美聯(lián)儲-貝弗里奇曲線告訴我們軟著陸的可能性是什么？What Does the Beveridge Curve Tell Us about the Likelihood of Soft Landings

FinanceandEconomicsDiscussionSeries

FederalReserveBoard,Washington,D.C.

ISSN1936-2854(Print)

ISSN2767-3898(Online)

WhatDoestheBeveridgeCurveTellUsabouttheLikelihoodof

SoftLandings?

AndrewFigura,ChrisWaller

2024-073

Pleasecitethispaperas:

Figura,Andrew,andChrisWaller(2024).“WhatDoestheBeveridgeCurveTellUsabouttheLikelihoodofSoftLandings?,”FinanceandEconomicsDiscussionSe-ries2024-073.Washington:BoardofGovernorsoftheFederalReserveSystem,

/10.17016/FEDS.2024.073

.

NOTE:StafworkingpapersintheFinanceandEconomicsDiscussionSeries(FEDS)arepreliminarymaterialscirculatedtostimulatediscussionandcriticalcomment.TheanalysisandconclusionssetfortharethoseoftheauthorsanddonotindicateconcurrencebyothermembersoftheresearchstafortheBoardofGovernors.ReferencesinpublicationstotheFinanceandEconomicsDiscussionSeries(otherthanacknowledgement)shouldbeclearedwiththeauthor(s)toprotectthetentativecharacterofthesepapers.

WhatDoestheBeveridgeCurveTellUsabouttheLikelihoodofSoftLandings?*

AndrewFiguraandChrisWaller

February1,2024

Abstract

Anyassessmentofthelikelihoodandcharacteristicsofasoftlandinginthelabormarketshouldtakeintoaccountthecurrentstateofthelabormarketandthelikelydynamicsinthelabormarketgoingforward.ModernlabormarketmodelscenteredaroundtheBeveridgecurveareauseful

toolinthisassessment.WeuseasimplemodeloftheBeveridgecurvetoinvestigatewhat

conditionsarenecessaryforasoftlandinginthelabormarkettooccurandwhatthelikelihoodoftheseconditionswasduringtheheightofthepandemic-periodinflation.Wefindthatasoft

landingwasaplausibleoutcomeatthattime.Sincethen,theevolutionofthelabormarkethasborneoutthatprediction.

*ThispaperwaspreparedfortheSt.LouisFed-JEDC-SCG-SNBGerzenseeConferenceonOctober13-142023,

andprovidesaninterpretationoftheU.S.labormarketatthattime.TheanalysisandconclusionssetfortharethoseoftheauthorsanddonotindicateconcurrencebyothermembersoftheresearchstaffortheBoardofGovernors.

WewouldliketothankDavidRatner,CharlesFleischman,theeditorsoftheconferencevolumeoftheJournalof

EconomicDynamicsandControlinwhichthispaperappears,andananonymousrefereeforhelpfulcomments.Allerrorsareourown.

1

Anyassessmentofthelikelihoodandcharacteristicsofasoftlandinginthelabormarketshouldtakeintoaccountthecurrentstateofthelabormarketandthelikelydynamicsinthelabormarketgoingforward.ModernlabormarketmodelscenteredaroundtheBeveridgecurvearea

usefultoolinthisassessment.WeuseasimplemodeloftheBeveridgecurvetoinvestigatewhatconditionsarenecessaryforasoftlandinginthelabormarkettooccurandwhatthelikelihoodoftheseconditionswasduringtheheightofthepandemic-periodinflation.Wefindthatasoft

landingwasaplausibleoutcomeatthattime.Sincethen,theevolutionofthelabormarkethasborneoutthatprediction.

