IMF-投票權(quán)輪換、委員會成員行為與金融市場反應(yīng)：來自美國聯(lián)邦公開市場委員會的證據(jù)（英）-2022.5

VotingRightRotation,Behavior

ofCommitteeMembersand

FinancialMarketReactions:

EvidencefromtheU.S.Federal

OpenMarketCommittee

MichaelEhrmann,RobinTietzandBaukeVisser

WP/22/105

IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)

andarepublishedtoelicitcommentsandtoencouragedebate.

TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)

anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutive

Board,orIMFmanagement.

Apreliminaryversionofapartofthispaperhasbeencirculatedunderthe

title“Strategicinteractionsinpreparingforcommitteemeetings”.It

presentstheauthors’personalopinionsanddoesnotnecessarilyreflect

theviewsoftheECB,theEurosystemortheIMF.

2022

MAY

?2022InternationalMonetaryFund

WP/22/105

IMFWorkingPaper

ResearchDepartment

VotingRightRotation,BehaviorofCommitteeMembersandFinancialMarketReactions:Evidencefrom

theU.S.FederalOpenMarketCommittee

PreparedbyMichaelEhrmann,RobinTietzandBaukeVisser*

AuthorizedfordistributionbyMariaSoledadMartinezPeria

May2022

IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.

ABSTRACT:WhichFederalReserveBankpresidentsvoteontheU.S.monetarypolicycommitteedependsonamechanical,yearlyrotationscheme.Rotationiswithoutexclusion:nonvotingpresidentsdoattendandparticipateinthemeetingsofthecommittee.Wetesttwohypothesesaboutthedependenceofpresidents'behavioronvotingstatus.(i)Losscompensation:presidentscompensatethelossoftherighttovotewithanincreaseduseofspeechesandcontributions.(ii)Motivation:presidentscomplementtherighttovotewithanincreaseduseofspeechesandcontributions.Theevidencefavorsthemotivationhypothesis.Also,inyearsthatpresidentsvote,theirspeechesmovefinancialmarketslessthaninyearstheydonotvote.Wearguethatthisvotediscountisconsistentwithpresidents’communicationbehavior.

JELClassificationNumbers:D71,D72,E58

Keywords:

votingrightrotation,monetarypolicycommittee,centralbankcommunication,FOMC,financialmarketresponse

Author’sE-MailAddress:michael.ehrmann@,rtietz@,bvisser@ese.eur.nl

*WewouldliketothankPhilippeAndrade,HansGersbach,KlodianaIstrefi,MoritzJanas,RobinLumsdaine,MichaelMcMahon,EllenMeade,MikePowell,AdriaanSoetevent,AnnetteVissing-Jorgensen,andparticipantsattheEEAAnnualCongress2019,the2019DeutscheBundesbankMacroWorkshop,the2021InternationalAssociationforAppliedEconometricsAnnualConference,the2021BehaviouralFinanceWorkingGroupConference,the2021QatarCentreforGlobalBankingandFinanceConferenceandatseminarsattheECBandErasmusUniversityRotterdamforhelpfulcommentsandconversations.

2EHRMANN,TIETZANDVISSER

1.INTRODUCTION

Votingrightrotationisawaytosharepowerandexpeditedecisionmakingwhencommitteesbecomelarge.Itimpliesthatamemberexperiencesperiodswithandwithouttherighttovote.Itisused,amongothers,attheSecurityCounciloftheUnitedNations,theGeneralAssemblyoftheWorldHealthOrganization,andatthemonetarypolicycommitteesofthemostprominentcentralbanks,theFederalOpenMarketCommittee(FOMC)oftheU.S.FederalReserveSystemandtheGoverningCounciloftheEuropeanCentralBank.

WestudythebehavioralconsequencesofvotingrightrotationattheFOMC,

1

andthesurprisingreactionoffinancialmarkets.ThepermanentmembersoftheFOMCarethe7membersoftheBoardofGovernorsoftheFederalReserveandthepresidentoftheFederalReserveBankofNewYork.Thepresidentsoftheremaining11regionalReserveBanksbelongto1of4groups.Eachgrouphas1seatontheFOMC.Withineachgroup,theReserveBankpresidentstaketurnsinasimplesuccessionthathasremainedunchangedsince1943.Rotationthuscreatesexogenousvariationinvotingstatus.ReserveBankpresidentsareappointedforfive-yearperiodsandcanberenewed.Asaresult,theytypicallyexperienceyearswithandwithouttherighttovote.

