforecasts(Greenbook),and1-year-aheadgeneralpriceinflationforecastsfromtheMichigansurvey(Michigan).WiththeexceptionoftheMichiganforecasts,,,although(notsurprisingly),theGreenbookforecastsarepreparedwithintheFed,whiletheMichigansurveyaimstocapturetheviewsofthegeneralpublic,theseseriesrepresentdifferentgroupsofagentswith(presumably)-abilityofthefourinflationforecasts,whichcover1968Q4–2005Q4,1970Q1–,,thetotalsamplesizeavailableandthenumberofinstrumentsusedmakeitinfeasibletorepeatthepartitioningmorethanonceinourapplication.
CHENGSIZHANG,,ANDDONGHEONKIM::SPF1Q,SPF1Y,Greenbook,andMichigandenote,respectively,SPF1-quarter-aheadGDP(before1992Q1,GNP)inflationforecasts,SPF1-year-aheadGDP(before1992Q1,GNP)inflationforecasts,GreenbookquarterlyGDPinflationforecasts,andthegeneralpriceinflationforecasts(1-year-ahead)fromtheMichigansurvey.(Datasources:ThefirstthreeforecastseriesareobtainedfromthewebsiteoftheFederalReserveBankofPhiladelphia,andtheMichigandataarecollectedfromthewebsiteofSurveyofConsumersoftheUniversityofMichigan.)forthe1-quarter-and1-year-ahead(respectively)SPFforecasts,and1968Q3–1999Q43and1968Q3–2005Q2(respectively),,inparticu-lar,–Ploberger(1994)unknownstruc-turalbreaktestsfortheNKPCmodel(8)(10),Overall,(c,α,α),and0fbαdefinetherelevantsubvectorβforjointstabilitytests,whiletheotherresults %ofthesample(orsubsample)-Supandp-Expdenotep-valuesassociatedwiththecorrespond-ingAndrews–Plobergerteststatisticsforthenullhypothesisofnostructuralchange,-sampleresultsinTable1,p-valuesforthebreakteststatisticstestingoverallmodelstabilityarealwayshighlysignificant,,theNKPCisstatisticallyunstableover1968–2005andthestrongestevidenceofchangerelatestotheintercept,,,earlierMichigansurveydata(.,priorto1967)maybeofdistinctlylowerquality(Rudebusch2002)andhence,asinRudebusch(2002),theMichigandata(meanvalues)-valuesassociatedwithαinregressionsusingSPF1QandSPF1Ymayybeunreliablesincetheimpliedbreakdate2000Q1correspondstothepolarpointinthesearchinginterval.
678:MONEY,CREDITANDBANKINGTABLE1ANDREWS–PLOBERGERTESTS(GMM)FORTHENKPCUSINGSURVEYINFLATIONFORECASTSWholesamplePre-1981Q1Post-1981Q1Post-1975Q11975Q2–———(c,α,α)ααα bα———(c,α,α)ααα bα———(c,α,α)ααα bα———(c,α,α)ααα bα:Thewholesampleis1968Q4–2005Q4,1970Q1–2005Q4forthe1-quarter-and1-year-ahead(respectively)SPFforecasts,1968Q3–1999Q4and1968Q3–2005Q2(respectively),(8).ThebaselineIVsetfortheNKPCincludestwolagsofeachofinflationforecasts,short-terminterestrate(3-monthTreasurybillrate),theoutputgap,unemploymentrate,andM2growth,plusthelagsofinflationincludedinthemodel(andaconstant).p-Expandp-Supdenotep-valuesofAndrews–%oftherelevant(sub)-consistentcovariancematrix(HCCME)isusedforalltests.