Introduction

TheCovidpandemicgeneratedhistoricallyunprecedenteddisruptionstotheU.S.and

globaleconomies.IntheU.S.,outputshrankatanannualrateofnearly20percentinthefirst

halfof2020.U.S.unemploymentshotuptoalmost15percent,thehighestratesince1940,in

justtwomonths,andthendeclinedrapidly,astheeconomystartedtoquicklyrecoverinresponsetotherapiddevelopmentofvaccines,largefiscalstimulus,andveryaccommodativemonetary

policy.Astheeconomyreopened,jobvacanciessurgedtounprecedentedlevels.Atthesame

time,laborsupplyremainedwellbelowpre-pandemicnorms,asbothimmigrationandlabor

forceparticipationdroppedsharplyduringthepandemicinresponsetoborderclosuresand

Covidfears,resultinginseverelaborshortages.

1

Withdemandwellinexcessofsupplyinmanyareasoftheeconomy,PCEinflation,whichhadbeenmiredbelow2percentformuchofthe

precedingdecadeshotuptonearly7percentinthesecondquarterof2022,whilewageinflation(asmeasuredbytheEmploymentCostIndex)doubledrelativetopre-pandemiclevels.

1SeeMontes,SmithandDajon(2022).

2

Inthisextremelychallengingmacroeconomicenvironment,someobserversarguedthatitwouldnotbepossibletobringinflationbackdownto2percentwithoutcausingarecession.

2

Akeytoolforevaluatingthisargumentandforassessingthepossibleoutcomesforthelabor

marketandthemacroeconomy,moregenerally,wastheBeveridgecurve.

TheBeveridgecurveisnamedafterWilliamBeveridge,whofirstnotedthecorrelation

betweenunemploymentandjobvacanciesthatthecurvedescribes.BlanchardandDiamond

(1989,1990)emphasizedtheutilityoftheBeveridgecurveinunderstandingthevariousshocksaffectingthelabormarketandthemacroeconomyandbroughttheBeveridgecurvetothe

forefrontoflaborandmacroeconomicanalysis.Pissarides(2000)describesthetheoretical

foundationoftheBeveridgecurvethatisthebasisforthesimplemodelwedescribebelow.Themodelhasbecomeaworkhorseofmodernlaborandmacroeconomics.

TheorysuggeststhatmovementsalongthedownwardslopingBeveridgecurveareduetochangesinlabordemandandthatshiftsintheBeveridgecurveareduetochangesintheextentoffrictionsinthelabormarketarising,forexample,fromchangesintheefficiencyofmatchingtheunemployedwithjobvacanciesorchangesinthepaceofreallocationofworkersacrossjobopportunities.

3

Inouranalysis,twoaspectsoftheBeveridgecurvemeritparticularattention.First,asemphasizedbyPetrosky-Nadeau,Zhang,andKuehn(2016),theBeveridgecurveishighlynon-linear.Innormaltimes,whenthelabormarketisnotattheextremesoftheBeveridgecurve,

non-linearitiescantypicallybesafelyignored.Butthefirstpartof2022wasahighlyunusual

2SeeBlanchard,DomashandSummers(2022).

3Shiftscanalsoarisefromchangesinlaborsupplyorotherfactors.SeeElsby,Michaels,andRatner(2015)foracomprehensivediscussionofmovementsinvacanciesandunemployment.

3

economicenvironment,makingitnecessarytoconsiderthenon-linearityoftheBeveridgecurve.Second,achangeinthepaceoflayoffsactstoshifttheBeveridgecurve,and,thus,itis

importanttoconsiderwhethersuchshiftswilloccuraslabordemandchanges.Inpastcyclicaldownturns,layoffshaveincreased,pushingouttheBeveridgecurveatthesametimeasa

vacancyratedeclinemovestheeconomydowntheBeveridgecurve.Asaresult,the

constellationsofobservationsinunemployment-vacancyspaceappearstotraceoutarelativelyflatrelationship.However,inasituationwherelabordemandandvacanciesareextremelyhighandlaborshortagesabound,itispossibleforvacanciestodeclinewithoutlayoffsincreasing,or

attheleast,itisadvisabletoconsiderthispossibility.Inthiscase,theconstellationofvacancy-unemploymentobservationswould,instead,traceouttherelativelysteepunshiftedBeveridge

curve.

