IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.2023JUNBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesHaNguyenWP/23/142IthankRabahArezki,BasBakker,AdolfoBarajas,RudolfsBems,AndrewBerg,MaiDao,MercedesGarcia-Escribano,HuiHe,ToanPhan,KoralaiKirabaeva,VladimirKlyuev,AntonKorinek,RuyLama,EmanueleMassetti,RodolfoMaino,JoeProcopio,NoomanRebei,NikolaSpataforaandMaryamVaziriforveryhelpfulcommentsandfeedback,andRuchunLiforeditorialhelp.IamgratefultoBerkayAkyapiandEmanueleMassettiforintroducingtomeclimatedataviaGoogleEarthEngine.©2023InternationalMonetaryFundWP/23/142IMFWorkingPaperInstituteofCapacityDevelopmentBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesPreparedbyHaNguyenAuthorizedfordistributionbyMercedesGarcia-EscribanoJune2023IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.ABSTRACT:Usingquarterlytemperatureandemploymentdatabetween1990and2021,thispaperuncoversnuancedevidenceontheimpactofseasonaltemperaturewithinUScounties:higherwintertemperatureincreasesprivatesectoremploymentgrowthwhilehighersummertemperaturedecreasesit.Theimpactsofhighertemperatureinmildseasons,fallandspring,arestatisticallyinsignificant.Moreover,thenegativeimpactofhighersummertemperaturepersistswhilethepositiveimpactofhighertemperatureinthewinterismoreshort-lived.Thenegativeeffectsofahottersummerarepervasiveandpersistentinmanysectors:mostsignificantlyin“Construction”and“LeisureandHospitality”butalsoin“Trade,Transport,andUtilities”and“FinancialActivities.”Incontrast,thepositiveeffectsofawarmerwinterarelesspervasive.Theemploymenteffectofahottersummerhasbeenmoresevereinrecentdecades.RECOMMENDEDCITATION:Nguyen,H.(2023).BeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCounties.IMFWorkingPapers,2023/142JELClassificationNumbers:C33,C55,E24,O44,Q54Keywords:Climatechange;temperature;employment;UScountiesAuthor’sE-MailAddress:Hnguyen7@imf.orgWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesPreparedbyHaNguyen11IthankRabahArezki,BasBakker,AdolfoBarajas,RudolfsBems,AndrewBerg,MaiDao,MercedesGarcia-Escribano,HuiHe,ToanPhan,KoralaiKirabaeva,VladimirKlyuev,AntonKorinek,RuyLama,EmanueleMassetti,RodolfoMaino,JoeProcopio,NoomanRebei,NikolaSpataforaandMaryamVaziriforveryhelpfulcommentsandfeedback,andRuchunLiforeditorialhelp.IamgratefultoBerkayAkyapiandEmanueleMassettiforintroducingtomeclimatedataviaGoogleEarthEngine.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCounties2ContentsI.Introduction..................................................................................................................................................4II.ATheoreticalFramework............................................................................................................................6III.DataandEmpiricalSpecification................................................................................................................9Data................................................................................................................................................................9EmpiricalSpecifications...............................................................................................................................10IV.MainFindings.............................................................................................................................................11AnnualRegressions.....................................................................................................................................11MainFindings...............................................................................................................................................12V.SummerandWinterImpactsAcrossState’sClimate.............................................................................15VI.OntheMechanismsoftheSummerTemperatureEffects.....................................................................18VII.OntheMechanismsoftheWinterTemperatureEffects.........................................................................21VIII.EffectsofTemperaturebyDecade...........................................................................................................23IX.RobustnessChecks...................................................................................................................................26NotUsingCountyEmploymentWeights......................................................................................................26DroppingExtremeEmploymentGrowth.......................................................................................................27DroppingRecessionQuarters......................................................................................................................27ControllingforNaturalDisasters..................................................................................................................28ControllingforPrecipitation..........................................................................................................................29X.Conclusions................................................................................................................................................30References.........................................................................................................................................................30FiguresFigure1:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature......14Figure2:AverageSummerImpactbyState.......................................................................................................16Figure3:AverageWinterImpactbyState..........................................................................................................17Figure4:TheEffectonEmploymentGrowthofHigherSummerTemperaturebySector..................................19Figure5:TheEffectofHigherWinterTemperaturebySector............................................................................21Figure6:AverageEmploymentSharesintheSummerandWinterinaCounty................................................23Figure7:AverageAnnualIncreaseinSummerTemperaturebyStateover1990and2021.............................24Figure8:DynamicImpactofYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(RegressionsareUnweighted)............................................................................................................................26IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCounties3Figure9:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(TopandBottom1percentileofEmploymentGrowthDataareDropped)..................................................................27Figure10:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(EmploymentGrowthDataforRecessionaryQuartersareDropped).................................................................28Figure11:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(ControllingforNaturalDisasters).......................................................................................................................29Figure12:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(ControllingforPrecipitation)...............................................................................................................................29TablesTable1:SummaryStatistics...............................................................................................................