IMF-超越年平均值：季节性气温对美国各县就业增长的影响（英）

BeyondtheAnnualAverages:Impactof

SeasonalTemperatureonEmploymentGrowthinUSCountiesHaNguyenWP/23/142IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.2023JUN©2023InternationalMonetaryFundWP/23/142IMFWorkingPaperInstituteofCapacityDevelopmentBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUS

CountiesPreparedbyHaNguyen*Authorizedfordistributionby

MercedesGarcia-EscribanoJune2023IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.

TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,or

IMFmanagement.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:Impactof

SeasonalTemperatureonEmploymentGrowthinUSCounties.

IMF

Working

Papers,

2023/142JELClassificationNumbers:Keywords:C33,C55,E24,O44,Q54Climatechange;temperature;employment;UScountiesHnguyen7@Author’sE-MailAddress:IthankRabahArezki,BasBakker,AdolfoBarajas,RudolfsBems,AndrewBerg,MaiDao,MercedesGarcia-Escribano,HuiHe,ToanPhan,KoralaiKirabaeva,VladimirKlyuev,AntonKorinek,RuyLama,EmanueleMassetti,RodolfoMaino,JoeProcopio,NoomanRebei,NikolaSpataforaandMaryamVaziriforveryhelpfulcommentsandfeedback,andRuchunLiforeditorialhelp.IamgratefultoBerkayAkyapiandEmanueleMassettiforintroducingtomeclimatedataviaGoogleEarthEngine.WORKINGPAPERSBeyondtheAnnualAverages:Impactof

SeasonalTemperatureonEmploymentGrowthinUSCountiesPreparedby

HaNguyen11

IthankRabahArezki,BasBakker,AdolfoBarajas,RudolfsBems,AndrewBerg,MaiDao,MercedesGarcia-Escribano,HuiHe,ToanPhan,KoralaiKirabaeva,VladimirKlyuev,AntonKorinek,RuyLama,EmanueleMassetti,RodolfoMaino,

JoeProcopio,NoomanRebei,NikolaSpataforaandMaryamVaziriforveryhelpfulcommentsandfeedback,andRuchunLiforeditorialhelp.IamgratefultoBerkayAkyapiandEmanueleMassettiforintroducingtomeclimatedataviaGoogleEarthEngine.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesContentsI.

Introduction

4II.

ATheoreticalFramework

6III.

DataandEmpiricalSpecification9Data9EmpiricalSpecifications

10IV.

MainFindings

11AnnualRegressions

11MainFindings12V.

SummerandWinterImpactsAcrossState’sClimate15VI.

OntheMechanismsoftheSummerTemperatureEffects18VII.

OntheMechanismsoftheWinterTemperatureEffects21VIII.Effects

ofTemperaturebyDecade

23IX.

RobustnessChecks26NotUsingCountyEmploymentWeights

26DroppingExtremeEmploymentGrowth27DroppingRecessionQuarters27ControllingforNaturalDisasters

28ControllingforPrecipitation

29X.

Conclusions30References30FiguresFigure1:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature14Figure2:AverageSummerImpactbyState16Figure3:AverageWinterImpactbyState

17Figure4:TheEffectonEmploymentGrowthofHigherSummerTemperaturebySector19Figure5:TheEffectofHigherWinterTemperaturebySector21Figure6:AverageEmploymentSharesintheSummerandWinterinaCounty

23Figure7:AverageAnnualIncreaseinSummerTemperaturebyStateover1990and2021

24Figure8:DynamicImpactofYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(RegressionsareUnweighted)262IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesFigure9:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(TopandBottom1percentileofEmploymentGrowthDataareDropped)27Figure10:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(EmploymentGrowthDataforRecessionaryQuartersareDropped)28Figure11:DynamicImpactonYoYEmploymentGrowthtoaOneDegreeFahrenheitHigherTemperature(ControllingforNaturalDisasters)29Figure12:DynamicImpactonYoYEmploymentGrowthtoaOneDegree

FahrenheitHigherTemperature(ControllingforPrecipitation)29TablesTable1:SummaryStatistics

10Table2:ImpactofAnnualAverageTemperatureonYoYGrowthofAnnualAverageEmployment

11Table3:ImpactofTemperatureonYoYPrivateEmploymentGrowth

12Table4:RelationshipbetweenEmploymentEffectandaState’sClimate18Table5:ImpactofTemperaturebyDecade253IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesI.

