53 Pages
English

Inattentive Professional Forecasters

-

Gain access to the library to view online
Learn more

Description

Niveau: Supérieur, Doctorat, Bac+8
Inattentive professional forecasters? Philippe Andrade Banque de France & CREM Herve Le Bihan Banque de France First draft: January 2009 This draft: November 2011 Abstract We use the ECB Survey of Professional Forecasters to characterize the dynamics of expec- tations at the micro level. We emphasize the following two new facts: forecasters (i) fail to systematically update their forecasts and (ii) disagree even when updating. We also find, as previous work on other surveys does, that forecasters have predictable forecast errors and dis- agree. We argue that these facts are altogether qualitatively supportive of recent theories in which agents are inattentive when they form their expectations. More precisely, they are in line with a model where agents imperfectly process information due to both sticky information a la Mankiw-Reis, and noisy information a la Sims. However, building and estimating such an expectation model, we find that it cannot quantitatively replicate the error and disagreement observed in the SPF data. Given how inaccurate they are, professionals agree too much to be consistent with the inattention model we test. Keywords: Expectations, imperfect information, inattention, forecast errors, disagreement, business cycle JEL classification: D84, E3, E37 ?We thank our discussants Bartosz Mackowiak, Ernesto Pasten and Gregor Smith as well as Carlos Carvalho, Olivier Coibion, Christian Hellwig, Anil Kashyap, Noburo Kiyotaki, Juan Pablo Nicolini, Giorgio Primiceri, Sergio Rebelo, Jonathan Willis, Alexander Wolman, Michael Woodford, Tao Zha and seminar participants at the Banque de France, ECB, New-York

