IMF-揭示温度的局部影响 Shedding Light on the Local Impact of Temperature 2024_第1页
IMF-揭示温度的局部影响 Shedding Light on the Local Impact of Temperature 2024_第2页
IMF-揭示温度的局部影响 Shedding Light on the Local Impact of Temperature 2024_第3页
IMF-揭示温度的局部影响 Shedding Light on the Local Impact of Temperature 2024_第4页
IMF-揭示温度的局部影响 Shedding Light on the Local Impact of Temperature 2024_第5页
已阅读5页,还剩47页未读 继续免费阅读

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

SheddingLightontheLocalImpactof

Temperature

DaHoang,DuongLe,HaNguyen,andNikolaSpatafora

WP/24/178

IMFWorkingPapersdescriberesearchin

progressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.

TheviewsexpressedinIMFWorkingPapersare

thoseoftheauthor(s)anddonotnecessarily

representtheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.

NAT

2024

AUG

©2024InternationalMonetaryFundWP/24/178

IMFWorkingPaper

AfricanDepartmentandInstituteforCapacityDevelopment

SheddingLightontheLocalImpactofTemperature

PreparedbyDaHoang,DuongTrungLe,HaNguyen,andNikolaSpatafora

AuthorizedfordistributionbyLucEyraudandMercedesGarcia-EscribanoAugust2024

IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicit

commentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseofthe

author(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.

ABSTRACT:Weuseanewdatasettoestimatetheimpactoftemperatureoneconomicactivityatamore

geographicallyandtemporallydisaggregatedlevelthantheexistingliterature.Analyzing30-kilometergridcellsatamonthlyfrequency,temperaturehasanegative,highlystatisticallysignificant,andquantitativelylarge

effectonoutput:a1°Cincreaseinmonthlytemperatureisassociatedwitha0.77percentreductionin

nighttimelights,aproxyforlocaleconomicactivity.Theeffectsofevenatemporaryincreaseintemperaturepersistforalmostoneyearaftertheshock.Increasesintemperaturehaveanespeciallylarge,negativeimpactongrowthinpoorercountries,indicatingthattheyaremorevulnerabletotheimpactofclimatechange.

RECOMMENDEDCITATION:Hoang,Da,DuongTrungLe,HaNguyen,andNikolaSpatafora.2024.

“SheddingLightontheLocalImpactofTemperature”.IMFWorkingPaperNo.24/178,InternationalMonetaryFund,Washington,DC.

JELClassificationNumbers:

Q54;O47

Keywords:

Climatechange;temperature;economicgrowth;nighttimelights

Author’sE-MailAddress:

dhoang@,

dle6@,

hnguyen7@,

nspatafora@

IMFWORKINGPAPERSHigh-FrequencyImpactofTemperatureonEconomicActivity

INTERNATIONALMONETARYFUND3

WORKINGPAPERS

SheddingLightontheLocalImpactofTemperature

PreparedbyDaHoang,DuongTrungLe,HaNguyen,andNikolaSpatafora

1

1Theauthor(s)wouldliketothankGuyAmouzouAgbe,RobertBeyer,LucEyraud,VladimirKlyuev,MercedesGarcia-Escribano,YaoJiaxiong,ThibaultLemaire,EmanueleMassetti,LucaRicci,GregorSchwerhoff,CanSever,FilipposTagklis,andseminar

participantsattheIMFandWorldBankforvaluablecommentsandfeedback.

IMFWORKINGPAPERSHigh-FrequencyImpactofTemperatureonEconomicActivity

INTERNATIONALMONETARYFUND4

Contents

I.Introduction 5

II.DataandEmpiricalSpecification 7

II.1Data 7

II.2TheoreticalMotivation 9

II.3EmpiricalSpecification 10

ContemporaneousEffectsofTemperature 10

DynamicEffectsofTemperature 11

III.Findings 12

III.1ContemporaneousEffects 12

III.2DynamicEffects 15

IV.RobustnessChecks 16

V.Conclusions 19

References 21

Appendix 23

A.SupplementaryResults 23

B.DataDescription:theGoogleEarthEngine 29

IMFWORKINGPAPERSHigh-FrequencyImpactofTemperatureonEconomicActivity

INTERNATIONALMONETARYFUND5

I.Introduction

Climatechangeisakeyglobalchallenge.Temperaturesarerising.Droughts,floods,wildfires,andmassive

stormsareoccurringmorefrequentlywithdevastatingeffects.Understandingtheimpactofrisingtemperature,themostbasicmanifestationofclimatechange,oneconomicactivityisfundamentaltoadaptationand

mitigationefforts.