ThenextsectiondescribesasimplemodeloftheBeveridgecurveandthesteadystate

unemploymentrate.Thenweexamineconditionsinwhichitispossiblefortheeconomyto

experienceasoftlandingstartingfromeconomicconditionssimilartothoseprevailinginthe

firsthalfof2022andtakingintoaccounthowthispossibilitychangesastheBeveridgecurve

becomesflatterandlayoffsincrease.Next,wedescribetheevolutionofthelabormarketfromearly2022onandcompareittotheconditionsnecessaryforasoftlanding.Thebehaviorofa

varietyoflabormarketvariables—vacancies,layoffs,quits,andwages—haveallbeeninaccordwiththatnecessaryforasoftlanding.Finally,weshowthatthesignificantdropinvacancies

withlittlechangeinunemploymentthathascharacterizedtheU.S.labormarketsincethe

beginningof2022hasalsobeenafeatureofregionalU.S.labormarketsandadvancedforeigneconomylabormarketsthatalsoexperiencedasurgeinvacanciesduringthepandemic.Thus,recenteventsinthenationalU.S.labormarketarenotafluke.Instead,steepnessinthe

4

Beveridgecurveatextremelyhighlevelsofvacanciesseemstobeamoregeneralfeatureofadvanced-economylabormarkets.

SimplemodeloftheBeveridgecurve

WefirstpresentourBeveridgecurveframework,whichisquitestandard.Itisstandardintheliteraturetoaccountforthedynamicsinthelabormarketbyaccountingfortheflowsofindividualsinandoutofunemployment.Consequently,thechangeinunemployment,?U,isgivenby

ΔU=E*S?U*f

Flowsintounemploymentequaltheseparationrate,S,timesthelevelofemployment,E.Flowsoutofunemploymentequaltherateofjobfinding,f,timesthenumberofunemployed.Wenotethatoursimplemodelabstractsfromtheeffectsofflowsintoandoutofthelaborforce,on-the-jobsearch,andjob-to-jobflows.Sinceforsimplicitywenormalizethelaborforcetoequal1,

employmentequals1minusunemployment,U.Insteadystate,flowsintounemploymentmustequalflowsoutofunemployment,therightsideoftheaboveequation.Thus,wecanderivetheBeveridgecurvefromasteady-stateequationforunemployment,showninequation(1).

4

(1?U)*S=U*f(1)

4See,forexample,Pissarides(2000).

5

Rearrangingthisequationyieldsanexpressionforthesteady-stateunemploymentrate,equation(2).

5

(2)

Becauseflowsintoandoutofunemploymentarequitehigh,theactualunemploymentrate

convergestothesteady-stateunemploymentratequickly,andthesteady-stateunemploymentratetypicallytrackstheactualrateclosely.

6

ThejobfindingratecanberelatedthroughamatchingfunctiontotheV-Uratio.A

matchingfunction,showninequation(3),positsthatthenumberofhiresisanincreasing

functionofboththenumberofjobvacanciesandthenumberofunemployedindividuals

searchingforjobs:Themorefirmstherearelookingforworkersandthemoreworkerstherearelookingforjobs,themorematches,orhires,therewillbe.

H=M(V,U)=μVσU1?σ(3)

Asisstandardinmuchoftheliterature,weassumethematchingfunctiontakesaCobb-Douglasform.

7

Therearetwokeyparametersinthematchingfunction:μandσ.Theparameterμ

measuresmatchingefficiency.Matchingefficiencyrepresentsfactorsthatcanincrease(or

5Ifthelaborforceisallowedtovary,thentheexpressionissimilarbutsomewhatmorecomplicated.BarnichonandFigura(2015b)andElsby,Hobijn,Sahin(2015)describetheinfluenceofchangesinlaborforceparticipationon

unemployment.

6Formoreondecomposingunemploymentratemovements,seeShimer(2012),Elsby,MichaelsandSolon(2009),FujitaandRamey(2012),andAhnandCrane(2020).

