The transmission of Federal Reserve tapering news to emerging financial … · 2016. 6. 1. ·...

The Transmission of Federal ReserveTapering News to Emerging Financial

Markets∗

Joshua Aizenman,a Mahir Binici,b and Michael M. HutchisoncaUniversity of Southern California and NBER

bCentral Bank of TurkeycUniversity of California, Santa Cruz

This paper evaluates the impact of tapering “news”announcements by Federal Reserve senior policymakers onfinancial markets in emerging economies. We apply a panelframework using daily data, and find that emerging-marketasset prices respond most to statements by Federal ReserveChairman Bernanke, and much less to other Federal Reserveofficials. We group emerging markets into those with “robust”fundamentals (current account surpluses, high internationalreserves, and low external debt) and those with “fragile” fun-damentals and, intriguingly, find that the exchange rates of therobust group (and, to a lesser extent, equity prices and CDSspreads) were more adversely affected by tapering news thanthe fragile group. The cumulative effects of tapering announce-ments after a month, however, appear to be quite similar forboth robust and fragile emerging markets.

We also show that more financially developed economies aremore affected by tapering news, and a plausible interpretationis that more financially developed economies are more exposed,at least in the short term, to external news announcements.

JEL Codes: F3, F36, G14.

∗We thank the editor of this journal and three anonymous referees forvery helpful suggestions on an earlier draft of this paper. Author e-mails:[email protected]; [email protected]; and [email protected].

317

318 International Journal of Central Banking June 2016

1. Introduction

The quantitative easing (QE) policies of the U.S. Federal Reservein the years following the crisis of 2008–9 included monthly secu-rities purchases of long-term Treasury bonds and mortgage-backedsecurities totaling $85 billion in 2013. The cumulative outcome ofthese policies has been the unprecedented increase of the monetarybase, mitigating the deflationary pressure of the crisis. The resul-tant lower interest rates and flattened yield curve improved finan-cial conditions and helped stimulate real economic activity, yet theQE policy raised pertinent questions regarding the timing and thenature of the exit strategy (Williams 2011, 2012). These issues cameto the fore in 2013, with vigorous and intensifying debate amongpolicymakers and market participants about the exit strategy fromthe massive monetary stimulus. The growing frequency of publicstatements by the Federal Reserve’s governors and presidents, com-bined with occasional press releases, have been the focus of thefinancial media, changing expectations and moving market prices.This process culminated on December 18, 2013, when the FederalReserve decided at the Federal Open Market Committee (FOMC)meeting (as announced in the public statement) to taper its quanti-tative easing policy by $10 billion per month, to $75 billion. Chair-man Bernanke also projected the program to wind down steadilythrough 2014 and conclude by year-end, assuming the economyremained healthy. Subsequent announcements gave news of a grad-ually reduced QE, and the pattern continued after the appointmentof Chair Yellen.

An important feature of quantitative easing and unprecedent-edly low U.S. interest rates is that it led to large short-term capitalinflows to a number of emerging markets, which in turn led severalto impose capital controls, such as Brazil, Indonesia, South Korea,and others (Ahmed and Zlate 2013). Quantitative easing led the U.S.dollar to be the funding currency in large-scale carry-trade activitywith emerging markets as the target currencies. The concern withtapering is the flip side: potentially disruptive large-scale capitaloutflows from emerging markets as carry-trade activity is unwoundin expectation of tapering (and, eventually, reduction in the FederalReserve balance sheet through sales of assets, not just reduction inthe pace of purchases) and hints at future interest rate increases.

Vol. 12 No. 2 The Transmission of Federal Reserve Tapering 319

Large capital outflows could create disruptions in financial marketsand eventually real economic activity in emerging markets.

This paper evaluates whether tapering announcements have dis-rupted financial markets in emerging economies. We investigate theimpact of tapering “news” announcements by Federal Reserve seniorpolicymakers on financial asset prices in emerging markets. Theemerging markets’ financial asset prices of interest are national stockmarkets, exchange rates, and credit default swaps (CDS) spreads.These reflect a broad spectrum of the potential effects of tapering,where we would expect greater likelihood of tapering, and hence cap-ital outflows, from emerging markets, to cause a fall in equity mar-kets, depreciation of exchange rates, and an increase in CDS spreads(reflecting greater uncertainty and risk in sovereign bond markets).In terms of Federal Reserve “news,” we focus on statements fromFederal Reserve Chairman Bernanke, Federal Reserve Board gover-nors, and Federal Reserve Bank presidents, as well as FOMC state-ments and minutes. Our presumption is that it is important to dif-ferentiate between announcements/statements by the Chair, who isthe public face and most important Federal Reserve policymaker,and other Federal Reserve policymakers (governors and presidents).

We employ daily data during November 2012 and October 2013.Statements about the likelihood of future tapering, or scaling backthe large-scale asset purchase program (LSAP), began to emerge inlate 2012, marking the beginning of our sample period. However,during this period there were also frequent and forceful statementsby Federal Reserve officials about the need to continue quantita-tive easing, so these statements were also included as “news” inour investigation both to address issues of symmetry and to judgemarket impacts.

The methodology of the paper is a quasi-event study, akin toDooley and Hutchison (2009), tracing the impact of evolving nar-rative about the expectation of future tapering, as revealed to thepublic through the news media, on key emerging-market prices. Weuse a panel fixed-effect framework using daily data with a varietyof models to evaluate the impact of “news” on the three assetsprices (stock market, exchange rate, and CDS spreads). Moreover,we explore whether market responses to tapering “news” are differ-ent across emerging markets, perhaps differentiated by the relativestrength of their economic fundamentals. In particular, we exploit


the heterogeneity among the emerging markets, evaluating the asso-ciation between asset price movements and key characteristics asso-ciated with “fragility” or “robustness” of a country, where thesecharacteristics are defined by their current account, internationalreserve, and foreign indebtedness positions.

Previewing results, we find that emerging-market asset pricesrespond most to statements by Federal Reserve Chairman Bernanke,and much less to the frequent, divergent, and sometimes inconsistentstatements by other Federal Reserve officials. This finding is consis-tent with the power of the Chairman to set and affect the agenda,and with the advantage of more frugal and clear communication. Wegroup emerging markets into those with “robust” fundamentals (tencountries) and those with “fragile” fundamentals (sixteen countries)and, intriguingly, find that stronger countries, on average, were gen-erally more adversely exposed in the short run to tapering news thanthe countries with weak fundamentals, especially in the deprecia-tion of their currencies. In particular, the exchange rate depreciatedin both groups immediately following tapering news from Chair-man Bernanke, yet the depreciations of the stronger group werethree times as large as the weaker group. Looking at these linkagesin detail, we find that higher current account balances, higher for-eign exchange rate reserves, and lower external debt are associatedwith greater exchange rate depreciation following Bernanke taper-ing announcements. Over the period of a month, however, we findasset price movements to tapering announcements are quite similarfor both the fragile and robust groups. Finally, we find that morefinancially developed emerging markets were also more affected byBernanke tapering announcements.

A possible interpretation of these findings is that countrieswith weaker fundamentals and less financial development were lessexposed to the inflows triggered by quantitative easing, and exchangerates responded less, in line with the conjecture that being closer tofinancial autarky provides deeper insulation from financial news. Theflip side is that tapering news had less impact on exchange rates inthese countries. Yet, these findings are also consistent with a less san-guine interpretation, reflecting financial markets’ initial inattentionto tail risks, overlooking the vulnerability of the weaker emergingmarkets to the adverse implications of higher future global interestrates. Indeed, in the last quarter of 2013, financial markets refocused


attention on the fragile emerging markets, with depressed financialasset prices of the “Fragile Five”—Brazil, India, Indonesia, SouthAfrica, and Turkey—a subgroup of the weaker emerging markets.1

Interestingly, however, it appears that differential responses betweenthe fragile and robust group tend to dissipate over time.

The methodology of our paper complements Eichengreen andGupta (2014), who used data for exchange rates, foreign reserves,and equity prices between April and August 2013 to analyze whowas hit and why. They concluded that better fundamentals did notprovide insulation. Instead, countries with larger markets experi-enced more pressure on the exchange rate, foreign reserves, andequity prices. Our focus on the event methodology allows us to tracethe immediate impact effect of the coming news on expectation ofadjustment, as reflected in the changes of key prices triggered by thenews.

The next section presents the data and methodology of the study.Section 3 presents the main results. Section 4 presents extensions androbustness checks of the basic models. Section 5 concludes.

