通貨膨脹回?fù)?-進(jìn)口競(jìng)爭(zhēng)和勞動(dòng)力市場(chǎng)的作用 Inflation Strikes Back - The Role of Import Competition and the Labor Market

NBERWORKINGPAPERSERIESLATIONSTRIKESBACKTHEROLEOFIMPORTCOMPETITIONANDTHELABORMARKETyAmitiihKarahanAy?egül?ahinWorkingPaper31211http//papers/w31211NATIONALBUREAUOFECONOMICRESEARCHCambridgeMA8May23WearegratefultoValerieRameyfordetailedcommentsandsuggestions.WethankDavidDam,WillSchirmer,AidanToner-Rodgers,AidanWang,andJuliaWuforexcellentresearchassistance.TheviewsexpressedhereinarethoseoftheauthorsanddonotnecessarilyreflecttheviewsoftheNationalBureauofEconomicResearch.NBERworkingpapersarecirculatedfordiscussionandcommentpurposes.Theyhavenotbeenpeer-reviewedorbeensubjecttothereviewbytheNBERBoardofDirectorsthataccompaniesofficialNBERpublications.?2023byMaryAmiti,SebastianHeise,FatihKarahan,andAy?egül?ahin.Allrightsreserved.Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicitpermissionprovidedthatfullcredit,including?notice,isgiventothesource.InflationStrikesBackTheRoleofImportCompetitionandtheLaborMarketMaryAmiti,SebastianHeise,FatihKarahan,andAy?egül?ahinNBERWorkingPaperNo.31211May23JELNoE,E31STRACTU.S.inflationhasrecentlysurged,withinflationreachingitshighestreadingssincetheearly1980s.Weexaminethedriversofthisriseininflation,focusingonsupplychaindisruptions,laborsupplyconstraints,andtheirinteraction.Usingacalibratedtwo-sectorNewKeynesianDSGEmodelwithmultiplefactorsofproduction,foreigncompetition,andendogenousmarkups,wefindthatsupplychaindisruptionscombinedwithariseinthedisutilityofworkraisedinflationbyabout2percentagepointsinthe2021-22period.Weshowthatthecombinedshockincreasedpriceinflationinthemodelby0.6percentagepointmorethanitwouldhaveriseniftheshockshadhitseparately.Thisamplificationarisesbecausethejointshocktolaborandimportedinputpricesmakessubstitutingbetweenlaborandintermediateslesseffectivefordomesticfirms.Moreover,thesimultaneousforeigncompetitionshockallowsdomesticproducerstoincreasetheirpass-throughintopriceswithoutlosingmarketshare.Wethenshowthatthebenefitofaggressivemonetarypolicyinthemodeldependsonthesourceoftheriseininflation.Iftheriseininflationisdemand-driven,thenaggressivemonetarytighteningcancontaininflationwithoutarecessionlater.Incontrast,aggressivepolicycanhavealargenegativeeffectonthelabormarketwheninflationisdrivenbysupplychainandlabormarketdisruptions.Weuseaggregateandindustry-leveldataonproducerprices,wages,andinputpricestoprovidecorroboratingevidenceforthekeyamplificationchannelsinthemodel.MaryAmitiFatihKarahanResearchDepartmentAFederalReserveBankofNewYork296HarlingenRd33LibertyStBelleMead,NJ08502NewYork,NY10045yfkarahan@andCEPRmary.amiti@Ay?egül?ahinDepartmentofEconomicsSebastianHeiseUniversityofTexasatAustin2225FederalReserveBankofNewYorkSpeedwayResearchandStatisticsGroupAustin,TX7871233LibertyStreetandNBERNewYork,NY10045aysegul.sahin@sebastian.heise@11IntroductionChairPowell—PressConferenceonJune15,2022tedheytdnsse1Foranecdotalevidence,seeNYTimesDailyBusinessBriefing:SupplyChainSnagsContinuedtoDriveUpPricesinDecember,January12,2022andNYTimes:CouldWagesandPricesSpiralUpwardinAmerica?February17,2022.2Forexample,Amitietal.