................10Table2:ImpactofAnnualAverageTemperatureonYoYGrowthofAnnualAverageEmployment..................11Table3:ImpactofTemperatureonYoYPrivateEmploymentGrowth...............................................................12Table4:RelationshipbetweenEmploymentEffectandaState’sClimate.........................................................18Table5:ImpactofTemperaturebyDecade........................................................................................................25IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesI.IntroductionClimatechangeisthebiggestchallengeforhumankind.Temperatureisrising.Theglobalaveragetemperatureisalreadyabout1.2degreeCelsiushigherthanthepre-industriallevel.Droughts,wildfires,andmassivestormsarestartingtooccurmorefrequentlywithdevastatingeffects.Understandingtheimpactofrisingtemperature,themostbasicmanifestationofclimatechange,oneconomicactivityisfundamentaltoadaptationandmitigationefforts.Theeconomicliteraturehasgenerallyfoundthathighertemperaturehurtseconomicactivity.Earlyliteratureexaminestherelationshipbetweenaveragetemperatureandaggregateeconomicvariables(e.g.,SachsandWarner,1997;Gallup,Sachs,andMellinger,1999).Itfindsthathottercountriestendtobepoorer.However,thisrelationshipmightbedrivenbyomittedvariablessuchascountryinstitutions.Recentliteratureusesfluctuationsintemperaturewithinacountryoraregiontocontrolforslow-movingcharacteristics(seeforexample,Delletal.,2012;Cashinetal.,2017;Colacitoetal.,2019;LettaandTol,2019;Acevedoetal.,2020;Kahnetal.,2021).1Itfindsthathighertemperaturereducestheeconomicgrowthofpoorcountries(Delletal.,2012;Acevedoetal.,2020)andtheUS(Colacitoetal.,2019).Thenegativeeffectsrunthroughreducedtotalfactorproductivitygrowth(LettaandTol,2019),andreducedinvestmentandlaborproductivity(Acevedoetal.,2020;KalkuhlandWenz,2020).Burkeetal.(2015)documentthenon-lineareffectoftemperature:economicgrowthriseswithaverageannualtemperatureuntilaround13degreesCelsiusanddropsafterthat.Thispaperexaminesthedynamiceffectsoftemperatureontheprivatesector’semploymentgrowthatalocallevel,namelyUScounty,andhighfrequency,namelyquarterly.Goingtothecountyandquarterlylevelsallowsformoreprecisetemperaturemeasurement.Therefore,itcouldestimatetheeffectsoftemperaturemorepreciselyanduncoverthesubtleeffectofseasonaltemperature.Thispaperfocusesonjobgrowthasthemaineconomicoutcome.JobsarefeaturedprominentlyintheUS'sdiscussionsofclimatechangemitigations.Manyworrythatclimatechangemitigationeffortswillhurtjobs(AFP,2022).Thispaperfindsthathighertemperature,onaverage,hurtsjobsintheUS.Usingdatabetween1990and2021,thispaperdiscoversopposingeffectsofhighertemperatureinthewinterandsummer.Onaverage,withinacounty,highersummertemperaturereducesprivatesectoremploymentgrowth,whilehigherwintertemperatureincreasesit.Theimpactsofhighertemperatureinmildseasons,fallandspring,arestatisticallyinsignificant.Thefindingsshowcasetheheterogenousandnuancedeffectsoftemperatureshocks.Thispaperfindsinterestingdynamiceffectsofseasonaltemperature.Highersummertemperaturehurtseconomicactivityinthecurrentandfollowingquarters.Atemporaryone-degreeFahrenheit(F)highersummertemperaturedecreasesyear-over-year(YoY)employmentgrowthofthatsummerby0.063percent.ItalsodecreasesYoYemploymentgrowthofthefollowingfallandwinterby0.08and0.075percent,respectively.Incontrast,thepositiveimpactofhighertemperatureinthewinterismoreshort-lived.Atemporaryone-degreeFahrenheitwarmerwinterboostsYoYemploymentgrowthinthatwinterby0.05percentbuthasstatisticallyinsignificanteffectsonemploymentgrowthinthefollowingspringandsummer.Insum,thenegativeimpactsof1AlsoseerecentsurveysbyDelletal.(2014)andAuffhammer(2018)IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountieshighertemperatureinthesummerarelargerandmorepersistentthanthepositiveimpactsofhighertemperatureinthewinter.Therefore,theaverageemploymenteffectofhighertemperatureacrossseasonsisnegative.Theeconomicliteraturetypicallyexaminestheimpactofannualaveragetemperatureonannualeconomicoutcomes(e.g.,seeDeschênesandGreenstone,2007;Delletal.,2012;Burkeetal.,2015;Acevedoetal.,2020;KalkuhlandWenz,2020;Akyapietal.,2022).However,sincetemperaturecanvarygreatlywithinayear,fromfreezingwinterstoscorchingsummers,thispaperarguesthatseasonaltemperatureisabetterapproximationofweatherthanannualtemperature.2Moreimportantly,theeconomicstructuresofdifferentseasonscouldbeverydifferent.Forexample,construction,travel,andtourismareexpectedtoriseinsummerandfallinwinter.Therefore,examiningtheeffectsofseasonaltemperatureonseasonaleconomicactivitycouldoffernewinsightstocomplementtheexistinganalysesusingannualaveragetemperatureandannual-averageeconomicoutcomes.Inaddition,workingwiththecountry-averagetemperatureisalsonotidealsinceevenwithinacountry,temperaturecanvarygreatly.Acountry,orevenaUSstate,mayhaveseveralclimatezones.Acaseoflocalizedtemperature,suchasatthecountylevel,canbemadehere.Nevertheless,granularanalysescomewiththeirchallengesandissues.First,andthemostobviousissueisthelackofhigh-frequencyeconomicdataatthelocallevel.OnereasonwhyemploymentgrowthischosenasthemainvariableofinterestisthattheUShasreliablequarterlydataatthecountylevel(moreonthatinsectionIII).Thesecond,andmoreconceptualissueislabormobilityatthelocallevel.Atthecountrylevel,labormobilityisrelativelyrestricted.Atleastintheshort-run,workershavetostayinacountryandtrytofindworkwithatemperatureshock.Butananalysisatthelocallevel,suchasUScounty,implieslabormobilityismuchlessrestricted.Peoplecouldmoveinandoutofacountytoworkinanothercountyinresponsetoatemperatureshock.Therefore,theeffectsoftemperatureonemploymentwithlabormobilitycanbelargerthanwithout.DeryuginaandHsiang(2017)andColacitoetal.(2019)examinetheimpactsofseasonaltemperatureatUScountyandstatelevels,respectively.However,theystilluseannualeconomicoutcomes,whichcouldmaskinterestingdynamiceffectsofseasonaltemperature.Thisanalysiscomplementstheiranalysesbynotfocusingonannualeconomicoutcomesbutonthehigh-frequencyimpactoftemperatureonquarterlyemploymentgrowthinUScounties.Byadoptingthislocalandhigh-frequencyempiricalframeworktogether,itunveilsnovelandinterestingdynamiceffectsofseasonaltemperature.Itcouldalsoshedlightonthemechanismsbydocumentingtheeffectsineachindustryandhowtheypropagateoverthenextquarters.Inotherwords,byobservingtemperature’simpactsondifferentsectorsatahighfrequency,insteadofbeingdilutedbytheannualaverages,thepapercanprovideadditionalinsightsintothemechanisms.Thispaperfindsthatthenegativeeffectsofahottersummerarepervasiveandpersistentinmanysectors:mostsignificantlyin“Construction”and“LeisureandHospitality”butalsoin“Trade,TransportandUtilities”and“FinancialActivities.”Employmentgrowthinthesesectorsmaygetdirectlyhitbyrisingtemperature.Itisalsopossiblethatsomeoftheloweremploymentgrowthisindirectlyaffectedduetoinput-outputlinkagesbetweendifferentsectorsortheaggregatedemandeffect.Forexample,jobgrowthin“FinancialActivities”couldbedampenedduetoaslowerfinancialservicedemandfrom“Construction.”Incontrast,thepositiveeffectsofa2Forexample,highestdailytemperatureinWashingtonD.C.(UnitedStates)in2021rangesfromthemid-30sFahrenheitinthewintertothemid-90sFahrenheitinthesummer.TheaverageannualtemperatureforWashingtonD.C.isabout70-degreeFahrenheit.Ifweusethisannualaverageof70-degreeFahrenheitinouranalyses,wemightbemistakenthatWashingtonD.C.’sweatherismoremoderatewhileinfact,ithasacoldwinterandahotsummer.