7Bok,Petrosky-Nadeau,Valletta,andYilma(2022)showthatassumingamatchingfunctionasinDenHaan,Ramey,andWatson(2000)alsogeneratesaverysteepBeveridgecurveasthelabormarkettightens.

6

decrease)hireswithoutachangeinlabormarkettightness.Iftheworkerssearchingforjobsarewellsuitedforthejobsthatareavailable,matchingefficiencywillbehigh;ontheotherhand,ifmanysearchingworkersarenotwellsuitedfortheavailablejobs,matchingefficiencywillbe

low.

8

Theparameterσcapturestherelativeimportanceofvacanciesforcreatinghires.Ifσisrelativelylow,vacanciesarerelativelylessproductiveatcreatingmatchesthanunemployed

workers.

Dividingbothsidesofequation(3)byunemployment,wegetequation(4),which

expressesthejobfindingrateasafunctionoftheratioofvacanciestounemployment,orlabormarkettightness.

(4)

Becausewehavedataforboththeleftandrightsidesofequation(4),wecanestimateitand

obtainparametervaluesfortheelasticityofjobfindingwithrespecttolabormarkettightness,

σ—thekeyparametergoverningcurvatureoftheBeveridgecurve—andmatchingefficiency,μ.

Specifically,usingOLS,weregressl0g(f)onl0gusingJOLTSdataonjobopeningsforV,

BLSdataonunemploymentforU,andBLSdataontherateoftransitionsfromunemploymenttoemploymentfor『.Weusedatafrom2010to2019.

Wenowexplainthereasonsforthesechoicesofdataandsampleperiodandconsider

alternativesampleperiods.Weusedataonunemployment-to-employmenttransitions(u-to-e)asourmeasureofjobfindingbecausewebelieveitisthemostdirectlyrelevantmeasurewhen

8Changesintheintensityoffirmrecruitingorworkersearchcanalsoaffectjobfinding.SeeDavis,Faberman,andHaltiwanger(2012)onthebehaviorandeffectsofrecruitingeffortandMukoyama,Patterson,andSahin(2018)onthebehaviorandeffectsofworkersearch.

7

analyzingchangesinunemployment.First,theunderlyingdatausedtoestimateu-to-e

transitionsarethesameasareusedtocomputetheunemploymentrate,theCurrentPopulationSurvey(CPS).Inaddition,asshowninequation(1),inoursimplemodel,flowsfrom

unemploymenttoemploymentisthetheoreticallycorrectmeasure.Incontrast,hiresfromthe

JOLTSsurveyusedbysomeresearchers,includeflowsfromoutofthelaborforcetoemploymentandjob-to-jobflows,whichdonotdirectlyaffectunemployment.

AspointedoutbyShimer(2012),therawtransitionratespublishedbytheBLSsuffer

fromtimeaggregationbias.Inparticular,someoftheunemployedindividualswhofindajob

betweenmonthtandmontht+1willbelaidoffbeforetheirlaborforcestatusisobservedin

periodt+1.Asaresult,thetruehazardrateformovingfromunemploymenttoemployment

betweentandt+1willbeunderstatedbythepublishedtransitionrates.Inaddition,thedegreeof

biaswillvaryasthelayoffratevariesoverthecycle.Tocorrectfortimeaggregationbias,weusethemethodproposedbyShimer(2012)toestimatethehazardrate.

9

Weusethe2010-2019sampleperiodbecausethereisaone-timelevelshiftinmatchingefficiencyaroundthetimeoftheGreatRecession,asshowninfigure1.Figure1showstheratio

ofthejobfindingratetowhereweuseourpreferredestimateof0.38,whichas

describedbelowisquitestableoverthisperiod.Asshowninequation(4),thisratioisequaltomatchingefficiency.MatchingefficiencydropspermanentlyduringtheGreatRecession.Wecanincludeadummyvariabletocontrolforthisbreak,allowingustoextendthesamplebacktoDecember2000,whenJOLTSdataonjobopeningsfirstbecameavailable.Wecanalsouse

9Elsby,Michaels,andSolon(2009)describesomepotentialproblemswiththecontinuoustimeassumption

underlyingthemethodinShimer(2012)—asdecisionsabouthiringandlayoffsarenotmadeonacontinuous

around-the-clockbasis—andsuggestthatthecontinuous-timemethodmayoverstatesomewhattimeaggregationbias.