CHENGSIZHANG,,ANDDONGHEONKIM:679TheremainingcolumnsinvestigatethepossibilityofmultiplebreaksandimplementtherefinementprocedureofBai(1997)andBaiandPerron(1998).Specifically,,,itshouldbenotedthatafterallowingforlags,,theasymptoticAndrews–,,,theevidencepointstoa1975break,,theOveralland/or(c,α,α)resultswiththeotherforecastseriestypicallyindicateabreakin0fbtheNKPCin1981,,threepossiblebreaksarerevealedinthesecases,namelyaround1975,1981,,,orthepost-1975or1975–2000subsamplesareused,,theevidenceforabreakin1975largelydisappears,,asalreadynoted,,thestructuralbreaktestsoverthepost-1981subsampleconsistentlysuggestafurtherbreakaroundthebeginningofthe2000s,,thispotentialbreakisclosetotheendofthesearchingintervals,,weconditiononabreakin1981andestimatetheNKPCof(8)recursivelybyGMM,startingwithaninitial(10-year)sampleperiodof1981Q2–α,α,,itisevidentthatα,ingeneral,dominatesαbefore2001,,whentherecursivesampleincorporatesobservationsafter2001(,the1975–,thisdateissupportedbythewholesampleandpost-1975results.
680:MONEY,-1Michigan (right scale):Estimatesreportedrelateto(8).Therecursiveestimationstartsfrom1981Q2–1990Q1;“forward”and“backward”denoteestimatesofαandα,),adistinctreductionoccursintheestimatesofαwhileafsizableincreaseoccursinα.ThelastgraphinFigure3plotsrecursiveestimatesofbintercept,,therecursiveestimatesofFigure3confirmtheendof2000astheappropriatebreak
CHENGSIZHANG,,ANDDONGHEONKIM:681TABLE2DUMMYTESTSFORTHENKPCUSINGSURVEYINFLATIONFORECASTS:∑ddddpi=c+αEpi+αpi+α pi+αy+cd+αdEpi+αdpi+αdy+εt0ftt+1bt−1 bit−iyttttt+1tt−1ttt0fbyi=1ddddddc˜α˜α˜α˜Allα=α=–()()()()–()()()()––()()()()α+α=1fb3∑dddpi−pi=c+α(Epi−pi)+α pi+αy+cd+αd(Epi−pi)+αdy+εtt−10ftt+1t−1 bit−iyttttt+1t−1ttt0yfi=1dddc˜α˜α˜–()()()–()()()––()()()NOTES:(seeTable1)andZd(.,Zmultipliedbythedummyvariable).,wedefinedummyvariablesbasedonachangepointin2001Q1toinvestigatethenatureofthisstructuralbreak,withresultsshowninTable2usingthesubsample1981Q2–α bisufferapost-1981structuralbreak,,theresultsdonotprovideclearevidenceonwhichcoefficientschange,α+α=1inordertoalleviatethefb7collinearity,andtheseresultsclearlyindicatethatthestructuralshiftin2001Q1canbeassociatedwiththeintercepttermof(8).Therefore,thenextsubsectioninvestigatesthenatureofchangesintheNKPCwhenbreaksinthecoefficientsarerecognizedatthebeginningof1981and(fortheintercept),.
682:MONEY,-1981c˜α˜α˜α˜()()()()SPF1Y–()()()()Michigan–()()()()()()()()()()()()()()()()()()()()()()()()∑-1981:pi=c+αEpi+αpi+α pi+αy+cd+εt0ftt+1bt−1bit−tti= iyt10dc˜c˜α˜α˜α˜p-autop-overWeakIVHausman0fby0SPF1Q–()()()()()SPF1Y–()()()()()Michigan–()()()()()Greenbook–—()—()()()ImposingconvexrestrictionSPF1Q–()()()()()SPF1Y–()()()()()Michigan–()()()()()Greenbook–—()—()()()NOTES:Pre-1981referstosamplesfromtheavailablestartingdatesforeachinflationforecastseriesto1980Q4,withtheequationestimatedinPanelAgivenby(8).Forthepost-1981periodinPanelB,thereportedestimatesareconditionalona2001Q1interceptbreak(exceptforGreenbook).-auto,p-over,WeakIV,andHausmanrefertop-valuesofIVserialcorrelationtest(uptoorder4),Hansen’s(1982)J-test,StockandYogo’s(2003)weakIVtest(asaruleofthumb,%leveland30%biasofOLSoverIVestimator),andDurbin–Wu–Hausmantest(withthenullofconsistencyoftheOLSestimator;HCCMErobust),,(8)overpre-andpost-1981periodsfortheintercept,forward-andbackward-lookinginflationcoefficients,and8theoutputgap,,theDurbin–Wu–,pre-andpost-1981correspondtosubsamplesendingin1980Q4andstartingin1981Q2,respectively,excludingthebreakpoint(1981Q1),theestimatedquarterofabreakissimilarlyexcludedfromsubsampleestimationsthroughoutthepaper.