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountieswarmerwinterarelesspervasive,onlyin“Construction,”“Leisureandhospitality,”and“NaturalResourcesandMining.”Itisalsomoreshort-lived.Therichnessofcounty-leveldataallowsfortheexaminationoftheeffectbyUSstate–whichisanotherimportantcontribution.Thispaperdiscoversarelationshipbetweenthenegativeeffectsofahottersummerwithastate’ssummerclimate:hotterstateshavemoreseverenegativeimpactsofhighersummertemperature.Somecoolerstates(e.g.,AlaskaandMassachusetts)evenbenefitfromthehighersummertemperature.Ontheotherhand,therelationshipbetweentheimpactofhigherwintertemperatureandastate’swinterclimateisnotasclear.Animportantpointofdiscussionishowwouldthesefindingsonshort-termresponseshelpuspredictthelong-termresponsestohotterclimates?Ithasbeenarguedthattheshort-runresponsestotemperaturefluctuationsarelikelynotthesameasthelong-runresponsestoclimatechange(seethediscussioninBurkeandEmerick,2016,forexample).Thisisareasonableargument.First,thefuturemagnitudeofclimatechangeisuncertain,dependingonhumankind’smitigationefforts.Inaddition,therecouldbearoleofadaptation.Adaptationefforts,suchasmorewidespreaduseofdrought-resistantseedsorair-conditioning,mightsoftentheimpactofrisingtemperatureinthefuture.Ifso,theshort-runimpactsmayoverstatethelong-runimpactsofclimate(seealsoMassettiandMendelsohn,2018).Conversely,therisingtemperaturemaycausepermanenteffectsonemployment(suchasemigrationoutofthehotareas).Inthatcase,theshort-termimpactsoftemperaturefluctuationmightunderstatethelong-runimpactsofclimatechange.Thispapercontributestothisdiscussionwithtwosetsoffindings.First,awarmerwinterhelpseconomicactivities,whileahottersummerhurtsthem.Inaddition,thenegativeeffectsofahottersummerinhotterstatesarelargerandmorepersistent.Thefindingssuggestthatcolderregionsorcountriesmaybenefitfromclimatechangewhilehotteronesmaybehurtwithoutsignificantadaptationefforts.Theseheterogeneouseffectspresentachallengeforaunifiedefforttofightclimatechange(whethertheyareglobaleffortsorthoseintheUS).Second,thispaperdiscoversmoresevereimpactsofsummertemperatureintheUSinrecentdecades(2000-2009and2010-2021)thanin1990-1999.Aone-degreeFahrenheithottersummerinthe2010sreducesemploymentgrowthinthesummerandthefollowingfallbyabout0.1percentmorethanitdidinthe1990s.ThisfindingimpliesadaptationeffortsintheUShavenottakenholdorsignificantlyalteredtheeffectsoftemperatureshocks.Thisfindinghasimplicationsforothercountries.EvenfortheUS,whichisadevelopedcountrywithgoodadaptationcapacityandwithagenerallymildclimate,weobservenegativeimpactsofhighertemperatureinthesummer.Forpoorer,hottercountries,theeffectsofrisingheat,withoutsignificantadaptationefforts,arelikelymuchmoresevere.Thepaperisorganizedasfollows.SectionIIpresentsasimpletheoreticalframeworktomotivatetheempiricalspecification.SectionIIIpresentsdataandthemainempiricalspecification.SectionIVpresentsthemainfindingsontheoverallimpacts.SectionVpresentstheimpactbytheUSstateandpatternsbetweentheimpactmagnitudeandastate’sclimate.SectionsVIandVIIexaminethesectoralimpactsofahottersummerandawarmerwinter.SectionVIIIpresentstheeffectsbydecade.SectionIXpresentsrobustnesschecks.SectionXconcludes.II.ATheoreticalFrameworkThissectionpresentsatheoreticalmotivationfortheempiricalsetup,wheretemperaturecanhavebothagrowtheffectandaleveleffectonemployment.InspiredbytheframeworkpresentedinDelletal.(2012),Iletemploymentinquarter𝑞afunctionofthecurrentquarter’sproductivityandlastquarter’semployment:IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCounties𝐿𝑞=𝜃𝑠𝑒𝜙𝑠𝑇𝑞𝐴𝑞𝜌𝑠𝐿𝑞−1𝜆𝑠(1)where𝑞denotesquarter,𝑠denotestheseason(i.e.,summer,fall,winter,orspring).𝐿𝑞,𝑇𝑞and𝐴𝑞areemployment,temperatureandproductivityinquarter𝑞.𝐿𝑞−1isemploymentinthepreviousquarter.Employmentinaquartercanbedrivenbythecurrentquarter’sproductivityandemploymentinthepreviousquarter.Capitalisomittedforsimplicity.Employmentatquarter𝑞candependonemploymentatquarter𝑞−1becausehiringmaytaketime.𝑒𝜙𝑠𝑇𝑞denotestheleveleffectofthequarter’stemperatureonemployment.Employmentcanbeaffectedbythequarter’saveragetemperature.Apositive/negative/zero𝜙𝑠impliesthathighertemperaturehasapositive/negative/zeroleveleffectonemployment.Notethattheparameters𝜃𝑠,𝜙𝑠,𝜌𝑠,𝜆𝑠areseasonspecific.Thatis,theparametersaredifferentforwinter,spring,summer,andfall.Forexample,highertemperaturemayhavedifferent(orevenopposite)leveleffectsinthesummerversusinthewinter,hence𝜙𝑠𝑢𝑚𝑚𝑒𝑟shouldbedifferentto𝜙𝑤𝑖𝑛𝑡𝑒𝑟.Seasonalproductivitygrowthisasfollows:log(𝐴𝑞)−log(𝐴𝑞−4)=𝑔𝑠+𝛿𝑠𝑇𝑞+𝜔𝑠𝑇𝑞−4(2)Equation(2)statesthatseasonalproductivitygrowthdependsonthisquarter’stemperatureaswellasthetemperatureofthesameseasonlastyear(4quartersago).𝛿𝑠and𝜔𝑠arealsoseasonspecific.𝑇𝑞,thisquarter’stemperature,couldhaveapositiveornegativeeffectonseasonalproductivitygrowth.𝑇𝑞−4,temperatureofthesamequarterinthepreviousyear,mayaffectthisquarter’sproductivitygrowthviatwochannels.Firstisthebaseeffect.Forexample,alower𝑇𝑞−4couldlower𝐴𝑞−4,whichboostslog(𝐴𝑞)−log(𝐴𝑞−4)duetothebaseeffect.Secondistheproductivitytransmissioneffect.Alower𝑇𝑞−4couldlower𝐴𝑞−4whichcouldinsteaddepressseasonalproductivitygrowthforthefollowingyear.Therefore,onthenet,itisnotclearthat𝜔𝑠isexpectedtohaveapositiveornegativevalue.Equations(1)and(2)statethattemperaturecouldhavealeveleffectonemployment(via𝑒𝜙𝑠𝑇𝑞).Itcouldalsohaveagrowtheffectonemploymentviaseasonalproductivitygrowthspecifiedinequation(2).3Now,let’srearrangeequations(1)and(2)toderiveanempiricalspecification.Foreaseofexposition,let’sstartequation(1)forthesummer𝐿𝑞=𝜃𝑠𝑢𝑚𝑚𝑒𝑟𝑒𝜙𝑠𝑢𝑚𝑚𝑒𝑟𝑇𝑞𝐴𝑞𝜌𝑠𝑢𝑚𝑚𝑒𝑟𝐿𝑞−1𝜆𝑠𝑢𝑚𝑚𝑒𝑟(3)andsubstitute𝐿𝑞−1=𝜃𝑠𝑝𝑟𝑖𝑛𝑔𝑒𝜙𝑠𝑝𝑟𝑖𝑛𝑔𝑇𝑞−1𝐴𝑞−1𝜌𝑠𝑝𝑟𝑖𝑛𝑔𝐿𝑞−2𝜆𝑠𝑝𝑟𝑖𝑛𝑔(notethatsince𝑞isthesummer,𝑞−1isthespring).(3)becomes𝐿𝑞=𝜃𝑠𝑢𝑚𝑚𝑒𝑟𝑒𝜙𝑠𝑢𝑚𝑚𝑒𝑟𝑇𝑞𝐴𝑞𝛽𝑠𝑢𝑚𝑚𝑒𝑟(𝜃𝑠𝑝𝑟𝑖𝑛𝑔𝑒𝜙𝑠𝑝𝑟𝑖𝑛𝑔𝑇𝑞−1𝐴𝑞−1𝜌𝑠𝑝𝑟𝑖𝑛𝑔𝐿𝑞−2𝜆𝑠𝑝𝑟𝑖𝑛𝑔)𝜆𝑠𝑢𝑚𝑚𝑒𝑟(4)3Althoughhumankind’sgreen-housegasemissionsinfluenceglobaltemperature,localtemperatureisconsideredexogenoustolocaleconomicactivities.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesKeepsubstituting𝐿𝑞−2=𝜃𝑤𝑖𝑛𝑡𝑒𝑟𝑒𝜙𝑤𝑖𝑛𝑡𝑒𝑟𝑇𝑞−2𝐴𝑞−2𝜌𝑤𝑖𝑛𝑡𝑒𝑟𝐿𝑞−3𝜆𝑤𝑖𝑛𝑡𝑒𝑟and𝐿𝑞−3=𝜃𝑓𝑎𝑙𝑙𝑒𝜙𝑓𝑎𝑙𝑙𝑇𝑞−3𝐴𝑞−3𝜌𝑓𝑎𝑙𝑙𝐿𝑞−4𝜆𝑓𝑎𝑙𝑙into(4),wecanseethat(1)takesthefollowinggeneralform:𝐿𝑞=𝜃𝑒𝜙0𝑇𝑞𝑒𝜙1𝑇𝑞−1𝑒𝜙2𝑇𝑞−2𝑒𝜙3𝑇𝑞−3𝐴𝑞𝜎0𝐴𝑞−1𝜎1𝐴𝑞−2𝜎2𝐴𝑞−3𝜎3𝐿𝑞−4𝜇(5)(5)statesthatemploymentisafunctionoftemperatureandproductivityofthisquarteraswellasthoseinthepreviousthreequartersandemploymentofquarter𝑞−4.Notethatallparametersof(5)areseasonspecific.Similar,forthesameseasoninthepreviousyear(i.e.,𝑞−4):𝐿𝑞−4=𝜃𝑒𝜙0𝑇𝑞−4𝑒𝜙1𝑇𝑞−5𝑒𝜙2𝑇𝑞−6𝑒𝜙3𝑇𝑞−7𝐴𝑞−4𝜎0𝐴𝑞−5𝜎1𝐴𝑞−6𝜎2𝐴𝑞−7𝜎3𝐿𝑞−8𝜇(6)Subtractlogof(5)bylogof(6):Δlog(𝐿𝑞)=∑𝜙𝜏Δ𝑇𝑞−𝜏3𝜏=0+∑𝜎𝜏Δlog(𝐴𝑞−𝜏)3𝜏=0+𝜇Δlog(𝐿𝑞−4)(7)whereΔlog(𝐿𝑞)=log(𝐿𝑞)−log(𝐿𝑞−4)isyear-over-yeargrowthinemployment;Δ𝑇𝑞−𝜏=𝑇𝑞−𝜏−𝑇𝑞−𝜏−4indicatesyear-over-yearchangeintemperature;andΔlog(𝐴𝑞−𝜏)=log(𝐿𝑞−𝜏)−log(𝐿𝑞−𝜏−4)indicatesyear-over-yeargrowthinseasonalproductivity.Substituting(2)into(7)yields:Δlog(𝐿𝑞)=∑𝜙𝜏Δ𝑇𝑞−𝜏3𝜏=0+∑𝜎𝜏{𝑔𝜏+𝛿𝜏𝑇𝑞−𝜏+𝜔𝜏𝑇𝑞−𝜏−4}3𝜏=0+𝜇Δlog(𝐿𝑞−4)(8)Rearrangetermsin(8)yields:Δlog(𝐿𝑞)=𝑔+∑𝛽𝜏𝑇𝑞−𝜏3𝜏=0+∑𝜋𝜏𝑇𝑞−𝜏7𝜏=4+𝜇Δlog(𝐿𝑞−4)(9)where𝑔=∑𝜎𝜏𝑔𝜏3𝜏=0;𝛽𝜏=𝜙𝜏+𝜎𝜏𝛿𝜏;𝜋𝜏=−𝜙𝜏+𝜎𝜏𝜔𝜏.Asbefore,allparametersareseasonspecific(thatis,theyvarydependingonwhetherthequarter𝑞issummer,fall,winter,orspring).Iaminterestedinthecoefficients𝛽0,𝛽1,𝛽2,𝛽3,representingtheeffectsoftemperatureinthisquarterandthreequartersagoonthisquarter’semployment.Intheempiricalsection,Iwillestimate𝛽0,𝛽1,𝛽2,𝛽3foreachseason.𝑇𝑞−4to𝑇𝑞−7andΔlog(𝐿𝑞−4)areconsideredcontrolvariables.44Asdiscussed,temperaturebetween𝑞−4and𝑞−7hasbaseeffectsaswellaspotentialproductivitytransmissioneffects.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesIII.DataandEmpiricalSpecificationDataEmploymentdata:Quarterlyemploymentdatabetween1990and2021attheUScountylevelarefromtheUSCensus’sQuarterlyCensusofEmploymentandWages(QCEW).TheQuarterlyCensusofEmploymentandWages(QCEW)programpublishesaquarterlycountof(formal)employmentandwagesreportedbyemployerscoveringmorethan95percentofUSjobs,availableatthecounty,metropolitan(MSA),state,andnationallevelsbyindustry.Majorexclusionsfromthedatasetincludeself-employedworkers,mostagriculturalworkersonsmallfarms,allmembersoftheArmedForces,electedofficialsinmoststates,mostemployeesofrailroads,somedomesticworkers,moststudentworkersatschools,andemployeesofcertainsmallnonprofitorganizations.QCEWincludesonlyabouthalfoftheU.S.agriculturalsector’semployment.Therefore,agriculturalemploymentisnotincludedintheanalysis.Dataforall50USstatesplusWashingtonD.Carecollected.Thenyear-over-yearpercentchangeinemploymentlevelatthecountyleveliscalculated.Growthinallprivateemploymentischosenasthekeyoutcome,butmoredisaggregatedemploymentisalsousedtoexaminethemechanismsoftheimpacts.ThisanalysisfocusesonprivateemploymentratherthanpublicemploymentbecauseprivateemploymenthasamuchlargershareintheUS.Inaddition,growthinprivateemploymentismorelikelytoreflecttheimpactsofrisingtemperature.Incontrast,growthinpublicemploymentcouldreflectadaptationeffortsbycountygovernments,whichisnotmyfocus.Temperaturedata:TemperaturedataforthecontiguousUSstatesarefromgridMET.gridMETisadatasetofdailyhigh-spatialresolution(about4km,1/24thdegree)surfacemeteorologicaldatacoveringthecontiguousUSfrom1979-yesterday.5Temperaturedataforthetworemainingstates,AlaskaandHawaii,arefromtheglobalERA5datasetwithlargergrids(about30km).ViatheplatformGoogleEarthEngine,IcollectmaximumdailytemperaturedatafromGridMETbetween1990andtheendof2021,matchingthetimecoverageoftheemploymentdata.However,ERA5dataareonlyavailabletoJuly9,2020.Hence,temperaturedataforAlaskaandHawaiiinthispaperonlygountilQ2of2020.Then,temperatureisaveragedacrossgridswithinacountytoconstructdailytemperaturedataattheUScountylevel.Next,temperatureisaveragedacrossdaystogeneratetemperatureatthequarterlyfrequency.TemperatureisinFahrenheit.Tomatchtemperaturedatawithquarterlyemploymentdata,thisanalysisdenotesaveragetemperatureforQuarter1(fromJanuarytoMarch)aswintertemperature,theaveragetemperatureforQuarter2(fromApriltoJune)asspringtemperature,theaveragetemperatureforQuarter3(fromJulytoSeptember)assummertemperature,theaveragetemperatureforQuarter4(fromOctobertoDecember)asfalltemperature.Precipitationdata:Precipitationdataarecollectedsimilarlytothewaytemperaturedataarecollected.First,totaldailyprecipitationdatabygridsarecollected,thenaveragedacrossgridswithinacountyandacrossthedayswithinaseasontogenerateseasonalaverageprecipitationforeachUScounty.PrecipitationdataforthecontiguousUSstatesbetween1990andendof2021arefromgridMET,whileprecipitationdataforAlaskaandHawaii’scountiesbetween1990andQ2-2020arefromERA5.5https://www.climatologylab.org/gridmet.htmlIMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesTable1presentsthesummarystatisticsofthetwosourcesofdata.Intermsofyear-over-yearquarterlyemploymentgrowthwithinacounty,themedianandmeanvaluesindicateitgrowsatcloseto1percentannuallybetween1990and2021.Thesimpleaveragetemperatureacrosscountiesis48.9degreesFahrenheit(forQ1orWinter),74.8degreesFahrenheit(forQ2orSpring),84degreesFahrenheit(forQ3orSummer),and56.3degreesFahrenheit(forQ4orFall).Themediansofseasonaltemperaturetakesimilarvalues.Intermsofyear-over-yearchangeinseasonaltemperaturewithinacounty,themeanandmedianvaluesofchangesinseasonaltemperaturedonotindicatethattemperatureconsistentlytrendsupovertimeacrossseasonsattheUScountylevel.Examiningthemediansmorecloselytoavoidextremevaluesmaybemoreuseful.ThemedianvaluesofYoYchangesinwinterandsummertemperaturebetween1990and2021arepositive(0.1degreesFahrenheitand0.057degreesFahrenheitperyear,respectively).However,themedianvaluesofYoYchangesinspringandfalltemperaturearenegative(-0.0279and-0.13degreesFahrenheitperyear,respectively).Thesesummarystatisticsshowcasetheheterogenous(andsometimesopposite)seasonaltemperaturechangesacrossseasons.Unit-roottestsofseasonaltemperatureinthecounty-quarterlypanelrejectthenullhypothesisthatseasonaltemperatureserieswithincountieshaveunitroots.Table1:SummaryStatisticsQuarterNNofcountiesMeanMinMedianMaxYoYEmploymentGrowth(%)Q196,7263,1410.937-92.1120.890841.506Q296,7473,1410.936-91.2540.966869.860Q396,7173,1410.890-91.0190.888752.892Q496,6573,1410.946-91.4510.943443.647Temperature(Fahrenheit)Q1(Winter)100,2613,14148.93-14.0448.8083.13Q2(Spring)100,2663,14174.8027.9175.1497.76Q3(Summer)100,2423,14184.0944.2684.49107.84Q4(Fall)100,2403,14156.34-3.2856.2585.41Yoychangeinseasonaltemperature(Fahrenheit)Q1(Winter)97,1153,141-0.1571-18.790.100619.84Q2(Spring)97,1213,1410.0039-11.73-0.027911.70Q3(Summer)97,0973,1410.0080-11.740.057312.14Q4(Fall)97,0963,1410.0689-15.95-0.131515.64EmpiricalSpecificationsTheregressionsarebasedoncounty-quarterlypaneldata.BasedonthetheoreticalmotivationinsectionII,themainempiricalspecificationisasfollows:∆𝐿𝑞,𝑐=𝑔+∑𝛽𝜏𝑇𝑞−𝜏,𝑐3𝜏=0+∑𝜋𝜏𝑇𝑞−𝜏,𝑐7𝜏=4+𝜇∆𝐿𝑞−4,𝑐+𝑓𝑒𝑐+𝑓𝑒𝑞+𝜖𝑞,𝑐(10)where𝑞denotesquarter,and𝑐denotescounty.∆𝐿𝑞,𝑐denotesyear-over-yearquarterlyemploymentgrowth(inpercent)atthecountylevel,forexample,betweenthisyear’ssummerandlastyear’ssummer.∆𝐿𝑞,𝑐canbegrowthinallprivatesectoremploymentorgrowthinsmallersectorssuchasconstruction,leisureandhospitality,andfinancialactivities.Sincethisisyear-over-yearquarterlyemploymentgrowth,ittakescareofseasonality.𝑇𝑞,𝑐denotescounty𝑐’stemperatureatquarter𝑞,and𝑇𝑞−𝜏,𝑐capturestemperature𝜏quartersago.𝛽𝜏(𝜏=0,1,2,3)capturestheeffectsoftemperatureinthiscurrentquarterandthepreviousthreequartersonyear-IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesover-yearemploymentgrowthinthisquarter.Thisspecificationiswelladoptedintheliterature,forexamplebyDelletal.(2012)andColacito(2019).ItisalsoderivedfromsimpletheoreticalmotivationinPartII,whichshowsthatequation(10)issufficientlyflexibletoallowagrowtheffectandaleveleffectoftemperature.∆𝐿𝑞−4,𝑐isacontrolvariable.Itdenotesyear-over-yearquarterlyprivateemploymentgrowthfourquartersago(thesamequarterinthepreviousyear).Thiscontrolisderivedfromthetheoreticalmotivationandisconsistentwiththeliterature(e.g.,Colacitoetal.,2019;Akyapietal.,2022).𝑇𝑞−𝜏,𝑐(𝜏=4,5,6,7)denotestemperature4to7quartersago.Theyarecontrolvariables.𝑓𝑒𝑐denotescountyfixedeffects;𝑓𝑒𝑞denotesyear-quarterfixedeffects.Hence,thisspecificationreflectstheimpactoftemperatureonemploymentgrowthwithinacounty.Theregressionsareweightedbytheconstantshare(ofnationalemployment)ofacounty’sprivateemploymenttogivehigherweightsformorepopulouscounties.Thisweightingistogivemorepopulouscountieslargerweights.Standarderrors𝜖𝑞,𝑐areclusteredatthestatelevel.IV.MainFindingsAnnualRegressionsThissectionpresentstheimpactofannualaveragetemperatureonYoYgrowthofannualaverageemploymentatthecountylevel.Table2showsthatwithinacounty,theimpactofannualaveragetemperatureongrowthofannualaverageemploymentisnegativebutnotstatisticallysignificant(p-value=0.208).Aswillbeclearbelow,usingannualaveragetemperatureandemploymentmasksinterestingimpactsofseasonaltemperature.Table2:ImpactofAnnualAverageTemperatureonYoYGrowthofAnnualAverageEmployment(1)YoYEmploymentGrowthYearAnnualAverageTemperature(y=0)-0.0443(0.0348)AnnualAverageTemperature(y=-1)-0.0203(0.0253)YoYEmploymentGrowth(y=-1)-0.0423(0.117)Constant4.969(3.343)Observations93,651Numberofcounties3,141R-squared0.275CountyFEyesYearFEyesStandarderrorsinparenthesesareclusteredatthestatelevel,p<0.01,p<0.05,p<0.1.Regressionsareweightedbytheconstantshareofacounty’sprivateemployment.