2. Data and Methodology

2.1 Data

Our objective is to evaluate the transmission of U.S. tapering “news”to financial markets in emerging markets. In undertaking this analy-sis, we consider announcements associated with support (or actions)of quantitative easing as well as tapering. During the period oftime under investigation—November 27, 2012 through October 3,2013—there were numerous statements in support of both policiesby Chairman Bernanke, Federal Reserve governors, Federal ReserveBank presidents, and the FOMC statement following meetings, aswell as FOMC minutes from past meetings. This allows us to evalu-ate the symmetry of the effects of the announcements, not only forQE and tapering but also for different sources of the statements. In

1The market inattention to tail risks was vividly illustrated by the euro cri-sis, where the pre-crisis sovereign spreads of Greece, Portugal, and Spain werecomparable to that of Germany and other euro-zone core countries (Aizenman,Hutchison, and Jinjarak 2013). See also Aizenman, Binici, and Hutchison (2013)for an overview of other issues in the pricing of risk during the euro crisis.


particular, we focus on six types of announcements from the Fed-eral Reserve: statements by Federal Reserve Chairman Bernanke,either in support of further quantitative easing (QE) or tapering(Tap); release of statements following the FOMC meetings eithersupporting further QE or Tap; and statements by governors of theFederal Reserve System (other than Bernanke) or by presidents ofthe Federal Reserve Banks either supporting QE or Tap.

To gather this information, we conducted two sets of newssearches on Bloomberg. The first search was for the keywords “QEFederal.” Additionally, to focus on the tapering aspect of the QEannouncements, another search was performed for the keywords“Federal Reserve Bank of,” “QE,” and “Fed Taper.” A filter wasapplied to select “News” for both searches. To verify whether theannouncement and speech “news” were coded consistently with howthey were perceived by the markets in the United States (not emerg-ing markets), we examined articles and other publications from theFactiva database. We utilized the date and topic filtering availablein the database to solicit descriptions of the perception as well asthe corresponding market reactions of the announcement/speech.We closely examined each of the original Bloomberg coding forannouncements to determine whether they matched the perceptionsfound in the Factiva publications. There were many more announce-ments/statements/speeches on these topics than were coded. Inorder to be coded as either supporting further (or continuing) QEor tapering, the announcements needed to be unambiguous.

Table 1 presents some summary statistics on our announce-ments as well as examples of announcement coding and the divisionbetween “tapering” and “quantitative easing.” We found three cases(seven cases) where Chairman Bernanke clearly indicated taperingwas a favored policy option (further or continuing QE was a favoredpolicy option); five cases where the FOMC statements/minutesclearly favored QE (and none clearly favored tapering); and a roughbalance between governors/presidents speaking unambiguously infavor of QE (twenty-two cases) or tapering (twenty-six cases).

We consider three main conduits through which these variablesmay affect emerging markets: daily changes in national stock mar-kets (log first differences), daily changes in foreign exchange rates(log first differences), and daily changes in CDS spreads (in basispoints). If tapering is expected to reduce financial flows to emerging

Vol. 12 No. 2 The Transmission of Federal Reserve Tapering 323Tab

le1.

Num

ber

and

Exam

ple

sof

Tap

eran

dQ

E“N

ews”

Eve

nts

(Nov

ember

27,20

12–O

ctob

er3,

2013

)

Sou

rce

ofEve

nts

Chai

rman

Ber

nan

keG

over

nor

/Pre

siden

tFO

MC

Agg

rega

te

Tap

erin

gQ

ETap

erin

gQ

ETap

erin

gQ

ETap

erin

gQ

E

No.

ofE

vent

s3

726

220

529

33

Exa

mpl

esof

Ann

ounc

emen

ts(C

ongr

essi

onal

Tes

tim

ony,

Spee

ches

,FO

MC

Stat

emen

ts,et

c.)

Ben

Tap

(Ber

nank

eTap

erin

g)M

ay22

,201

3:B

erna

nke

said

the

Fede

ralR

eser

veco

uld

“tak

ea

step

dow

nin

our

pace

ofpu

rcha

ses”

inth

e“n

ext

few

mee

ting

s”in

test

imon

yto

the

Join

tE

cono

mic

Com

mit

tee

ofC

ongr

ess.

Ben

QE

(Ber

nank

eQ

uant

itat

ive

Eas

ing)

Janu

ary

24,20

13Tes

tim

ony

toSe

nate

Ban

king

Com

mit

tee:

“We

dono

tse

eth

epo

tent

ialco

sts

ofth

ein

crea

sed

risk

-tak

ing

inso

me

finan

cial

mar

kets

asou

twei

ghin

gth

ebe

nefit

sof

prom

otin

ga

stro

nger

econ

omic

reco

very

.In

flati

onis

curr

entl

ysu

bdue

d,an

din

flati

onex

pect

atio

nsap

pear

wel

lan

chor

ed.”

Ber

nank

eus

edhi

ste

stim

ony

topu

shba

ckag

ains

tco

lleag

ues

onth

eFO

MC

who

favo

red

curt

ailin

gth

e$8

5bi

llion

inm

onth

lybo

ndbu

ying

.G

ov/P

res

Tap

(Fed

eral

Res

erve

Gov

erno

ror

Fede

ralR

eser

veB

oard

Pre

side

ntTap

erin

g)

June

3,20

13:Fe

dera

lR

eser

veB

ank

ofSa

nFr

anci

sco

Pre

side

ntJo

hnW

illia

ms

said

polic

ymak

ers

may

star

tre

duci

ngth

epa

ceof

bond

purc

hase

sov

erth

ene

xtth

ree

mon

ths

and

pote

ntia

llyen

dqu

anti

tati

veea

sing

byye

ar-e

nd.H

esa

idth

atw

ith

cont

inue

d“g

ood

sign

s”on

jobs

and

confi

denc

ein

a“s

ubst

anti

alim

prov

emen

t”he

coul

dse

eas

“ear

lyas

this

sum

mer

som

ead

just

men

t,m

aybe

mod

est

adju

stm

ent

dow

nwar

d,in

our

purc

hase

prog

ram

.”(c

ontinu

ed)


Tab

le1.

(Con

tinued

)

Sou

rce

ofEve

nts

Chai

rman

Ber

nan

keG

over

nor

/Pre

siden

tFO

MC

Agg

rega

te

Tap

erin

gQ

ETap

erin

gQ

ETap

erin

gQ

ETap

erin

gQ

E

No.

ofE

vent

s3

726

220

529

33

Exa

mpl

esof

Ann

ounc

emen

ts(C

ongr

essi

onal

Tes

tim

ony,

Spee

ches

,FO

MC

Stat

emen

ts,et

c.)

Gov

/Pre

sQ

E(F

eder

alR

eser

veG

over

nor

orFe

dera

lR

eser

veB

oard

Pre

side

ntQ

uant

itat

ive

Eas

ing)

Mar

ch27

,20

13:T

wo

regi

onal

Fede

ralR

eser

vepr

esid

ents

said

they

wan

tth

eFe

dera

lR

eser

veto

keep

buyi

ngbo

nds

thro

ugh

the

end

of20

13,w

hile

ath

ird

offici

alsa

idth

ece

ntra

lba

nkis

n’t

doin

gen

ough

tosp

urec

onom

icgr

owth

.“W

esh

ould

cont

inue

our

larg

e-sc

ale

asse

tpu

rcha

ses

ofTre

asur

yan

dm

ortg

age-

back

edse

curi

ties

thro

ugh

this

year

—al

thou

ghth

eam

ount

may

need

tobe

adju

sted

upor

dow

n,de

pend

ing

onho

wth

eec

onom

icsi

tuat

ion

evol

ves,

”Fe

dera

lR

eser

veB

ank

ofB

osto

nP

resi

dent

Eri

cR

osen

gren

said

ina

spee

chin

Man

ches

ter,

New

Ham

pshi

re.“T

his

isa

poin

tw

hen

we

have

tobe

pati

ent

and

let

our

polic

ies

wor

k,”

wit

hst

imul

us“fi

ring

onal

lcy

linde

rs,”

Fede

ralR

eser

veB

ank

ofC

hica

goP

resi

dent

Cha

rles

Eva

nssa

idto

repo

rter

s.FO

MC

QE

(FO

MC

Qua

ntit

ativ

eE

asin

g)M

ay1,

2013

FO

MC

Mee

ting

Stat

emen

t:T

heFe

dera

lR

eser

vesa

idit

will

keep

buyi

ngbo

nds

ata

mon

thly

pace

of$8

5bi

llion

whi

lest

andi

ngre

ady

tora

ise

orlo

wer

purc

hase

sas

econ

omic

cond

itio

nsev

olve

.T

hepu

rcha

ses

will

rem

ain

divi

ded

betw

een

$40

billi

ona

mon

thof

mor

tgag

e-ba

cked

secu

riti

esan

d$4

5bi

llion

am

onth

ofTre

asur

yse

curi

ties

.


markets, then we would expect national equity markets to fall,exchange rates to depreciate, and (perhaps) CDS spreads to rise.We focus on twenty-six emerging markets, shown in table 2. (The“fragile” and “robust” groups and statistics are explained below.)Emerging markets were required to have stock market and CDSprice data for the entire sample period to be included in the sample.

