IMF-外部沖擊、政策和實(shí)際匯率的尾部變動(dòng)（英）

ExternalShocks,Policies,andTail-ShiftsinRealExchangeRatesNicolásE.

Magud

and

SamuelPienknaguraWP/23/129IMFWorkingPapersdescriberesearchinprogressby

theauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.Theviewsexpressed

inIMFWorkingPapersarethoseoftheauthor(s)anddo

notnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.2023JUN?2023InternationalMonetaryFundWP/23/129IMFWorkingPaperWesternHemisphereDepartmentExternalShocks,Policies,

andTail-Shiftsin

RealExchangeRatesNicolásE.MagudandSamuelPienknagura

*Authorizedfordistributionby

GustavoAdlerJune2023IMFWorkingPapersdescriberesearchinprogressby

theauthor(s)andarepublishedto

elicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,

or

IMFmanagement.ABSTRACT:

We

use

panel

quantile

regressions

to

study

extreme

(rather

than

average)

movements

in

thedistribution

of

the

real

effective

exchange

rate

(REER)

of

small

open

economies.

We

document

that

globaluncertainty

(VIX)

and

global

financial

conditions

(U.S.

monetary

policy)

shocks

have

a

strong

impact

on

thedistribution

oftheREERchanges,with

largerimpactsinthetailsofthe

distribution,andespeciallyineconomieswith

shallower

FX

markets,

lower

central

bank

credibility,

and

higher

credit

risk

(i.e.,

weaker

macro

fundamentals).Foreign

exchange

intervention

(FXI)

partially

offsets

the

impact

of

these

shocks,

especially

in

the

left

tail

(largedepreciations)andparticularlyineconomieswith

weakerfundamentalsbut,moreimportantly,whenFXIisusedsporadically.

Thus,

our

results

highlight

the

importance

of

deepening

FX

markets,

improving

central

bankcredibility,

and

strengthening

macro

fundamentals

against

the

potential

dynamic

trade-offs

of

overreliance

on

apolicy

that

would

exacerbate

the

previously

mentioned

frictions.

While

our

results

point

to

low

effectiveness

ofcapital

flow

management

in

preventing

large

REER

movements,

they

seem

to

enable

more

impactful

foreignexchangeinterventionintheimmediateaftermathofshocks.RECOMMENDEDCITATION:Magud,NicolasE.andPienknagura,Samuel(2023),“ExternalShocks,Policies,andTail-ShiftsinRealExchangeRates,”IMFWorkingPaper23/129(InternationalMonetaryFund:Washington,DC)JELClassificationNumbers:Keywords:F31,F38,F41Realexchangerates;externalshocks;foreignexchangeintervention;capitalcontrolsAuthor’sE-MailAddress:nmagud@;spienknagura@*

We

thank

Yoshiki

Ando

for

data

preparation

and

Jing

Xie

for

superb

research

assistance,

and

Gustavo

Adler,

Cian

Allen,

AndrewBerg,

Mariarosaria

Comunale,

Pablo

Druck,

Andrés

Fernández,

Alejandro

Guerson,

Geoff

Heenan,

Luis

Jácome,

Ruy

Lama,Leonardo

Martinez,

Pau

Rabanal,

Felipe

Saffie,

Rodrigo

Valdés,

Carlos

Vegh,

and

Frank

Wu

for

excellent

comments

and

discussions.The

views

expressed

in

the

paper

are

those

of

the

authorsand

do

notnecessarily

represent

the

views

ofthe

IMF,

its

Executive

Board,orIMFmanagement.IMFWORKINGPAPERSExternalShocks,Policies,

andTail-ShiftsinRealExchangeRatesWORKINGPAPERSExternalShocks,Policies,andTail-ShiftsinRealExchangeRatesPreparedbyNicolásE.