8

datafromtheHelpWantedIndextoextendourvacancymeasurebackevenfurther,asdescribedinBarnichon(2010).Asshownintable1,fortheperiods2001-2007and2001-2019(excluding

2008and2009)theestimateofσisidenticaltotheestimatefortheperiod2010-2019.

10

Forthe20yearsbefore2000,theestimateisslightlylarger.Weprefertheestimatefromthemostrecentdataasitusesaconsistentestimateofjobopenings(JOLTSjobopenings).Inaddition,themorerecentsampleperiodlikelyreflectsalabormarketthatismoresimilartothecurrentone.In

sum,ourestimatesofσaresimilaracrossvarioussampleperiodssuggestingthatfortheU.S.labormarketthisparameterisquitestableataroundourpreferredestimateof0.38.Whilewefeeltheseestimatesarereasonablyrobust,wealsoconsiderbelowhowouranalysiswould

changeunderalternativevalues.

11

OtherworkusingtheBeveridgecurvecalibratesσbasedonpreviousestimatesinthe

literature.Butthisparameterislikelynotthesameindifferentlabormarketsorfordifferent

measuresofjobfinding.SinceouranalysisappliesspecificallytotheU.S.labormarketandu-to-etransitions,webelieveitismostappropriatetoestimatethisparameterusingrelatively

recentU.S.data,especiallysincehigh-qualitydataisavailable.

12

Withestimatesofmatching-functionparametersinhand,wecanplugtheexpressionforjobfindingintoequation(2),thesteady-stateunemploymentrate,yieldingequation(5).

10Theresultsintable1arebasedonOLSregressionsofthelogofthehazardrateofjobfindinginperiodtonthelaggedlogoftheV-Uratio.Forthe2000-2007and2010-2019estimates,Newey-Weststandarderrorsareusedtocomputeconfidenceintervals.

11Onereasontothinkofthisestimateasanupperboundisthatthecompositionoftheunemployedvariesoverthecycleinawaythatincreasesthecyclicalityofjobfinding.Ifonecontrolsforthesecompositionchanges,estimatesofσdecline.SeeBarnichonandFigura(2015a)andAhnandHamilton(2020)oncompositionalchangesin

unemploymentoverthecycle.

12Shimer(2012)proposesanothermethodtomeasurethejobfindingratebasedonthenumberofshort-term

unemployed.Butthismethodincludesflowsfromunemployedtooutofthelaborforce,whichcanbehavequite

differentlyfromjobfinding.Asaresult,weprefermeasuresofjobfindingbasedonunemployment-to-employmenttransitions.

9

(5)

Equation(5)showshowvacanciesaffecttheunemploymentrate.Toillustratethisrelationship,wesolveequation(5)fordifferentvaluesofVandS,holdingthematchingefficiencyparameterconstant.Theresultisshowninfigure2,whichplotsfourcurvesshowingtheeffectof

vacanciesonunemploymentforfourdifferentseparationrates.Theseparationsrateinthefar-rightcurveischosentoapproximatevaluesduringtheheightofthepandemicwhenthe

unemploymentratesurgedwellabove10percent.

13

Theseparationsrateforthefar-leftcurveischosentocorrespondtovaluesaroundthebeginningof2022,whentheunemploymentrate

movedunder4percent.Thetworemainingcurvesareequallyspacedbetweenthesetwolabor-marketpoles.

Attheonsetofrecessionsvacanciestypicallyplungeandseparationsincrease.The

consensusintheliteraturethathasinvestigatedtheroleofseparationsandjobfindingindrivingcyclicalfluctuationsinunemploymentisthatseparationscontributetotheinitialsurgeinthe

unemploymentratebutthatthepersistentincreaseinunemploymentduringrecessionsislargelydrivenbyapersistentreductioninjobfinding.