  • forecasters

  • professionals

  • expectation

  • aggregate forecast

  • cross-section dispersion

  • generate disagreement

  • rather than

  • rational expectation

  • heterogeneity among

  • forecast errors


Subjects

Informations

Published by
Reads 27
Language English
InattentiveprofessionalforecastersPhilippeAndradeHerve´LeBihanBanquedeFrance&CREMBanquedeFranceFirstdraft:January2009Thisdraft:November2011AbstractWeusetheECBSurveyofProfessionalForecasterstocharacterizethedynamicsofexpec-tationsatthemicrolevel.Weemphasizethefollowingtwonewfacts:forecasters(i)failtosystematicallyupdatetheirforecastsand(ii)disagreeevenwhenupdating.Wealsofind,aspreviousworkonothersurveysdoes,thatforecastershavepredictableforecasterrorsanddis-agree.Wearguethatthesefactsarealtogetherqualitativelysupportiveofrecenttheoriesinwhichagentsareinattentivewhentheyformtheirexpectations.Moreprecisely,theyareinlinewithamodelwhereagentsimperfectlyprocessinformationduetobothstickyinformationa`laMankiw-Reis,andnoisyinformationa`laSims.However,buildingandestimatingsuchanexpectationmodel,wefindthatitcannotquantitativelyreplicatetheerroranddisagreementobservedintheSPFdata.Givenhowinaccuratetheyare,professionalsagreetoomuchtobeconsistentwiththeinattentionmodelwetest.Keywords:Expectations,imperfectinformation,inattention,forecasterrors,disagreement,businesscycleJELclassification:D84,E3,E37WethankourdiscussantsBartoszMac´kowiak,ErnestoPaste´nandGregorSmithaswellasCarlosCarvalho,OlivierCoibion,ChristianHellwig,AnilKashyap,NoburoKiyotaki,JuanPabloNicolini,GiorgioPrimiceri,SergioRebelo,JonathanWillis,AlexanderWolman,MichaelWoodford,TaoZhaandseminarparticipantsattheBanquedeFrance,ECB,New-YorkFed,PhiladelphiaFed,San-FranciscoFed,UniversityParis1andattheconferencesESEM2009,AFSE2010,CESIfoon“MacroeconomicsandSurveyData”andSED2010forusefulcomments.Allremainingerrorsareours.WearealsogratefultoSylvieTarrieuforsuperbresearchassistanceaswellastoClaudiaMarchiniandIevaRubenefortheirhelpwiththeSPFdata.ThispaperdoesnotreflectnecessarilytheviewsoftheBanquedeFrance.e-mails:philippe.andrade@banque-france.fr,herve.lebihan@banque-france.fr1
1IntroductionModelsinwhichimperfectinformationandtheformationofexpectationsactasatransmissionmechanismofeconomicfluctuations—inthespiritofFriedman(1968),Phelps(1968)andLucas(1972)—haverecentlyregainedinterestinthemacroeconomicliterature.1Imperfectinformation,inparticular,hasbeenrelatedtotheinattentionofagentstonewinformation,abehaviorthatcanberationalizedbycostlyaccesstoinformationandlimitedprocessingcapacities.2Oneappealofthesemodelsistoprovideanalternativechanneltostickypricestoexplainthepersistenteffectsoftransitoryshocks—andinparticular,monetaryshocks—ontheeconomy.Moreover,thisapproachcanparsimoniouslyaccountforpatternsofindividualexpectationsobservedinsurveydatathatareatoddswiththestandardperfectinformation,rationalexpectationframework,suchaspredictableforecasterrorsandforecastsdifferingacrossforecasters.3Inthispaper,weexploitthepaneldimensionofsuchasurveyofforecasts—namelytheECBSurveyofProfessionalForecasters(SPF)—toproducenewmicrofactscharacterizingtheformationofexpectations.TheECBSPFisaquarterlypanelstartingin1999surveyingaround90forecastingunitsineitherpublicorprivateinstitutionsandallowstotracksequencesofforecastsmadebythesameinstitution.Wethenelaborateonthesenewfactstoassesswhethermodelsofinattentionaccuratelydescribethebehaviorofforecasters.Wefocusontwotypesofinattentionmodelsthathavebeendiscussedintherecentliterature.Ontheonehand,stickyinformationmodelsdevelopedbyMankiw&Reis(2002)andReis(2006a,b),inwhichagentsupdatetheirinformationsetinfrequentlybutgetperfectinformationoncetheydo.OntheotherhandnoisyinformationmodelsproposedbyWoodford(2002),Sims(2003)andMac´kowiak&Wiederholt(2009),inwhichagentscontinuouslyupdatetheirinformationbuthaveanimperfectaccesstoitateachperiod.1See,amongothers,Woodford(2002),Hellwig&Veldkamp(2008),Angeletos&La’O(2009)andLorenzoni(2009).Imperfectinformationisalsocrucialintherecentwelfareanalysisofinformation.See,amongothers,Morris&Shin(2002),Angeletos&Pavan(2007)orAdamor&Weill(2009).Veldkamp(2009)andMankiw&Reis(2010)providesurveys.2SeeMankiw&Reis(2002),Sims(2003),Reis(2006a,b),andMac´kowiak&Wiederholt(2009).3SeeforexampleMankiw,Reis&Wolfers(2003).2
Assomerelatedpreviouswork,weuseprofessionalforecastdatatotestimperfectinformationmod-els.Expertforecastersmaynotberepresentativeoflesssophisticatedagents,sinceprofessionalsobviouslyallocatesubstantiallymoretime,human,collectingandcomputingresourcestothetaskofforecastingmacroeconomicvariables.However,Carroll(2003)showsthattheopinionofprofes-sionalforecastersspreadstofirmsandhouseholds,andhencealsoinfluencestheirexpectationsanddecisions.Furthermore,weexpectprofessionalforecasterstobetheagentsinthebestpositiontopayattentiontotherelevantmacroeconomicinformation.Asaresult,theextentofattentiontonewsamongprofessionalforecasterscanbeseenasanupperboundforotheragents’attentiontoaggregateconditions.Ourpaperhastwomaincontributions.Thefirstoneistodocumenttwonewfactsrelatedtoforecastrevisions,namelythat(i)forecastersdonotsystematicallyupdatetheirforecastsevenwhennewinformationisreleased,andthat(ii)forecasterswhoupdatealsodisagreeontheirforecasts.Theoriginalityofourapproachistoexploitthesequencesofindividualforecastsforagivenevent(sayinflationattheendofagivenyear)providedbytheECBSPFtoconstructadirectmicro-dataestimateofthefrequencyofupdatingaforecast,which,toourknowledge,hasnotbeendocumentedinsurveydatabefore.Theresultsshowthat,onaverage,eachquarteronly75%ofprofessionalforecastersupdatetheir1-yearor2-yearforecasts.Thisfirstresultisinlinewiththepredictionsofasticky-informationmodel.Furthermore,inthissetupthefrequencyofupdatinghasastructuralinterpretationandcorrespondstothekeyparameteroftheattentiondegree.Inadditiontounfrequentupdating,wealsouncoverthatforecasterswhoupdatetheirinformationsetsdisagreeabouttheirforecasts.Consequently,disagreementamongexpertsisnotonlyrelatedtodifferencesintheinformationsetsofforecasterswhoupdatedandofthosewhodidnot,butalsotothefactthat,whentheyupdate,theyusedifferentinformationsets.Thissecondresultisinlinewiththepredictionsofanoisy-informationmodel.Wealsofind,aspreviousworkrelyingonsurveydatadoes,thatforecastsofexpertsexhibitpredictableerrorsandthatforecastersdisagreeastheyreportdifferentpredictionsforthesamevariableatthesamehorizon.Theselattertwo3
characteristicsareinlinewithbothstickyandnoisyinformationmodels.Thesecondmaincontributionofthispaperistoperformaformalempiricalassessmentofinat-tentionmodels.Thefactsweunderlinequalitativelysupportamodelfeaturingatthesametimetwotypesofinattention;namely,sticky-informationa`laMankiw-Reis,andnoisy-informationa`laSims.WedevelopsuchanexpectationmodelandthenusetheSPFdatatocarryoutaMinimumDistanceEstimation(MDE).Moreprecisely,theestimationprincipleistomatchthemomentscharacterizingtheforecasterrorsandthedisagreementgeneratedbythistheoreticalmodelwiththeirempiricalcounterpartobservedintheECBSPFdata.Wefindthatthisinattentionmodelfailstoquantitativelyreproducetheobservedpersistenceoftheaverageforecastingerrorstogetherwiththerelativelysmalldisagreementbetweenforecasters.FittingthesmoothnessobservedintheaverageSPFforecastswouldrequireamuchlowerattentiondegreethanourmicrodataestimates.SuchalowattentionwouldinturnleadtomuchmoredisagreementthanobservedintheSPFdata.Therefore,elementsothersthanthetypeofinattentionincludedinourexpectationmodelareneededtoreconciletherelativelylowdisagreementamongprofessionalsandtherelativelyhighpersistenceoftheaggregateforecastingerror.Ourpaperrelatestothevastliterature,mostlyrelyingonUSdata,thatstudiesthebehaviorofsurveyforecastsandcomparesitwiththeimplicationsoftheoreticalexpectationmodels.Numerousstudies(seePesaran&Weale,2006forarecentsurvey)foundsystematicaggregateforecasterrorsanddisagreementinthesedata,atoddswiththeperfectinformationrationalexpectationframe-work.Here,weprovideevidenceofdisagreementandpredictableforecasterrorsforEuropeanSPFdataandforarecentsampleperiod.Wealsocomplementtheseresultsbyprovidingnewevidenceontheinfrequencyofindividualexpectationsrevision.OurworkisalsorelatedtoMankiwetal.(2003),Branch(2007),Coibion&Gorodnichenko(2008)andPatton&Timmermann(2009),whorelyonthecharacteristicsofsurveyexpectationstoassessinattentionand,moregenerally,imperfectinformationtheories.Mankiwetal.(2003)andBranch(2007)focusonthecross-sectiondistributionofforecaststocalibratethesticky-information4
attentionparametermentionedabove.Bycomparison,weunderlinetheimportanceofinvestigatingtheconsistencyofthisparametervalueswithboththecross-sectiondispersionofforecastsandtheaggregateforecasterrors.Furthermore,weimproveontheirapproachbyconsideringamodelthatcanexplainthedisagreementamongforecasterswhoupdatetheirinformationset.Lastly,ratherthancalibrateit,weestimatetheattentionparameterusingalternativelydirectmicro-dataestimatesandaMDEprocedure.Coibion&Gorodnichenko(2008)lookattheconditionalresponsetovariousstructuralshocksoftheaggregateerroranddisagreementimpliedbysurveystodisentanglethesticky-informationandthenoisy-informationmodelsofinattention.Theyfindmixedsupportinfavorofthetwo,justaswedo.Thedistinctivefeatureofouranalysisisthatweestimateamodelfeaturingsimultaneouslythetwotypesofinattention.Patton&Timmermann(2009)relyontheevolutionofforecastsoverdifferentforecasthorizonstostressthatdifferencesintheinterpretationofinformation—ratherthandifferentinformationsets—arethemainculpritforforecasters’disagreement.Theyleavethesourceofthesedifferentinterpretationsunexplained.Themodelweconsiderisanalternativeapproachtogeneratedisagreementthatdoesnotrelyon“deep”heterogeneityamongforecasters.Ourpaperismoreovercloselylinkedtoseveralrecentcontributionsthatrelyonaggregatetimeseriestoestimatetheattentiondegreeinasticky-informationmodelofinflationdynamics(Kiley2007,Do¨pkeetal.2008,Coibion2010).Whensignificant,theresultsimplyafrequencyofupdatingaforecastrangingfrom10%to30%,wellbelowthefigureof75%weobtain.Partofthediscrepancycouldbeexplainedbythefactthatwerelyonapanelofprofessionals,sincetheseagentsmayupdatemorefrequentlytheirforecaststhanothertypes.However,manyoftheserecentstudies(forinstanceDo¨pkeetal.2008,Coibion,2010)alsouseprofessionalforecasters’expectationstoperformtheirestimation.Thediscrepancyisthusalsorelatedtothemethodology.Studiesthatusemacroeconomictimeseriestypicallyrelyonauxiliaryassumptionsabouttheeconomy,andarepotentiallysubjecttoaggregationbiases.Bycontrast,weprovideadirect,arguablymorereliable,micro-dataestimateofthisparameterthatiskeyinstickyinformationmodels.Althoughwefind5