Can Macroprudential Measures Make Cross-Border Lending ...Vol. 15 No. 1 Can Macroprudential Measures...

Can Macroprudential Measures MakeCross-Border Lending More Resilient?

Lessons from the Taper Tantrum∗

Elod Takatsa and Judit Temesvaryb

aBank for International SettlementsbFederal Reserve Board

We study the extent to which macroprudential regulatorymeasures were successful in alleviating the negative shock thatthe taper tantrum of 2013 imposed on bilateral cross-borderlending flows. We use a novel data set combining the BISstage 1 enhanced banking statistics on bilateral cross-borderlending flows with the IBRN’s macroprudential database. Ourresults suggest that macroprudential measures implementedin borrowers’ host countries prior to the taper tantrum signif-icantly reduced the negative effect of the tantrum on cross-border lending growth. The shock-mitigating effects of host-country macroprudential rules are present both in lending tobanks and in lending to non-banks, and are stronger for lendingflows to borrowers in advanced economies and to the non-banksector in general. Source (lending) banking system measuresdo not affect bilateral lending flows, nor do they enhance theeffect of host-country macroprudential measures. Our resultsimply that policymakers may consider applying macropruden-tial tools to mitigate international shock transmission throughcross-border bank lending.

JEL Codes: F34, F42, G21, G38.

∗We are thankful for comments from Ingo Fender, Victoria Nuguer, Den-nis Reinhardt, Gretchen Weinbach, participants at seminars at the Bank forInternational Settlements, the Federal Reserve Board, and the London School ofEconomics, and participants at the IFABS Summer 2017 Conference in Oxford,England, the 2018 International Journal of Central Banking conference in Ams-terdam, and the 2018 CEBRA conference in Frankfurt. The views expressedin this paper are solely those of the authors and do not necessarily reflect theviews of the Bank for International Settlements or the Board of Governors ofthe Federal Reserve System. Author contact: Takats: Bank for International Set-tlements, Centralbahnplatz 2, Basel, CH-4002 Switzerland; [email protected]: Federal Reserve Board, 1801 K Street, Washington, DC 20006 USA;[email protected].

61

62 International Journal of Central Banking March 2019

1. Introduction

Since the financial crisis, many jurisdictions have used macropru-dential policies, applying prudential tools in order to limit systemicrisk (IMF–FSB–BIS 2016). Yet, evidence on the effectiveness of suchtools remains scarce—in part due to the relative absence of recentepisodes of financial disruptions which would allow for an assessmentof what remains a fairly new set of tools. A notable exception is theso-called taper tantrum—a large U.S. dollar funding shock in inter-national financial markets, which developed after Federal ReserveBoard Chairman Ben Bernanke’s indication in mid-2013 that theFed would start tapering off its quantitative easing program—whichcaused a broad-based but heterogeneous reduction in cross-borderfinancial flows (Avdjiev and Takats 2014). As such, this episode is anideal testing ground for assessing the role of macroprudential toolsin lending stabilization.

The taper tantrum has raised a series of questions for policy-makers and researchers alike: Did cross-border bank lending remainmore stable in those borrowers’ host countries which had moreactively applied macroprudential tools beforehand? Was lendingmore resilient from those source (lending) banking systems whichhad applied more macroprudential tools? Were source or host meas-ures more effective? Were advanced and emerging markets affecteddifferently? Which macroprudential tools affected cross-border banklending more? Did regulations stabilize lending to banks and non-bank borrowers alike? Did source lending system and host-countrymeasures interact?

In this paper, we answer these questions by using a novel dataset. We combine the “stage 1 enhancements” to the Bank for Inter-national Settlements’ (BIS) international banking statistics with theInternational Banking Research Network’s (IBRN) prudential regu-latory database. Because we are interested in the workings of macro-prudential tools, we focus on macroprudential elements in the IBRNdatabase (and not on (micro)prudential measures such as capitalrequirements).