14

Inmildrecessions,suchasin2001,the

recessionaryspikeinseparationsissubduedandfadesfairlyquickly.However,duringdeep

recessions,suchastheGreatRecessionin2008-2009,thespikeinseparationsinmuchmorepronouncedandcanbesomewhatpersistent.Becausethespikeinrecessionsisgenerally

relativelyshort-lived,evenduringtheGreatRecession,changesinthejobfindingrateaccountformostofthevariationintheunemploymentrateoverthebusinesscycle.

13Theactualseparationsratesurgedtoanincredible18percentinAprilof2020,andthenfellto5percentinMay,4percentinJune,3percentinJuly,andaround2percentfortheremainderoftheyear.

14See,forexample,Shimer(2012),FujitaandRamey(2009),andElsby,Michaels,andSolon(2009).

10

Eachofthecurvesinfigure2isconvex;asthenumberofvacanciesincreasesrelativeto

thenumberofindividualslookingforwork,itbecomesharderforfirmstofilljobswithsuitable

workers,andmorejobsremainvacant.Putdifferently,asfirmscompeteforadiminishing

numberofworkers,eachvacancyhasasmallerandsmallereffectonunemploymentand

employment.Thisisexactlythesituationmanyemployersfacedduringtheheightofthelaborsupplyshortagesduringthepandemic.

Buteventhoughadditionalvacanciesdon’tincreaseemploymentbymuch,theycontinuetostronglyaffectwagegrowth.Intuitively,asfirmscompeteforadiminishingsupplyoflabor,wagesgetbidup.StandardNashbargainingassumptionsresultinthefollowingequationfor

steady-statewages,inwhichtheeffectofvacanciesislinearandhencedoesnotdiminishasvacanciesincrease,unliketherelationshipbetweenvacanciesandunemployment.

15

Here,thewage,w,isequaltotheweightedaverageoftheworker’sreservationwage,z,anda

secondtermcapturingtheeffectoftheproductivityofthejob,p,andfrictionsinthelabor

market,c,wherecistheperperiodcostofmaintaininganopenvacancy.Therelativeweightsonthesetwotermsisdeterminedbytheworker’sbargainingpower,β.

Turningagaintofigure2,thecombinationofmovementsinseparationsandvacanciesisshownbytheblackcurveinfigure2,whichisfittoactualvaluesofVandU.Decreasesintheseparationsratereducetheunemploymentratewithoutchangingvacancies,impartingaflatnesstothefittedcurve,relativetothesteepercurvesthatonlyreflecttheeffectofvacancies.Ifwe

wanttojustfocusontheeffectofvacancies,thenweshouldbelookingatthesteepcurves.

15See,forexample,Pissarides(2000).Ofcourse,therearereasonstobelievethatwagesarestickyanddonotadjustimmediatelytothelevelimpliedbythissteady-stateequation.

11

Possibilityofasoftlandinginthefirstquarterof2022

Next,wearguethatwhethertheeconomycanexperienceasoftlanding(whichwedefineasanoticeablysmallerincreaseinunemploymentthanhasoccurredinpreviousrecessions)ifthevacancyratedeclinessignificantlydependsontwoimportantfactors:(1)theslopeofthe

Beveridgecurve,whichdependsimportantlyonthecurrentpositionofthelabormarketalongtheBeveridgecurve,and(2)whetherlayoffsincreasesignificantly.

16

Weconsiderwhatthesetwoimportantfactorssuggestedforthepossibilityofasoftlandinginthehighlyunusual

circumstancesofearly2022.

RegardingtheslopeoftheBeveridgecurve,ahighV-Uratio,suchasexistedinthefirstpartof2022,impliesthatthelabormarketisonaverysteepportionoftheBeveridgecurve.ThestylizedBeveridgecurveinFigure3illustratesthispoint.WhentheV-Uratioisequaltotherayfromtheoriginlabeled“HighV-Uratio”,theslopeoftheconvexBeveridgecurveismuch

steeperthanwhentheV-Uratioisequaltotheraylabeled“AverageV-Uratio”.Asteepercurveimpliesthattheunemploymentratewillchangelessforagivenreductioninvacancies(holdingseparationsandmatchingefficiencyconstant).