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesMainFindingsTable3showsthedynamicimpactoftemperatureonyear-over-year(allprivate)employmentgrowthforeachseason(columns[1]to[4])andforthewholesample(column[5]).Themainvariablesofinterestaretemperaturethisquarteranduptothreequartersago.Forexample,regression(1)showstheimpactoftemperatureforthecurrentwinter(Q1ofthecurrentyear),thepreviousfall(i.e.,Q4ofthepreviousyear),theprevioussummer(i.e.,Q3ofthepreviousyear)andthepreviousspring(i.e.,Q2ofthepreviousyear)onprivateemploymentgrowthinthecurrentquarter(Q1ofthecurrentyear).Thedynamicimpactofaseason’stemperaturecanbepickedupbycollectingthecoefficientsacrosscolumns[1]to[4].Table3:ImpactofTemperatureonYoYPrivateEmploymentGrowth(1)(2)(3)(4)(5)YoYEmploymentGrowthWinter(Q1)Spring(Q2)Summer(Q3)Fall(Q4)PooledKeyvariablesTemperature(currentquarter)0.0480.025-0.063-0.0170.015(0.017)(0.034)(0.036)(0.02)(0.015)Temperature(q-1),0.00820.00390.0017-0.08-0.011(0.015)(0.013)(0.022)(0.03)(0.009)Temperature(q-2),-0.075-0.0110.0069-0.027-0.022(0.025)(0.018)(0.015)(0.032)(0.012)Temperature(q-3),-0.00860.0089-0.000940.007-0.0047(0.0137)(0.0314)(0.0304)(0.0112)(0.0088)ControlvariablesTemperature(q-4)0.0022-0.07-0.041-0.0084-0.031(0.0139)(0.018)(0.018)(0.027)(0.005)Temperature(q-5),-0.030.065-0.0037-0.048-0.0053(0.0261)(0.0103)(0.0124)(0.0209)(0.0066)Temperature(q-6),-0.022-0.0930.0560.0320.0071(0.0234)(0.0267)(0.00785)(0.0190)(0.0064)Temperature(q-7),0.0016-0.07-0.050.039-0.004(0.0137)(0.0273)(0.0302)(0.0121)(0.008)YoYEmploymentGrowth(q-4),0.104-0.0280.0240.1150.049(0.028)(0.081)(0.061)(0.028)(0.0379)ConstantyesyesyesyesyesObservations93,50693,53293,50293,427373,967R-squared0.3720.5540.4570.4430.468Numberofcounties3,1413,1413,1413,1413,141CountyFEyesyesyesyesyesYear-SeasonFEyesyesyesyesyesStandarderrorsinparenthesesareclusteredatthestatelevel,p<0.01,p<0.05,p<0.1.Regressionsareweightedbytheconstantshareofacounty’sprivateemployment.Thedarkgrayhighlightsshowthedynamiceffectsofsummertemperatureonsummerandsubsequentseasons.Thelightgrayhighlightsshowthedynamiceffectsofwintertemperature.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure1visualizesthesamefindingsin5charts.Theshadedareasrepresent90%confidenceintervals.Table3andFigure1revealthattheaverageeffectofhighertemperatureintheUSissmall(column5).However,theaverageeffectmaskslargeandopposingeffectsofhighertemperatureindifferentseasons.Temporarilyhighersummer(Q3)temperaturehurtsemploymentgrowthsignificantlyandpersistently.HighertemperatureinasummernotonlyreducesYoYprivateemploymentgrowthinthatsummerbutalsoYoYemploymentgrowthinthefollowingfallandwinter.Theeffectofatemporarilyhottersummerdissipatesinthefollowingspring(seeFigure1).Thecoefficientof0.063(column3)impliesthatatemporaryone-degreeFahrenheithighersummertemperaturedecreasesYoYemploymentgrowthofthatsummerby0.062percent.ItalsoreducesYoYemploymentgrowthofthefollowingfallandwinterby0.08and0.075percent,respectively.Tosummarize,atemporarilyhottersummerhastemporarybutpersistenteffectsonemploymentgrowth.Therefore,ithasapermanenteffectontheemploymentlevel.Notethatthisfindingisnotdrivenbyapossiblecorrelationoftemperatureinthesubsequentseasonsbecausethetemperatureinthesubsequentseasonsisalreadycontrolled(seeTable3).Temporarilyhigherwinter(Q1)temperaturehelpsemploymentgrowth,butonlysignificantlysoforthecurrentwinter.Thecoefficientof0.048(column1)impliesthataone-degreeFahrenheitwarmerwinterboostsYoYemploymentgrowthinthatwinterby0.048percentbuthasstatisticallyinsignificanteffectsonYoYemploymentgrowthinsubsequentseasons.Temperatureinthemilderseasons,fallandspring,doesnotaffectemploymentsignificantly.Theirmagnitudeisalsorelativelysmall.Notethattheimpactofahottersummerislesspreciselyestimatedthantheimpactofawarmwinter.The90%confidenceintervalislarger.Asthenextsectionshows,theimpactofahottersummervariesmoreacrossstates.Somestatesarehitheavilybyahottersummer,whileothersbenefit.Sincethenegativeemploymenteffectofatemporaryone-degreeFahrenheithottersummerislargerandmorepersistentthanthepositiveemploymenteffectofatemporaryone-degreeFahrenheitwarmerwinter,theaggregateeffectacrossseasonsisdominatedbythesummereffects(seecolumn5andFigure1).Theaverageemploymenteffectofaone-degreeFahrenheithighertemperatureacrossallseasonsispositivebutinsignificantforthecurrentquarter.However,itturnsnegativeandmorestatisticallysignificantforthenextthreequarters,especiallyq+1andq+2(seeFigure1).IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure1:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperatureNote:Thechartsshowthedynamicemploymentgrowtheffectsofaone-degreeFahrenheithighertemperature.Winter=JanuarytoMarch(Q1);Spring=ApriltoJune(Q2);Summer=JulytoSeptember(Q3);Fall=OctobertoDecember(Q4).Theshadedareasrepresent90%confidenceintervals.Standarderrorsareclusteredatthestatelevel.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesV.SummerandWinterImpactsAcrossState’sClimateOneoftheimportantquestionsishowtheimpactvarieswithclimatezones.Theexistingliteraturepointedoutthattheimpactofhighertemperatureismoresevereforhottercountries.ThissectiontacklesthisquestionbyexamininghowhighertemperatureinthesummerandwinteraffectsUSstatesdifferently.Inotherwords,howtheimpactsdifferforhotterversuscoolerstates.Therichdataatthecountylevelallowgoingtothisgranularlevel,sinceeachstatehasseveralcounties.6ThispaperfindsthathighersummertemperaturehurtsemploymentgrowthinmanyUSstates,butmostsoinhotterstatesintheSouth,suchasArizona(AZ),NewMexico(NM),andMissouri(MS).Coolerstatesevenbenefitfromahottersummer.Ontheotherhand,thereisnoclearrelationshipbetweentheimpactofhigherwintertemperaturewithastate’swinterclimate.Thisanalysisproceedsasfollows.First,sinceastatehasseveralcounties,thebaselineregression(10)isrunforeachstateforthe50statesandtheDistrictofColumbia.Thentheaverageimpactofaone-degreeFahrenheithighersummertemperatureinthecurrentquarterandthenextthreequartersiscalculated(whichistheaverageofthe4coefficients,likethosehighlightedindarkgrayinTable3).PanelAofFigure2showsthespatialdistributionofthesummerimpact.Theredcolorshowsstateswithamorenegativeimpactofahottersummer,whilethebluecolorshowsstateswithamorepositiveimpactofahottersummer.Arizona(AZ)andNewMexico(NM)standoutastwostateswithlargenegativeeffectsofahottersummer,whileAlaskaandNewEnglandstateslooktobenefitfromahottersummer.Toexaminetherelationshipbetweenastate’ssummerclimateandthesummerimpact,inPanelBofFigure2,Iplottheaveragesummerimpactagainstastate’saveragesummertemperature(between1990and2021).Thescatterplotisshownwithconstantstate-levelemploymentweights(panelB1).Forreaders’reference,statenamesareprovidedinpanelB2.NotethatthescatterplotsexcludeDelaware,whichhasonlythreecounties,andtheDistrictofColumbiaduetoimpreciseestimates.7PanelBshowsadownward-slopingrelationshipbetweentheaverageemploymentgrowthimpactinthesubsequentfourquartersandastate’ssummerclimate.Hotterstateshaveamoreseverenegativeemploymentimpactofahottersummer.Employmentgrowthinmanycoolstates(Alaska(AK),Vermont(VT),andNewHampshire(NH))benefitfromahottersummer.Howeachstatelosesorbenefitsfromahottersummerisatopicforfutureresearch.6Thegranularanalysisforeachstatewouldbeimpossiblewithanotherunitofanalysis,suchascommutingzone.7Theestimatesofaone-degreeFahrenheithottersummerforDelawareandWashingtonD.C.are-0.9and-1.26,respectively.