2

AttheFOMC,rotationiswithoutexclusion:nonvotingpresidentscustomarily“attendthemeet-ingsoftheCommittee,participateinthediscussions,andcontributetotheCommittee’sassessmentoftheeconomyandpolicyoptions.”

3

Thatis,theonlydifferencewithvotingpresidents(FOMCmembers)isthatnonvotingpresidents(FOMCparticipants)donotvote.

Weuse20yearsofmeetingtranscriptsandtextsofspeeches,rangingfrom1994to2013,toanswerthreequestions.Howdoesthevotingstatusofpresidentsinfluencethespeechesthattheydeliverbetweenmeetingsandthecontributionsthattheymakeduringmeetings?Howdoesthereactionoffinancialmarketstoaspeechdependonapresident’svotingstatus?Andistheobservedvotediscount,theweakermarketreactiontoaspeechinyearswiththerighttovote,consistentwiththeobserveddifferenceinpresidents’behavioracrossyearswithandwithouttherighttovote?

Weconsidertwohypothesesthatexplainwhyvotingstatuschangesbehavior.Theypredictop-positebehavioralresponses.First,votingstatusmightchangebehaviorbecauseinyearswithouttherighttovote,apresidentseekstocompensateforthelossofformalpowerbystrategicallyintensi-fyingtheargumentativeuseofintermeetingspeechesandmeetingcontributions.Accordingtothisloss-compensationhypothesis,votingontheonehand,andspeechesandcontributionsontheotherhandaresubstitutestoshapethefinalcommitteedecision.Therivalhypothesismaintainsthatvot-ingstatuschangesbehaviorbecauseinyearswiththerighttovote,presidentsaremoremotivatedtopreparethemselvesforthemeeting,astheireffortsaremorelikelytopayoffthankstotheirformal

1TheFOMCdecidesonthetargetforthefederalfundsrateandauthorizesopenmarketoperations–thebuyingandsellingofU.S.governmentsecurities–bytheFederalReservetoreachthattarget.ThefederalfundsrateistheinterestrateatwhichcommercialbankslendbalancesattheFederalReservetoothercommercialbanksovernight.Itaffectsawiderangeofotherinterestrates(

Drechsler,SavovandSchnabl

,

2017

).

2InAppendix

C

,weexplainhowtheFOMCemerged.

3ThequoteistakenfromtheFOMCwebsite,

/monetarypolicy/fomc.htm

,accessedSep-

tember29,2020.

VOTINGRIGHTROTATION3

right.Accordingtothemotivationhypothesis,votingontheonehandandspeechesandcontribu-tionsontheotherhandarecomplements.Thiscomplementarityleadsthemtostrategicallyintensifytheuseofspeechesandcontributionsduringyearswiththerighttovote.

4

Weproxythisuseofspeechesandcontributionsbythetoneandnumberofintermeetingspeechesandthetoneandlengthofmeetingcontributions.

Asweexplaininsection

2

,aReserveBankpresidentisexpectedtobringintelligenceabouttheregionaleconomicconditionstothediscussiononmonetarypolicyattheFOMC.Moreover,thepresidentofaReserveBankisthechiefexecutiveofficeroftheBankandisaccountabletotheBank’sboardofdirectors,whichhasastrongregionalbase.Insection

2

wealsoshowthatstudentsoftheFOMCfindevidencethatbesidesnationaleconomicvariablesalsoregionalvariables—inparticular,district-levelunemployment—affectspeechesandevenpolicypreferencesstatedinthemeetingandactualvotescast.

Againstthisbackground,presidentsaresaidtostrategicallyintensifytheirargumentativeuseofintermeetingspeechesandmeetingcontributionsifspeechesandcontributionsbecomemorere-sponsivetothedifferencebetweendistrict-levelunemploymentandU.S.-widelevelofunemploy-ment.Theloss-compensationhypothesispredictsthatthishappensinyearswithouttherighttovote,themotivationhypothesisinyearswiththerighttovote.