CHENGSIZHANG,,ANDDONGHEONKIM:683Hausmanstatistics(heteroskedasticityrobust)inthelastcolumnconfirmtheneedforGMMestimation,sincethenullhypothesisofconsistencyofOLSinestimating(8)istypicallyrejectedatconventionallevels(althoughthisisnotthecaseforGreenbookdataortheMichiganseriesoverthepre-1981period).PanelAprovidesevidencethatthebackward-lookingbehaviorplaysamoreim-portantrolethantheforward-lookingcomponentbefore1981,-spondingestimatespost-1981conditionalonthebreakintheinterceptin2000Q1,withoutandwiththerestrictionα+α=1imposed,-fbsults,andconditionalontheinterceptshift,,,itisalsonotablethattheresultsofTable3indicatethattheestimatedslopecoefficientfortheoutputgap,α,typicallyyfallssubstantially(oftenbyaroundahalf)(2001),Staiger,Stock,andWatson(2001),andRoberts(2006),,,theestimatesofαandαprovideourmainfindingfromTablefb3,namely,thatthebackward-lookinginflationbehaviorisstrongover1968–1981,,firstbyusingthestylizedNKPCmodel(9)inconjunctionwithobservedinflationforecastseriesandsecondbyapplyingarationalexpectationsapproximationinthecontextofthemoregeneraldynamicspecificationof(8),Osborn,andKim(2006),empiricalestimationsforthestylizedformulationoftheNKPC(9)generallymanifestserialcorrelation,,theresultsinTable1indicatethatα,thisignorestheevidenceofabreakinothercoefficientsin1981Q1andthereisnoindicationinthepost-1981analysisinTable2thatα.
684:MONEY,,weemployabase-lineIVsetforthestylizedNKPC,whichincludestwolagseachofsurveyinflationforecasts,theoutputgap,unemploymentrate,andshort-terminterestrate,,weimplementtheAndrews–Plobergerstructuralbreaktestsandtheresults(notreportedhere)suggestthatingeneralthepatternandtimingofthestructuralbreaksin(9)aresimilartothosefor(8).Thatis,1975,1981,and2001appeartobethreepossiblestructuralbreakdates,withthe1981dategenerallysignificantwhenthe1975breakisrecognized,-1981subsampleareagainusedtoinvestigatethenatureofthebreakin2001,withresults(afterimpositionoftheconvexrestriction)-lookingbehaviorinthestylizedmodel(9)dominatesbefore1981whiletheforward-lookingcomponentisquantitativelyverysmallinallcases,-1981period,theresultsinPanelBreinforcethoseembeddedinPanelBofTable3:thatis,conditioningonaninterceptbreakin(9)11in2001,,withouttheconvexrestriction,,,whiletheestimatesforα,thelattercoefficientsarebnotstatisticallysignificant(at5%),inlinewithTable3,,theIVserialcorrelationtest(uptoorder4)–,,theStockandYogo(2003)weakIVstatisticsuggeststhattheinstrumentsherearemuchlessstrongthanthebaselineIVsetformodel(8),,,usingtheMichigandata,abreakintheforward/,-1981inTable4appearseriallyuncorrelated,,wedonotincludelaggedinflationintheIVforthestylizedmodel.