MagudandSamuelPienknagura

11

We

thank

Yoshiki

Ando

for

data

preparation

and

Jing

Xie

for

superb

research

assistance,

and

Gustavo

Adler,

Cian

Allen,

AndrewBerg,

Mariarosaria

Comunale,

Pablo

Druck,

Andrés

Fernández,

Alejandro

Guerson,

Geoff

Heenan,

Luis

Jácome,

Ruy

Lama,Leonardo

Martinez,

Pau

Rabanal,

Felipe

Saffie,

Rodrigo

Valdés,

Carlos

Vegh,

and

Frank

Wu

for

excellent

comments

and

discussions.The

views

expressed

in

the

paper

are

those

of

the

authorsand

do

notnecessarily

represent

the

views

ofthe

IMF,

its

Executive

Board,orIMFmanagement.INTERNATIONALMONETARYFUND3IMFWORKINGPAPERSExternalShocks,Policies,

andTail-ShiftsinRealExchangeRatesI.

IntroductionFinancial

integration

can

have

pro-growth

effects,

but

it

can

also

lead

to

short-term

challenges.Integration

eases

countries’

financing

needs,

which

can

foster

long-term

investment

and

helpsmoothdomesticandexternal

shocks.

Itcanalso

helpincreasesexportsandimports,whichcanraise

growth

prospects.

However,

globally

integrated

economies,

especially

small

openeconomies

(SOEs),

are

exposed

to

transitory

external

financial

shocks,

which

can

result

inexcessive

movements

in

key

economic

variables.

In

particular,

financial

integration

exposescountries

to

global

uncertainty,

as

reflected

in

higher

VIX

levels,

and

to

changes

in

global

financialconditions,

such

as

those

stemming

from

U.S.

monetary

policy

actions.

These

shocks

can

leadto

temporary

swings

in

capital

inflows,

increase

funding

costs,

and

result

in

excessive

movementsinexchangerates.The

challenges

and

opportunities

brought

about

by

global

integration

have

sparked

a

lively

debateabout

the

appropriate

design

of

policies

to

sustainably

reap

the

benefits

of

a

more

interconnectedglobal

economy.

One

side

of

that

debate,

which

has

garnered

growing

attention

in

both

academicand

policy

circles,

regards

the

usefulness

and

effectiveness

of

foreign

exchange

intervention

(FXI)and

capital

flow

management

(CFM)

policies

in

mitigating

the

adverse

effects

of

excessivemovements

of

key

economic

variables

in

response

to

temporary

shocks.

Part

of

such

debate,however,

oftentimes

lacks

a

proper

definition

of

the

specific

objective

that

such

policies

aim

atachieving—akeyconsiderationinpolicydesign.In

this

paper,

taking

as

the

policy

objective

to

smooth

exchange

rate

volatility

(as

SOE

typicallyaimat2),

first

wefocus

on

the

impactthat

financial

external

(VIXandUSMonetarypolicy)

shockshave

on

the

distribution

of

real

exchange

rate

changes,

and

especially

extreme

movements

inrealexchangerates.

Then,

weexplorethe

rolethat

FXIandCFMs

(bothof

inflowsandoutflows)have

in

mitigating

the

impact

on

these

shocks.

Especially,

we

care

for

whether

and

how

thesepolicies

mitigate

the

effects

of

external

financial

shocks

in

the

tails

of

the

distribution

of

realexchangeratemovements,whatwedubthe

extremes.Keytoourcontributionis

thatwedepartfromthe

existingliterature,

whichmostly

focusesontheaverage

effect

of

shocks

on

real

exchange

rates,

to

focus

on

the

extremes

of

the

distribution,where

policy

becomes

more

relevant.

We

care

about

this

because,

by

definition,

the

averageimpact

of

shocks

may

hide

important

tail

risks

stemming

from

shocks.

To

this

end,

we

track

the2

The

realexchangerate

is

anendogenousrelativeprice

notonlyassociatedwith

competitiveness,but

also

a

country’sincomein

theglobal

economy.

It

is

highly

correlated

with

the

nominal

exchange

rate,

but

unlike

the

latter,

is

less

contaminated

by

economicpolicies

such

as

a

country’s

exchange

rate

regime,

inflation

control

strategies,

and

other

policies.

It

may

also

summarize

typicalmeasures

of

welfare

such

as

GDP

or

consumption.

Theoretically,

Lahiri

and

Vegh

(2001)

show

that

under

some

circumstancesintervening

in

foreign

exchange

markets

to

reduce

excessive

REER

volatility

may

be

the

optimal

policy

of

a

central

bankin

a

smallopeneconomy.INTERNATIONALMONETARYFUND4IMFWORKINGPAPERSExternalShocks,Policies,

andTail-ShiftsinRealExchangeRatesdynamicimpactofshocksacross

specific

deciles

of

the

real

exchange

rate

distributionby

meansof

panel

quantile

regressions.