Theeconomicliteraturehasgenerallyfoundthathighertemperaturehurtseconomicactivity,particularlyforhotandpoorcountries.Theearlyliteratureexaminedtherelationshipbetweenaveragetemperatureandaggregateeconomicvariables(e.g.,SachsandWarner,1997;Gallup,Sachs,andMellinger,1999).Itfoundthathotter

countriestendtobepoorer.However,thisrelationshipmightbedrivenbyomittedvariables,suchascountryinstitutions.Themorerecentliteratureusesfluctuationsintemperaturewithinacountrytocontrolforslow-

movingcountrycharacteristics.

2

Itfindsthathighertemperaturereduceseconomicgrowthinpoorcountries(Delletal.,2012;Acevedoetal.,2020)andtheUnitedStates(Colacitoetal.,2019).Thenegativeeffectsrunthroughreducedtotalfactorproductivitygrowth(LettaandTol,2019),reducedinvestmentandlabor

productivity(Acevedoetal.,2020;KalkuhlandWenz,2020)andreducedsectoralproductivity(LeporeandFernando,2023).Burkeetal.(2015)documentthenon-lineareffectoftemperature:economicgrowthriseswithaverageannualtemperatureuntilapproximately13°C,atwhichpointtherelationshipreverses.

Theliteraturetypicallyexaminestheimpactofaverageannualtemperatureonannualeconomicoutcomes.

3

However,temperatureatagivenlocationmayvarygreatlyacrossseasons.Forinstance,temperaturein

Washington,D.C.,variesgreatlywithintheyear

(AppendixFigure1)

;theannualaverageofcloseto21°Ccan

bemisleadinglymoderatebecauseWashingtonD.C.hasacoldwinterandahotsummer.Averaging

temperatureovertheentireyearmaythereforemissimportantfluctuations,suggestingtheimportanceofcarryingouttheanalysisatafrequencythatishigherthanannual.

Similarly,averagingtemperatureoveralargegeographicarea,suchasacountryorevenaprovince,maymissimportantdetails.Bothtemperatureanditsfluctuations,andthestructureofeconomicactivityanditssensitivitytotemperature,mayvarygreatlyacrossspace.Hence,examiningtheimpactoftemperaturerequiresanalysisatamoregranularspatiallevel.

Inthispaper,weexplicitlyacknowledgethatbothtemperatureanditseffectsarehighlyheterogeneousand

localized.Tothatend,weanalyzethelinkbetweentemperatureandeconomicactivityusingdatafinely

disaggregatedacrossbothtimeandspace.Weexaminetheimpactofmonthlytemperatureoneconomic

activitywithin30-kilometerby30-kilometergridcells.Goingtothegridandmonthlylevelsallowsustoestimatetheeffectsoftemperaturemorepreciselyacrossdifferentclimatezones.Thisanalysisoffersnewinsightsto

2See,forinstance,Kahnetal.,2021;Acevedoetal.,2020;Colacitoetal.,2019;LettaandTol,2019;Cashinetal.,2017;andDelletal.,2012.SeealsothesurveysbyChangetal.(2023),Auffhammer(2018),andDelletal.(2014).

3See,forinstance,Bergetal.,2023;Newelletal.,2021;Acevedoetal.,2020;KalkuhlandWenz,2020;Burkeetal.,2015;Delletal.,2012;andDeschênesandGreenstone,2007.Akyapietal.(2022)identify,amongalargenumberofannualclimateindicators,thosemostpredictiveofeconomicdamage.

INTERNATIONALMONETARYFUND6

complementtheexistingliterature,whichfocusesonannualaveragetemperatureandeconomicoutcomesatthecountrylevel.

4

Granularanalysescomewiththeirchallenges.Thefirstandmostobviousissueisthelackofeconomicdataatthelocallevel.

5

Inthispaper,localeconomicactivityisproxiedbytheintensityofnighttimelights.Alarge

literaturepointstothevalueofnighttimelightsasanindicatorforeconomicactivity(ChenandNordhaus,2011;HuandYao,2022;Martinez,2022;Asheretal.,2021).Relatedtoourpaper,Felbermayretal.(2022)examinetheimpactofstorms,excessiveprecipitation,droughts,andcoldspellsoneconomicactivityproxiedby

nighttimelights.