CHENGSIZHANG,,ANDDONGHEONKIM:-1981c˜α˜α˜α˜()()()()SPF1Y–()()()()()()()()()()()()()()()()()()()()Michigan–()()()()()()()()-1981:pi=c+αEpi+αpi+αy+cd+εt0ftt+1bt−1yttt0dc˜c˜α˜α˜α˜p-autop-overWeakIVHausman0fby0SPF1Q–()()()()()SPF1Y–()()()()()Michigan–()()()()()Greenbook–—–())—()()()ImposingconvexrestrictionSPF1Q–()()()()()SPF1Y–()()()()()Michigan–()()()()()Greenbook–—––()—()()()NOTES:TheequationestimatedinPanelAisgivenby(9).TheIVsetincludestwolagsofeachofsurveyinflation,outputgap,unemploymentrate,andshort-terminterestrate(3-monthTreasurybillrate).-autoandp-overrefertop-valuesofIVserialcorrelationtest(uptoorder4)andHansen’s(1982)J-test,,,itismorecommontoexaminetheNKPCusingarationalexpectationsapproximation,asinGaliandGertler(1999)orGali,Gertler,andLopez-Salido(2005).Theseauthorsproposethatthismodelshouldbeestimatedusingrealmarginalcostratherthananoutputgapvariableandarguethatlaggedinflationquantitativelyplaysanegligiblerolewhilefutureinflationis
686:MONEY,(1999)andGali,Gertler,andLopez-Salido(2005).However,ouranalysisisbasedonthemodelintheformof(8),,weemploythesameIVsetsasinGaliandGertler(1999)andGali,Gertler,andLopez-Salido(2005),denoted13byGG-1999IVandGGL-2005IV,,whichgivesrisetoexactlythesamecoefficientestimatesasobtainedwhenEpiisreplacedtt+1bypiin(8).However,asdiscussedinZhang,Osborne,andKim(2006),thisrenderst++1addition,sinceGaliandGertler(1999)emphasizetheimportanceoflaborincomeshareofthenonfarmbusinesssectorastherealdrivingvariable,wealsoprovideempiricalresultsusingthelaborincomeshare(denotedNFB-LS),,weapplythetestofAndrewsandPloberger(1994),searchingforabreakoverthecentral70%ofthe(sub)sampleandapplyingtherepartitioningprocedureofBai(1997)andBaiandPerron(1998),wenowuseasampleperiodstartingin1960Q1,sincewearenotconstrainedbytheavailabilityofsurveydataandthelongerperiodprovidestheopportunitytoexaminewhetherthereisastructuralbreakintheNKPCin1975,,-1999IVinstrumentsetdoesnotprovidesignificantevidenceofanybreakwhenthewholeperiodisconsidered,,,andincontrasttoTable1,therationalexpectationsapproximationovertheextended(whole),thepre-1981subsampleresultssometimespointstoabreakinthe1960s,wheninflationbegantorise,(1997)refinementprocedure,thesampleisrepartitionedbasedonthepre-1981breakdateasestimatedfromtheoveralltestforeachspecificregression,withthisdateindicatedinthetablesimplyby1960s/-1960s/-1999IVincludesfourlagsofeachofthefollowingvariables:inflation,outputgap,laborincomeshare,wageinflation,commoditypriceinflation,andlong-shortinterestratespread;GGL-2005IVincludesfourlagsofinflationandtwolagsoflaborincomeshare,outputgap,+1errorproblem.
CHENGSIZHANG,,ANDDONGHEONKIM:687TABLE5ANDREWS–PLOBERGERTESTS(GMM)FORTHENKPCWITHRATIONALEXPECTATIONS1960–2005Pre-1981Q1Post-1981Q1Post-1960s/70s1960s/70s–(c,α,α)ααα bα(c,α,α)ααα bα(c,α,α)ααα bα(c,α,α)ααα bα:Thestructuralbreaktests(HCCMErobust)areimplementedoverthecentral70%oftherelevant(sub)sample,withtheequationestimatedgivenby(8).