In terms of methodology, we employ a difference-in-differenceanalysis to compare the growth rate of bilateral cross-border banklending right before and immediately after the taper tantrum shock.More precisely, we assess whether the slowdown in lending growth

Vol. 15 No. 1 Can Macroprudential Measures Make Cross-Border 63

during the taper tantrum was mitigated by the cumulative macro-prudential measures undertaken in the period leading up to thetantrum (that is, from the beginning of our sample in 2000:Q1until 2013:Q1—the quarter just prior to the beginning of the tapertantrum).

We find broad evidence that macroprudential measures appliedbeforehand stabilized cross-border lending flows during the tapertantrum. Most importantly, macroprudential tools applied in bor-rowers’ host countries stabilized the growth of cross-border banklending inflows significantly. The impact is stronger for lending toborrowers in advanced economies than for lending to borrowersin emerging markets. The impact is generally present for lendingto both bank and non-bank borrowers, but the results on lendingto non-bank borrowers are more broadly robust. We do not findevidence that macroprudential policies applied in source (lending)banking systems affected the growth of cross-border bank lendingoutflows during the tantrum, or that source (lending banking sys-tem) and host-country macroprudential measures would have signif-icantly interacted with each other in affecting cross-border lendingflows.

The lending-stabilizing impact of macroprudential tools was notonly statistically but also economically significant. In the overallsample, an additional macroprudential tightening action over the2000:Q1–2013:Q1 period attenuated the tantrum effect on cross-border bank lending by around 1 percentage point. This translatesto a differential tantrum impact on cross-border flows of around8 percentage points, when we compare host countries at the 90thversus the 10th percentile of macroprudential regulatory stringency.For advanced economies, the respective figures are around 2.5 per-centage points and 13 percentage points. Our findings are robustto the addition of a range of macroeconomic controls, and to vari-ations in the construction of macroprudential indexes and lend-ing flows. The results are also present in most subsets of macro-prudential tools, though, as expected, are not significant for allcomponents.

The results are relevant for policymakers: They suggest thatmacroprudential tools not only contribute to the stability of thedomestic financial system but also enhance cross-border bank financ-ing inflows to an economy. Our result confirming the strong and


significant effects of macroprudential tools applied in host countriessuggests that, though international coordination can play a use-ful rule, it remains paramount that countries “keep their housein order.” The significant shock-alleviating role of macropruden-tial rules in lending to the non-bank sector of emerging marketsmay be of particular policy relevance. Importantly, while we do notfind significant evidence that the stabilizing potential depends onthe interaction between source and host macroprudential measures,this does not necessarily imply that externalities do not exist. Forinstance, more-stable host-country financial systems might make thesource (lending) banking systems also more resilient, by providingstable bank financing demand.

The paper proceeds as follows. In section 2 we link our work tothe growing volume of related literature. In section 3 we describeour data in detail, and we present the econometric methodology insection 4. We detail the results in section 5, and discuss the robust-ness of our findings in section 6. In section 7, we discuss the maincaveats, and we conclude with policy implications in section 8.

2. Related Literature

Our work is at the intersection of macroprudential policies andspillovers in international financial stability and regulation. Theresearch on macroprudential policies dates back to Crockett (2000)and Borio (2003) and is reviewed in detail by Galati and Moessner(2011), among others. The policy discussion, as recently shown, forinstance, in IMF–FSB–BIS (2016), suggests that macroprudentialpolicies have an international dimension.

In particular, several research pieces point in the direction thatmacroprudential policies could spill over across borders throughthe activities of internationally active banks. However, our paperis the first to look at the stabilizing role of macroprudential policiesduring financial stress. This focus sets our paper aside from mostpapers in the literature which look at how macroprudential or regula-tory policies might affect lending under normal financial conditions.In this line of work, Houston, Lin, and Ma (2012), for instance,find that regulations can affect international banks’ activities bytransferring funds to less-regulated markets. Ongena, Popov, and


Udell (2013) find evidence for risk shifting when lending standardsdiffer across jurisdictions.1 Berrospide et al. (2017) and Temesvary(2018) find regulatory spillovers into bilateral bank flows to andfrom the United States. In addition, Ocampo (2011) studies coun-tercyclical regulations in developing countries, and Laeven, Rat-novski, and Tong (2014) examine the importance of bank ruleswhich address systemic risk. One particular channel of regulatoryspillovers, as Karolyi, Sedunov, and Taboada (2017) show, couldwork through financial stability. The presence of this latter channelraises the possibility that macroprudential policies, which aim tostrengthen financial stability, might also spill over to other countriesvia international bank lending. Buch and Goldberg (2016) summa-rize related cross-country evidence in the framework of the IBRNnetwork.