IntroductionClimate

change

is

the

biggest

challenge

for

humankind.

Temperature

is

rising.

The

global

average

temperatureisalreadyabout1.2degree

Celsiushigherthanthepre-industriallevel.Droughts,wildfires,andmassivestormsare

starting

to

occur

more

frequently

with

devastating

effects.

Understanding

the

impact

of

rising

temperature,the

most

basic

manifestation

of

climate

change,

on

economic

activity

is

fundamental

to

adaptation

and

mitigationefforts.The

economic

literature

has

generally

found

that

higher

temperature

hurts

economic

activity.

Early

literatureexamines

the

relationship

between

average

temperature

and

aggregate

economic

variables

(e.g.,

Sachs

andWarner,

1997;

Gallup,

Sachs,

and

Mellinger,

1999).

It

finds

that

hotter

countries

tend

to

be

poorer.

However,

thisrelationshipmightbedrivenbyomittedvariablessuchascountryinstitutions.Recentliteratureusesfluctuationsin

temperature

within

a

country

or

a

region

to

control

for

slow-moving

characteristics

(see

for

example,

Dell

et

al.,2012;Cashin

etal.,2017;Colacitoetal.,2019;LettaandTol,2019;Acevedoetal.,2020;Kahnetal.,2021).1

Itfinds

that

higher

temperature

reduces

the

economic

growth

of

poor

countries

(Dell

et

al.,

2012;

Acevedo

et

al.,2020)

and

the

US

(Colacito

et

al.,

2019).

The

negative

effects

run

through

reduced

total

factor

productivity

growth(Letta

and

Tol,

2019),

and

reduced

investment

and

labor

productivity

(Acevedo

et

al.,

2020;

Kalkuhl

and

Wenz,2020).

Burke

et

al.

(2015)

document

the

non-linear

effect

of

temperature:

economic

growth

rises

with

averageannualtemperatureuntilaround13

degreesCelsiusanddropsafterthat.This

paper

examines

the

dynamic

effects

of

temperature

on

the

private

sector’s

employment

growth

at

a

locallevel,

namely

US

county,

and

high

frequency,

namely

quarterly.

Going

to

the

county

and

quarterly

levels

allowsfor

more

precise

temperature

measurement.

Therefore,

it

could

estimate

the

effects

of

temperature

moreprecisely

and

uncover

the

subtle

effect

of

seasonal

temperature.

This

paper

focuses

on

job

growth

as

the

maineconomic

outcome.

Jobs

are

featured

prominently

in

the

US's

discussions

of

climate

change

mitigations.

Manyworry

that

climate

change

mitigation

efforts

willhurtjobs

(AFP,

2022).

This

paper

finds

that

higher

temperature,onaverage,hurtsjobsintheUS.Using

data

between

1990

and

2021,

this

paper

discovers

opposing

effects

of

higher

temperature

in

the

winterand

summer.

On

average,

within

a

county,

higher

summer

temperature

reduces

private

sector

employmentgrowth,

while

higher

winter

temperature

increases

it.

The

impacts

of

higher

temperature

in

mild

seasons,

fall

andspring,

are

statistically

insignificant.

The

findings

showcase

the

heterogenous

and

nuanced

effects

oftemperatureshocks.This

paper

finds

interesting

dynamic

effects

of

seasonal

temperature.

Higher

summer

temperature

hurtseconomic

activity

in

the

current

and

following

quarters.

A

temporary

one-degree

Fahrenheit

(F)

higher

summertemperature

decreases

year-over-year

(YoY)

employment

growth

of

that

summer

by

0.063

percent.

It

alsodecreases

YoY

employment

growth

of

the

following

fall

and

winter

by

0.08

and

0.075

percent,

respectively.

Incontrast,

the

positive

impact

of

higher

temperature

in

the

winter

is

more

short-lived.

A

temporary

one-degreeFahrenheit

warmer

winter

boosts

YoY

employment

growth

in

that

winter

by

0.05

percent

but

has

statisticallyinsignificant

effects

on

employment

growth

in

the

following

spring

and

summer.