2.2 Methodology

We employ panel fixed-effect estimation techniques to estimate theimpact of U.S. Federal Reserve announcements on financial marketprices in emerging markets. We estimate a panel data model of thefollowing form:

ΔPit = β0 + β1Ben Tapt + β2Ben QEt + β3GovPres Tapt+ β4GovPres QEt + β5FOMC QEt + μi + εit, (1)

where ΔPit is the log first difference in foreign exchange rates,national equity markets, or the change in CDS spread for countryi at time t. Ben Tapt (Ben QEt) are statements supporting taper-ing (further quantitative easing) by Chairman Bernanke at time t;GovPres Tapt (GovPres QEt) are statements by Federal ReserveBoard governors or Federal Reserve Bank presidents in support oftapering (further quantitative easing); and FOMC QEt are officialstatements following FOMC meetings supporting further quantita-tive easing. (No official FOMC statements supporting tapering wereidentified during this sample period.)

We find no evidence of dynamic adjustment in asset price move-ments (log first differences in equity markets and exchange rates, firstdifferences in CDS spreads) and considered all news events emanat-ing from the United States as strictly exogenous for emerging mar-kets. We therefore estimate the model as a static panel with countryfixed effects. We report robust standard errors clustered with coun-tries, and do not find any evidence of serial correlation in residuals.To take time differences between U.S. and Asian markets countriesinto account, all Federal Reserve announcements (independent vari-ables) are lagged one day for South Korea, Malaysia, Philippines,Thailand, China, Indonesia, India, and Pakistan. For the rest of thecountries, news events are entered in the model contemporaneously.


Table 2. Emerging-Market Sample

CAB/GDP Reserves/ ExternalCountry (%) GDP (%) Debt/GDP (%)

Robust Group

Peru −4.89 31.11 25.72Israel 2.33 28.68 37.00Korea 4.61 26.86 37.74Malaysia 3.49 43.11 32.87Philippines 2.51 27.04 30.34Thailand 0.11 41.09 37.04Bulgaria 1.15 32.66 72.56Russia 2.89 22.44 31.56China 2.50 39.32 8.99Hungary 2.22 34.22 136.53

Fragile Group

Turkey −7.38 12.85 43.30South Africa −6.07 11.82 34.40Argentina −0.75 7.13 24.25Brazil −3.38 16.75 19.74Chile −4.58 14.55 42.61Colombia −3.22 10.82 22.18Mexico −1.34 12.37 29.90India −4.41 15.02 20.85Indonesia −3.41 12.55 25.86Pakistan −0.97 3.10 24.49Ukraine −7.29 12.37 78.49Czech R. −1.76 21.75 52.33Latvia −1.14 23.82 140.82Lithuania −0.26 15.15 79.00Poland −3.03 20.03 75.37Romania −1.96 22.95 77.38

Notes: Current account balance (in percent of GDP) and international reserves (inpercent of GDP) data is taken from the International Monetary Fund’s World Eco-nomic Outlook Database, and external debt (in percent of GDP) is from the CentralIntelligence Agency’s World Factbook. A country is grouped under “robust” or “frag-ile” based on whether it has current account surplus, high reserves, and low externaldebt. Thus, if at least two of three (or three of three) of the above criteria holds,then a country is in the “robust group”; otherwise, it is in the “fragile group.” Lowinternational reserve level is defined as a reserves/GDP ratio below 20 percent, andlow external debt is defined as an external debt/GDP ratio less than 34 percent.Both cut-off points are the median values of their respective samples.


Recognizing the substantial heterogeneity among the twenty-sixemerging markets, we grouped the emerging markets according tothree fragility/strength criteria: current account deficits or surpluses;low or high international reserves; and high or low external debts.2

We estimated the effects of “news” on the full sample as well as thetwo subgroups (“robust” and “fragile”). We also grouped countriesby their level of financial development (“low” or “high”). As dis-cussed in the introduction, we anticipate different reactions to newsdepending on the state of the emerging-market economy.

3. Results

3.1 Full-Sample Results

Table 3 reports the impact of the tapering and QE news on thestock market indices, exchange rates, and sovereign spreads for thefull sample of twenty-six emerging markets from November 27, 2012to October 3, 2013. These regression results summarize the changeof market prices in the twenty-four-hour window following the news.

Tapering. Bernanke’s tapering news was associated with signif-icant drops in stock market indices and exchange rate depreciations,but had no significant impact on sovereign spreads. This may reflectthe expectation of reduced capital inflows and carry-trade activ-ity to emerging markets, with less investment in equity markets.Not surprisingly, tapering news doesn’t seem to affect CDS spreadson sovereign debt. By contrast, numerous and frequently quite vig-orous statements in support of tapering by Federal Reserve Bankpresidents had little discernible effect on emerging-market financialprices—equities, exchange rates, or CDS spreads—during our sampleperiod. (No tapering news was attributable to Federal Reserve gover-nors during this period, only Federal Reserve Bank presidents.) Thismay be because one of the Federal Reserve Bank presidents—Fisherof the Federal Reserve Bank of Dallas—made many public state-ments advocating tapering (nine of our twenty-six events) that mayhave been discounted by financial markets due to their frequency and

2Low international reserve level is defined as a reserves/GDP ratio below 20percent, and low external debt is defined as an external debt/GDP ratio less than34 percent. Both cut-off points are the median values of their respective samples.


Table 3. Full Sample—Asset Market Reactionto Federal Reserve Announcements

Stock Market Exchange Rate CDS Spreads(1) (2) (3)

Ben Tap −0.328∗ 0.219∗∗∗ −0.550(0.164) (0.064) (4.075)

Ben QE 0.128 −0.141∗∗∗ −1.419(0.128) (0.037) (1.122)

Gov/Pres Tap 0.035 −0.010 0.564(0.043) (0.012) (0.717)

Gov/Pres QE −0.176∗∗∗ −0.097∗∗∗ −3.689(0.052) (0.018) (4.417)

FOMC QE 0.164∗∗ −0.252∗∗∗ −3.535∗(0.079) (0.045) (1.936)

Constant 0.047∗∗∗ 0.033∗∗∗ 0.421(0.009) (0.004) (0.497)

Observations 5,590 5,375 5,156R-squared 0.004 0.012 0.001No. of Countries 26 25 24

Notes: The table presents panel fixed-effect estimation results with robust standarderrors in parentheses. ***, **, and * denote p < 0.01, p < 0.05, and p < 0.1, respec-tively. Independent variables are log first difference of stock market index, foreignexchange rate, and change in CDS spreads in specification 1, 2, and 3, respectively.Exchange rate estimation excludes China; CDS spread estimation excludes Chinaand Pakistan.

predictability of the message. (The other most vocal advocate fortapering was Philadelphia Federal Reserve Bank President Plosser,with five news events during the sample period.) No explicit andunambiguous news in support of tapering in FOMC announcementswas coded during this period.

Quantitative Easing. Bernanke’s QE news, symmetrically,was associated with strong exchange rate appreciation in table 3.Exchange rates also significantly appreciated in response to QEnews contained in FOMC statements and in announcements by gov-ernors/presidents, where the FOMC had the largest impact andgovernors/presidents had the least impact. The support for QEwas broader than tapering, including governors (Vice Chair Yellenand Governors Duke and Stein) and many presidents. President


Rosengren (Boston) and President Bullard (St. Louis) spoke themost frequently (four times each) in support of continuing QE. Inaddition, consistent with the very strong impact of FOMC QE newsexchange rate appreciation, these statements also had a large impacton pushing up stock market prices.

3.2 Differential Effects of Tapering

As discussed above, we expect tapering news to have different effectsdepending on the strength of a country’s international “fundamen-tals.” Full-sample results, in turn, could mask differential effects dueto group heterogeneity associated with disparate fundamentals. Toaddress this issue, tables 4–6 report the impact of the tapering andQE news on market prices, allowing comparison between countrieshaving current account deficits/surpluses (table 4), low/high inter-national reserves (table 5), and high/low external debts (table 6).(Table 2 presents the external positions of countries and the groupingbetween “robust” and “fragile” countries.) We also group countriesin terms of these fundamentals, robust and fragile, where the robustgroup are countries that meet at least two “strong” criteria (cur-rent account surplus, higher reserves, and low debt), and the fragilegroup meet at least two “weak” criteria (current account deficits,low reserves, and high debt). The countries in each group are shownin table 2, and table 7 reports the regression results for these twogroups.

This comparison reveals asymmetric and divergent patterns,depending on whether the news is emanating from Bernanke’s state-ments, as opposed to governors/presidents and the FOMC, andbetween the groups with robustness or fragile fundamentals. In par-ticular, Bernanke’s tapering news had much larger exchange ratedepreciation effects on countries with (i) current account surpluses asopposed to deficits (four times larger, table 4), (ii) high internationalreserves contrasted with low reserves (three times larger, table 5),and (iii) low external debt rather than high debt (about half as muchlarger, table 6).3 The analysis of the two groups—fragile or robust,

3The differences between the groups are statistically significant in the casesof exchange rates (chi-square of 4.56 and probability of 0.033) and internationalreserves.


Tab

le4.

Diff

eren

tial

Effec

ts—

Curr

ent

Acc

ount

Defi

cit

and

Surp

lus

Cou

ntr

ies

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

CA

CA

CA

CA

CA

CA

Defi

cit

Surp

lus

Defi

cit

Surp

lus

Defi

cit

Surp

lus

(1a)

(1b)

(2a)

(2b)

(3a)

(3b)

Ben

Tap

−0.