The channels of macroprudential spillovers seem in many aspectsanalogous to how monetary policy spillovers occur. One channelis cross-border bank lending (Cetorelli and Goldberg 2012; Forbesand Warnock 2012; Miranda-Agrippino and Rey 2015); another isthe currency denomination of cross-border bank loans (Ongena,Schindele, and Vonnak 2014; Alper et al. 2016; Avdjiev, Subelyte,and Takats 2016; Avdjiev and Takats 2016; Takats and Temesvary2017).

In terms of data use, Avdjiev et al. (2017), part of the IBRNresearch effort discussed in Buch and Goldberg (2016), is the closestto our work, as they also combine BIS data (though not the stage 1enhancements) and the IBRN prudential database. However, theirfocus is very different from ours. They investigate the continuousimpact of prudential tools on cross-border bank lending over time(i.e., whether prudential tools enacted in one quarter affect cross-border bank lending in the following quarter, on average). Theydo not focus on, as we do, whether the accumulation of prudentialregulation increases resilience of cross-border bank lending during

1Somewhat similarly, Aiyar et al. (2014) or Forbes, Reinhardt, and Wieladek(2017) argue that stronger capital regulations might reduce cross-border lendingsupply. However, this research differs from ours in that we explicitly exclude pru-dential regulation from our database (with the exception of robustness checks),as we explain in detail later.


financial stress—and how rules adopted in one country affect cross-border bank lending and potential spillovers in this regard.

In terms of theoretical work, Agenor et al. (2017) provide a modelfor how macroprudential policies might spill over internationally.However, this theoretical model focuses on spillovers into financialflows during normal times, while we focus on the stability of financialflows during a period of financial stress.

Finally, our work also builds on research which argues thatnational borders and economically relevant decisionmaking unitsoften diverge, as we also describe in Takats and Temesvary (2017).The discussion dates back to Cecchetti, Fender, and McGuire (2010)and Fender and McGuire (2010), who show that the lending bank’snationality tends to be more relevant than its residence in identi-fying the decisionmaking unit. This argument is further developedfrom a policy perspective in Committee on the Global Financial Sys-tem (2011). Building on these findings, Avdjiev, McCauley, and Shin(2015) coin the term of the (absence of) triple coincidence in inter-national finance. This term refers to the phenomenon that nationalborders, the conventional units of international economic analysis,often do not coincide with the economically relevant decisionmak-ing unit. Following these lessons, we focus on “lending banking sys-tems” as opposed to “lending countries,” so that we can follow thedecisionmaking unit as precisely as possible.

3. Data Description

3.1 Data on Bilateral Cross-Border Lending Flows

We collect data on bilateral cross-border bank lending flows fromthe stage 1 enhancements to the Bank for International Settle-ment’s international banking statistics (BIS IBS) which is detailed,for instance, in Avdjiev, McGuire, and Wooldridge (2015). The datacover around 30 trillion U.S. dollars in total cross-border claims. Themain advantage of using the stage 1 enhancements is that this dataset allows us to link lending banking systems with the host countriesof borrowers, while retaining information on the currency composi-tion of the lending flows (table 9 in the appendix). Importantly, thisdata set uniquely allows us to correctly measure the change in bank


Figure 1. Changes in Lending Flows: 2013–14

Source: BIS banking statistics.Note: See online file at http://www.ijcb.org for color version of this figure.

claims between lending banking systems and host countries, withoutthe confounding impact of currency valuation movements. We selectthose source (lending) banking system–host-country pairs for whichthe IBRN data set provides detailed prudential regulatory informa-tion for both the source banking system and the host country. Thebilateral cross-border lending data set is also broken down by target(counterparty) sector, separately covering claims on host countries’banking and non-banking (including non-bank financials, such asinsurance) sectors.