In

sum,

the

negative

impacts

of1

AlsoseerecentsurveysbyDelletal.(2014)andAuffhammer(2018)IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountieshigher

temperature

in

the

summer

are

larger

and

more

persistent

than

the

positive

impacts

of

higher

temperatureinthewinter.Therefore,theaverageemploymenteffectofhighertemperatureacrossseasonsisnegative.The

economic

literature

typically

examines

the

impact

of

annual

average

temperature

on

annual

economicoutcomes

(e.g.,

see

Deschênes

and

Greenstone,

2007;

Dell

et

al.,

2012;

Burke

et

al.,

2015;

Acevedo

et

al.,

2020;Kalkuhl

and

Wenz,

2020;

Akyapi

et

al.,

2022).

However,

since

temperature

can

vary

greatly

within

a

year,

fromfreezing

winters

to

scorching

summers,

this

paper

argues

that

seasonal

temperature

is

a

better

approximation

ofweather

than

annual

temperature.2

More

importantly,

the

economic

structures

of

different

seasons

could

be

verydifferent.

For

example,

construction,

travel,

and

tourism

are

expected

to

rise

in

summer

and

fall

in

winter.Therefore,

examining

the

effects

of

seasonal

temperature

on

seasonal

economic

activity

could

offer

new

insightsto

complement

the

existing

analyses

using

annual

average

temperature

and

annual-average

economicoutcomes.

In

addition,

working

with

the

country-average

temperature

is

also

not

ideal

since

even

within

a

country,temperature

can

vary

greatly.

A

country,

or

even

a

US

state,

may

have

several

climate

zones.

A

case

of

localizedtemperature,suchasatthecountylevel,canbemadehere.Nevertheless,granularanalysescomewiththeirchallengesandissues.First,andthemostobvious

issueisthelack

of

high-frequency

economic

data

at

the

local

level.

One

reason

why

employment

growth

is

chosen

as

themain

variableof

interest

isthat

the

US

has

reliablequarterly

data

atthe

countylevel(more

on

thatin

sectionIII).The

second,

and

more

conceptual

issue

is

labor

mobility

at

the

local

level.

At

the

country

level,

labor

mobility

isrelatively

restricted.

At

least

in

the

short-run,

workers

have

to

stay

in

a

country

and

try

to

find

work

with

atemperature

shock.

But

an

analysis

at

the

local

level,

such

as

US

county,

implies

labor

mobility

is

much

lessrestricted.

People

could

move

in

and

out

of

a

county

to

work

in

another

county

in

response

to

a

temperatureshock.Therefore,theeffectsoftemperature

onemploymentwithlabormobilitycanbelargerthanwithout.Deryugina

and

Hsiang

(2017)

and

Colacito

et

al.

(2019)

examine

the

impacts

of

seasonal

temperature

at

UScounty

and

state

levels,

respectively.

However,

they

still

use

annual

economic

outcomes,

which

could

maskinteresting

dynamic

effects

of

seasonal

temperature.

This

analysis

complements

their

analyses

by

not

focusingon

annual

economic

outcomes

but

on

the

high-frequency

impact

of

temperature

on

quarterly

employment

growthin

US

counties.

By

adopting

this

local

and

high-frequency

empirical

framework

together,

it

unveils

novel

andinteresting

dynamic

effects

of

seasonal

temperature.

It

could

also

shed

lighton

the

mechanisms

by

documentingthe

effects

in

each

industry

and

how

they

propagate

over

the

next

quarters.

In

other

words,

by

observingtemperature’simpactsondifferentsectorsatahighfrequency,insteadofbeingdilutedbytheannualaverages,thepapercanprovideadditionalinsightsintothemechanisms.This

paperfinds

thatthe

negative

effects

of

a

hotter

summerare

pervasive

andpersistent

in

many

sectors:

mostsignificantly

in

“Construction”

and

“Leisure

and

Hospitality”

but

also

in

“Trade,

Transport

and

Utilities”

and“Financial

Activities.”

Employment

growth

in

these

sectors

may

get

directly

hit

by

rising

temperature.

It

is

alsopossible

that

some

of

the

lower

employment

growth

is

indirectly

affected

due

to

input-output

linkages

betweendifferent

sectors

or

the

aggregate

demand

effect.

For

example,

job

growth

in

“Financial

Activities”

could

bedampened

due

to

a

slower

financial

service

demand

from

“Construction.”

In

contrast,

the

positive

effects

of

a2

Forexample,highestdailytemperatureinWashingtonD.C.