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure2:AverageSummerImpactbyStatePanelA:SpatialdistributionsofthesummerimpactPanelB:Summerimpactandastate’ssummerclimatePanelB1:Withstate’sconstantemploymentsharePanelB2:WithstatenamesNote:Thechartsshowtheemploymentgrowtheffectsofaone-degreeFahrenheithottersummer.Theverticalaxisistheaverageemploymentgrowthimpactforthesubsequentfourquarters(t=0untilt=+3).ThescatterplotexcludesDelawareandWashingtonDCbecauseofimpreciseestimates(Delawarehasonlythreecounties;WashingtonDChasone).IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure3:AverageWinterImpactbyStatePanelA:SpatialdistributionsofthewinterimpactPanelB:Winterimpactandastate’swinterclimatePanelB1:Withstate’sconstantemploymentsharePanelB2:WithstatenamesNote:Thechartsshowtheemploymentgrowtheffectsofaone-degreeFahrenheitwarmerwinter.Theverticalaxisistheaverageemploymentgrowthimpactforthesubsequentfourquarters(t=0untilt=+3).ThescatterplotexcludesDelawareandWashingtonDCbecauseofimpreciseestimates(Delawarehasonlythreecounties;WashingtonDChasone).IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesSimilarly,theaverageimpactofhigherwintertemperatureonprivateemploymentgrowthinthecurrentquarterandthenextthreequartersiscalculated.PanelAofFigure3showsthespatialdistributionofthewinterimpact.Theredcolorshowsstateswithmorenegativeimpactsofawarmerwinter,whilethebluecolorshowsstateswithmorepositiveimpacts.Theredoesnotseemtobeaparticularpatterninthespatialdistribution.Floridastandsoutasaclearbeneficiaryofawarmwinter,possiblyduetothebetterattractionofitswarmerbeachesandothertouristdestinations.Toexamineapotentialrelationshipbetweenastate’swinterclimateandthewintereffect,Iplottheaverageemploymentimpactagainsttheaveragewintertemperaturebetween1990and2021.ThescatterplotisshowninpanelB1ofFigure3.Inaddition,statenamesareprovidedinpanelB2ofFigure3.PanelB1showsaflatrelationshipbetweentheaverageemploymentgrowthimpactandastate’swinterclimate.Table4presentsregressionstoformalizetherelationshipshowninPanelsB1ofFigures2and3.Inthefirstcase,thefindingisrobustwithorwithoutapotentialoutlier(Alaska(AK)).Theregressionsconfirmasignificantlinearrelationship:stateswithhigheraveragesummertemperaturehavemorenegativeaverageemploymenteffects.Inthesecondcase,thereisnorelationshipbetweenastate’swinterclimateandthewintereffect.Table4:RelationshipbetweenEmploymentEffectandaState’sClimateAverageEmploymentEffectofone-degreeFahrenheithottersummerAverageEmploymentEffectofone-degreeFahrenheitwarmerwinterVARIABLESIncludingAlaska(AK)ExcludingAlaska(AK)AverageSummerTemp(1990-2021)-0.0110-0.0111(0.0055)(0.0059)AverageWinterTemp(1990-2021)0.00292(0.00290)Constant0.9060.919-0.104(0.468)(0.498)(0.143)Observations494849R-squared0.1200.1160.051Robuststandarderrorsareinparentheses.p<0.01,p<0.05,p<0.1.Regressionsareweightedbytheconstantshareofastate’sprivateemployment.TheregressionsexcludeDelawareandWashington,D.C.becauseofimpreciseestimates(theyaresmallstateswithfewcounties).VI.OntheMechanismsoftheSummerTemperatureEffectsThissectionexaminesthemechanismsofthesummertemperatureeffectsbyestimatingthedynamiceffectsbysectors.Figure4showsthat“Construction”(NAICScode1012),“LeisureandHospitality”(NAISCcode1026),“Trade,TransportationandUtilities”(NAICScode1021),and“FinancialActivities”(NAICScode1023)arethehardesthit.Thisfindingmakessensebecause“Construction”and“LeisureandHospitality”aremostexposedtooutsideheat.Employmentgrowthinthesesectorsforsubsequentquartersisalsohitfortwopossiblereasons.First,theactivitycouldbeautocorrelatedacrossseasons.Forexample,fewerconstructionstartsinthesummercanreduceconstructionemploymentinthefall.Second,anoveralldeclineineconomicactivityinthesubsequentseasonscanaffectthesesectors(theaggregatedemandeffects).IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesServicesarelessunaffected.However,employmentgrowthin“Financialactivities”(NAICScode1023)ishurt.Thisfindingisinterestingbecauseworkplacesforfinancialsectorsaregenerallyclimate-insulated.Oneexplanationisthatloweremploymentgrowthinothersectorscouldhurtemploymentgrowthinfinancialactivity,drivenbythedeclineinthedemandforfinancialservices.Forexample,declinesinconstructionactivitiescoulddepressdemandforfinancialservices.Affectedtoalesserextentareemploymentgrowthin“Trade,TransportationandUtilities”and“Professionalandbusinessservices”(NAICScode1024).Notethat“Construction,”“Trade,TransportationandUtilities,”and“LeisureandHospitality”havelargeemploymentshares(seeFigure6).“Manufacturing”isnotaffectedcontemporaneously(probablybecauseitismostlyindoor).Asectorthatbenefitsfromahottersummeris“NaturalResources&Mining.”However,ithasaverysmallemploymentshare(seeFigure6).Figure4:TheEffectonEmploymentGrowthofHigherSummerTemperaturebySectorIMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesNote:Thechartsshowthedynamicemploymentgrowtheffectsofaone-degreeFahrenheithottersummerbysector.Theshadedareasrepresent90%confidenceinterval.Standarderrorsareclusteredatthestatelevel.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesVII.OntheMechanismsoftheWinterTemperatureEffectsThesectorsthatbenefitfromawarmerwinterare“Construction,”“Leisureandhospitality,”and“NaturalResourcesandMining.”Nosectorissignificantlyhurtbyawarmerwinter.However,itisinterestingtonotethattheeffectofawarmerwinterislesspervasivethantheeffectofawarmersummer.Onlythreesectorsaresignificantlyaffected.Figure5:TheEffectofHigherWinterTemperaturebySectorIMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesNote:Thechartsshowthedynamicemploymentgrowtheffectsofaone-degreeFahrenheitwarmerwinterbysector.Theshadedareasrepresent90%confidenceintervals.Standarderrorsareclusteredatthestatelevel.Thepositiveeffecton“Construction”ismoreshort-livedthanthecaseofthesummer.Whiletheeffectforthecurrentquarterisverylarge,thespillovertoemploymentgrowthinthespringandsummerissmall(seeFigure5).Thisfindingcouldbebecausehousingprojectsdonottypicallystartinthewinter.Employmentinconstructionis,onaverage,thehighestinthesummer(11%morethanthatinthewinter).Similarly,theIMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesspillovereffectsofaone-degreeFahrenheitwarmerwinterfor“Leisureandhospitality”insubsequentseasonsarealsosmall.Othersectorsthathavepersistentnegativeimpactsfromahottersummer(suchas“Financialactivities,”“Trade,transportation,andutilities”)donothavesuchpersistentpositiveeffectsfromawarmerwinter.Thesesectorshavelargeemploymentshares(seeFigure6).Thesefindingsexplainwhytheemploymentgrowtheffectofawarmerwinter,asshowninFigure1,isshort-lived.Figure6:AverageEmploymentSharesintheSummerandWinterinaCountyVIII.EffectsofTemperaturebyDecadeThissectionexaminestheeffectsoflong-termclimatechanges,whichoccuratdifferentratesfordifferentcountriesandregions.Forexample,Figure7showstheUSstate'saverageannualsummertemperatureincrease.Iobtainthisestimatebyrunningthefollowingregressionforeachstate𝑇𝑐=𝛽0+𝛽𝑠𝑦𝑒𝑎𝑟+𝑓𝑒𝑐+𝜖𝑐(2)where𝑇𝑐isseasonaltemperature(e.g.,summer)forcountyc,𝑦𝑒𝑎𝑟istimetrend.𝑓𝑒𝑐iscountyfixedeffects.𝛽𝑠isthestate-levelannualincreaseintheseasontemperature.Figure7belowshowsstate-averageannualincreaseinsummertemperaturebetween1990and2021.Itrangesfrom0.0022F-degreeto0.1degreesFahrenheityearly(or3degreesFahrenheitover30years).ThestateswiththemostdramaticincreasesarecolderintheNorth(Alaska,Wisconsin,Minnesota,NorthDakota,andSouthDakota).However,somehotstatesintheSouth(Texas,NewMexico)alsoseesubstantialtemperatureincreases.TherangeoftheincreaseinthesummertemperatureisconsistentwiththesummarystatisticsshowninTable1.