Visualization of the cross-border bank lending data offers twoinsights. First, the taper tantrum episode constituted a slowdownin cross-border bank lending flows (figure 1). Flows (measured hereas two-quarter changes in cross-border lending volumes) to bothadvanced-economy and emerging market borrowers declined dur-ing the tantrum, followed by a recovery—mostly driven by flows toadvanced-economy borrowers. Second, a comparison of post-taper-tantrum cross-border bank lending growth in borrowing countriescharacterized by high versus low levels (stringency) of macropru-dential tools suggests that on average borrower countries with moreaccumulated macroprudential tools saw higher cross-border lendinginflows (figure 2).

Given the substantial heterogeneity in the bilateral banking data,we winsorize the observations at the 5th and 95th percentile, as is


Figure 2. Average Post-Tantrum Lending Flows

Source: BIS banking statistics.Note: See online file at http://www.ijcb.org for color version of this figure.

common in the related literature (for instance, Avdjiev and Takats2016; Takats and Temesvary 2017). The reason is that the growthrate of smaller-volume bilateral lending pairs is very volatile, some-times amounting to changes of several hundred percentage points.For instance, small idiosyncratic shocks, such as a new foreign directinvestment (FDI) project, can substantially affect these smaller-scalerelationships.

3.2 Data on Prudential Measures

Our database combines information on country-level macropruden-tial measures enacted during the period up to the taper tantrum,with detailed information on bilateral cross-border lending flowsbefore and after the tantrum. Our data on country-level regula-tory measures come from the macroprudential database employed bythe 2016 IBRN project, also incorporating the 2013 Global MacroPrudential Instruments (GMPI) survey (Cerutti et al. 2016; Avd-jiev et al. 2017; Berrospide et al. 2017; Cerrutti, Claessens, andLaeven 2017). Table 1 and table 10 (in the appendix) summarizeand describe these indexes.

The IBRN database contains a mix of macroprudential measuresand also standard (micro)prudential minimum capital requirements.More precisely, while eight out of the nine IBRN categories (sscb res,sscb cons, sscb oth, concrat, ibex, ltv cap, rr foreign, and rr local)are clearly macroprudential, the ninth index on capital requirements(cap req) is less clearly macroprudential, as we explain in more detail


Tab

le1.

Var

iable

Defi

nitio

ns

and

Sum

mar

ySta

tist

ics

Vari

able

sU

nit

Defi

nit

ion

Mea

nSD

Min

.p25

p50

p75

Max.

N

Dep

ende

ntVar

iabl

es

Chan

gein

Bilat

eral

Len

din

gFlo

ws—

Tot

al

%C

han

gein

tota

lbilat

eral

ban

kle

ndin

gflow

sfo

rea

chso

urc

e–hos

tpai

r,fr

ombef

ore

toaf

ter

the

taper

tant

rum

;B

ISban

king

stat

isti

cs

5.23

148

.03

−21

8.3

−9.

343

3.72

821

.519

51,

875

Chan

gein

Bilat

eral

Len

din

gFlo

ws—

Ban

ks

%D

efined

asab

ove,

for

clai

ms

onban

ks2.

309

81.5

3−

334.

8−

16.9

53.

312

30.8

626

81,

599

Chan

gein

Bilat

eral

Len

din

gFlo

ws—

Non

-ban

ks

%D

efined

asab

ove,

for

clai

ms

onnon

-ban

ks5.

1942

.6−

150.

6−

12.4

32.

792

25.3

412

51,

743

Reg

ula

tory

Inde

xes

Cum

ula

tive

Sou

rce

Pru

c6In

dex

Inte

gers

Cou

ntry

index

byti

me

tan

dco

unt

ryc,w

hic

his

equal

to1

ifth

esu

mof

eigh

tm

acro

pru

den

tial

inst

rum

ents

(ssc

bre

s,ss

cbco

ns,

sscb

oth,

concr

at,ib

ex,lt

vca

p,

rrfo

reig

n,an

drr

loca

l,as

defi

ned

inta

ble

9)is

grea

ter

than

oreq

ual

to1,

iseq

ual

to–1

ifth

esu

mof

the

inst

rum

ents

isle

ssth

anor

equal

to–1

,an

dis

0ot

her

wis

e

0.99

82.