(UnitedStates)in2021rangesfromthemid-30sFahrenheitinthewintertothemid-90sFahrenheitinthesummer.TheaverageannualtemperatureforWashingtonD.C.isabout

70-degreeFahrenheit.Ifweusethisannualaverageof70-degreeFahrenheitinouranalyses,wemightbemistakenthatWashingtonD.C.’sweatherismore

moderatewhileinfact,ithasacoldwinterandahotsummer.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountieswarmer

winter

are

less

pervasive,

only

in

“Construction,”

“Leisure

and

hospitality,”

and

“Natural

Resources

andMining.”Itisalsomoreshort-lived.The

richness

of

county-level

data

allows

for

the

examination

of

the

effect

by

US

state

–

which

is

another

importantcontribution.

This

paper

discovers

a

relationship

between

the

negative

effects

of

a

hotter

summer

with

a

state’ssummer

climate:

hotter

states

have

more

severe

negative

impacts

of

higher

summer

temperature.

Some

coolerstates

(e.g.,

Alaska

and

Massachusetts)

even

benefit

from

the

higher

summer

temperature.

On

the

other

hand,therelationshipbetweentheimpactofhigherwintertemperatureandastate’swinterclimate

isnotasclear.An

important

point

of

discussion

is

how

would

these

findings

on

short-term

responses

help

us

predict

the

long-term

responses

to

hotter

climates?

It

has

been

argued

that

the

short-run

responses

to

temperature

fluctuationsare

likely

not

the

same

as

the

long-run

responses

to

climate

change

(see

the

discussion

in

Burke

and

Emerick,2016,

for

example).

This

is

a

reasonable

argument.

First,

the

future

magnitude

of

climate

change

is

uncertain,dependingon

humankind’s

mitigationefforts.Inaddition,therecouldbearoleofadaptation.Adaptationefforts,such

as

more

widespread

use

of

drought-resistant

seeds

or

air-conditioning,

might

soften

the

impact

of

risingtemperature

in

the

future.

If

so,

the

short-run

impacts

may

overstate

the

long-run

impacts

of

climate

(see

alsoMassetti

and

Mendelsohn,

2018).

Conversely,

the

rising

temperature

may

cause

permanent

effects

onemployment

(such

as

emigration

out

of

the

hot

areas).

In

that

case,

the

short-term

impacts

of

temperaturefluctuationmightunderstatethelong-runimpactsofclimatechange.This

paper

contributes

to

this

discussion

with

two

sets

of

findings.

First,

a

warmer

winter

helps

economic

activities,whileahottersummerhurtsthem.Inaddition,thenegativeeffectsofahottersummerinhotterstatesarelargerand

more

persistent.

The

findings

suggest

that

colder

regions

or

countries

may

benefit

from

climate

change

whilehotteronesmaybehurtwithoutsignificantadaptationefforts.Theseheterogeneouseffectspresentachallengefor

a

unified

effort

to

fight

climate

change

(whether

they

are

global

efforts

or

those

in

the

US).

Second,

this

paperdiscovers

more

severe

impacts

of

summer

temperature

in

the

US

in

recent

decades

(2000-2009

and

2010-2021)than

in

1990-1999.

A

one-degree

Fahrenheit

hotter

summer

in

the

2010s

reduces

employment

growth

in

thesummerandthefollowingfallbyabout0.1percentmorethan

itdidinthe1990s.Thisfindingimpliesadaptationefforts

in

the

US

have

not

taken

hold

or

significantly

altered

the

effects

of

temperature

shocks.

This

finding

hasimplications

for

other

countries.

Even

for

the

US,

which

is

a

developed

country

with

good

adaptation

capacityand

with

a

generally

mild

climate,

we

observe

negative

impacts

of

higher

temperature

in

the

summer.

For

poorer,hottercountries,theeffectsofrisingheat,withoutsignificantadaptationefforts,arelikelymuchmoresevere.The

paper

is

organized

as

follows.

Section

II

presents

a

simple

theoretical

framework

to

motivate

the

empiricalspecification.

Section

III

presents

data

and

the

main

empirical

specification.

Section

IV

presents

the

main

findingson

the

overall

impacts.

Section

V

presents

the

impact

by

the

US

state

and

patterns

between

the

impact

magnitudeand

a

state’s

climate.

Sections

VI

and

VII

examine

the

sectoral

impacts

of

a

hotter

summer

and

a

warmer

winter.SectionVIIIpresentstheeffectsbydecade.SectionIXpresentsrobustnesschecks.SectionXconcludes.II.