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure7:AverageAnnualIncreaseinSummerTemperaturebyStateover1990and2021Animportantpointofdiscussionishowwouldthefindingsonshort-termresponseshelpuspredictthelong-termresponsestoahotterclimate?Ithasbeenarguedthattheshort-runresponsestotemperaturefluctuationsarelikelynotthesameasthelong-runresponsetoclimatechange(seethediscussioninBurkeandEmerick,2016,forexample).Forexample,adaptationefforts,suchasmorewidespreaduseofdrought-resistantseedsorair-conditioning,mightsoftentheimpactofrisingtemperatureinthefuture.Ifso,theshort-runimpactsmayoverstatethelong-runimpactsofclimate.Conversely,therisingtemperaturemaycausepermanenteffectsonemployment(suchasemigrationoutofthehotareas).Inthatcase,theshort-termimpactsoftemperaturefluctuationmightunderstatethelong-runimpactsofclimatechange.Thispapercontributestothisdiscussionbyexaminingtheshort-termimpactsbydecade.Wedosobyinteractingthetemperaturevariableswithdecadaldummyvariables.Thepointistocomparethedynamiceffectsoftemperatureovertime(1990-1999;2000-2009;2010-2021).Theargumentisthatiftheeffectsoftemperature,especiallythenegativeeffectsofthesummertemperature,areweakerinrecentdecades,thisimpliessomedegreeofadaptationhasbeenconductedtorespondtoatemporarilyhottersummer.Thispaperfindsthatahottersummer'snegativeimpactgraduallyworsens.First,thenegativeimpactofahottersummerisalsoworseduring2000-20thanduring1990-1999(thebaseline).Table5showstheinteractionbetweenthetemperaturevariableswiththe2000–2009-decadedummyandthe2010–2021-decadedummy.Theinteractionbetweenthesummertemperatureandthe2000–2009-decadedummyisnegativebutnotstatisticallysignificant(-0.0718incolumn3and-0.0838incolumn4).IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesTable5:ImpactofTemperaturebyDecade(1)(2)(3)(4)(5)YoyEmploymentGrowthWinterSpringSummerFallPooledTemperature(0)0.1190.04150.0105-0.05540.0175(0.0418)(0.0505)(0.0576)(0.0330)(0.0205)Temperature(q-1)-0.1020.05030.0252-0.00420-0.00964(0.0386)(0.0412)(0.0294)(0.0474)(0.00860)Temperature(q-2)-0.0649-0.08840.0446-0.0378-0.0255(0.0464)(0.0453)(0.0346)(0.0361)(0.0175)Temperature(q-3)0.05470.0618-0.08750.0205-0.00319(0.0159)(0.0609)(0.0497)(0.0244)(0.00975)Temperature(0)2000-2009decade-0.112-0.0482-0.07180.0755-0.000768(0.0428)(0.0378)(0.0623)(0.0447)(0.0117)Temperature(q-1)2000-2009decade0.193-0.0659-0.0317-0.08380.000449(0.0482)(0.0527)(0.0390)(0.0555)(0.00569)Temperature(q-2)2000-2009decade0.01530.133-0.05810.01600.00586(0.0569)(0.0639)(0.0473)(0.0373)(0.0104)Temperature(q-3)2000-2009decade-0.108-0.03930.120-0.0405-0.0104(0.0464)(0.0486)(0.0601)(0.0318)(0.00866)Temperature(0)2010-2021decade-0.09200.00322-0.1040.0483-0.00476(0.0445)(0.0466)(0.0426)(0.0453)(0.00977)Temperature(q-1)2010-2021decade0.166-0.0703-0.0151-0.102-0.00251(0.0469)(0.0565)(0.0356)(0.0351)(0.00469)Temperature(q-2)2010-2021decade-0.04160.135-0.05850.02660.00427(0.0425)(0.0668)(0.0472)(0.0279)(0.00890)Temperature(q-3)2010-2021decade-0.0638-0.1010.133-0.01030.00371(0.0381)(0.0555)(0.0575)(0.0358)(0.00833)Temperature(q-4toq-7)YesYesYesYesYesEmploymentGrowth(q-4)YesYesYesYesYesConstantyesyesyesyesyesObservations93,50693,53293,50293,427373,967R-squared0.3760.5550.4600.4450.468Numberofcounties3,1413,1413,1413,1413,141CountyFEyesyesyesyesyesYear-SeasonFEyesyesyesyesyesStandarderrorsareclusteredatthestatelevelandinparenthesesp<0.01,p<0.05,p<0.1.Regressionsareweightedbytheconstantshareofacounty’sprivateemployment.Second,thenegativeimpactofahottersummerismuchworseduring2010-2021thanduring1990-1999(thebaseline).Theinteractionbetweenthesummertemperatureandthe2010–2021-decadedummyisnegativeandstatisticallysignificant(-0.104incolumn3and-0.102incolumn4).Thefirstcoefficientmeansthataone-degreeFahrenheithottersummerin2010-2021reducessummeremploymentgrowthby0.1percentmorethanitdidinthe1990-1999decade.Thesecondcoefficientmeansaone-degreeFahrenheithottersummerduringIMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCounties2010-2021reducesemploymentgrowthinthefollowingfallby0.1percentmorethanitdidinthe1990-1999decade.Ahottersummeralsohurtemploymentgrowthinthefollowingwinterandspring,buttheeffectsareweakerandlessstatisticallysignificant.TheevidenceofaworseningimpactofhighersummertemperatureintheUSindicatesalimiteddegreeofadaptationtofluctuationsinsummertemperatureoritslimitedimpact.Logically,betweendifferentseasonaltemperature,thehottersummerposesthemostsubstantialchallengestoeconomicactivity.Therefore,ifonethinksadaptationshavebeeneffective,oneexpectsthattheimpactofatemporaryone-degreeFahrenheithighersummertemperaturewouldbesmallerovertimeduetoadaptationefforts.Inaddition,thereisamorepositiveimpactofawarmerfall(Q4)onemploymentgrowthinrecentdecadesthanduring1990-1999(thebaseline),notonlyforthatfallbutalsoforsubsequentseasons.Theinteractionsbetweenthetemperaturevariableswiththe2000-2009andthe2010-2021dummyvariablesarepositiveandlargelystatisticallysignificant.Ontheotherhand,awarmwinter(Q1)'spositiveimpacthasbeendampenedinrecentdecades.Thefourinteractionsbetweenthetemperaturevariableswiththe2000-2009dummyvariablearenegative,butonlyoneinteractionisstatisticallysignificant.Similarly,oneoffourinteractionsbetweenthetemperaturevariableswiththe2010-2021dummyvariableisstatisticallysignificant.Itisnotclearwhytheimpactsoffall(Q4)andwinter(Q1)temperaturehavemovedinoppositedirectionsinrecentdecades.IX.RobustnessChecksNotUsingCountyEmploymentWeightsThissectionprovidesarobustnesschecktothemainfindingswheretheempiricalregressionsarenotweightedbyacounty’semployment.Notethatthebaselineregressionsareweightedbythe(long-termaverage)constantshareofacounty’semploymenttogivehigherweightstomorepopulouscounties.Thisapproachisstandardintheliterature.Nevertheless,thissectionshowsthatthebaselineresultsarerobustwithouttheemploymentweights.Inotherwords,givingallcountiesthesameweightregardlessoftheiremploymentsizedoesnotchangethemainfinding.Figure8:DynamicImpactofYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(RegressionsareUnweighted)IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure8presentsthedynamicimpactofwinterandsummertemperaturewithoutcountyemploymentweights.Itshowsthattheimpactofaone-degreeFahrenheithighersummertemperaturepersistsinsubsequentquarters,whiletheeffectsofaone-degreeFahrenheithigherwintertemperaturearemoreshort-lived.DroppingExtremeEmploymentGrowthThissectionprovidesarobustnesschecktothemainfindingswhencountiesandquarterswithextremeemploymentchangesaredropped.Apotentialconcernisthatextremeemploymentchangesinacountyandquartercouldbecausedbyeventsthatarenotdirectlyrelatedtotemperatureofthatyear.Anexampleisaplantopeningorclosing.Totheextentthattheextremeemploymentchangescouldbeuncorrelatedtotemperatureofthequarter,droppingthemmightnotchangetheestimatesoftemperature’seffects.IdropYoYemploymentgrowthatthecounty-quarterlevelinthetopandbottomonepercentileofthedistribution(i.e.,thoselargerthan20.56percent-the99-percentilethreshold-andthosesmallerthan-17.27percent-the1-percentilethresholdinthedata).Notethattemperaturedataoftheseobservationsarenotdropped.Onlyemploymentgrowthdataare.ThedynamicimpactsonYoYemploymentgrowthtoaone-degreeFahrenheithighersummertemperature,andaone-degreeFahrenheitwintertemperatureremainsimilartothebaselinefindings(seeFigure9).Figure9:