022

−3

01

28

1,87

5

(con

tinu

ed)

70 International Journal of Central Banking March 2019Tab

le1.

(Con

tinued

)

Vari

able

sU

nit

Defi

nit

ion

Mea

nSD

Min

.p25

p50

p75

Max.

N

Reg

ula

tory

Inde

xes

Cum

ula

tive

Hos

tPru

c6In

dex

Inte

gers

Defi

ned

asso

urc

ePru

c6in

dex

,fo

rhos

tre

gula

tion

s1.

445

3.52

7−

50

13

161,

875

Cum

ula

tive

Sou

rce

Pru

CIn

dex

Inte

gers

Defi

ned

asPru

c6in

dex

,al

soin

cludin

gca

pre

q,as

defi

ned

inta

ble

9

2.00

22.

349

−2

12

313

1,72

3

Cum

ula

tive

Hos

tPru

CIn

dex

Inte

gers

Defi

ned

asso

urc

ePru

Cin

dex

,fo

rhos

tre

gula

tion

s2.

591

4.53

5−

60

24

221,

723

Cum

ula

tive

Sou

rce

sscb

Index

Inte

gers

Sum

ofss

cbre

s,ss

cbco

ns,

and

sscb

oth

(as

defi

ned

inta

ble

9)in

the

count

ryin

whic

hth

eso

urc

eban

king

syst

emis

loca

ted

0.19

10.

882

−2

00

05

1,72

3

Cum

ula

tive

Hos

tss

cbIn

dex

Inte

gers

Sam

eas

sourc

ess

cb,fo

rhos

tco

unt

ry0.

375

1.18

8−

30

00

51,

723

Add

itio

nal

Var

iabl

es

Bilat

eral

Len

din

gFlo

ws—

Tot

al%

Chan

gein

the

nat

ura

llo

gari

thm

ofto

talbilat

eral

ban

kle

ndin

gcl

aim

sfo

rea

chso

urc

e–hos

tpai

r,fr

ombef

ore

toaf

ter

the

taper

tant

rum

;B

ISban

king

stat

isti

cs

1.08

721

.44

−83

−5.

637

−0.

823

4.71

219

11,

875

Bilat

eral

Len

din

gFlo

ws—

Ban

ks%

Defi

ned

asab

ove,

for

clai

ms

onban

ks1.

238

30.0

6−

83−

8.34

7−

1.74

36.

673

191

1,62

8

Bilat

eral

Len

din

gFlo

ws—

Non

-ban

ks%

Defi

ned

asab

ove,

for

clai

ms

onnon

-ban

ks1.

819

.18

−60

.4−

5.95

8−

0.17

77.

062

90.2

1,78

4

(con

tinu

ed)


Tab

le1.

(Con

tinued

)

Vari

able

sU

nit

Defi

nit

ion

Mea

nSD

Min

.p25

p50

p75

Max.

N

Wei

ghts

inEst

imat

ions

Shar

eof

aSou

rce–

Hos

tPai

rin

Agg

rega

teB

ilat

eral

Tot

alC

laim

s

Rat

ioB

ilat

eral

cros

s-bor

der

clai

ms

from

ban

king

syst

emi

tohos

tco

unt

ryj

atti

me

t,div

ided

byag

greg

ate

cros

s-bor

der

clai

ms

atti

me

t(s

um

med

acro

ssal

lso

urc

e–hos

tpai

rs)

0.04

80.

045

6.64

E-0

60.

014

0.03

20.

067

0.21

1,87

5

Shar

eof

aSou

rce–

Hos

tPai

rin

Agg

rega

teB

ilat

eral

Cla

ims

onB

anks

Rat

ioD

efined

assh

are

ofto

talcl

aim

s,fo

rcl

aim

son

ban

ks0.

047

0.04

68.

66E-0

60.

013

0.03

0.06

90.