Overall,wefindanegativeandhighlystatisticallysignificanteffectoftemperatureonthegrowthofnighttimelights.Relatedly,ourheterogeneityanalysisindicatesthatthissignificanteffectoftemperatureisfoundalsoinlocationswithrelativelyloweraveragetemperatures.Thissignificanteffectisobservedincountriesacrosstheincomedistribution,althoughitisstatisticallylargerinpoorerareas.

Underourfullyspecifiedmodel,a1°Cincreaseinmonthly-averagetemperaturereducesthe

contemporaneousyear-over-yeargrowthofmonthlynighttimelightsby0.77percent.CommonestimatesoftheelasticityofnighttimelightswithrespecttoGDPinturnsuggestthiswouldbeassociatedwithareductionin

GDPgrowth,relativetobaseline,ofbetween0.5and0.77percentagepoints.Thismagnitudeislargerthanthefindingsoftenreportedintheliteraturethatusesmoreaggregateddata.

6

Ourestimatesaremoreinlinewith

Acevedoetal.(2020),whichreportthata1°Cincreaseinannualtemperaturereducesoutputgrowthby0.6–1.2percentagepointsforlow-andmiddle-incomecountries,althoughtheeffectislargelynotstatistically

significantforadvancedeconomies.

Akeyquestionishowtheseeffectschangeacrossdifferentclimatezones.Sofar,thereislittleevidence

becauseofthelackofgranulardata.Theliteraturetendstofindstrongereffectsinpoorercountriesandin

warmerareas(Delletal.,2012;Burkeetal.,2015,Acevedoetal.,2020).Inanearlyandinfluentialpaper,

Burkeetal.(2015),usingcountry-averageandannualaveragetemperaturedata,findunevenimpactsof

temperatureonGDPgrowth:positiveforcountrieswithaveragetemperaturebelow13°C,andnegativefor

countrieswithaveragetemperaturesabovethat.However,recentresearchsuggeststhatthisfindingofuneveneffectsisnotrobust.Forinstance,Nathetal.(2023)showthattheuneveneffectsdisappearwhencontrollingforlaggedtemperature.Thefragilefindingswithannual-andcountry-averagedataonceagainindicatethe

needformoregranularanalyses.

Animportantandrelatedissueisthattheliteratureusingaggregatedataisoftenincapableofdisentanglingtheimpactofrisingtemperatureinhotlocationsfromthatinpoorerregions;thatis,itisuncleariftheimpact

reflectsahotclimateorloweconomicdevelopment.Werevisitthisdebateinthispaper.Usingverygranulardata,wecancontrolforbothfactors.Weassignanygivengridcellinanygivenmonthtoaspecific

4Nguyen(2024)examinestheimpactofUS-countylevelseasonaltemperatureoncountry-leveljobgrowth.Thepaperfindsthatahottersummerhurtsjobgrowth,butawarmerwinterhelps.

5Someeffortshavebeenmadetocomputecell-levelcrosscellproduct(seeGeigeretal.,2017,butonlyfor10-yearincrements).

6Forexample,usingcountryaverageandannualaveragetemperaturedata,Delletal.(2012)findthatonaverage,acrossall

countries,a1°Cincreaseintemperaturereducesacountry’sannualGDPgrowthby0.3percentagepoints,andthedeclineisnotstatisticallysignificant.

INTERNATIONALMONETARYFUND7

temperaturebin,dependingonitsaveragemonthlytemperatureinthatmonth.

7

Wethenexaminetheimpactoftemperatureacrossdifferenttemperaturebins.Wefindthattheimpactoftemperatureongrowthinnighttimelightsisnegativeandstatisticallysignificant,exceptincolderareas.Inotherwords,ourfindingssupportthe

commonhypothesisthattemperatureaffectseconomicactivitymorenegativelyinrelativelywarmareas.

Atthesametime,wefindthat,whencontrollingfortemperature,incomelevelsalsomatter.Withineach

temperaturebin,wefindthattemperaturehasmorenegativeeffectsingridcellsthatarepartofalow-ormiddle-incomecountry.Ourresultthereforesuggeststhattemperaturedoeshavearelativelymoresevereeffectoneconomicactivityinpoorerareas.

Anotherimportantpointrelatestothelong-termeffectsoftemperatureoneconomicactivity.Couldatemporarytemperatureshockinducelong-lastinggrowtheffects?Toanswerthisquestion,werelyonlocalprojection

modelstoanalyzetheimpulseresponseofeconomicactivity,proxiedbynighttimelights,followinganincreaseintemperature.Wefindthatevenatemporaryshocktomonthlytemperatureshasarelativelypersistenteffectoneconomicactivity,whichremainsstatisticallysignificantforalmostayear,althoughtheeffectwanesasthetimehorizonexpands.