688:MONEY,CREDITANDBANKINGconfirmthe1981Q1break,withtheexceptionoftheGG-2005IVinstrumentsetandthelaborshare,,ofcourse,beanindicationofabreakaround2000,,asimilarbreak(nowin2001Q1),however,
CHENGSIZHANG,,ANDDONGHEONKIM:689TABLE6SUBSAMPLEESIMATESOFTHENKPCWITHRATIONALEXPECTATIONSα˜α˜α˜()()()()()()()()()()()()–()()()()()()–()()()––()()()()()()()()()()()()()()()–()()()()()()()()()()()()NOTES:Theequationestimatedisgivenby(8).-1999IVindicatesthattheinstrumentalvariablesusedarethoseofGaliandGertler(1999),namely,fourlagsofeachofinflation,outputgap,laborincomeshare,wageinflation,commoditypriceinflation,andlong-shortinterestratespread;GGL-2005IVindicatesthattheIVusedarethoseofGali,Gertler,andLopez-Salido(2005),namely,twolagsofeachoflaborincomeshare,outputgap,wageinflation,-1999IV,,,,theestimateonlaggedinflationinthepre-1981regressionusingthelaborincomeshareiseffectivelyequaltothatonfutureinflation,whichindicatesthatGaliandGertler’s(1999)-2005IVproducesdifferentestimatesforthekeycoefficientsandtheestimatesofαarelargerthanαfb
690:MONEY,CREDITANDBANKINGpre-1981,,usingthelaborincomeshare,thepointestimateforα,(negative)-1999IVandGGL-2005IV,(usingsurveyinflationexpectations),,Ball(1999)employsthePhillipscurveofinflationstickinessinconjunctionwiththesimpledynamicaggregatedemandequationy=−βr+λy+ε,(15)tt−1t−1twhererdenotesthedeviationoftherealinterestratefromitssteady-statelevelandεα=0andα=1intheNKPC,(1999a,1999b),however,setα=1andfα=0andconclude(throughsimulations)thatBall’’s(1999)analysisandpolicyrecommendationmaybeapplicabletothehighandvolatileinflationregimeof1968–1981,whiletheproposalofMcCallumandNelson(1999a,1999b)appearsmoreappealingoverthemostrecenttwodecadessincetheempiricalresultsshowthatforward-lookingbehaviorhasbeenmoredominantafter1981.
CHENGSIZHANG,,ANDDONGHEONKIM:691Inaddition,thefindingofstructuralchangeintheforward-andbackward-lookingbehaviorsofinflationin1981alsolendssomeinsightontheoptimization-basedmonetarypolicyruleanalysisasinRotembergandWoodford(1997)andLevin,Wieland,andWilliams(1999),,ourresultsindicatethatforward-andbackward-lookingbehaviorintheNKPCmaybecloselyrelatedtothedegreeofinflationpersistence,sincethetimingofthebreaksidentifiedinthispaperareinlinewithchangesininflationpersistencedocumentedinTaylor(2000),Willis(2003),andZhang(Forthcoming).Thismayimplythatduringhighlypersistentinflationperiods,pastinflationcontainsmorerelevantinformationforfirms’,thefindinghereisalsoinagreementwithErcegandLevin(2003),–Greenspanperiod(after1984).Therefore,changesininflationdynamics,andspecificallyforward-versusbackward-lookingbehavior,maybecloselyassociatedwithdifferentmonetarypoli-ciespursuedbeforeandaftertheVolcker–,Gali,andGertler(2000)findthat,associatedwithdifferinganti-inflationstancesoftheFederalRe-serve,beforetheVolcker–GreenspanyearstheFedappearsveryreluctanttorespondtochangesinexpectedinflationwhileduringtheVolcker–,theempiricalresultsinthecurrentstudysuggestthatchangesinthemonetarypolicyrulemayalsohaveinducedchangesintheforward-andbackward-lookingbehaviorsininflationdynamics,,’,post-1981,theFedhasemphasizedpricestabilitymorestronglythanoutputstabilizationasthegoalofmonetarypolicy(see,.,Clarida,Gali,andGertler2000).Thismighthavebeentransmittedtoprice-settingbehavioroffirms,,-1980s(.,KimandNelson1999,McConnellandPerez-Quiros2000)andfirmsmightincorporatethisstabilizedoutputintotheirprice-settingbyrespondinglesstotheoutputgap(orlaborshare),ourresultsareinlinewiththoseofKimandKim(Forthcoming),andbackward-lookingbehaviordecreases,,aninteresting