WeconsiderhowthesteepnessoftheBeveridgecurveaffectsthechangeinthe

unemploymentratefora2.4percentagepointdeclineinthevacancyrateusingequation(5)

alongwithourpreferredparametersandanassumptionofnochangeinmatchingefficiencyoverthepandemic.A2.4percentagepointdeclineinthevacancyratewouldreturnthevacancyratefromitspeaklevelduringinthefirsthalfof2022ofabout7percenttothepeaklevelpriortothepandemicofabout4.6percent.Wecomputethattheincreaseintheunemploymentrate

16Morespecifically,wewoulddefineasoftlandingasatrough-to-peakincreaseinthequarterlyunemploymentrateoflessthan1.5percentagepoints.Trough-to-peakincreasesintheunemploymentratearoundrecessions(outsideofthepandemic)haverangedfrom1.9percent(1960-1961recession)to5.4percent(GreatRecession).

12

producedbysuchareductioninthevacancyratewhenstartingfromavacancyrateof7percentis0.85percentagepoint,asisshowninthetop-middlecellofTable2a.Thisresultsuggeststhatitispossibleforthevacancyratetonormalizefromitspre-pandemicextremeswhilecausing

onlyarelativelymodestincreaseinunemploymentthatiswellbelowwhatwasexperiencedinpriorrecessions.

Thetop-middlecellofTable2bshowsthechangeintheunemploymentrateforasimilar2.4percentagepointchangeinthevacancyratewhenstartingfromamorenormalvacancyrateof4.6percent.Now,theincreaseintheunemploymentrateisalittleover2percent,aboutthe

increasethathasoccurredinpreviousmildrecessions,suchastherecessionsof1960-1961and2001.Thedifferenceinunemploymentrateschanges,alittleunder1percentversusalittleover2percent,illustratestheimportanceoftakingintoconsiderationtheconvexityoftheBeveridgecurveandwhetherthelabormarketisonasteeporamorenormalpositiononthecurvewhen

consideringthepossibilityofasoftlanding.

Ofcourse,wedon’tknowpreciselywhattheslopeoftheBeveridgecurveis.Inour

modeltheslopedependsontheparameterσ.Whileourestimatessuggestavalueof.38,it’s

possibleσcouldbelarger,ascommoncalibrationsassume,orsmaller,astakingintoaccountcyclicalchangesinthecompositionofunemploymentwouldimply(seefootnote11).Figure4showsthreeBeveridgecurveswiththreedifferencevaluesofσ:ourpreferredvalue,.38,20

percentlowerthanthisestimate,and20percenthigherthanthisestimate.Asshowninthefigure,asσincreases,theBeveridgecurvebecomesflatterandviceversa.Thedifferent

columnsofTables2aand2bconsiderhowthecomputedchangesintheunemploymentratewouldchangeforthesedifferentvaluesofσ.

13

It’salsopossiblethatmatchingefficiencyhasdeclinedduringthepandemic.For

example,Blanchard,Domash,andSummers(2022)estimatethatmatchingefficiencydecreased20percentduringthepandemic.DecreasesinmatchingefficiencywillshiftouttheBeveridgecurvebutalsochangeitsslope.ThedifferentrowsinTables2aand2bshowchangesinthe

unemploymentrateinresponsetoa2.4percentagepointreductioninthevacancyratefor

differentlevelsofmatchingefficiency.Inparticular,weconsiderbotha20percentlowerlevelofmatchingefficiency(relativetothepre-pandemiclevel)anda10percentlowerlevel,whichwefindtobemoreconsistentwithourestimatedmatchingfunction.Inourcalculationsofthechangeintheunemploymentrate,matchingefficiencyisassumedtohavedeclinedpriortothebeginning