II.DataandEmpiricalSpecification

II.1Data

Dataforthemainanalysisvariables—temperatureandnighttimelights—arecollectedfromtwomainsources.ThetemperatureandprecipitationdataarecollectedfromtheERA5MonthlyAggregatesdataset(C3S,2017;Gorelicketal.,2017).ERA5isthefifth-generationglobalclimatereanalysisproducedbyEuropeanCentreforMedium-RangeWeatherForecasts(ECMWF).ERA5MonthlyprovidesaggregatedvaluesforeachmonthforsevenERA5climatereanalysisparameters:2mairtemperature(inKelvin,convertedtoCelsius),2mdewpointtemperature,totalprecipitation(inmeters),meansealevelpressure,surfacepressure,10mu-componentofwindand10mv-componentofwind.ThedatasetisavailablefromJanuary1979toJune2020andcoverstheentireglobewithagridcellresolutionofabout30kilometers(thatis,withdataon30-kilometerwidesquare

areas).Wecollectmonthlymaximumairtemperatureat2mandmonthlytotalprecipitation.

WecollectnighttimelightsdataasmonthlyaverageradiancecompositeimagesfromtheVisibleInfrared

ImagingRadiometerSuite(VIIRS)Day/NightBand(DNB)(Elvidgeetal.,2013).ThisversionisanalternativeconfigurationoftheVIIRSDNB,whichcorrectsforstraylight(dataneartheedgesoftheswathareexcluded).Dataimpactedbycloudcoverisalsoexcluded.And,sincetheusefulnessofnighttimelightstoproxyeconomicactivitycanbeaffectedbysnow,weexcludeobservationswheresnowcover(collectedfromERA5dataset)

exceeds50percentinamonth.

8

ThedatasetisavailablefromJanuary2014toJune2022withagridcell

7Specifically,eachgridcellinanygivencalendarmonthisassignedtooneofsixpossibletemperaturebins(lessthan0°C,0°C–10°C,10°C–20°C,20°C–30°C,30°C–35°C,andgreaterthan35°C),dependingonitsaveragetemperatureinthatmonth.Forexample,thegridcellscoveringWashingtonD.C.haveanaverageJanuarytemperaturebetween0°Cand10°C,andanaverageJunetemperaturebetween20°Cand30°C.

8Snowcoverinthepresenceofmoonlightcanmakeanareaappearverybright(Zhangetal,2023).Theresultsarealsorobusttoexcludingobservationswheresnowcoverexceeds40percentor60percent.

IMFWORKINGPAPERSHigh-FrequencyImpactofTemperatureonEconomicActivity

INTERNATIONALMONETARYFUND8

resolutionofabout500meters.Tomergetemperatureandprecipitationdatawithnighttimelightdata,we

collectthemonthlydatafortheoverlappingperiodfromJanuary2014toJune2020(78months)usinggridcellsof30kilometersby30kilometers,whichareavailableforbothdatasets.Finally,wecollectgrid-cellleveldataonpopulation(CIESINandSEDAC,2018).AlldataarecollectedviaGoogleEarthEngine.

Table1

presentssummarystatisticsofnighttimelightsandtemperature.Acrossallcell-monthobservations,

themedianyear-on-year(YoY)growthrateofnighttimelightisapproximately4percent,withsignificant

variation(94percentagepoints).Theaveragetemperatureis28.5°C.Temperaturehasincreasedby0.10°C(or0.46percent)annuallyforthetypical(median)cell.

Figure1

illustratesthespatialandseasonalvariationsintemperatureacrosstheglobe;bluetonesdenotelowertemperatures,redtoneshighertemperatures.

Table1:SummaryStatistics.

YoYGrowthinNighttimeLights

Temperature(°C)

YoYChangeinTemperature(°C)

Totalmonthly

precipitation(meter)

Observations(cell

Xmonth)

8,438,699

12,517,650

10,585,842

12,517,650

Numberofcells

197,166

200,648

200,648

200,648

Mean

0.1167

28.5310

0.1293

0.0758

Std

0.9360

9.3457

3.0510

0.1017

Min

-24.2884

-9.9429

-22.2279

0.0000

P1

-2.5226

4.8381

-8.1817

0.0000

P5

-1.1159

10.6698

-4.7809

0.0000

P25

-0.2597

22.7557

-1.4121

0.0089

Median

0.0405

30.3510

0.0980

0.0434

P75

0.4174

34.8485

1.6333

0.1001

P95

1.6501

42.0870

5.2489

0.2760

P99

3.1880

45.4611

8.6147

0.4515

Max

14.4289

53.7555

20.9840

4.0605

Note:Summarystatisticsofnighttimelight(YoYgrowth)andtemperature(inlevelandYoYchange).Unitsarepercentagepoints,exceptfortherows“Observations(cellXmonth)”and“Numberofcells”.Eachgridcellisa30-kilometerby30-kilometersquare.MonthlyobservationsfromJanuary2014toJune2020(78months).