21,

663

Shar

eof

aSou

rce–

Hos

tPai

rin

Agg

rega

teB

ilat

eral

Cla

ims

onN

on-b

anks

Rat

ioD

efined

assh

are

ofto

talcl

aim

s,fo

rcl

aim

son

non

-ban

ks0.

045

0.04

70

0.01

30.

030.

064

0.3

1,77

9


below. The precise description of the nine indexes is detailed in table10 in the appendix.2

For our formal analysis, we create a new index of macropruden-tial tools (which we denote by Pruc6). This Pruc6 index is analogousto the pre-defined PruC index of the IBRN database, but excludeschanges in capital requirements. More precisely, Pruc6 is a countryindex by time t and country c, which is equal to 1 if the sum of eightdistinct macroprudential instruments (sscb res, sscb cons, sscb oth,concrat, ibex, ltv cap, rr foreign, and rr local) is greater than orequal to 1, is equal to −1 if the sum of the instruments is less thanor equal to −1, and is 0 otherwise. In other words, our Pruc6 indexis similar to the pre-defined Pruc index, only it excludes the impactof capital requirements (cap req).

The rationale for excluding the cap req (capital requirements)index is threefold. First is focus: we hone in on the efficiency ofmacroprudential instruments, while capital requirements generallybelong to the category of microprudential instruments rather thanmacroprudential ones. Second is policy stance: changes in capitalrequirements mostly signal the adoption of the Basel III regime andthereby international harmonization. Hence, cumulative differencesin country-specific capital indexes are more likely to reflect differ-ences in prudential regulation at the start rather than at the end ofthe observation period. Third is the role of expectations: given thatthe adoption of Basel III was anticipated along country-specific time-lines, the problem of expectations is the strongest for capital require-ments. Nonetheless, despite these issues, we repeat our analysis forthe pre-defined indexes to check robustness.

While the IBRN database is the most comprehensive and thor-ough database of country-specific macroprudential instruments,

2In addition, the IBRN database contains three pre-defined indexes. First,PruC is a country index by time t and country c, which is equal to 1 if the sumof nine distinct macroprudential instruments (shown in table 10) is greater thanor equal to 1, is equal to −1 if the sum of the instruments is less than or equalto −1, and is 0 otherwise. Second, PruC2 denotes a country index by time t andcountry c, which is equal to 1 if the sum of the nine instruments is greater thanor equal to 1, is equal to −1 if the sum of the instruments is less than or equalto −1, and is 0 otherwise. In this case, all individual instruments are adjustedto have maximum and minimum changes of 1 and −1. Third, sscb is the sumof changes in sector-specific capital buffers across the residential, consumer, andother sectors.


Figure 3. Average PruC6 Index in Advanced-Economyand Emerging Market Borrowers: 2000–14

Source: IBRN prudential regulatory database.Note: See online file at http://www.ijcb.org for color version of this figure.

there are several issues associated with using this data, which weaddress in our analysis. First, there is the possibility of a confound-ing effect in that stricter macroprudential policy settings may sig-nal (unobserved) vulnerability. Second, a cross-sectional analyticalconcern is that the IBRN macroprudential database does not meas-ure the absolute level of policy stance at any given time: it showscumulative actions taken relative to the start of the panel. Third,a time-series limitation is that the confounding effect of expecta-tions complicates the identification of the timing of the regula-tory spillovers. Fourth, we must consider the issue of asymmetryin that not all macroprudential actions have the same strength andeffects.

For our analysis, we use a cumulative index that encompasses theentire available time series starting from 2000:Q1 up until the quar-ter prior to the beginning of the taper tantrum: 2013:Q1. The useof macroprudential tools was more active among emerging marketsthan among advanced economies, suggesting separate examinationof the two groups of countries (figure 3). Importantly, our use ofcumulative changes mitigates potential problems arising from theuncertainty of timing (i.e., whether or not the announcement wasexpected beforehand).

According to these measures, we see a general tightening ofmacroprudential policies across the countries in our sample overtime, as shown in the summary statistics of table 1 as well as


figure 3 (see also IMF–FSB–BIS 2016). These averages, however,hide cross-sectional differences—with some countries, for instance,substantially tightening and others significantly easing their macro-prudential policy stance during this period.