(2021)andLaBelleandSantacreu(2022)documentthathigherimportpriceswereassociatedwithhigherproducerpricesforU.S.outputintherecentperiod.3SeeforexampleCrumpetal.(2022)andFabermanetal.(2022).ReservationwageincreasesaredocumentedintheSurveyofConsumerExpectations(SCE)oftheNYFed:/microeconomics/sce/labor#/expectations-job-search16.2Figure1:CPIInflation(a)TimeSeriesofCPIInflation(b)CoreCPIPriceGrowthinExpansionsCPIInflation(%)CumulativeCumulativeCoreCPIGrowth(%)401975Q2?401975Q2?1979Q21992Q3?2000Q42009Q42019Q31982Q4?1989Q12003Q2?2006Q42020Q22022Q4302005101520QuarterssincePeakUnemploymentRateSource:BLSandauthors’calculations.Note:TheleftfigureplotstheannualrateofCPIinflation.TherightfigureplotsthecumulativecoreCPIpricegrowth(Allitemslessfoodandenergy,seasonallyadjusted)againsttime,startingatthequarterofpeakunemploymentofagivenrecession,forthepastsixexpansions.sndks3seetvannietalDeSoyresetal4yorlRelatedLiterature.Ourpaperiscloselyrelatedtorecentworkondriversofinflationsconodand5oftheriseininflationbothintheUKandtheUS.Crumpetal.(2022)re-examinethealsoc2WhyisThisTimeDifferent?ind6ion(a)CoreGoods(b)CoreServicesCumulativeCumulativePriceGrowth(%)403020005101520Quarters051015201975Q2?1979Q21992Q3?2000Q42009Q4?2019Q31982Q41975Q2?1979Q21992Q3?2000Q42009Q4?2019Q3CumulativeCumulativePriceGrowth(%)403020005101520QuarterssincePeakUnemploymentRate1975Q2?1979Q21992Q3?2000Q42009Q4?2019Q31982Q41975Q2?1979Q21992Q3?2000Q42009Q4?2019Q3Source:BLSandauthors’calculations.Note:TheleftpanelplotsthecumulativecoregoodsCPIpricegrowth(Allitemslessfoodandenergy,seasonallyadjusted)againsttime,startingatthequarterofpeakunemploymentofagivenexpansion.Therightpanelplotsthecumulativecoreservicespricegrowth(Allitemslessenergyservices)againsttime.tRe-emergenceofGoodsInflation.TheleftpanelofFigure2showsthecumulativeeConsumptionshareofgoodsincreasedrelativetoservices.Adefiningfeatureofond7454035300m10m10m12000m12010m12020m1Source:BEAandauthors’calculations.Note:Thisfigureplotsthemonthlyshareofrealpersonalconsumptionexpendituresattributedtogoods.sLaborsupplyandwillingnesstoworkdeclined.TheCOVIDpandemicwasamajor8Average4-QuarterChange(1)WageGrowth(ECI)(2)Importprices(excl.petroleum)?0.1%(3)-Industrialsuppliesexcl.petroleum(4)IntermediatePPI(excl.foodandenergy)(5)CoreCPISource:BLSandauthors’calculations.Notes:Tableshowstheaverage4-quarterchangeinthelistedvariableforeachofthepastfoureconomicexpansionsshownintheheader.WagesandInputPricesHaveRisenSharply.Twoimportantfactorsthatareoftenfthegns9an3ANewKeynesianDSGEModelwithTwoSectorsandImportCompetition3.1TheHouseholdSector∞E0工BU,t=0sln(βt+1)=(1?ωβ)ln(β)+ωβln(βt)+?,Uτ(C,?,?)=(C?Ht)1?σ?(?)1+φ?(?)1+φ.κ+1=(1?ωκ)κs+ωκκ+?(1)CPf,t+btB+Qt+1A+1≤W?+W?+B?1+A+Pf,tΠ,ureConsumptionandSavingsBehavior.Thesolutiontothehouseholdconsumption-(Ct?hCt?1)?σ=βtEt[(Ct+1?hCt)?σ],plyDecisionsandWageSettingHouseholdsarewagesettersinthelaborl=()?ηsL,y sL=[\01(?)η1dτ]ηs?1Wts=(\01(W)1?ηsdτ)1?ηs(ηs?1)(Ct?hCt?1)?σw=κηs(L)φ?