IMFWORKINGPAPERSHigh-FrequencyImpactofTemperatureonEconomicActivity

INTERNATIONALMONETARYFUND9

Figure1:Averagetemperaturebyselectedcalendarmonth,acrosstheworld.JanuaryApril

JulyOctober

Source:C3S,2017;Gorelicketal.,2017;Elvidgeetal.,2013;andIMFstaffcalculations.

Note:Themapsexcludecellswithzeroormissingnighttimelights.

II.2TheoreticalMotivation

Tostudytherelationshipbetweentemperatureandeconomicgrowth,wefirstpresentatheoreticalmotivation,basedonDelletal.(2012).Thiswillinformoursubsequentempiricalspecification.Leteconomicactivity

changeaccordingto

yt=eβTtA(1)

whereYdenoteseconomicactivity,Tdenotestemperature,Aisaproductivityterm,andtdenotesthemonth.Here,temperaturehasadirectleveleffectoneconomicactivity,representedbyβ.Inaddition,productivity

affectseconomicactivity,asrepresentedbyα.Andtemperaturemayalsoaffectannualproductivitygrowth:

log(At)−log(At−12)=g+λTt+σTt−12(2)

whereλdenotestheshort-runeffectoftemperatureonproductivitygrowth,and(λ+σ)thelong-runeffect.Thespecificationallowsforeitherpositiveornegativeeffectsoftemperatureonproductivitygrowth.Takinglogs

andyear-over-yearfirstdifferencesof

(1)

,andcombiningwith

(2)

,yields

log(yt)−log(yt−12)=β(Tt−T12)+α(g+λTt+σTt−12)

IMFWORKINGPAPERSHigh-FrequencyImpactofTemperatureonEconomicActivity

INTERNATIONALMONETARYFUND10

⇒log(yt)−log(yt−12)=αg+(β+αλ)Tt+(ασ−β)Tt−12(3)

So,temperatureaffectsthegrowthrateofoutput;theshort-rungrowtheffectisgivenby(β+αλ),andthelong-runeffectby[α(λ+σ)].Equation

(3)

formsthebasisfortheempiricalspecification.

II.3EmpiricalSpecification

ContemporaneousEffectsofTemperature

Inlinewiththetheoreticalmotivation,themainempiricalregressiontakestheform

log(Lc,t)−log(Lc,t−12)=β0+β1Tc,t+β2Tc,t−12+σ1pc,t+FEt+FEc,m+Ec,t(4)

whereLdenotesnighttimelights,Tdenotestheaveragetemperatureofthemonth,cindexesthe(30-kilometerby30-kilometer)gridcells,andtdenotesthemonth.Changesinnighttimelightsproxyforchangesineconomicactivity.FEtaremonth-specificfixedeffectsthatcaptureglobalfactorsaffectingthegrowthofnighttimelightsacrossallcellsinaparticularmonth.FEc,marecell–by–calendar-monthfixedeffects,whichcontrolsforthecalendar-month-specificgrowthofnighttimelightsforeachgridcell.

9

Theyaddressdifferentseasonalityfordifferentcells.Forinstance,theyhandlethefactthatthenorthernandsouthernhemisphereshaveoppositeseasons,orthatnighttimelightsmaygrowespeciallyfastduringsummermonthsforcellslocatedincoldclimatezones.Ourpreferredspecificationfurthercontrolsforprecipitationasacontemporaneouscovariateclimatefactor,wherepc,tdenotesavectorcomprisingtwodummiesindicatingwhetheragridcellexperiencedabnormallyhigh(onestandarddeviationabovethelong-termaverageprecipitationinaparticularcalendarmonth)orlow(onestandarddeviationbelow)inmontht.

10

Weuseequation

(4)

toanalyzethecontemporaneouseffectoftemperatureontheyear-over-yeargrowthofnighttimelights.Ttisthemainexplanatoryvariable,capturingmontht’stemperature.Thekeycoefficientofinterestisβ1.