229

−0.

516

0.11

4∗0.

441∗

∗−

2.76

73.

877∗

∗

(0.1

98)

(0.2

95)

(0.0

56)

(0.1

36)

(6.0

41)

(1.5

59)

Ben

QE

0.20

3∗∗

−0.

013

−0.

129∗

∗∗−

0.16

7∗−

2.30

00.

321

(0.0

79)

(0.3

42)

(0.0

39)

(0.0

86)

(1.6

51)

(0.3

40)

Gov

/Pre

sTap

−0.

001

0.10

3−

0.01

70.

002

1.05

4−

0.42

8(0

.060

)(0

.088

)(0

.016

)(0

.018

)(1

.052

)(0

.264

)G

ov/P

res

QE

−0.

152∗

−0.

222∗

∗∗−

0.10

8∗∗∗

−0.

073∗

∗∗−

5.96

40.

968∗

∗

(0.0

72)

(0.0

65)

(0.0

25)

(0.0

19)

(6.5

52)

(0.3

28)

FO

MC

QE

0.22

3∗∗

0.05

0−

0.24

7∗∗∗

−0.

264∗

∗∗−

4.69

3−

1.13

1(0

.101

)(0

.123

)(0

.064

)(0

.045

)(2

.820

)(0

.836

)C

onst

ant

0.04

7∗∗∗

0.04

7∗∗∗

0.04

2∗∗∗

0.01

40.

678

−0.

102∗

(0.0

12)

(0.0

12)

(0.0

05)

(0.0

08)

(0.7

39)

(0.0

50)

Obs

erva

tion

s3,

679

1,91

13,

679

1,69

63,

460

1,69

6R

-squ

ared

0.00

40.

006

0.00

90.

024

0.00

20.

014

No.

ofC

ount

ries

189

188

178

Note

s:T

heta

ble

pres

ents

pane

lfix

ed-e

ffec

tes

tim

atio

nre

sult

sw

ith

robu

stst

anda

rder

rors

inpa

rent

hese

s.**

*,**

,an

d*

deno

tep

<0.

01,p

<0.

05,an

dp

<0.

1,re

spec

tive

ly.In

depen

dent

vari

able

sar

elo

gfir

stdi

ffer

ence

ofst

ock

mar

ket

inde

x,fo

reig

nex

chan

gera

te,

and

chan

gein

CD

Ssp

read

sin

spec

ifica

tion

1,2,

and

3,re

spec

tive

ly.

Exc

hang

era

tees

tim

atio

nex

clud

esC

hina

;C

DS

spre

ades

tim

atio

nex

clud

esC

hina

and

Pak

ista

n.


Tab

le5.

Diff

eren

tial

Effec

ts—

Inte

rnat

ional

Res

erve

s:Low

-an

dH

igh-R

eser

veC

ountr

ies

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

Low

Hig

hLow

Hig

hLow

Hig

hR

eser

ves

Res

erve

sR

eser

ves

Res

erve

sR

eser

ves

Res

erve

s(1

a)(1

b)

(2a)

(2b)

(3a)

(3b)

Ben

Tap

−0.

344

−0.

314

0.12

20.

308∗

∗∗−

4.94

53.

166∗

∗∗

(0.2

73)

(0.2

04)

(0.0

78)

(0.0

97)

(8.8

56)

(1.0

12)

Ben

QE

0.23

8∗∗

0.03

5−

0.08

0∗−

0.19

7∗∗∗

−2.

699

−0.

341

(0.1

07)

(0.2

20)

(0.0

44)

(0.0

56)

(2.4

19)

(0.3

75)

Gov

/Pre

sTap

−0.

017

0.08

0−

0.03

00.

007

1.67

5−

0.37

3∗

(0.0

68)

(0.0

54)

(0.0

21)

(0.0

13)

(1.5

21)

(0.1

74)

Gov

/Pre

sQ

E−

0.21

0∗−

0.14

6∗∗

−0.

112∗

∗∗−

0.08

3∗∗∗

−8.

914

0.75

4∗∗∗

(0.0

97)

(0.0

52)

(0.0

34)

(0.0

15)

(9.6

08)

(0.2

19)

FO

MC

QE

0.25

1∗0.

090

−0.

230∗

∗−

0.27

3∗∗∗

−6.

822

−0.

755

(0.1

34)

(0.0

91)

(0.0

90)

(0.0

32)

(4.0

54)

(0.5

36)

Con

stan

t0.

062∗

∗∗0.

034∗

∗∗0.

056∗

∗∗0.

013∗

∗0.

998

−0.

067∗

(0.0

14)

(0.0

11)

(0.0

06)

(0.0

05)

(1.0

88)

(0.0

34)

Obs

erva

tion

s2,

580

3,01

02,

580

2,79

52,

361

2,79

5R

-squ

ared

0.00

60.

003

0.00

80.

019

0.00

20.

011

No.

ofC

ount

ries

1214

1213

1113

Note

s:Se

eno

tes

tota

ble

4.


Tab

le6.

Diff

eren

tial

Effec

ts—

Inte

rnat

ional

Deb

tLev

els:

Hig

h-

and

Low

-Deb

tC

ountr

ies

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

Hig

hLow

Hig

hLow

Hig

hLow

Deb

tD

ebt

Deb

tD

ebt

Deb

tD

ebt

(1a)

(1b)

(2a)

(2b)

(3a)

(3b)

Ben

Tap

0.02

1−

0.67

7∗∗∗

0.18

0∗∗

0.26

1∗∗

0.77

8∗−

2.12

1(0

.224

)(0

.204

)(0

.082

)(0

.102

)(0

.403

)(9

.095

)B

enQ

E−

0.08

00.

337∗

∗−

0.20

7∗∗∗

−0.

069

0.18

7−

3.31

0(0

.219

)(0

.113

)(0

.052

)(0

.047

)(0

.443

)(2

.315

)G

ov/P

res

Tap

0.04

30.

027

0.00

8−

0.03

0−

0.12

71.

388

(0.0

68)

(0.0

55)

(0.0

11)

(0.0

22)

(0.2

48)

(1.5

45)

Gov

/Pre

sQ

E−

0.20

6∗∗

−0.

145∗

−0.

081∗

∗∗−

0.11

4∗∗∗

0.66

3∗∗

−8.

877

(0.0

76)

(0.0

73)

(0.0

16)

(0.0

33)

(0.2

26)

(9.6

82)

FO

MC

QE

0.00

40.

325∗

∗∗−

0.23

9∗∗∗

−0.

266∗

∗∗−

0.74

4−

6.83

4(0

.116

)(0

.092

)(0

.041

)(0

.086

)(0

.554

)(4

.046

)C

onst

ant

0.04

8∗∗∗

0.04

6∗∗∗

0.01

0∗0.

058∗

∗∗0.

095∗

∗0.

804

(0.0

15)

(0.0

09)

(0.0

05)

(0.0

07)

(0.0

39)

(1.0

84)

Obs

erva

tion

s2,

795

2,79

52,

795

2,58

02,

795

2,36

1R

-squ

ared

0.00

30.

011

0.01

50.

011

0.00

20.

002

No.

ofC

ount

ries

1313

1312

1311

Note

s:Se

eno

tes

tota

ble

4.


le7.

Diff

eren

tial

Effec

ts—

Rob

ust

and

Fra

gile

Cou

ntr

ies

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

Fra

gile

Rob

ust

Fra

gile

Rob

ust

Fra

gile

Rob

ust

(1a)

(1b)

(2a)

(2b)

(3a)

(3b)

Ben

Tap

−0.

234

−0.

478

0.11

1∗0.

409∗

∗−

3.28

94.

008∗

∗

(0.2

11)

(0.2

66)

(0.0

60)

(0.1

24)

(6.4

32)

(1.3

81)

Ben

QE

0.20

5∗∗

0.00

6−

0.13

4∗∗∗

−0.

154∗

−2.

316

0.05

8(0

.084

)(0

.306

)(0

.041

)(0

.077

)(1

.764

)(0

.399

)G

ov/P

res

Tap

0.00

40.

085

−0.

016

−0.

001

1.16

7−

0.45

2∗

(0.0

63)

(0.0

80)

(0.0

17)

(0.0

17)

(1.1

18)

(0.2

34)

Gov

/Pre

sQ

E−

0.16

3∗∗

−0.

196∗

∗−

0.11

0∗∗∗

−0.

073∗

∗∗−

6.40

30.

928∗

∗

(0.0

76)

(0.0

63)

(0.0

26)

(0.0

17)

(6.9

87)

(0.2

91)

FO

MC

QE

0.21

0∗0.

089

−0.

250∗

∗∗−

0.25

6∗∗∗

−4.

877

−1.

218

(0.1

07)

(0.1

16)

(0.0

68)

(0.0

40)

(3.0

06)

(0.7

40)

Con

stan

t0.

058∗

∗∗0.