For

simplicity,

and

as

a

broad

depiction

of

shifts

in

the

distributionof

the

changes

in

the

REER

in

the

aftermath

of

shocks,

we

assess

the

response

of

the

medianvalue,

the

10th

percentile

and

the

90th

percentile,

which

allows

us

to

study

the

behavior

beyondthecentralpartof

thedistributionand,

importantly,

aroundthe

twotails.Focusing

on

both

the

median

and

the

tails

of

the

distribution

enables

us

to

show

how

mitigatingpolicies

can

be

deemed

as

more

effective

and

less

costly

at

different

deciles—such

as

in

collapsesor

crises,

periods

during

which

containment

policies

may

be

needed

the

most.

In

fact,

we

showthat

the

mitigating

power

of

FXI

and

CFM

varies

substantially

across

the

exchange

ratemovements’

distribution.

We

stress

this

point,

which,

to

be

the

best

of

our

knowledge,

has

beenmostly

unexplored,

because

a

large

part

of

the

received

literature

finds

FXI

to

be

ineffective.

Incontrast,

we

argue

that,

in

part,

such

lack

of

statistical

significance

derives

from

focusing

onaverages

rather

than

exploiting

the

information

in

the

tails

of

the

distribution.

In

addition,

potentialasymmetries

in

the

impact

of

positive

vs.

negative

shocks

on

different

deciles

of

the

REERdistributionamplifythepreviously

mentionedconsiderations.Specifically,

we

find

that

an

increase

in

global

uncertainty

(as

given

by

the

VIX)

shifts

the

threedeciles

we

focus

on

to

the

left,

suggesting

an

overall

weakening

of

the

REER.

However,

the

effectis

stronger

atthe

10th

percentile

of

the

distribution,

that

is,

currencies

that

weaken

relatively

more.Therefore,

our

results

imply

that,

other

things

equal,

a

VIX

shock

increases

markedly

theprobability

of

observing

large

depreciations.

As

an

example

of

these

non-linearities,

and

as

isdiscussed

in

more

details

in

the

rest

of

the

paper,

Figure

1

shows

the

response

of

variousquantiles

of

the

distribution

of

REER

changes

to

a

one

standard

deviation

change

to

the

VIX.

Notethat

although

the

average

real

depreciation

is

about

0.6

percent,

the

real

depreciation

of

the

1stdecile

is

about

1.1

percent

(almost

twice

the

impact),

while

that

of

the

9th

decile

is

0.2

percent.U.S.

monetarypolicy

shocksalsodepreciate

SOEs’currencies(left

shift

in

the

decileswestudy),but

the

impact

is

stronger

on

the

median

and

in

the

90th

percentile

(stronger

currencies)—thus,reducingthe

probabilityofobservinglargeappreciations.We

also

find

that

negative

shocks

are

morerelevant

than

positive

shocks.

We

observe

that

onlyincreases

in

the

VIX

have

a

statistically

significant

impact

on

the

three

deciles

that

we

focus

on(with

no

effect

from

a

lower

VIX).

And

only

a

tightening

of

U.S.

monetary

policy

results

inprotracted

and

substantial

depreciations

in

small

open

economies

(left

shift

of

REER

deciles),withonlymarginalappreciationinresponseto

monetaryeasing.33

Of

course,

countries’

REER

do

not

systematically

weaken.

We

speculate

that

this

could

be

associated

with

domestic

shocks,

policies,and

policy

frameworks

contributing

to

secularly

strengthening

REER

distributions

in

the

spirit

of

Balassa-Samuelson.

To

the

extentpossible

we

control

for

variables

plausibly

related

to

such

trend.

Terms-of-trade

shocks,

which

can

potentially

impact

the

realexchangerates,

is

foundnotto

affectit,

however.