692:MONEY,CREDITANDBANKINGdifferenceisthatKimandKim(Forthcoming)alwaysfindforward-lookingbehaviortobemoreimportantthanthebackward-bookingcomponent,,inasystemofNewKeynesianequations,Ireland(2004)showsthatthecoefficientestimateonlaggedinflationoftheNKPCisstatisticallyinsignificantover1980Q1–2003Q1,(2005)alsofindsnobackward-lookingpricingbehaviorover1987Q4–(1997)isnotabletorejectthenullhypothesisthatinflationdynamicsover1979–1994arepurelybackward-lookingatconventionallevelsofsignificance,,,althoughtemporary,,theFederalOpenMarketCommittee(FOMC),,theshort-terminterestrate(theFFR)%%bytheendof2001andwasheldatahistoricallylowleveluntil2004,,,asnotedinSection1andseenin(4),(albeittemporarily)suchthattheFedhadahigherimplicitinflationtargetintheearly2000sthaninthe1990s,thenahigherlevelofinflationwillresult,-and
CHENGSIZHANG,,ANDDONGHEONKIM:693backward-lookingbehaviors,,however,,-lookingbehavioremphasizedbyGaliandGertler(1999)andGali,Gertler,andLopez-Salido(2005)appearstoplayarelativelysmallroleininflationdynamicsoverthe1968–81periodwhilethisbehaviorbecomesmoredominantafter1981,andtheconversephenomenonappliestothebackward-lookingbehaviorthatisstressedbyFuhrerandMoore(1995),Fuhrer(1997),EstrellaandFuhrer(2002,2003),andRuddandWhelan(2005).Thefindinghereiscon-sistentwithrecentresearchofIreland(2004),Bindelli(2005),andKimandKim(Forthcoming),asdocumentedinClarida,Gali,andGertler(2000),maycontaintheseedsofchangesininflationpersistenceandthechangesoftheforward-andbackward-lookingbehaviorsininflationdynamicsaround1981,,,,duringwhichdepressedeconomicactivitywasexacerbatedbytheSeptember11terroristattacks,resultinginchangesineconomicagents’perceptionofthecentralbank’,whichisconfirmedusingbothsurveyinflationexpectationsdataandarationalexpectationsapproximation,,whentheGreenbookforecastsoftheFedareemployed,(2000)findthattheFedhasinformationnotavailabletootheragentswhenproducingitsinflationforecasts,,,otherinflationforecastsmaybemorerelevantinthiscontextthanGreenbookforecasts.
694:MONEY,CREDITANDBANKINGAPPENDIX:DERIVATIONOFTHEEXTENDEDNKPCThefollowingdescribesthederivationoftheextendedNKPC,assuminganeco-nomicenvironmentsimilartoCalvo’s(1983)model,inwhichfirmsareabletorevisetheirpricesinanygivenperiodwithafixedprobability(1−θ).AsinGaliandGertler(1999),weassumeboth“forward-”and“backward-looking”firmscoexistintheeconomywithaproportionofωand(1−ω),,weextendtherecentpricingbehaviorofthebackward-lookingfirmstoincorporateaweightedprocessofpastinflation,’s(1983)modelandlog-linearapprox-imations,itispossibletoobtainthe(log)aggregatepricelevelas∗p=θp+(1−θ)p,(A1)tt−1t∗(relativetFBtotheaggregateprice)canbeexpressedasaconvexcombinationofpandp:tt()()∗FBp−p=(1−ω)p−p+ωp−p.(A2)ttttttNext,followingWoodford(2003,),thepricingbehavioroftheforward-lookingfirmscanbewrittenas[]∞T∑∑FT−tp−p=(1−θβ)(θβ)Epi+ζy,(A3)ttjttT=tj=t+1whereβdenotesasubjectivediscountfactor,ζisintroducedbytheprocedureoflog-linearization(seeWoodford2003foradiscussionofeconomicimplicationsofζ),(A3)givest()FFp−p=θβEpi+(1−θβ)ζy+θβEp−p.(A4)ttt+1ttt+1tt+1Weassumethatfirmsadjusttheirpricingbehaviorbyaweightedaverageofpastinflation,∗p=p+ρ(L)pi,(A5)t−1tt−12q−1whereρ(L)=ρ+ρL+ρL+···+ρLispolynomialinlagoperatorwith123qρ(1)=,wefindq=(A1)–(A5)givestheextended(theoretical)NKPCpi=αEpi+α(L)pi+αy,(A6)tftt+1bt−1yt