ψwπ(1+π)+Etβtψwπ(1+π),(4)ion3.2FinalOutputFirmYf,t=(Y)γ(Y)γ,(5)ln(γ)=(1?ωγ)ln(γM)+ωγln(γ)+?whereMisthesteadystateshareofexpendituresinthegoodssectorand?isanof θYt=(\01y,t(j)dj)θ?1.(6)y,t(j)=γ()?θ()Yf,t,(7)whereP,t=(′01P,t(j)1?θdj)1/(1?θ)isthesectoralpriceindex.TheconsumerpriceindexPf,t=()γ ()γ(P)γ(Pt)γ.ry,ty,t(j)=(\01x(j,i)di)v?1, v verevistheelasticityofsubstitutionbetweenindustriesandxjiisthequantityofingibyPx,t(j,i).Givendemand(7)andsolvingforasymmetricequilibriumwithj=j\andpEtwherep,t≡P,t/Pf,t,p≡P,t/Pf,t,sectoralinflationisπ=P,t/P,t?1?1,andwehave3.4IntermediateGoodsFirmssx(j,i)=(Ns)(k=1x(j,i,(k=1FNsF+之x(j,i,k)k=1) μμ?1x(j,i)=x(j).AsinAtkesonandBurstein(2008),weassumethatμ>ν>1sothatitisDemandx(j,i,k)=(Pts北)?μ,(12)andPs北,t(j,i)=(Ns)1(k=1k=1Ps北,t(j,i,k)1?μ+Ps北,t(j,i,(k=1k=1Production ρsx(j,i,k)=[(AtL(j,i,k))ρ1+ΛD(j,i,k)ρ1]ρs?1,ywhichiscommonacrosssectorsandLjikandD(j,i,k)arelaborandintermediateinputsusedbythefirm.Theparameterρsisofh ξD(j,i,k)=[Mts(j,i,k)+Z(j,i,k)]ξ?1tsln(p,imp,t+1)=(1?ωP)ln(p,imp)+ωPln(p,imp,t)+?,Px,dom,t=()γ()γ(P)γ(Pt)γ,MC,t=[(Wts/At)1?ρs+Λs(P,input,t)1?ρsJ1/(1?ρs),P,input,t=[(Px,dom,t)1?ξ+(P,imp,t)1?ξJ1/(1?ξ).costs,mc,t≡MC,t/Pf,t,givenbyln(mc,t+1)=(1?ωF)ln(mc)+ωFln(mc,t)+?,Px,t(j,i,k)=MC,t,whereE(j,i,k)=μ(1?S(j,i,k))+νS(j,i,k)istheeffectiveelasticityofsubstitutionS(j,i,k)=().Equation(21)highlightsthatfirmssetavariablemarkupMt(j,i,k)≡E(j,i,k)/(E(j,i,k)?3.5MonetaryAuthorityrRt=?Rt?1+(1??)R+(1??)[Φππt+Φy(ln(Yf,t)?ln(Yf))]+?,(23)3.6AggregationYt=()?μ(γ()Ct+γ(π)2Ct+γ()Z+γ()Z\).Z=Λs()?ξ()?ρsYt.Yg,t=Yg+Yg.L=As?1()?ρsYt,Note:Thisfiguresummarizeskeycomponentsofthemodel.Householdsarerepresentedbygraycells,thecentralbankinorange,whilethegreenandbluecellsshowthemanufacturingandservicessector,respectively.Thedarkblueboxhighlightsforeigncompetitors.A4QuantitativeAnalysisined4.1CalibrationStandardparametersbasedontheDSGEliterature.Wesetstandardvaluesfornσ=1.371,φ=2.491,β=0.99,ηM=ηS=3,andh=0.595.gAssumingasteadystatemarkupoftheyfindavalueofψp=72.Atthatvalue,aLaborshareingoodsandservicessectors.Ourmodelimpliesthatthelaborshareλs≡(24) (ws/A)1λs≡(24)(ws/A)1?ρs+Λs(p,input)1?ρs,tedParameterDescriptionValueSourceσRiskaversionSmetsandWouters(2003)φInverseelasticityoflaborsupply2.491SmetsSmetsandWouters(2003)βSteadystatediscountfactor0.99SmetsSmetsandWouters(2003)Wagemarkup3SmetsSmetsandWouters(2003)hHabitparameter0.595SmetsSmetsandWouters(2003)ψpAdjustmentcostspricesKeenKeenandWang(2007)ψwAdjustmentcostswagesAssumedsameasprices?Taylorrulepersistence0.8Carvalhoetal.(2021)ΦπTaylorruleweightoninflationCarvalhoCarvalhoetal.(2021)ΦyTaylorruleweightonoutput0.95CarvalhoCarvalhoetal.(2021)λM,λSLaborsharegoods(services)0.31(0.6)CensusBureau,authors’calculationsγM,γSConsumptionsharegoods(services)0.35(0.65)BureauofEconomicAnalysisκM,κSSteadystatedisutilityoflaboringoods(services)122(1)CensusBureau,authors’calculationsD,DND,DDomesticfirmsgoods(services)13(20)AtkesonandBurstein(2008)FNMFForeignfirmsgoods7CensusBureau,authors’calculationsαM,αSImportedinputshareingoods(services)0.17(0.04)CensusBureau,authors’calculationsθElasticityfinalgoods6Christianoetal.