029∗

∗0.

042∗

∗∗0.

018∗

∗0.

706

−0.

062

(0.0

13)

(0.0

11)

(0.0

05)

(0.0

07)

(0.7

89)

(0.0

47)

Obs

erva

tion

s3,

464

2,12

63,

464

1,91

13,

245

1,91

1R

-squ

ared

0.00

40.

005

0.00

90.

023

0.00

20.

015

No.

ofC

ount

ries

1710

179

169

Note

s:T

his

tabl

epr

esen

tspa

nel

fixed

-effec

tes

tim

atio

nre

sult

sw

ith

robu

stst

anda

rder

rors

inpa

rent

hese

s.**

*,**

,an

d*

deno

tep

<0.

01,p

<0.

05,an

dp

<0.

1,re

spec

tive

ly.In

depen

dent

vari

able

sar

elo

gfir

stdi

ffer

ence

ofst

ock

mar

ket

inde

x,fo

reig

nex

chan

gera

te,an

dch

ange

inC

DS

spre

ads

insp

ecifi

cati

on1,

2,an

d3,

resp

ecti

vely

.A

coun

try

isgr

ouped

unde

r“r

obus

t”or

“fra

gile

”ba

sed

onw

heth

erit

has

curr

ent

acco

unt

surp

lus,

high

rese

rves

,an

dlo

wex

tern

alde

bt.T

hus,

ifat

leas

ttw

oof

thre

e(o

rth

ree

ofth

ree)

ofth

eab

ove

crit

eria

hold

s,th

ena

coun

try

isin

the

“rob

ust

grou

p”;o

ther

wis

e,it

isin

the

“fra

gile

grou

p.”

Exc

hang

era

tees

tim

atio

nex

clud

esC

hina

;C

DS

spre

ades

tim

atio

nex

clud

esC

hina

and

Pak

ista

n.


shown in table 7—also indicates that exchange rate depreciation isstatistically significant in both cases at the time of Bernanke’s taper-ing announcements but more than three times larger in the robustgroup (0.409) compared with the fragile group (0.111).4

By contrast, tapering announcements by Federal Reserve gover-nors or presidents had little or no discernible impact on exchangerates in emerging markets, regardless of whether they were classi-fied as robust/weak in fundamentals, had current account surplusesor deficits, had high/low international reserves, or had low/highexternal debt. This is consistent with the full-sample results.

Bernanke’s tapering news increased CDS spreads very substan-tially for countries with robust fundamentals (4.0 points; table 7),and especially for those countries with current account surpluses(3.8 points; table 4) and high international reserves (3.2 points;table 5), while having little or no effect on fragile countries (exceptfor increasing spreads marginally, 0.77 points, for those with highexternal debt positions). And, similar to the full-sample results,tapering announcements by Federal Reserve governors/presidentshad little effect on CDS spreads, lowering them overall by only –0.45(table 7; significant at the 10 percent level), apparently attribut-able to the small decline in CDS in the high-reserves group (–0.37;table 5).

Bernanke’s tapering news lowered equity market values in emerg-ing markets for the full sample (–3.28 percent; table 3) and in thecountries with low external debt positions (table 6). And, similar tothe full-sample results, group results based on fundamentals indi-cated that tapering announcements from governors/presidents didnot have any significant impact on stock markets.

Intriguingly, tables 4–7 imply that Bernanke’s tapering newsdepreciated exchange rates most in countries typically associ-ated with strong international positions—countries running currentaccount surpluses, having high reserves, and having low debt—andour aggregate of robust countries. A possible interpretation is thatfragile economies were less exposed to financial flows in search ofhigher yields during the earlier QE years; thereby they were expectedto be less exposed to the immediate impact of the tapering news.

4This difference is statistically significant, with a chi-square of 4.31 (probabil-ity 0.038).


3.3 Differential Effects of Quantitative Easing

The effects of QE news also showed substantial variation dependingon who made the statement and the international position of thecountry. Bernanke’s QE news was associated with strong exchangerate appreciation in the full sample (–0.14; table 3) as well as inboth fragile and robust groups (table 7), where the impact effectswere virtually identical (–0.13 and –0.15, respectively). Apprecia-tion was an even more dominating characteristic of countries withstrong current account positions, high international reserves, andhigh international debt.5

Significant exchange rate appreciations were also associated withFOMC QE statements in the full sample, in both fragile and weakgroups (table 7), and in all of the subsamples (table 4). Interest-ingly, no asymmetric effects were found between those with weak orstrong fundamentals—the coefficients on the exchange rate responseto FOMC QE support news were virtually identical (and significant)across subgroups.

Differential effects emerge again in stock market and CDSresponses to QE. Bernanke’s and the FOMC’s QE news were associ-ated with higher stock prices of the fragile group but did not affectthe stock prices of the strong group. By contrast, QE news from gov-ernors/presidents was associated with lower stock market indices forboth fragile and robust groups (table 7), and coefficient estimatesare similar in magnitudes regardless of international fundamentalpositions. In terms of CDS spreads, QE news from Federal Reservegovernors/presidents had no discernible effect in the full sample butappears to raise spreads for the robust group of countries (table 7).This result stems from the rise in CDS spreads in countries with highinternational reserve positions. However, somewhat at odds with theother results, spreads climbed in those countries with relatively highexternal debt positions. The differential impacts of the news sourcereflect the much higher frequency of news from governors/presidents(twenty-six tapering, twenty-two QE) in comparison with news fromBernanke (three tapering and seven QE) and FOMC QE news (five).

5The difference between depreciation in countries with high and low inter-national debt positions is statistically significant with a chi-square value of 3.25(probability 0.072).


On balance, the market was focused and reacted more to the scarcerand more coherent news from Bernanke and the FOMC than tothe frequent and diffused news from governors/presidents. Gover-nors’/presidents’ news reflected the inner debate among heteroge-neous and non-coordinated views of the Federal Reserve’s senior offi-cials, whereas Bernanke’s and FOMC QE news were viewed as muchclearer signals regarding the stance of the Federal Reserve’s policies.These interpretations are also supported by the results reported forthe full sample (table 3).

3.4 Robust and Fragile Fundamentals and Tapering Dynamics

Emerging-market countries with robust (fragile) fundamentals—those with current account surpluses (deficits), high (low) inter-national reserves, and low (high) external debt—were most (less)affected by tapering announcements, particularly with respect toexchange rates. Robust (fragile) economies seemed to have beenmore (less) exposed to financial flows in search of higher yieldsduring the earlier QE years, and also more (less) exposed whentapering started, i.e., markets may have expected capital flow rever-sals to occur mainly in countries that had experienced the largestinflows initially. However, this may simply be the impact effect.Fragile economies could have experienced tapering effects with somedelay.

This interpretation is consistent with figures 1–6. Figures 1, 2,and 3 show the impact of Bernanke’s tapering announcements onstock prices, exchange rates, and CDS spreads, respectively, forthe fragile and strong groups. Though we find the impact effectsof Bernanke’s tapering news depreciated exchange rates most inthe robust group, the fragile group experienced more depreciationafter several months (figure 1). And while Bernanke’s tapering newscaused CDS spreads to rise substantially more in the robust group,CDS spreads eventually climbed more in the fragile group (figure 3).Stock prices did not show significant differential impact effects acrossthe fragile and robust groups, but over time it appears that therobust group has performed less well.

More formally, table 8 considers the differential impact ofannouncements on cumulative asset price changes over twenty-onebusiness days (approximately one month). The negative cumulative


Figure 1. Foreign Exchange Rate and Tapering “Events”

.98

11.

021.

041.

061.

08

Nov1 Feb1 May1 Aug1 Nov1

Robust Group Fragile Group

Notes: Vertical dashed lines indicate Bernanke’s tapering announcement dates(March 20, 2013; May 22, 2013; June 19, 2013) as described in the data section.For each country, the nominal exchange rate (national currency per U.S. dollar)index is constructed by setting the value equal to “1” at the beginning of oursample, November 27, 2012. The (unweighted) average for the “robust group”(Peru, Israel, South Korea, Malaysia, Philippines, Thailand, Bulgaria, Russia,and Hungary) and the “fragile group” (Turkey, South Africa, Argentina, Brazil,Chile, Colombia, Mexico, India, Indonesia, Pakistan, Ukraine, Czech Repub-lic, Latvia, Lithuania, Poland, and Romania) is taken on each day over thesample.

effect of Bernanke’s tapering news on fragile economies is somewhatlarger on equity prices than in robust economies. The cumulativeimpact on exchange rate depreciation in fragile and robust economiesfollowing Bernanke’s tapering announcements is very similar. Bycontrast, fragile countries experienced twice the average cumulativeincrease in CDS spreads (45 basis points) as did robust countries (18basis points) in response to Bernanke’s tapering announcements.The cumulative impacts of tapering announcements by the Fed-eral Reserve governors/presidents is more mixed—relatively smalldeclines in equity markets for the fragile group, greater exchangerate depreciation for the fragile group, and a much larger (thoughnot statistically significant) rise in CDS spreads for the fragilegroup.