11

Weviewtheestimatedeffectoflocaltemperatureonlocaleconomicactivityascausal,sincelocaltemperatureisbestviewedasexogenouswithrespecttolocaleconomicactivity.Inparticular,thereislittleriskofreversecausalityfromlocaleconomicactivitytolocaltemperature.

12

ThecontrolvariableTt−12capturesthe

9Essentially,FEc,mincludes12fixedeffectspergridcell,eachcorrespondingtoacalendarmonthinayear.

10Morespecifically,theprecipitationdummiesaredefinedasfollow:

High_Pc,t=1ifpc,t>average(pc,t)+std(pc,t),and=0otherwise;andLow_Pc,t=1ifpc,t<average(pc,t)−std(pc,t),and=0otherwise;

wherepc,tdenotesthetotalprecipitationlevelofthemonth,andbothaveragesandstandarddeviationsarecell–calendar-monthspecific.

11Whichcorrespondstotheparameters(β+αλ)inequation(3).

12Inprinciple,someomitted,climate-relatedvariablemightbothbesystematicallycorrelatedwithlocaltemperature,andaffectlocaleconomicactivity.Themostobvioussuchvariableisprecipitation;asdiscussed,wethereforecontrolforthis.

IMFWORKINGPAPERSHigh-FrequencyImpactofTemperatureonEconomicActivity

INTERNATIONALMONETARYFUND11

impactoftemperatureinthesamemonthofthepreviousyear.Standarderrorsareclusteredattheprovince(specifically,first-tieradministrative)levelineachcountrywheredataareavailable.

Wethenextendtheanalysis,investigatingseparatelytheimpactofpositiveandnegativechangesintemperature:

log(Lc,t)−log(Lc,t−12)=β0+β1Tc,t+β2Tc,t−12+y1psitivec,t+y2Tc,t×psitivec,t+σ1pc,t+FEt+FEc,m+Ec,t(5)

wheretheindicatorPositivec,t=1if(Tc,t−Tc,t−12)≥0,and=0otherwise.Here,thecontemporaneouseffectoftemperatureontheyear-over-yeargrowthofnighttimelightsisgivenbyβ1fornegativetemperatureshocks,and(β1+y2)forpositivetemperatureshocks.

Inananalogousmanner,wealsoinvestigateseparatelytheimpactoflargepositiveandlargenegativechangesintemperature,definedasthosechangesthataremorethanonestandarddeviationawayfromtheaveragetemperaturechange:

log(Lc,t)−log(Lc,t−12)=β0+β1Tc,t+β2Tc,t−12+δ1Tc,t×Higℎpsitivec,t+δ2Tc,t×HigℎNegativec,t+μ1Higℎpsitivec,t+μ2HigℎNegativec,t+σ1pc,t+FEt+FEc,m+Ec,t(6)

wheretheindicators

HighPositivec,t=1if(Tc,t−Tc,t−12)>average(Tc,t−Tc,t−12)+std(Tc,t−Tc,t−12),and=0otherwise;and

HighNegativec,t=1if(Tc,t−Tc,t−12)<average(Tc,t−Tc,t−12)−std(Tc,t−Tc,t−12),and=0otherwise;andbothaveragesandstandarddeviationsarecell-specific.

DynamicEffectsofTemperature

Wealsowanttoexaminethedynamicimpactoftemperatureonnighttimelights,includingthelong-runeffect.Tothisend,weemployaflexibleempiricalspecification,adoptingthelocallinearprojectionsmethodology

(Jordà,2005).Impulseresponseanalysisusinglocalprojectionshasbecomeapopularalternativetothe

traditionalstructuralvectorautoregressive(SVAR)models.Inparticular,localprojectionsofferthefollowing

benefitsoverthetraditionalapproach:(i)theyaremorerobusttomisspecification,sincetheydonotconstraintheshapeofimpulseresponsefunctions;(ii)theycanaccommodatenonlinearandmoreflexiblespecifications;

and(iii)theyareeasiertoimplement,becausetheyonlyrequirestandardlinearregressions(Jordà,2005).Ourspecificationtakestheform:

log(Lc,t+ℎ)−log(Lc,t−1)=αc,ℎ+βℎTc,t+Xc,t+σ1pc,t+FEt+FEc,m+Ec,t(7)

wherehdenotestheprojectionhorizon,andlog(Lc,t+ℎ)−log(Lc,t−1)capturesthegrowthrateofnigh

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论