(2005)μElasticityacrossfirms3AtkesonandBurstein(2008)νElasticityacrossindustries1AtkesonandBurstein(2008)Eermediates2(1.5)Chan(2021)ξElasticitydomesticv.foreign2Feenstraetal.(2018)Consumptionshareingoodsandservicessectors.ThesteadystateconsumptionγM=0.35andγS=0.65.offirmsbasedonWethereforesetN=13,N=7,andN=20.4Wedefinethemanufacturingandminingsectorsasthegoodssector,andsetutilities,construction,wholesaleandretailtrade,transportation,andallotherservicesasservices.Relativeimportprice.Tocalibratethesteadystaterelativeimportprice,p,imp,we(px,dom)1?ξ+(px,dom)1?ξ+(p,imp)1?ξ,αs≡ytalhe4.2MacroeconomicsEffectsofSupplyChainDisruptions,RiseinDisutilityofWorkandGoods-FavoredDemandShockspresentingworkersinpanGiventheel4.2.1SupplyChainDisruptionsandLaborSupply.esdsFigureshowstheeffectofajointdisutility,importedinputprice,andforeigncompetitorcostshock.Eachpanelshowsthepercentdeviationofavariablefromitssteadystatevalueagainstthenumberofquarterspassedsincetheinitialshock.Forwageandpriceinflationandthemarketshare,weshowpercentagepointchangesfromsteadystate.Fortheconsumptionshareofgoodsweshowitsactualvalue.noSubstitutioneffect.Figure6tracesouttheeffectoftheimportedinputpriceshockheeeFigureshowstheeffectofaninputpriceshockonkeyvariables.Eachpanelshowsthepercentdeviationofavariablefromitssteadystatevalueagainstthenumberofquarterspassedsincetheinitialshock.Forwageandpriceinflationandthemarketshare,weshowpercentagepointchangesfromsteadystate.Fortheconsumptionshareofgoodsweshowitsactualvalue.Figureshowstheeffectofashocktoforeigncompetitors’realmarginalcostonkeyvariables.Eachpanelshowsthepercentdeviationofavariablefromitssteadystatevalueagainstthenumberofquarterspassedsincetheinitialshock.Forwageandpriceinflationandthemarketshare,weshowpercentagepointchangesfromsteadystate.Fortheconsumptionshareofgoodsweshowitsactualvalue.cethecFigureshowstheamplificationeffectofajointshockcomparedtothethreesupplyshocksseparately.Theleftpanelshowsimpulseresponsesofaveragewageinflation,constructedbyaveragingacrossthetwosectors’wagegrowthusingeachsector’sshareinconsumption.Therightpanelshowsimpulseresponsesofconsumerpriceinflation.Thedashedlinestraceouttheimpulseresponsestoaseparateimportpriceshock,labordisutilityshock,andcompetitionshock.Theblacklinesumsovertheseimpulseresponses.Theredlineshowsthejointeffectofallthreeshockssimultaneously.panelofd(a)SensitivitytoρM,ρS(b)SensitivitytoμFigureplotsthedifferencebetweentheimpulseresponsesofconsumerpriceinflationandaveragewageinflationforthejointshockrelativetotheimpulseresponseofthesummedseparateshocksonimpact(thedifferencebetweenthereddashedandtheblacksolidlinefromFigure8inquarterone).TheleftpanelshowsthisamplificationasafunctionofthevalueofρMandρS.Therightpanelshowstheamplificationasafunctionofμ.4.2.2Goods-FavoredDemandShocken4.3AggressiveMonetaryPolicywithSupplyandDemandShocksOurbaselinecalibrationspecifiestheTaylorrulewith8,aweightoninflationofothsupplyanddemandns5Seeforexample,/2021/11/16/on-inflation-its-past-time-for-team-transitory-to