Wefindthat,byandlarge,patternsinthedatasupportthemotivationhypothesisandrejecttheloss-compensationhypothesis,bothwhenwestudytheintermeetingspeechesandthemeetingcontributions.

Votingstatusmatterstofinancialmarkets.Wemeasurethemarketreactiontopresidents’speechesbytheabsolutedailychangeinconstantmaturityTreasuryyields,withmaturitiesvaryingfromthreemonthstofiveyears.Wefinda‘votediscount:’themarketrespondssystematicallylesstospeechesinyearspresidentsvotethaninyearstheydonotvote.Theestimatedcoefficientislargecomparedtotheaveragemarketreaction.Thisfindingisrobustandbecomesstrongerifweextendthesampleperiodofspeechesto2018.

Thispatternmightseemsurprising–afterall,formallyapresidentismoreinfluentialinyearswiththerighttovotethanwithout.Wearguethatthisvotediscountisconsistentwiththedifferenceinspeechbehaviorduetovotingstatus.Essentially,asspeechesgivenbypresidentsinyearswiththerighttovotearemoreaffectedbyregionalconditionsthatnationalmarketscarelessabout,marketssuchastheU.S.Treasurymarket,reactlesstothosespeeches.Infurthersupportofthisargument,wefindthatthevotediscountisespeciallylargeforspeechesgivenbeforethepublicationoftheBeigeBookwithdistrict-levelinformationorafterameetingthatendedwithdissentingvotes;thesearepreciselythespeechesthatrespondmoststronglyintoneandnumbertotheeconomicsituationatthedistrict-levelwhenapresidenthasvotingstatus.

4WeformalizethesehypothesesinamodelinAppendix

A

.

4EHRMANN,TIETZANDVISSER

Theremainderofthispaperisorganizedasfollows.Section

2

developsourhypothesesonthebehavioraleffectsofvotingstatus.Section

3

presentstheeconometricmodelsthatwetest.Section

4

discussesourdataandsection

5

presentstheempiricalanalysis.Section

6

discussestherelatedliterature.Section

7

concludes.Appendix

A

providesasimplemodelthatformalizesthehypotheses.Appendix

B

containsadditionaltablesandAppendix

C

providesinformationabouthowtheFOMCcameabout.

2.TWOHYPOTHESESABOUTHOWTHERIGHTTOVOTEAFFECTSPRESIDENTS’BEHAVIOR

AttheFOMC,nonvotingpresidentsparticipateinthemeeting,alongwithvotingpresidents.Tostudythepossiblebehavioralconsequencesofvotingstatus,onemustthusacknowledgethepossi-bilitythat,evenaftergroupdeliberation,presidentsmaycontinuetohavedifferentviewsaboutthebetterdecision.Whyelsewouldarighttovotematter?

5

Weinvestigatewhetherapresident’svotingstatuschangestheeffectthateconomicvariablesaboutherdistricthaveonherspeechesbetweenmeetingsandcontributionsduringthedelibera-tionstageofmeetings.ThatBankpresidentsputparticularemphasisonregionalconditionswhenpreparingforthemeetingandduringthemeetingis,weargueinsection

2.1

,bydesignoftheFed-eralReserveSystem.Districtlevelvariablesmayaffectbehaviorforatleasttworeasons.Theycandirectlyenterapresident’spreferencesasshemaycarenotjustwhetherthedecisionmatchestheoverallU.S.economy,butalsoherdistrict’seconomy.Anddistrict-levelvariablesmaybepartoftheinformationsetsheusestoassestheU.S.economy.Weareagnosticabouttherelativeimportanceofthesereasons.Insection

2.2

,wereviewtheempiricalliteratureontheinfluenceofregionaleco-nomicvariablesonpresidents’behaviorandevenpolicypreferences.Weformulatethehypothesesinsection

2.3

andtheeconometricmodelsthatweestimateinsection

3

.

2.1.TheregionaldimensionoftheFederalReserveSystem.TheFederalReserveSystemconsistsofaBoardofGovernors,locatedinWashington,D.C.,and12FederalReserveBanks,oneforeachFederalReservedistrict.TheFOMC’smandateis“topromoteeffectivelythegoalsofmaximumem-ployment,stableprices,andmoderatelong-terminterestrates”intheU.S.economy.