A

TheoreticalFrameworkThissectionpresentsatheoretical

motivationfortheempiricalsetup,wheretemperaturecanhavebothagrowtheffectandaleveleffectonemployment.InspiredbytheframeworkpresentedinDelletal.(2012),Iletemploymentinquarter

푞

afunctionofthecurrentquarter’sproductivityandlastquarter’semployment:IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCounties휙

푇푠휆푠푞−1퐿

=

휃

푒푠푞퐴

퐿(1)푞푠푞where푞

denotesquarter,푠

denotestheseason(i.e.,summer,fall,winter,orspring).퐿

,

푇

and퐴

are푞푞푞employment,temperatureandproductivityinquarter

푞.퐿푞−1

isemploymentinthepreviousquarter.Employmentinaquartercanbedrivenbythecurrentquarter’sproductivityandemploymentinthepreviousquarter.Capitalisomittedforsimplicity.Employmentatquarter푞

candependonemploymentatquarter푞

−

1becausehiringmaytaketime.푒휙푠푇푞

denotestheleveleffectofthequarter’stemperatureonemployment.Employmentcanbeaffectedbythequarter’saveragetemperature.Apositive/negative/zero휙

impliesthathighertemperaturehasa푠positive/negative/zeroleveleffectonemployment.Notethattheparameters휃

,

휙

,

휌

,

휆

areseasonspecific.Thatis,theparametersaredifferentforwinter,푠푠푠푠spring,summer,andfall.Forexample,

highertemperaturemayhavedifferent(orevenopposite)leveleffectsinthesummerversusinthewinter,hence휙

shouldbedifferentto휙푤푖푛푡푒푟푠푢푚푚푒푟.Seasonalproductivitygrowthisasfollows:log(퐴

)

−

log(퐴

)

=

푔

+훿

푇

+휔

푇푠

푞−4(2)푞푞−4푠푠푞Equation(2)statesthatseasonalproductivitygrowth

dependsonthisquarter’stemperatureaswellasthetemperatureofthesameseasonlastyear(4quartersago).

훿

and휔

arealsoseasonspecific.

푇

,thisquarter’s푠푠푞temperature,couldhaveapositiveornegativeeffectonseasonalproductivitygrowth.

푇푞−4,temperatureofthesamequarterinthepreviousyear,

mayaffectthisquarter’sproductivitygrowthviatwochannels.Firstisthebaseeffect.Forexample,alower푇couldlower퐴푞−4

,whichboostslog(퐴

)

−

log(퐴푞−4)

duetothebase푞푞−4effect.Secondistheproductivitytransmissioneffect.Alower푇couldlower퐴푞−4

whichcouldinstead푞−4depressseasonalproductivitygrowthforthefollowingyear.Therefore,on

thenet,itisnotclearthat휔

is푠expectedtohaveapositiveornegativevalue.Equations(1)and(2)statethattemperature

couldhavealeveleffectonemployment(via푒휙

푇

).Itcouldalso푠푞haveagrowth

effectonemploymentviaseasonalproductivitygrowthspecifiedinequation(2).3Now,let’srearrangeequations(1)and(2)toderiveanempiricalspecification.Foreaseofexposition,let’sstartequation(1)forthesummer휆푠푢푚푚푒푟푒휙푠푢푚푚푒푟푇푞퐴푞푠푢푚푚푒푟퐿푞−1(3)퐿

=

휃푞푠푢푚푚푒푟휆푠푝푟푖푛푔푞−2andsubstitute퐿푞−1

=

휃푠푝푟푖푛푔

푒휙푠푝푟푖푛푔푇푞−1퐴

푠푝푟푖푛푔퐿(notethatsince푞

isthesummer,푞

−

1

isthespring).푞−1휆푠푢푚푚푒푟휆푠푝푟푖푛푔(3)becomes퐿푞

=

휃푠푢푚푚푒푟푒휙푠푢푚푚푒푟푇

퐴훽푠푢푚푚푒푟

(휃푠푝푟푖푛푔푒휙푠푝푟푖푛푔푇푞−1퐴푠푝푟푖푛푔퐿푞−2)(4)푞푞푞−13

Althoughhumankind’sgreen-housegasemissionsinfluenceglobaltemperature,localtemperatureisconsideredexogenoustolocaleconomicactivities.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesKeepsubstituting퐿푞−2