Figure 2. Stock Market Index and Tapering “Events”

100

105

110



Notes: Vertical dashed lines indicate Bernanke’s tapering announcement dates(March 20, 2013; May 22, 2013; June 19, 2013) as described in the data section.For each country, the stock market index is constructed by setting the value equalto “100” at the beginning of our sample, November 27, 2012. The (unweighted)average for the “robust group” (Peru, Israel, South Korea, Malaysia, Philippines,Thailand, Bulgaria, Russia, and Hungary) and the “fragile group” (Turkey, SouthAfrica, Argentina, Brazil, Chile, Colombia, Mexico, India, Indonesia, Pakistan,Ukraine, Czech Republic, Latvia, Lithuania, Poland, and Romania) is taken oneach day over the sample.

Overlapping events is a major drawback when measuring longer-duration effects using our methodology. To address this issue, weestimated the effects of Bernanke’s tapering news alone on cumu-lative asset price changes for non-overlapping announcements. Non-overlapping events in this instance led to a nineteen-day post-eventwindow for the asset price changes and are presented in table 9. Theresults are virtually identical to the results presented in the previoustable.

Figures 4, 5, and 6 report the stock market indices, exchangerates, and sovereign spreads of Brazil and Turkey (relatively fragilecountries) and Hungary (a relatively robust country), from Novem-ber 27, 2012 to October 3, 2013. Again, toward the end of the sampleperiod, the fragile countries were not insulated from the taperingnews. Indeed, over time the fragile countries were hit harder thanthe stronger countries.


Figure 3. CDS Spreads and Tapering “Events”

8010

012

014

0



Notes: Vertical dashed lines indicate Bernanke’s tapering announcement dates(March 20, 2013; May 22, 2013; June 19, 2013) as described in the data section.For each country, the CDS spread index is constructed by setting the value equalto “100” at the beginning of our sample, November 27, 2012. The (unweighted)average for the “robust group” (Peru, Israel, South Korea, Malaysia, Philippines,Thailand, Bulgaria, Russia, and Hungary) and the “fragile group” (Turkey, SouthAfrica, Argentina, Brazil, Chile, Colombia, Mexico, India, Indonesia, Pakistan,Ukraine, Czech Republic, Latvia, Lithuania, Poland, and Romania) is taken oneach day over the sample.

The impact of the tapering news focused first on the strong coun-tries, probably in anticipation of large short-run outflows of past hotmoney inflows. Market attention shifted over time to the possibilitythat fragile countries would find it harder to adjust to the higher U.S.interest rates induced by future tapering, leading to large adjust-ments in the last quarter of 2013. Arguably, the initial large effectof the tapering news, affecting mostly the strong countries, may alsoreflect financial markets’ initial inattention to tail risks, overlook-ing the vulnerability of the weaker emerging markets to the adverseimplications of higher future global interest rates. Indeed, in the lastquarter of 2013, financial markets refocused attention on the fragileemerging markets and hammered the prices of the “Fragile Five”—Brazil, India, Indonesia, South Africa, and Turkey—a subgroup ofthe weaker emerging markets.


Tab

le8.

Diff

eren

tial

Effec

tsO

ver

One-

Mon

thEve

nt

Win

dow

and

Ass

etP

rice

Chan

ges—

Rob

ust

and

Fra

gile

Cou

ntr

ies

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

Fra

gile

Rob

ust

Fra

gile

Rob

ust

Fra

gile

Rob

ust

(1a)

(1b)

(2a)

(2b)

(3a)

(3b)

Ben

Tap

−3.

378∗

∗−

2.64

4∗∗

1.43

7∗∗∗

1.55

8∗∗∗

45.4

17∗

18.4

84∗∗

∗

(1.1

94)

(0.8

33)

(0.3

12)

(0.2

57)

(24.

965)

(2.6

12)

Ben

QE

0.90

31.

488∗

∗−

0.06

20.

471

9.82

93.

129

(0.6

47)

(0.5

96)

(0.2

75)

(0.3

27)

(15.

972)

(2.6

65)

Gov

/Pre

sTap

−1.

377∗

∗∗−

2.30

2∗∗∗

0.83

8∗∗∗

0.27

5∗∗

68.0

365.

615∗

∗∗

(0.3

52)

(0.6

51)

(0.2

32)

(0.1

15)

(57.

096)

(1.0

92)

Gov

/Pre

sQ

E−

1.21

6∗∗∗

−1.

424∗

∗−

0.70

6∗∗

−0.

190

−34

.842

0.89

9(0

.406

)(0

.628

)(0

.288

)(0

.337

)(3

7.06

5)(1

.423

)FO

MC

QE

0.96

72.

001∗

−1.

282∗

∗∗−

0.84

3∗∗∗

−31

.166

−10

.358

∗∗∗

(0.5

80)

(0.9

83)

(0.2

71)

(0.2

27)

(21.

313)

(2.9

17)

Con

stan

t3.

601∗

∗∗3.

168∗

∗0.

338

−0.

301

−33

.890

−7.

843∗

∗∗

(0.7

68)

(1.0

97)

(0.2

84)

(0.4

17)

(25.

013)

(1.9

90)

Obs

erva

tion

s3,

784

2,32

63,

784

2,09

13,

537

2,09

1R

-squ

ared

0.12

40.

179

0.16

90.

188

0.02

60.

267

No.

ofC

ount

ries

1710

179

169

Note

s:Se

eno

testo

tabl

e7.

Eve

ntw

indo

wsar

eex

tend

edto

twen

ty-o

neda

ys(s

tart

ing

from

the

day

ofan

noun

cem

ents

),an

dcu

mul

ativ

eas

set

pric

ech

ange

sov

erth

esa

me

even

tw

indo

war

eco

nsid

ered

.


Tab

le9.

Diff

eren

tial

Effec

tsO

ver

Non

-Ove

rlap

pin

g(O

ne-

Mon

th)

Eve

nts

—R

obust

and

Fra

gile

Cou

ntr

ies

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

Fra

gile

Rob

ust

Fra

gile

Rob

ust

Fra

gile

Rob

ust

Ben

Tap

−4.

246∗

∗∗−

3.99

8∗∗∗

1.66

1∗∗∗

1.63

8∗∗∗

46.0

49∗∗

21.1

73∗∗

∗

(1.1

52)

(0.9

92)

(0.3

64)

(0.2

94)

(20.

701)

(2.8

92)

Con

stan

t2.

170∗

∗∗1.

447∗

∗∗−

0.09

4−

0.43

4∗∗∗

−9.

695∗

−5.

847∗

∗∗

(0.2

97)

(0.2

60)

(0.0

94)

(0.0

77)

(5.3

49)

(0.7

59)

Obs

erva

tion

s3,

784

2,32

63,

784

2,09

13,

537

2,09

1R

-squ

ared

0.11

40.

129

0.09

10.

127

0.01

00.

198

No.

ofC

ount

ries

1710

179

169

Note

s:Se

eno

tes

tota

ble

7.T

heeff

ects

ofB

erna

nke’

sta

per

ing

new

sar

ees

tim

ated

alon

eon

cum

ulat

ive

asse

tpr

ice

chan

ges

for

non-

over

lapp

ing

anno

unce

men

tsov

era

nine

teen

-day

pos

t-ev

ent

win

dow

.


Figure 4. Selected Countries: Stock Marketand Tapering “Events”

8090

100

110

120

130


Turkey BrazilHungary

Notes: Vertical dashed lines indicate Bernanke’s tapering announcement dates(March 20, 2013; May 22, 2013; June 19, 2013) as described in the data section.For each country, the stock market index is constructed by setting the value equalto “100” at the beginning of our sample, November 27, 2012.

Figure 5. Selected Countries: Foreign Exchange Rateand Tapering “Events”

.91

1.1

1.2



Notes: Vertical dashed lines indicate Bernanke’s tapering announcement dates(March 20, 2013; May 22, 2013; June 19, 2013) as described in the data section.For each country, the nominal exchange rate (national currency per U.S. dollar)index is constructed by setting the value equal to “1” at the beginning of oursample, November 27, 2012.


Figure 6. Selected Countries: CDS Spreadsand Tapering “Events”

5010

015

020

0



Notes: Vertical dashed lines indicate Bernanke’s tapering announcement dates(March 20, 2013; May 22, 2013; June 19, 2013) as described in the data section.For each country, the CDS spread index is constructed by setting the value equalto “100” at the beginning of our sample, November 27, 2012.