CHENGSIZHANG,,ANDDONGHEONKIM:695where−1α=θβψ,(A7)f−1α(L)=ω{θ+(1−θ)ρ(L)−(1−θ)θβρ(L)]}ψ,(A8)b −1α=(1−ω)(1−θ)(1−βθ)ζψ,(A9)yψ=θ(1+ωθβ)+ω(1−θ)(1+θβρ),(A10)12ρ(L)=ρ+ρL+ρL.(A11) 234Reparameterizing(A6)andtakingaccountofastochasticerroryieldstheNKPCmodel(8),,Klaus,andMarioPadula.(2003)“InflationDynamicsandSubjectiveExpectationsintheUnitedStates.”EuropeanCentralBank,,GeorgeS.,andRonSmith.(1991)“ThePhillipsCurve,thePersistenceofInfla-tion,andtheLucasCritique:EvidencefromExchange-RateRegimes.”AmericanEconomicReview,81,1254–,.(1993)“TestsforParameterInstabilityandStructuralChangewithUnknownChangePoint.”Econometrica,61,821–,.,andWernerPloberger.(1994)“OptimalTestsWhenaNuisanceParameterIsPresentOnlyundertheAlternative.”Econometrica,62,1383–,Andrew,.(2001)“ArePhillipsCurvesUsefulforForecastingInflation?”FederalReserveBankofMinneapolisQuarterlyReview,25,2–,Jushan.(1997)“EstimatingMultipleBreaksOneataTime.”EconometricTheory,13,315–,Jushan,andPierrePerron.(1998)“EstimatingandTestingLinearModelswithMultipleStructuralChanges.”Econometrica,66,47–,Laurence.(1993)“HowCostlyIsDisinflation?TheHistoricalEvidence.”FederalReserveBankofPhiladelphiaBusinessReview,November/December,17–,Laurence.(1994)“CredibleDisinflationwithStaggeredPrice-Setting.”AmericanEco-nomicReview,84,282–,Laurence.(1999)“EfficientRulesforMonetaryPolicy.”InternationalFinance,2,63–,Bergljot,VincenzoCassino,AileenGosselin-Lotz,andLauraPiscitelli.(2005)“TheNewKeynesianPhillipsCurveintheUnitedStatesandtheEuroArea:AggregationBias,StabilityandRobustness.”,Luca.(2005)“SystematicMonetaryPolicyandPersistence.”Mimeo,Universite´deLausanne.
696:MONEY,CREDITANDBANKINGBrainard,WilliamC.,.(2000)“MakingPolicyinaChangingWorld.”InEconomicEvents,Ideas,andPolicies:The1960sandafter,,–,DC:,GuillermoA.(1983)“StaggeredPricesinaUtilityMaximizationFramework.”JournalofMonetaryEconomics,12,383–,Fabio.(2005)“MonetaryPolicyandtheEvolutionofUSEconomy.”Workingpaper,,LawrenceJ.,MartinEichenbaum,.(2005)“NominalRigiditiesandtheDynamicEffectsofaShocktoMonetaryPolicy.”JournalofPoliticalEconomy,113,1–,Richard,JordiGali,andMarkGertler.(1999)“TheScienceofMonetaryPolicy:ANewKeynesianPerspective.”JournalofEconomicLiterature,37,1661–,Richard,JordiGali,andMarkGertler.(2000)“MonetaryPolicyRulesandMacroe-conomicStability:EvidenceandSomeTheory.”QuarterlyJournalofEconomics,115,147–,Timothy,.(2001)“.”NBERMacroeconomicsAnnual,331–,Dean.(1993)“Introducing:TheSurveyofProfessionalForecasters.”BusinessReview,FederalReserveBankofPhiladelphia,November,3–,Russell,.(1993),UK:,ChristopherJ.,.(2003)“ImperfectCredibilityandInflationPer-sistence.”JournalofMonetaryEconomics,50,915–,Arturo,.(2002)“DynamicInconsistencies:CounterfactualIm-plicationsofaClassofRational-ExpectationsModels.”AmericanEconomicReview,92,1013–,Arturo,.(2003)“MonetaryPolicyShiftsandtheStabilityofMonetaryPolicyModels.”ReviewofEconomicsandStatistics,85,94–,JeffreyC.(1997)“The(Un)importanceofForward-LookingBehaviorinPriceSpeci-fications.”JournalofMoney,Credit,andBanking,29,338–,JeffreyC.(2006)“IntrinsicandInheritedInflationPersistence.”InternationalJournalofCentralBanking,12,49–,JeffreyC.,.(1995)“InflationPersistence.”QuarterlyJournalofEconomics,110,127–,Jordi,andMarkGertler.(1999)“InflationDynamics:AStructuralEconometricAnalysis.”JournalofMonetaryEconomics,44,195–,Jordi,MarkGertler,-Salido.(2005)“RobustnessoftheEstimatesoftheHybridNewKeynesianPhillipsCurve.”JournalofMonetaryEconomics,52,1107–,LeslieG.(1994)“TestingforSerialCorrelationbyVariableAdditioninDynamicModelsEstimatedbyInstrumentalVariables.”ReviewofEconomicsandStatistics,76,550–,Marvin,.(1997)“TheNewNeoclassicalSynthesisandtheRoleofMonetaryPolicy.”NBERMacroeconomicsAnnual,231–,BruceE.(1997)“ApproximateAsymptoticPValuesforStructuralChangeTests.”JournalofBusinessandEconomicStatistics,15,60–80.