6

AstheU.S.A.isalargecountry,eachFederalReserveBanksystematicallycollectsinformationaboutitsdistrict.Partofthisinformationisobtained“throughreportsfrom[FederalReserve]BankandBranchdirec-tors,plusphoneandin-personinterviewswithandonlinequestionnairescompletedbybusinesses,communitycontacts,economists,marketexperts,andothersources.”

7

VariousprocessesareinplacetobringthisinformationtobearonFOMCdecisionmaking.IntelligencebasedontheanecdotalinformationispresentedinTheBeigeBook.SummaryofCommentaryonCurrentEconomicConditions

5

Coughlan

(

2000

)showsthatifmembersofacommitteereceiveprivatesignalsaboutastatevariableandcandeliberateabouttheirprivatesignalsusingcheaptalkbeforetheycastavote,thenanequilibriumexistsinwhichtheytruthfullyrevealtheirsignalsandcastthesamevoteiftheirexantepreferencesarehomogeneousornottooheterogeneous.Inthisequilibrium,thevotingruleisimmaterial.Asaresult,havingornothavingtherighttovoteisalsoimmaterial,aslongasnomemberisexcludedfromdeliberation.

6FederalReserveAct,section2A.Monetarypolicyobjectives.

7ThequoteappearsinthecolophonoftheBeigeBook.

VOTINGRIGHTROTATION5

byFederalReserveDistrict,areportthatBoardgovernorsandReserveBankpresidentsreceiveabouttwoweeksbeforeanFOMCmeeting.RegionalinformationabouteconomicandcreditconditionsisalsoregularlyreviewedduringanFOMCmeeting.Intheeconomygo-round,thefirstpartofameeting,Bankpresidentsdiscussandcommentonregionalconditions.

ThegovernanceoftheFederalReserveSystemalsohasastrongregionalcomponent.Sinceitsinceptionin1913,presidentsofReserveBankshavebeenchiefexecutiveofficersoftheirBanks.Eachbankhasitsownboardofdirectors.Directorshavestrongtieswithregionalfinancialindustry,businessesandthecommunityingeneral.Presidentsareaccountabletotheseboards.

2.2.RegionaleconomicconditionsandbehaviorofBankpresidents.BoththeregionalnatureofgovernanceandtheroleofregionalinformationatFOMCmeetingssuggestthatpresidentsiden-tifywiththeirdistricts.ManystudentsoftheFOMChavefoundevidencethatregionaleconomicvariablesaffectpresidents’behaviorandpolicypreferences.

Sheng

(

2015

)providesevidencethatregionalconditionsaffecttheforecastsofrealGDP,inflationandunemployment.

ChappellJr,Mc

-

GregorandVermilyea

(

2008

),

JungandLatsos

(

2015

)and

Bennani,FarvaqueandStanek

(

2018

)findthatanincreaseinthedifferencebetweenregionalandU.S.-wideunemploymentratesisassociatedwithahigherprobabilityofavoicedpreferenceforalowerpolicyrate.While

Tootell

(

1991

)findsnoevidencethatregionalunemploymentexplainsactualvotes,

Gildea

(

1992

)and

MeadeandSheets

(

2005

)findthatmoreunemploymentinapresident’sdistrictisassociatedwithanincreasedproba-bilityofvotinginfavoroflowerpolicyrates,evenifthisimpliesadissentingvote.

8

WeareunawareofotherpapersthatstudywhetherachangeinaBankpresident’svotingstatusleadstoachangeinthedependenceofherbehaviororstatedpreferencesondistrictlevelvariables.

9

2.3.Hypotheses.Weinvestigatetheeffectofvotingstatusontherelationshipbetween,ontheonehand,thenumberandtoneofspeechesbetweenmeetingsandthelengthandtoneofcontribu-tionsduringmeetingsand,ontheotherhand,district-leveleconomicvariables.Speechesappearanaturalplacetostartananalysisofbehavioralchangesinintermeetingperiods.Theyarepub-liclyobservable.Testimonytotheirpotentiallylargeinfluenceonfinancialmarketsisa“blackoutperiodonmonetarypolicycommunications”aroundFOMCmeetingsthatparticipantsmustob-serve.