Weshow

this

inthe

Annex2.INTERNATIONALMONETARYFUND5IMFWORKINGPAPERSExternalShocks,Policies,

andTail-ShiftsinRealExchangeRatesTo

understand

the

transmission

of

these

external

shocks

to

the

REER,

we

study

whether

theeffects

of

these

shocks

differ

when

conditioningbythe

depth

of

the

foreign

exchange

market,

thecredibility

of

the

central

bank,

and

the

country’s

credit

risk.

We

find

that

in

fact

all

these

are

keyfactors.

Countries

with

shallower

foreign

exchange

markets,

less

credible

central

banks,

or

highercredit

risk

(i.e.,

weaker

fundamentals)

all

experience

larger

depreciations

when

hit

by

the

externalshocks,andimpactsareparticularlystrongerintheleftpart

of

the

REERchangedistribution.In

terms

of

policy

effectiveness,

we

find

that,

in

response

to

an

increase

in

global

volatility,

asexpected,

a

larger

FXI

yields

a

more

muted

change

in

the

three

deciles

that

we

study,

a

purevolume

effect4

while

FXI

seems

to

reduce

the

depreciation

pressures

of

U.S.

monetary

policyshocks

only

marginally,

and

especially

around

the

median.

Additionally,

we

split

the

sample

inepisodes

of

FXI

selling

international

reserves

(in

response

to

depreciation

pressures)

and

FXIpurchasing

reserves

(in

response

to

appreciation

pressures)—the

typical

central

bank

“receivedwisdom”—and

find

symmetry

in

that,

in

response

to

VIX

shocks

and

U.S.

monetary

policy

shocks,selling

international

reserves

mitigates

depreciation

pressures

while

purchasing

internationalreservesdampensappreciationpressures.A

key

insight

is

to

understand

the

channels

through

which

these

effects

operate

and

whetherthese

happens

in

a

linear

or

nonlinear

manner.

To

the

best

of

our

knowledge,

addressing

thisissue

empirically

is

still

missing

in

the

literature,

and

this

paper

helps

to

fill

this

gap

by

documentingnotorious

nonlinearities.

We

show

that,

in

response

to

VIX

shocks,

FXIameliorates

the

impact

onthe

real

exchange

rate

distribution

more

in

economies

with

shallower

FX

markets,

in

those

withlesscrediblecentralbanks,andthosewithhighercreditrisk.5

Thisis

inpartdueto

abaseeffect,associated

with

larger

response

of

real

exchange

rate

changes

in

economies

that

lack

deep

FXmarkets

or

with

central

banks

with

lower

credibility,

in

absence

of

intervention.

The

differentialeffectiveness

of

FXI

reflects

the

variation

in

the

degree

of

FX

market

shallowness

and

centralbank

credibility.

In

this

sense,

when

external

financial

shocks

hit

the

economy,

FXI

helps

tocompensate

financial

market

and

institutional

weaknesses,

especially

in

the

left

tail.

Also,

FXI

ismore

effective

in

situations

of

capital

outflows

events

than

in

onflows

events.

This

is

consistentwith

anecdotal

evidence

that

policymakers

in

small

open

economies

are

oftentimes

more

worriedabout

excessive

real

depreciation

than

appreciation

pressures

(e.g.,

Calvo

and

Reinhart,

2002)—andreinforcesthe

usefulnessofthequantileregressionapproachusedinthispaper.The

above

finding,

however,

does

not

imply

that

FXI

is

a

silver

bullet.

In

fact,

when

we

splitcountries

into

heavy

FXI

users

and

non-heavy

users,

we

find

that

FXI

is

substantially

moreeffective

in

mitigating

external

shocks

for

non-heavy

users

of

FXI

policy.

That

implies

that

FXI

isa

reasonable

policy

to

contain

the

(negative)

effects

of

VIX

and

U.S.

monetary

policy

shocks

for4

Theresults

arequalitatively

unaffectedwhenwenotonlycontrolforcontemporaneous

FXI,

but

alsoforfutureinterventions.5

Basuandothers

(2023)modelrelatednon-linearitiestheoreticallyforcountriesthatfaceexternalborrowingconstraints.INTERNATIONALMONETARYFUND6IMFWORKINGPAPERSExternalShocks,Policies,

andTail-ShiftsinRealExchangeRatescountries

that

do

not

rely

on

FXI

too

frequently.

Moreover,

we

interpret

the

latter

combined

resultsas

highlighting

the

importance

of

SOEs

developing

deep

FX

markets

and

building

central

bankcredibility.