-stand-down/Source:Author’scalculations.Figureshowstheeffectofagoods-biaseddemandshockonkeyvariables.Eachpanelshowsthepercentdeviationofavariablefromitssteadystatevalueagainstthenumberofquarterspassedsincetheinitialshock.Forwageandpriceinflationandthemarketshare,weshowpercentagepointchangesfromsteadystate.Fortheconsumptionshareofgoodsweshowitsactualvalue.esupplyoNote:Figureshowstheeffectofmonetarypolicyonoutcomesinresponsetoasupplyshock.Note:Figureshowstheeffectofmonetarypolicyonoutcomesinresponsetoademandshock.5EmpiricalEvidence:AggregateAnalysisised?0.4048121620QuartersEndyearofthe25?yearperiod?0.4048121620QuartersEndyearofthe25?yearperiod320(a)ResponseofFinishedPPItoWages048121620s(b)ResponseofFinishedPPItoInt.Prices0.0?0.2(c)ResponseofPPItoWagesat9Quarters4.03.02.000.0?1.0199019951990199520002005201020152020Endyearofthe25?yearperiod(d)ResponseofPPItoInt.Pricesat5Quarters19901995199019952000200520102015202040.20.0Source:BLSandauthors’calculations.Note:Thetoptwopanelspresenttheestimatedcoefficientsβhandγhandtheir90percentconfidenceintervalsfromspecification(26)runatquarterlyfrequency,forhorizonsh=0,...,20quarters.Inthetopleftpanel,wepresentthecoefficientsβhshowingpass-throughfromwageinflationandproducerpriceinflation.Wagesareaveragehourlyearningsofproductionandnon-supervisoryemployees.Producerpricesaretheproducerpriceindex(PPI)offinishedgoodslessfoodandenergy.Allvariablesaretransformedintoaquarterlyseriesbytakingasimpleaverageacrossthemonthsineachquarter.Inthetoprightpanel,weshowthecoefficientsγhofthepass-throughfromintermediatepricesintoproducerprices.IntermediatepricesarethePPIofintermediategoodslessfoodandenergyinputs.Thebottomtwopanelsestimatespecification(26)over25-yearrollingwindowsforh=9forwagesandh=5forintermediateprices,wheretheendingyearofthe25-yearperiodisindicatedonthex-axis.888π=α+βhπ+γhπ+工δjπ+工ζjπ+工ξjπ+?t,(26)j=1j=1j=?.0?0.1048048QuartersFigureshowstheestimatedcoefficientontheinteractiontermππandits90percentconfidenceintervalfromanaugmentedversionofspecification(26)whichincludesinteractionsrunatquarterlyfrequency,forhorizonsh=0,...,20quarters.6SincetheBLSstartedcollectingservicespricesonlyin2004,wedonothaveacomprehensiveseriescoveringbothgoodsandservices.6EmpiricalEvidence:Industry-levelAnalysisl6.1LinkingtheTheorytoDatamarginalcosttimesavariablemarkupMt(j,i,k)=Et(j,i,k)/(Et(j,i,k)?1).Takinglogsdln(Px,t(j,i,k))=dln(MC,t)+dln(Mt(j,i,k)),(27)dln(MC,t)=λs[dlnWts?dlnAt]+(1?λs)dlnP,input,t,dln(P,input,t)=(1?αs)dln(Px,dom,t)+αsdln(P,imp,t),(28)dln(Mt(j,i,k))=?Γt(j,i,k)[dlnPx,t(j,i,k)?dlnPx,t(j,i)],whereΓt(j,i,k)=?(?logMt(i,j,k)/?logP北,t(j,i,k))≥0istheelasticityofthemarkupdln(P北,t(j,i,k))=[dlnWts?dlnAt]+dlnPs北,input,t+dlnP北,t(j,i)+dlnP北,t(j,i).(29)1+Γt(j,i,k)sP北,t(j,i)allowsthefirmtoraiseitspricesitselfbyincreasingitsmarkup.TherelativeelasticityΓt(j,i,k).FirmswithahighermarkupelasticityputahigherweightonthendsatisfiesΓt(j,i,k)=0ifSt(j,i,k)=0.7SeeAppendixA.7forthederivationsinthissection.dln(Px,t(j,i,k))=[dlnWts?dlnAt]+dlnP,input,t+{(dlnWts?dlnAt)dlnP,input,t(30)??}+dlnPx,t(j,i).