10

FOMCmembersalsogiveinterviewsduringintermeetingperiods.Asthesearetypically

8Similarily,

Eichler,L?hnerandNoth

(

2018

)findthattheworseningofadistrict’sregionalbankingsectorhealthisassoci-atedwithahigherprobabilitythatthedistrict’spresidentvotestoeaseinterestrates.

9Somestudiesmakecomparisonsacrosspresidentswithandwithoutvotingstatus.Noneofthemmakewithin-presidentcomparisonsacrossyearswithandwithouttherighttovote.

Tillmann

(

2011

)findsthatvoterspredictinflationratesinlinewiththeconsensusview,whereasnonvotersoverpredictinflationratesiftheyfavortighterpolicyandunderpredictratesiftheyfavorlooserpolicy.

HayoandNeuenkirch

(

2013

)findthatpredictedregionalactivitydoesnotdeterminetheprob-abilitythataspeechisdeliveredbyvoters;nonvoters,however,aremorelikelytodeliveraspeechfavoringatighteningstancewheneconomicactivityispredictedtoexpandintheirdistricts.

Meade

(

2005

)findsthatnonvotingpresidentsvoicemoredisagreementwithchairmanGreenspan’sinterestrateproposalduringthemeetingthanvotingpresidents.Also,thepublicationofverbatimtranscriptsofFOMCmeetingsfrom1993onwardsmadenonvotingpresidentsmoreinclinedtovoicedisagreement,andvotingpresidentslessinclined(

Meade

,

2010

).

10See

Blinder,Ehrmann,Fratzscher,DeHaanandJansen

(

2008

),

EhrmannandFratzscher

(

2009

)and

FederalOpenMarket

Committee

(

2011

).Atpresent,theblackoutperiodstartsontheSaturdaybeforethemeetingandlastsuntiltheendofthedayfollowingtheendofthemeeting.

6EHRMANN,TIETZANDVISSER

shorterthanspeeches,theyseemlessamenabletotextualanalysis.Othercommunicationchannels,likephoneconversationswithgovernorsorotherpresidents,orinformalcontactswiththepressappearimpossibletoobserve.

11

Theeconomicvariablesthatwestudyaretheonesthathavebeenusedintheliterature:(i)theabsolutedifferencebetweendistrictandU.S.unemploymentasadeterminantofthenumberofspeechesandthelengthofcontributions,

12

and(ii)thedifferencebetweendistrictandU.S.unem-ploymentasadeterminantofthetoneofspeechesandcontributions.Thetwohypothesesthatwetestpredictoppositeeffectsofvotingstatusontheserelationships.AmodelinAppendix

A

formal-izesthehypothesesandderivestherelationshipsthatwetestempirically.

Beingpartofacommitteeprecludesthatanymembercanunilaterallydeterminethedecision.Preparingforthedecision-makingprocess,acostlyeffort,isthennotonlyanexerciseinidentifyingthebestoptionthroughthecollectionofintelligence;itisalsoanexerciseinconvincingothersofone’sarguments,one’sinterpretationofcurrentandfutureeconomicconditionsandone’spreferredchoice.Aspeechcanbeusedtoinfluencepolicymaking,eitherdirectly,asotherBankpresidentsandBoardgovernorsmayhearthespeechorreadatranscriptandareinfluencedbyit,orindirectly,asaspeechcaninfluenceeconomicactorslikefinancialmarketparticipants.Aspeechcanalsobeusedtotestandhonelinesofargumentandinterpretationofcurrenteventstobeusedattheupcomingmeeting.

13

OnthebasisoftheexistingliteratureweexpectthatpresidentsusespeechesmoreoftenthelargeristheabsolutedifferencebetweenthedistrictunemploymentrateandtheU.S.unemploymentrate,asmoreisatstake.Similarly,weexpectthatpresidentswillattempttoputmoreofastampontheeconomygo-roundduringthemeetingwhenthedifferenceinunemploymentratesincreases,asmoreisatstake.Weoperationalizethisbymeasuringthelengthofapresident’scontributions.