Otherwise,

relying

on

the

expectation

that

the

central

bank

will

always

be

there

tosmooth

FX

markets

would

reduce

incentives

to

invest

in

improving

such

macroeconomicfundamental—a

typical

example

of

moral

hazard.

Moreover,

not

doing

it

may

eventually

rendertheeffectivenessofFXI

weak,

withsubsequentrealandfinancialdeleteriouseffects.In

terms

of

magnitudes

and

costs

(as

measured

by

international

reserve

losses),

we

find

that

alarger

andmoredurableeffect

of

FXIonthe

10th

percentileoftheREERdistribution—thatis,

theweaker

extreme.

Thus,

the

arguments

for

these

types

of

interventions

are

stronger

when

theobjective

is

to

avoid

large

depreciations

(tail

outcomes).Second,

our

estimates

also

shed

light

onthe

feasibility

of

achieving

a

given

reduction

in

the

REER

decile

of

interest.

Reducing

in

10

percentthe

depreciation

of

the

10th

percentile

of

the

REER

in

the

aftermath

of

a

one

standard

deviationVIX

shock

would

requirea

sizeable

intervention,

of

about

2.8percent

of

GDP.

Thisis

roughly

15percent

of

the

median

holding

of

international

reserves

in

our

sample

in

2019,

and

equivalent

insize

to

the

decline

in

reserves

observed

for

the

median

country

during

the

GFC.

In

other

words,forthe

lefttail,for

eachone

percentof

GDPof

international

reservesusedto

interveneinforeignexchange

markets,

the

real

exchange

rate

depreciation

is

mitigated

in

close

to

3.5

percentagepoints.

ThiscontrastswithresultsinBlanchardandothers(2013),

whofind,

lookingataverages,that

one

percent

of

GDP

intervention

renders

a

2

percent

change

in

the

real

exchange

rate

(almosthalf).

This

highlights

that

the

elasticities

of

real

exchange

rate

movements

to

foreign

exchangeintervention

differalong

the

distribution.

The

latteris

especially

relevant

when

designing

a

policyof

foreign

exchange

intervention,

and

supports

methodologically

shifting

to

panel

quantileregressions,

asopposedto

standardpanel

(thatis,

averageeffectsonly)regressions.CFMs

do

not

seem

to

be

effective

in

reducing

the

effects

of

any

of

these

external

shocks,regardless

of

whether

those

controls

aim

at

stemming

excessive

inflows

(associated

with

realappreciation

pressures)

or

mitigating

capital

outflows

(in

event

of

rea

depreciation).

We

alsoexplore

whether

imposing

CFMs

enable

a

more

effective

FXI—under

the

logic

that

a

relativelyclosed

financial

account

strengthen

the

effectiveness

of

FXI.

There

seems

to

be

some

evidencethat

restricting

the

external

financial

account

does

contribute

to

enable

FXI

to

become

moreeffective,albeitmostlyinthe

immediateaftermathofthe

shock.6Weextend

the

analysistothe

impactof

theseexternal

shockson

threedecilesofthe

distributionofREERvolatility.

WefindthatonlychangesintheVIX

affect

thedecileswestudy.AhigherVIXresults

in

an

increase

in

all

deciles

we

study,

with

the

90th

percentile

of

the

empirical

distribution6

A

recent

literature

(e.g.,

Das

and

others,

2021)

studies

ex-ante

CFM

restrictions.

Given

the

nature

of

our

sample,

we

do

not

focusonsuchCFMsinthis

paper.INTERNATIONALMONETARYFUND7IMFWORKINGPAPERSExternalShocks,Policies,

andTail-ShiftsinRealExchangeRatesreacting

more

strongly.

This

suggests

a

higher

probability

of

higher

REER

volatility

in

theaftermathofshocks.

FXI

seemsto

helpcontaintheimpact

of

VIXshockson

REERvolatility.Finally,

we

show

that

initial

conditions

matter.

In

particular,

external

shocks

affect

more

net

debtorcountries

compared

to

net

creditor

countries.

This

highlights

the

importance

of

strong

macro

policyframeworksasawayto

copewithexternal

shocks.Related

literature.