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(TopandBottom1percentileofEmploymentGrowthDataareDropped)DroppingRecessionQuartersThisrobustnesscheckdropsYoYemploymentgrowthforquartersdefinedbytheNBERasrecessions.TheyareQ3andQ4of2002;2008andQ1andQ2of2009;Q2of2020.8Liketheideaoftherobustnessabove,employmentgrowthinthesequartersisarguablycausedbyreasonsunrelatedtotemperatureshocks.Notably,8Seehttps://fred.stlouisfed.org/series/USRECIMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesthebaselinespecificationcapturestheaggregateeffectsoftheseeconomy-wideshocksbytheyear-quarterfixedeffects.Nevertheless,therecessionscouldhavecounty-specificeffects,dependingoncountyemploymentcomposition.Countieswithmoredemand-elasticindustries(suchastourism)couldbedisproportionatelyaffectedbytherecessions.Intheunlikelyscenariothattemperatureinthesecountiesmovessystematicallyandyieldsthesameeffectsastherecession,theeffectoftemperaturecouldbecontaminated.Forexample,employmentgrowthincountiesheavilydependentontourismmightbemoreseverelyhurtbyrecessions.Atthesametime,ifthesummertemperaturehappenstoalsoriseinthesecounties,thenegativeeffectsofhottersummertemperatureonsummeremploymentgrowthmightbecontaminatedbytherecessions.Figure10presentsthedynamicimpactsonemploymentgrowthafteremploymentgrowthdatafortheserecessionaryquartersaredropped.Notethattemperaturedataforthesequartersarenotdropped.Again,theimpactsremainrobustandsimilartothebaselinefindingsinFigure1.Figure10:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(EmploymentGrowthDataforRecessionaryQuartersareDropped)ControllingforNaturalDisastersThisrobustnesscontrolsforFEMA-declarednaturaldisasters.WeobtainedtheFEMA-declareddisastersfromFEMAwebsite9atthecountylevel.Theyincludefire,severestorms,hurricanes,snowstorms,floods,tornados,severeicestorms,droughts,andvolcaniceruptions.Theymightormightnotbecorrelatedwithtemperature.Forexample,ahottertemperaturemayleadtofire.Iassignadummyvariableof1ifacountyexperiencedatleastoneFEMA-declarednaturaldisasterinaquarter.Figure11presentsthedynamicimpactsoftemperatureonemploymentgrowthafterthenaturaldisasterdummyvariableiscontrolledfor.Again,theimpactsremainrobustandsimilartothebaselinefindingsinFigure1.99https://www.fema.gov/openfema-data-page/disaster-declarations-summaries-v2IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure11:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(ControllingforNaturalDisasters)ControllingforPrecipitationThisrobustnesscheckcontrolsforaveragedailyprecipitationatthecountylevel.Icollectprecipitationdatafromthesamesourcesoftemperature(seeSectionIII-Data).Existingliterature,mostlyontheannualfrequency,showsthatprecipitationdoesnotaltertheeffectsoftemperatureoneconomicactivity.Therefore,inthisrobustnesscheck,Ifollowthebaselinespecificationinequation(10)butalsocontrolfortheaveragedailyprecipitationofthecurrentquarterandtheprevious7quarters.Thefindingisverysimilartothebaselinefinding.Figure12presentsthedynamicimpactsoftemperatureonemploymentgrowthafterprecipitationiscontrolledfor.Again,theimpactsremainsimilartothebaselinefindingsinFigure1.Figure12:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(ControllingforPrecipitation)IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesX.ConclusionsUsingtemperatureandemploymentdatabetween1990untilendof2021,thispaperexaminesthehigh-frequencydynamicimpactoftemperatureatalocalgeographicallevel,namelyUScounty.Itfindsthattheeffectsarenuanced:hottertemperatureinthesummerhurtsemploymentgrowth,whilehottertemperatureinthewinterhelps.However,theeffectsinthesummerarelargerandmorepersistent,hurtingemploymentgrowthinsubsequentquarters.Inaddition,theeffectsofahottersummerhavebeenmorenegativeinrecentdecadesandmorenegativeinhotterstates.Thefindingssuggestthatcolderregionsorcountriesmaybenefitfromclimatechangewhilehotteronesmaybehurtwithoutsignificantadaptationefforts.Thisfindingpresentsachallengeforaunifiedefforttofightclimatechange.Inaddition,astheeffectsofahottersummerhavebeenmorenegativeinrecentdecades,thisimpliesadaptationeffortsintheUShavenottakenholdorhavenotsignificantlyalteredtheeffectsofahottersummer.Thisfindinghasimplicationsforothercountries.EvenfortheUS,adevelopedcountrywithgoodadaptationcapacityandgenerallymildclimate,weobservenegativeimpactsofhighertemperatureinthesummer.Forpoorer,hottercountries,theeffectsofrisingheat,withoutsubstantialadaptationefforts,arelikelymuchmoresevere.Thispaperhighlightstheneedtostudytheimpactofclimatechangeatgranularlevels,inbothtimeandspacedimensions,touncoverthehighlyheterogenouseffectsofclimatechange.ReferencesAcevedo,Sebastian,MicoMrkaic,NatalijaNovta,EvgeniaPugachevaandPetiaTopalova.2020."TheEffectsofWeatherShocksonEconomicActivity:WhataretheChannelsofImpact?,"JournalofMacroeconomics,vol.65(C).AFP.2022.DespiteDisasters,ClimateIsATabooElectionIssueInUSCoalCountry(Oct12,2022)linkAkyapi,Berkay,MatthieuBellon,andEmanueleMassetti.2022."EstimatingMacro-FiscalEffectsofClimateShocksFromBillionsofGeospatialWeatherObservations."IMFWorkingPaper2022/156Auffhammer,Maximilian.2018."QuantifyingEconomicDamagesfromClimateChange."JournalofEconomicPerspectives,32(4):33-52.Burke,Marshall,SolomonHsiang,andEdwardMiguel.2015.GlobalNon-LinearEffectofTemperatureonEconomicProduction.Nature527,235–239.Burke,Marshall,andKyleEmerick.2016.“AdaptationtoClimateChange:EvidencefromUSAgriculture:Dataset.”AmericanEconomicJournal:EconomicPolicy8(3):106-140Cashin,Paul&Mohaddes,Kamiar&Raissi,Mehdi,2017."Fairweatherorfoul?ThemacroeconomiceffectsofElNiño,"JournalofInternationalEconomics,vol.106:37-54.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesColacito,Riccardo,BridgetHoffmann,andToanPhan.2019."Temperatureandgrowth:ApanelanalysisoftheUnitedStates."JournalofMoney,CreditandBanking51,no.2-3:313-368.Dell,Melissa,BenjaminF.Jones,andBenjaminA.Olken.2012“TemperatureShocksandEconomicGrowth:EvidencefromtheLastHalfCentury.”AmericanEconomicJournal:Macroeconomics,4,66–95.Dell,Melissa,BenjaminF.JonesandBenjaminA.Olken.2014."WhatDoWeLearnfromtheWeather?TheNewClimate-EconomyLiterature,"JournalofEconomicLiterature,vol.52(3),pages740-798,September.Deryugina,TatyanaandSolomonHsiang.2017."TheMarginalProductofClimate,"NBERWorkingPapers24072Deschênes,Olivier,andMichaelGreenstone.2007."Theeconomicimpactsofclimatechange:evidencefromagriculturaloutputandrandomfluctuationsinweather."AmericanEconomicReview97,No.1:354-385.Jordà,Òscar.2005."EstimationandInferenceofImpulseResponsesbyLocalProjections."AmericanEconomicReview,95(1):161-182.Gallup,JohnLuke,JeffreyD.Sachs,andAndrewD.Mellinger.1999.“GeographyandEconomicDevelopment.”InternationalRegionalScienceReview22(2):179–232.Henseler,MartinandIngmarSchumacher.2019."Theimpactofweatheroneconomicgrowthanditsproductionfactors,"ClimaticChange,vol.154(3):417-433.Kalkuhl,Matthias,andLeonieWenz.2020."Theimpactofclimateconditionsoneconomicproduction.Evidencefromaglobalpanelofregions."JournalofEnvironmentalEconomicsandManagement103:102360.Letta,MarcoandRichardS.J.Tol,2019."Weather,ClimateandTotalFactorProductivity,"Environmental&ResourceEconomics,vol.73(1):283-305Massetti,EmanueleandRobertMendelsohn.2018.“MeasuringClimateAdaptation:MethodsandEvidence.”ReviewofEnvironmentalEconomicsandPolicy,12(2):324-341.Sachs,JeffreyD.,andAndrewM.Warner.1997.“SourcesofSlowGrowthinAfricanEconomies.”JournalofAfricanEconomies6(3):335–76BeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesWorkingPaperNo.WP/2023/142