Presidentscanexertefforttomatchthetoneofspeechesandcontributionstothedistrictunem-ploymentratetoaidinthetransmissionofthemessage.

14

Onthebasisoftheexistingliteratureweexpectthatthetonewillbecomemorenegativethelargeristhedistrictunemploymentrate.

Wenowdiscusswhatdistinguishesthetwohypotheses.

15

2.3.1.Losscompensationhypothesis.Avotingrightisaninstrumenttofurtherone’sgoals;itis,how-ever,nottheonlyinstrument.Thishypothesisviewsspeechesbetweenmeetingsandcontributionsduringmeetingsontheonehandandtherighttovoteontheotherhandassubstitutestoincreasethelikelihoodthatthecommitteetakesthedecisiondesiredbythepresident.

11TheagendasofsixFederalReservegovernorsthat

MorseandVissing-Jorgensen

(

2020

))obtainedthroughaFreedomofInformationrequestdonotshowtheidentityofthepresidentswithwhomgovernorsspoke.

Vissing-Jorgensen

(

2021

)showsthatinformalcontactswiththepressandthepossibilityofleaksarearecurringtopicatFOMCmeetings.12Althoughmostregressionsconsiderunemployment,wealsolookattheroleplayedbydifferencesininflationandinfinancialsectorreturns.

13Thatpreparingspeechesandcontributionstomeetingstakesconsiderabletimeandeffortbecomesquiteclearwhenreading

Meyer

(

2004

),LaurenceMeyer’saccountofhisyearsasagovernoroftheFederalReserveBoard.

14Meyer,althoughnotaFedpresident,speaksof"fine-tuningtheremarks[he]wouldmakeattheFOMCmeeting"(

Meyer

,

2004

,p.75)andofworkinghardtogetthetoneofhisspeechesright(p.70).

15WepresentasimplemodelinAppendix

A

toderivetheconsequencesofthetwohypotheses.

VOTINGRIGHTROTATION7

Ifapresidentloosestherighttovote,shebecomesmoreinclinedtotakesubstituteactionsandgivesmorespeechesandmakeslongercontributionstocompensatefortheloss.Thisisespeciallythecase,themoredistrict-levelunemploymentdeviatesfromnationalunemployment,asmoreisatstake.Thus,thehypothesispredictsthatapresident’sinclinationtodeliveraspeechismoreresponsivetothedifferenceinunemploymentinyearswithouttherighttovotethaninyearswith.

Similarly,withouttherighttovote,apresidentwillmakemoreefforttomatchthetoneofspeechesandcontributionstotheeconomicsituationtoaidinthetransmissionofthemessage.Thehypoth-esispredictsthatinyearswithouttherighttovote,theresponsivenessofthesetonestodistrictunemploymentwillbegreaterthaninyearswiththerighttovote.

2.3.2.Motivationhypothesis.Havingtherighttovotemakesitmorelikelythatanyotheractivitytopromotethedesireddecisionpaysoff.Obtainingtherighttovotemotivatesamoreintenseuseofsuchactivities.Thishypothesisviewstherighttovoteontheonehandandspeechesandcontri-butionsontheotherascomplementarymeanstoincreasetheprobabilitythatthegroupmakesthedecisiondesiredbythepresident.

Ifapresidentobtainstherighttovote,shewillusespeechesandcontributionsespeciallywhenmoreisatstake.Thishypothesispredictsthatapresident’sinclinationtodeliveraspeechismoreresponsivetothedifferenceinunemploymentinyearswiththerighttovotethaninyearswithout.Similarly,hereffortstoattunethetoneofaspeecharestrengthenedbythehigherprobabilitythatsucheffortspayoffthankstothevotingright.Thehypothesispredictsthatinyearswiththerighttovote,theresponsivenessofthesetonestodistrictunemploymentwillbegreaterthaninyearswithouttherighttovote.

3.ECONOMETRICMODELSTOBEESTIMATED

Sections

3.1

to

3.3.2

presentthemaineconometricmodelsthatweestimatetotestthelosscom-pensationandmotivationhypotheses.Inaddition,insection

3.4

,welookintospecificcircumstanceswherethemarginalreturnsfrompresidents’effortspresumablyarehigher,allowingustotesthowthisaffectspresidents’behavior,irre