=

휃푤푖푛푡푒푟푒휙푤푖푛푡푒푟푇푞−2퐴푤푖푛푡푒푟푞−2퐿휆푤푖푛푡푒푟푞−3and퐿푞−3

=

휃

푒휙푓푎푙푙푓푎푙푙푇퐴퐿푓푎푙푙

휆푓푎푙푙푞−3

푞−4into(4),푞−3wecanseethat(1)takesthefollowinggeneralform:휙

푇

휙

푇휙

푇휙

푇푞−3

퐴푞0퐴123휇퐿

=

휃푒푞0푞푒1푞−1푒2푞−2푒3퐴퐴퐿(5)푞−1

푞−2

푞−3

푞−4(5)statesthatemploymentisafunctionoftemperatureandproductivityofthis

quarteraswellasthoseinthepreviousthreequartersandemploymentofquarter

푞

−

4.Notethatallparametersof(5)areseasonspecific.Similar,forthesameseasoninthepreviousyear(i.e.,

푞

−

4):휙

푇휙

푇휙

푇휙

푇0123휇퐿푞−4

=

휃푒0푞−4푒1푞−5푒2푞−6푒3푞−7퐴퐴퐴퐴퐿(6)푞−4

푞−5

푞−6

푞−7

푞−8Subtractlogof(5)bylogof(6):33Δlog(퐿

)

=

∑

휙

Δ푇

+∑

휎

Δlog(퐴푞−휏)+휇Δ

log(퐿푞−4)

(7)푞휏푞−휏휏휏=0휏=0whereΔlog(퐿

)

=

log(퐿

)

−

log(퐿푞−4)

isyear-over-yeargrowthinemployment;

Δ푇

=

푇

−

푇indicates푞−휏−4푞푞푞−휏푞−휏year-over-yearchangeintemperature;andΔlog(퐴푞−휏)

=

log(퐿푞−휏)

−

log(퐿푞−휏−4)

indicatesyear-over-yeargrowthinseasonalproductivity.Substituting(2)into(7)yields:33Δlog(퐿

)

=

∑

휙

Δ푇

+∑

휎

{푔

+훿

푇

+휔

푇푞−휏−4}+휇Δ

log(퐿푞−4)

(8)푞휏푞−휏휏휏휏푞−휏휏휏=0휏=0Rearrangetermsin(8)yields:37Δlog(퐿

)

=

푔

+∑

훽

푇

+∑

휋

푇

+휇Δ

log(퐿푞−4)

(9)푞휏푞−휏휏

푞−휏휏=0휏=4where푔

=

∑3

휎

푔

;훽

=

휙

+휎

훿

;휋

=

−휙

+휎

휔

.Asbefore,allparametersareseasonspecific(thatis,휏=0휏휏휏휏휏휏휏휏휏휏theyvarydependingonwhetherthequarter푞

issummer,fall,winter,orspring).Iaminterestedinthecoefficients훽

,

훽

,

훽

,

훽

,representingtheeffectsoftemperatureinthisquarterand

three0123quartersagoonthisquarter’semployment.Intheempiricalsection,Iwillestimate훽

,

훽

,

훽

,

훽

foreach0123season.

푇

to푇andΔ

log(퐿푞−4)

areconsideredcontrolvariables.4푞−7푞−44

Asdiscussed,temperaturebetween푞

−

4

and푞

−

7

hasbaseeffectsaswellaspotentialproductivitytransmissioneffects.IMFWORKINGPAPERSBeyondtheAnnualAverages:ImpactofSeasonalTemperatureonEmploymentGrowthinUSCountiesIII.

DataandEmpiricalSpecificationDataEmploymentdata:Quarterlyemploymentdatabetween1990and2021attheUScountylevelarefromtheUSCensus’sQuarterlyCensusofEmploymentandWages(QCEW).

TheQuarterlyCensusofEmploymentandWages(QCEW)programpublishesaquarterlycountof

(formal)employmentandwagesreportedbyemployerscoveringmorethan95percentofUS

jobs,availableatthecounty,metropolitan(MSA),state,andnationallevelsbyindustry.

Majorexclusionsfromthedatasetincludeself-employedworkers,mostagriculturalworkersonsmallfarms,allmembersoftheArmedForces,electedoffi