4. Extensions

4.1 Fragile and Robust Groupings

In the previous sections we distinguished between fragile and robusteconomies using the following criteria: (i) median values for externaldebt (above the median indicates fragile and below median indicatesrobust) and international reserves (above median indicates robustand below median indicates fragile), and (ii) current account sur-pluses (robust) and deficits (fragile). We modify these criteria tobe median values for all three fundamentals, i.e., changing our cri-teria for current accounts from surplus/deficits to above/below themedian values in the sample. In table 10 we present results where theclassification of a “robust” current account is any value above themedian (–1.24) and a “fragile” current account is any value below themedian. This gives somewhat different groupings for robust/fragile,with Argentina, Latvia, and Pakistan moving from the fragile tothe robust group. Results shown in table 10 are almost identical to

344 International Journal of Central Banking June 2016Tab

le10

.D

iffer

ential

Effec

ts—

Alter

nat

ive

Rob

ust

and

Fra

gile

Cou

ntr

yG

roups

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

Fra

gile

Rob

ust

Fra

gile

Rob

ust

Fra

gile

Rob

ust

(1a)

(1b)

(2a)

(2b)

(3a)

(3b)

Ben

Tap

−0.

244

−0.

412∗

0.14

1∗0.

303∗

∗3.

194∗

∗−

4.97

3(0

.261

)(0

.206

)(0

.072

)(0

.107

)(1

.393

)(8

.775

)B

enQ

E0.

271∗

∗∗−

0.01

4−

0.14

2∗∗

−0.

140∗

∗−

0.41

0−

2.60

3(0

.079

)(0

.241

)(0

.047

)(0

.060

)(0

.440

)(2

.400

)G

ov/P

res

Tap

0.02

30.

047

−0.

022

0.00

20.

056

1.16

6(0

.068

)(0

.055

)(0

.021

)(0

.013

)(0

.191

)(1

.568

)G

ov/P

res

QE

−0.

132

−0.

220∗

∗∗−

0.12

8∗∗∗

−0.

063∗

∗∗0.

657∗

∗∗−

8.86

7(0

.088

)(0

.056

)(0

.028

)(0

.017

)(0

.171

)(9

.682

)FO

MC

QE

0.19

70.

132

−0.

274∗

∗∗−

0.22

9∗∗∗

−2.

132∗

∗∗−

5.19

1(0

.131

)(0

.094

)(0

.081

)(0

.040

)(0

.538

)(4

.229

)C

onst

ant

0.02

00.

074∗

∗∗0.

042∗

∗∗0.

024∗

∗∗0.

230∗

∗∗0.

644

(0.0

15)

(0.0

09)

(0.0

05)

(0.0

05)

(0.0

37)

(1.0

82)

Obs

erva

tion

s2,

795

2,79

52,

795

2,58

02,

791

2,36

5R

-squ

ared

0.00

50.

005

0.01

00.

019

0.00

60.

002

No.

ofC

ount

ries

1313

1312

1311

Note

s:T

heta

ble

pres

ents

pane

lfix

ed-e

ffec

tes

tim

atio

nre

sult

sw

ith

robu

stst

anda

rder

rors

inpa

rent

hese

s.**

*,**

,an

d*

deno

tep

<0.

01,p

<0.

05,an

dp

<0.

1,re

spec

tive

ly.In

depen

dent

vari

able

sar

elo

gfir

stdi

ffer

ence

ofst

ock

mar

ket

inde

x,fo

reig

nex

chan

gera

te,an

dch

ange

inC

DS

spre

ads

insp

ecifi

cati

on1,

2,an

d3,

resp

ecti

vely

.A

coun

try

isgr

ouped

unde

r“r

obus

t”or

“fra

gile

”ba

sed

onw

heth

erit

has

high

curr

ent

acco

unt

bala

nce

(im

plyi

nggr

eate

rth

anm

edia

nva

lue

ofth

esa

mpl

eof

coun

trie

s),hi

ghre

serv

es,an

dlo

wex

tern

alde

bt.T

hus,

ifat

leas

ttw

oof

thre

e(o

rth

ree

ofth

ree)

ofth

eab

ove

crit

eria

hold

s,th

ena

coun

try

isin

the

“rob

ust

grou

p”;

othe

rwis

e,it

isin

the

“fra

gile

grou

p.”


those in table 4, however, and the main results are qualitatively notaffected by this change in the robust/fragile groups.

4.2 Financial Development

Another issue is whether the degree of financial development is acritical distinguishing characteristic between countries in how theyrespond to tapering announcements. That is, is this institutionalfeature important in how tapering and quantitative easing announce-ments affect asset prices in emerging markets? To address this issuewe use the World Economic Forum’s Financial Development Index,and use the median value of the sample to divide countries into“high” and “low” degrees of financial development. Table 11 reportsthe results in an analogous fashion to table 4.

The results indicate that countries with a high degree of finan-cial development had larger declines in equity markets in response toBernanke’s tapering announcements than did those with low degreesof financial development. They also experienced greater exchangerate depreciation and larger increases in CDS spreads than didcountries with low levels of financial development. The “taperingresponse” of more financially developed (less financially developed)emerging economies is broadly similar to those with the robust (frag-ile) fundamentals, providing an alternative explanation for why thesecountries were hardest hit by tapering announcements.

4.3 U.S. Interest Rates

Another important question is whether tapering and QE announce-ments are transmitted to emerging-market equity prices, exchangerates, and CDS spreads through U.S. interest rates only, or whetherthere is an additional transmission channel at work. To address thisissue, we included in the baseline regression (table 3) U.S. interestrates of various maturities. In particular, we considered one-month,three-month, six-month, one-year, five-year, ten-year, and twenty-year U.S. Treasury interest rates, each entered separately in theregression equations.

The response of equity prices to Bernanke tapering results wasvery stable regardless of which U.S. interest rate was included inthe regression, with coefficients ranging from –0.321 to –0.428, all


le11

.D

iffer

ential

Effec

ts—

Fin

anci

alD

evel

opm

ent

(WEF

Index

)

Sto

ckM

arke

tExch

ange

Rat

eC

DS

Spre

ads

Les

sM

ore

Les

sM

ore

Les

sM

ore

Dev

elop

edD

evel

oped

Dev

elop

edD

evel

oped

Dev

elop

edD

evel

oped

(1a)

(1b)

(2a)

(2b)

(3a)

(3b)

Ben

Tap

−0.

232

−0.

424∗

0.11

00.

337∗

∗∗−

3.78

62.

684∗

∗

(0.2

67)

(0.1

98)

(0.0

78)

(0.0

96)

(8.1

64)

(0.9

74)

Ben

QE

0.23

3∗0.

024

−0.

097∗

−0.

188∗

∗∗−

2.58

5−

0.25

4(0

.125

)(0

.224

)(0

.054

)(0

.050

)(2

.209

)(0

.387

)G

ov/P

res

Tap

0.01

70.

053

−0.

003

−0.

018

1.26

0−

0.12

8(0

.066

)(0

.058

)(0

.018

)(0

.017

)(1

.434

)(0

.110

)G

ov/P

res

QE

−0.

185∗

−0.

166∗

∗∗−

0.08

5∗∗∗

−0.

110∗

∗∗−

7.87

30.

495∗

∗

(0.0

94)

(0.0

49)

(0.0

25)

(0.0

27)

(8.8

53)

(0.2

23)

FO

MC

QE

0.20

40.

125

−0.

245∗

∗∗−

0.26

0∗∗∗

−6.

125

−0.

946

(0.1

37)

(0.0

84)

(0.0

68)

(0.0

62)

(3.7

66)

(0.5

34)

Con

stan

t0.

057∗

∗∗0.

037∗

∗∗0.

042∗

∗∗0.

024∗

∗∗0.

846

−0.

004

(0.0

14)

(0.0

11)

(0.0

05)

(0.0

06)

(0.9

97)

(0.0

35)

Obs

erva

tion

s2,

795

2,79

52,

795

2,58

02,

576

2,58

0R

-squ

ared

0.00

50.

004

0.00

80.

017

0.00

20.

009

No.

ofC

ount

ries

1313

1312

1212

Note

s:T

heta

ble

pres

ents

pane

lfix

ed-e

ffec

tes

tim

atio

nre

sult

sw

ith

robu

stst

anda

rder

rors

inpa

rent

hese

s.**

*,**

,an

d*

deno

tep

<0.

01,p

<0.

05,an

dp

<0.

1,re

spec

tive

ly.In

depen

dent

vari

able

sar

elo

gfir

stdi

ffer

ence

ofst

ock

mar

ket

inde

x,fo

reig

nex

chan

gera

te,an

dch

ange

inC

DS

spre

ads

insp

ecifi

cati

on1,

2,an

d3,

resp

ecti

vely

.Les

s(m

ore)

finan

cial

deve

lopm

ent

isba

sed

onth

eW

orld

Eco

nom

icFo

rum

(WE

F)

Fin

anci

alD

evel

opm

ent

Inde

x(2

012)

valu

eof

less

(mor

e)th

an3.

29(o

vera

llin

dex

valu

e).T

heW

EF

inde

xis

not

repor

ted

for

Lit

huan

ia,

Lat

via,

and

Bul

gari

a;th

us,

base

don

thei

rcr

edit

/GD

Pm

easu

res,

we

incl

ude

Lit

huan

iain

the

less

deve

loped

grou

p,an

dLat

via

and

Bul

gari

ain

the

mor

ede

velo

ped

grou

p.


of which were statistically significant (compared with the baselineregression coefficient of –0.327). Similarly, the baseline coefficient onthe exchange rate from Bernanke tapering was 0.129, compared withcoefficients ranging from 0.123 to 0.225 (six of seven are statisticallysignificant) when U.S. interest rates were included. Bernanke taper-ing did not have a significant impact on CDS spreads in the base-line regression, and was also not significant in the seven regressionswhere U.S. interest rates were included. The other results are alsovery similar to the baseline regressions, indicating that controllingfor U.S. interest rates does not change the transmission of tapering(or quantitative easing) “news” announcements on emerging-marketasset prices. (Detailed results are not included for brevity but areavailable from the authors upon request.)