The

importance

of

global

uncertainty/financial

shocks,

such

as

the

VIX,

inaffecting

the

dynamics

of

key

economic

variables

in

SOEs

has

been

thoroughly

documented,including

with

Rey

(2015),

as

well

as

the

key

contribution

of

U.S.

monetary

policy

shocks,

startingwith

Calvo

and

others

(1993).

Higher

uncertainty

or

tighter

U.S.

monetary

policy

result

in

netcapital

outflows

from

SOEs,

depreciating

the

real

exchange

rate.

Our

results

are

aligned

withmost

of

the

received

wisdom

in

that

we

take

VIX

and

US

monetary

policy

shocks

as

exogenousto

SOEs.

Our

contribution

(and

eventual

differences),

however,

lies

in

deviating

from

(previous)average

results

and

analyzing

the

impact

of

external

shocks

on

the

changes

in

the

distributionsofrealexchangerates.Sterilized

foreign

exchange

intervention

gained

theoretical

support

starting

with

Kouri’s

(1976)contribution

to

the

portfolio

balance

approach—other

channels

include

the

signaling

channel(Mussa,

1982)

and

the

microstructure

channel

(Evans

and

Lyons,

2002).

To

the

extent

thatdomestic

and

foreign

assets

are

not

perfect

substitutes,

Kouri

(1976)

shows

that

foreign

exchangeintervention

can

be

effective

in

affecting

the

exchange

rate

(and

in

a

sticky-prices

environment,also

the

real

exchange

rate).

Lahiri

and

Vegh

(2001)

provide

a

theoretical

argument

forintervening

in

foreign

exchange

markets

in

response

to

large

shocks,

but

not

otherwise,

as

it

isoptimalonlytomitigatelargershocks.

Recent

theoretical

developments,

building

on

theportfoliobalance

approach

haveshown

the

existenceof

optimal

FXI

effectivenesstocontainmovementsin

the

exchange

rate

triggered

by

capitalflows

(e.g.,

Cavallino,

2019).

Fanelli

and

Straub

(2021)study

the

mitigating

role

of

FXI

in

the

presence

of

partial

segmentation

between

domestic

andforeign

bonds

and

of

pecuniary

externalities

affecting

the

volatility

of

the

exchange

rate,

andBianchi

and

Lorenzoni

(2022)

explore

the

prudential

use

of

capital

controls

and

foreign

exchangeintervention.Despite

theoretical

arguments,

the

empirical

evidence

about

the

effectiveness

of

FXI

has

beenmixed.

Early

contributions

found

scarce

evidence

of

FXI

effectiveness,

mostly

focusing

onadvanced

economies

(see

Dominguez

and

Frenkel,

1993).

A

recent

body

of

work

finds

moresystematic

support

for

FXI

effectiveness.

Blanchard

and

others

(2015)

find

evidence

of

FXIaffecting

exchange

rates,

albeit

only

in

the

short

run.

Daude

and

others

(2016)

find

FXI

affectsthe

level

and

volatility

of

the

real

exchange

rate,

while

Adler

and

others

(2019)

find

that

FXI

iseffective

in

impacting

the

nominal

and

real

exchange

rates.

Using

daily

data,

Fratzscher

andothers

(2019)

find

that

FXIis

a

useful

tool

to

smooth

the

path

of

exchange

rates.

Adler

and

othersINTERNATIONALMONETARYFUND8IMFWORKINGPAPERSExternalShocks,Policies,

andTail-ShiftsinRealExchangeRates(2019)theoretically

show

the

output-inflation

volatility

trade-off

arising

from

FXI

by

central

bankswith

low

credibility.

While

our

results

do

not

focus

on

the

above-mentioned

trade-off,

we

do

findthat

credibility

of

the

central

bank

affects

the

effectiveness

of

FXI,

in

line

with

their

results.

Recentevidence

supportingthe

effectivenessofFXI

lends

credencetothe

fact

that

it

hasbeenawidelyused

policy

tool

(see

Chamon

and

others

2019,

who

document

the

use

of

FXI

in

Latin

Americancentral

banks

that

have

an

inflation

target).

Our

paper

contributes

to

this

literature

and

to

the

policydebate

by

distilling

conditions

under

which

FXI

could

be

an

effe