CHENGSIZHANG,,ANDDONGHEONKIM:697Hansen,BruceE.(2000)“TestingforStructuralChangeinConditionalModels.”JournalofEconometrics,97,93–,LarsPeter.(1982)“LargeSamplePropertiesofGeneralizedMethodofMomentsEstimators.”Econometrica,50,1029–,PeterN.(2004)“TechnologyShocksintheNewKeynesianModel.”ReviewofEco-nomicsandStatistics,86,923–,Henrik.(2002)“TargetingNominalIncomeGrowthorInflation?”AmericanEconomicReview,92,928–,Eric,andHerveLeBihan.(2005)“TestingfortheNewKeynesianPhillipsCurve:AdditionalInternationalEvidence.”EconomicModelling,22,521–,JohnP.,.(1998)“Taylor’sRuleandtheFed:1970-1997.”FederalReserveBankofSanFranciscoEconomicReview,3,3–,Chang-Jin,andYunmiKim.(Forthcoming)“NewEvidenceontheImportanceofForward-LookingandBackward-LookingComponentsinaNewKeynesianPhillipsCurve.”,Chang-Jin,.(1999)“
698:MONEY,CREDITANDBANKINGRoberts,JohnM.(2006)“MonetaryPolicyandInflationDynamics.”InternationalJournalofCentralBanking,September,193–,ChristinaD.,.(2000)“FederalReserveInformationandtheBehaviorofInterestRates.”AmericanEconomicReview,90,429–,DavidH.(2006):,JulioJ.,andMichaelWoodford.(1997)“AnOptimization-BasedEconometricModelfortheEvaluationofMonetaryPolicy.”NBERMacroeconomicsAnnual,297–,Jeremy,andKarlWhelan.(2005)“NewTestsoftheNewKeynesianPhillipsCurve.”JournalofMonetaryEconomics,52,1167–,GlennD.(2002)“AssessingNominalIncomeRulesforMonetaryPolicywithModelandDataUncertainty.”TheEconomicJournal,112,402–,ArgiaM.(2002)“PricesandUnitLaborCosts:ANewTestofPriceStickiness.”JournalofMonetaryEconomics,49,265–,ArgiaM.(2005)“DoExpectedFutureMarginalCostsDriveInflationDynamics?”JournalofMonetaryEconomics,52,1183–,Douglas,,.(2001)“Prices,Wages,.”,JamesH.,.(2007)“
CHENGSIZHANG,,ANDDONGHEONKIM:699Woodford,Michael.(2003)InterestandPrices:,NJ:,Tack.(1996)“NominalPriceRigidity,MoneySupplyEndogeneity,andBusinessCycles.”JournalofMonetaryEconomics,37,345–,Chengsi.(Forthcoming)“.”,Chengsi,,andDongHeonKim.(2006)“ObservedInflationForecastsandtheNewKeynesianPhillipsCurve.”DiscussionPaperNumber79,CentreforGrowthandBusinessCycleResearch,UniversityofManchester.