While we focused above on price transmission channels, operat-ing via the exchange rate and asset prices, some transmission mayoperate through quantities, impacting the emerging markets via cap-ital flows (Caruana 2012). Specifically, tapering information maychange the expected future cross-border bank landing (the creditchannel) and portfolio flows (market risk taking).6

4.4 Interactions with Fundamentals

An alternative approach to grouping countries with respect totheir common fundamentals (“robust” or “fragile”) is to interactannouncements with the level of the fundamental (e.g., BenTap ×Reserves/GDP). We conduct this analysis for exchange rates andthe five types of announcements, reporting the results in table 12.Table 13 reports the joint significance tests (of the announce-ment coefficient and the interaction term coefficient). If countries

6The value of aggregate cross-border bond and equity investment in emerging-market economies increased from $3.29 trillion at the end of 2007 to $4.46 trillionat the end of 2012, according to the International Monetary Fund’s CoordinatedPortfolio Investment Survey. About 85 percent of the increase was in the formof debt, with a much smaller part in equity. Thus, cross-border debt invest-ment has been a significant driver of credit growth in many emerging markets(Mohanty 2014). These developments may have increased the emerging mar-kets’ vulnerability. Such borrowing may be highly procyclical, increasing theborrowers’ exposure to exchange rate risk—at times of stress, sharp exchangerate depreciations may result in further selling pressure on emerging-marketeconomies.


le12

.In

tera

ctio

ns

with

Fundam

enta

ls—

Exch

ange

Rat

eEst

imat

ions

(1)

(2)

(3)

(4)

(5)

Ben

Tap

0.21

9∗∗∗

0.27

2∗∗∗

−0.

024

0.39

2∗∗∗

0.23

9∗

(0.0

64)

(0.0

69)

(0.0

84)

(0.1

10)

(0.1

28)

Ben

QE

−0.

141∗

∗∗−

0.15

1∗∗∗

−0.

054

−0.

004

−0.

000

(0.0

37)

(0.0

48)

(0.0

67)

(0.0

55)

(0.0

82)

Gov

/Pre

sTap

−0.

010

−0.

009

−0.

017

−0.

009

−0.

010

(0.0

12)

(0.0

13)

(0.0

26)

(0.0

24)

(0.0

32)

Gov

/Pre

sQ

E−

0.09

7∗∗∗

−0.

094∗

∗∗−

0.09

1∗−

0.08

1∗∗

−0.

066

(0.0

18)

(0.0

16)

(0.0

47)

(0.0

31)

(0.0

50)

FO

MC

QE

−0.

252∗

∗∗−

0.25

3∗∗∗

−0.

145

−0.

214∗

∗−

0.09

0(0

.045

)(0

.037

)(0

.101

)(0

.085

)(0

.096

)B

enTap

*C

AB

/GD

P0.

040∗

∗0.

026

(0.0

17)

(0.0

17)

Ben

QE

*C

AB

/GD

P−

0.00

7−

0.00

3(0

.012

)(0

.012

)G

ov/P

res

Tap

*C

AB

/GD

P0.

001

0.00

1(0

.003

)(0

.003

)G

ov/P

res

QE

*C

AB

/GD

P0.

002

0.00

3(0

.005

)(0

.006

)FO

MC

QE

*C

AB

/GD

P−

0.00

10.

010

(0.0

12)

(0.0

13)

Ben

Tap

*R

eser

ves/

GD

P0.

012∗

∗0.

014∗

∗∗

(0.0

04)

(0.0

04)

(con

tinu

ed)


le12

.(C

ontinued

)

(1)

(2)

(3)

(4)

(5)

Ben

QE

*R

eser

ves/

GD

P−

0.00

4−

0.00

1(0

.003

)(0

.003

)G

ov/P

res

Tap

*R

eser

ves/

GD

P0.

000

0.00

0(0

.001

)(0

.001

)G

ov/P

res

QE

*R

eser

ves/

GD

P−

0.00

0−

0.00

0(0

.002

)(0

.002

)FO

MC

QE

*R

eser

ves/

GD

P−

0.00

5−

0.00

7∗

(0.0

03)

(0.0

04)

Ben

Tap

*E

xter

nalD

ebt/

GD

P−

0.00

3∗∗

−0.

005∗

∗∗

(0.0

01)

(0.0

01)

Ben

QE

*E

xter

nalD

ebt/

GD

P−

0.00

3∗∗∗

−0.

003∗

∗∗

(0.0

01)

(0.0

01)

Gov

/Pre

sTap

*E

xter

nalD

ebt/

GD

P−

0.00

0−

0.00

0(0

.000

)(0

.000

)G

ov/P

res

QE

*E

xter

nalD

ebt/

GD

P−

0.00

0−

0.00

0(0

.000

)(0

.000

)FO

MC

QE

*E

xter

nalD

ebt/

GD

P−

0.00

1−

0.00

0(0

.001

)(0

.001

)C

onst

ant

0.03

3∗∗∗

0.03

3∗∗∗

0.03

3∗∗∗

0.03

3∗∗∗

0.03

3∗∗∗

(0.0

04)

(0.0

04)

(0.0

04)

(0.0

04)

(0.0

04)

Obs

erva

tion

s5,

375

5,37

55,

375

5,37

55,

375

R-s

quar

ed0.

012

0.01

30.

013

0.01

30.

015

Note

s:R

obus

tst

anda

rder

rors

are

inpa

rent

hese

s.**

*,**

,an

d*

deno

tep

<0.

01,p

<0.

05,an

dp

<0.

1,re

spec

tive

ly.


Table 13. Significance Test Results

Columns from Table 12

(2) (3) (4) (5)

Significance of Sum of the CoefficientsF-statistics 15 0.02 12.64 4.04Prob > F (p-value) (0.001) (0.883) (0.002) (0.056)

Joint Significance of the CoefficientsF-statistics 7.76 7.84 6.94 12.39Prob > F (p-value) (0.003) (0.002) (0.004) (0.000)

Notes: The table reports the significance test results for the Ben tapering and Bentapering interaction with the fundamentals from table 12, specifications 2–5.

with stronger fundamentals are affected more by tapering (consis-tent with the earlier analysis), then we would expect a positivecoefficient on BenTap × CAB/GDP (i.e., the larger the currentaccount surplus, the larger is exchange rate depreciation associ-ated with a Bernanke tapering announcement), a positive coefficienton BenTap × Reserves/GDP (i.e., the larger the level of reserves,the larger is exchange rate depreciation associated with a Bernanketapering announcement), and a negative coefficient on BenTap ×ExternalDebt/GDP (i.e., the larger the level of external debt, the lessis exchange rate depreciation associated with a Bernanke taperingannouncement).

Column 1 of table 12 reports the baseline results without inter-action terms (as in table 3) for comparison purposes. Column 2reports the interactions of tapering announcements with the currentaccount (CAB/GDP), column 3 reports interactions with reserves(Reserves/GDP), column 4 reports interactions with external debt(ExternalDebt/GDP), and column 5 combines the three sets of inter-action terms into one regression.

Focusing on Bernanke’s tapering announcements (BenTap) incolumns 2–4, the results indicate that larger current account sur-pluses are associated with greater exchange rate deprecation (coeffi-cient of 0.040), larger reserves are associated with greater exchangerate depreciation (0.012), and high external debt is associated with


less exchange rate depreciation. It is useful to calculate the estimatedeffect for the robust and fragile groups using mean and median val-ues (for each group) of the fundamentals. The estimated impact onexchange rates from BenTap for a country with a mean “robust”(“fragile”) current account is 0.37 (0.14), for reserves is 0.37 (0.16),and for external debt is 0.26 (0.25). The same basic results hold upin the combined regression (column 5), with the exception of theinteraction term with current account balances.

The bulk of these results are therefore consistent with our earlierresults, i.e., countries with stronger fundamentals appear to be moreaffected by tapering announcements.

4.5 Excluding Bernanke’s May 22 Tapering Announcement

A natural question that arises is whether the Bernanke taperingresults are entirely driven by the market shock of the May 22announcement. To address this issue, we reestimate the basic modelsestimated in table 3 with the modification that the Bernanke taper-ing announcements exclude the May 22 observation. Interestingly,the effect of the Bernanke tapering announcements excluding May22 is greater on equity markets (a

The transmission of Federal Reserve tapering news to emerging financial … · 2016. 6. 1. ·...

Documents

Transcript of The transmission of Federal Reserve tapering news to emerging financial … · 2016. 6. 1. ·...