3.3 Additional Macro Controls

In specifications without host-country fixed effects, controls formacroeconomic effects on host-country credit demand may beneeded. We run our benchmark regression without any specificmacroeconomic control variables. However, we include such con-trols, such as the credit-to-GDP gap, credit growth, GDP growth,current-account-to-GDP ratio, and institutional quality measures inour extensive robustness checks.

4. Estimation Methodology

We employ a difference-in-difference methodology to compare thecross-section of bilateral cross-border lending flows before the tapertantrum with the cross-section immediately following the tantrum,as a function of ex ante macroprudential measures taken in sourcebanking systems and host countries. In other words, we investigatethe second derivatives of bank claims: their acceleration or slowdown.

Our main hypothesis is that macroprudential tools in place, irre-spective of the specific tool, increase the resilience of source bank-ing systems and borrowers. Hence, macroprudential tools applied insource (lending) banking systems strengthen lending banks by mak-ing their operations more prudent, thereby stabilizing the supply ofcross-border lending during times of stress.3 Similarly, macropruden-tial tools applied in host countries make local bank and non-bankborrowers more resilient. They do so by directly stabilizing banks’loan demand (i.e., interbank lending) as banks become more robustto shocks, and indirectly stabilizing the demand from non-banks (asthe non-bank private sector also becomes more stable by curbingexcessive borrowing).

3For alternative hypotheses on the role of source regulations in shaping cross-border lending outflows, please see the literature review in section 2.


Importantly, we do not formulate hypotheses around specifictools and their impact on cross-border bank lending. That is, wedo not solely focus on direct effects, such as a tool specifically tar-geting lending denominated in a foreign currency that could improveinternational lending during stress. Rather, we are also interested inindirect effects. For instance, sector-specific capital buffers in realestate lending, a tool that is applied purely for domestic reasons,may also improve the resilience of the banking system in such a waythat it is able to lend more to cross-border borrowers as well.

Our main explanatory variable is our index of applied macropru-dential measures (Pruc6). As described above, we use the cumulativechanges from the beginning of the sample (2000:Q1) up until thequarter immediately preceding the tantrum (2013:Q1) to proxy forex ante bank macroprudential regulatory stringency at the countrylevel.

In all specifications, Δflows is the change in bilateral flows whichthemselves represent the log change of claims. Hence, our depen-dent variable is a second difference (i.e., change in the growth rateof claims). We always investigate the change before and after thetaper tantrum. More precisely, we measure the change in claimsfrom 2012:Q4 to 2013:Q1) compared with the change in claimsafter (from 2013:Q2 to 2013:Q3) the tantrum. In our regressions, weweigh the observations by the size of pre-tantrum (2013:Q1) bilateralexposures.

The shock-mitigating effects of macroprudential tools are two-sided: measures taken by both source banking systems and hostcountries can affect bank flows’ resilience. The basic difference-in-difference regressions take the following form:

Δflowsij = β0 + β1source reg indj + β2dem hosti + εij (1)

Δflowsij = γ0 + γ1host reg indi + γ2dem sourcej + νij . (2)

In equation (1), we examine the shock-mitigating effect of source(lending) banking system macroprudential measures, while control-ling for changes in the demand for credit at the host-country levelthrough the use of fixed effects. Similarly, in equation (2) we focuson the role of host-country macroprudential stringency in mitigat-ing the tantrum shock on bilateral cross-border bank inflows, whilecontrolling for the supply side of credit through source fixed effects.


In equations (1) and (2) we include fixed effects on the lendingor borrowing side, in order to control for unobserved country-levelcharacteristics. We drop these fixed effects in equation (3) to be ableto investigate regulatory effects on both sides simultaneously:

Δflowsij = δ0 + δ1host reg indi + δ2source reg indj + νij . (3)

In equation (3), we aim to assess the impact of both source- andhost-country-specific macroprudential tools simultaneously.

Next, we also investigate the interaction between source and hostmacroprudential measures. Formally, in equation (4) we include apolicy interaction term, and also apply source and host-country fixedeffects:

Δflowsij = θ0 + θ1demhosti + θ2demsourcej

+ θ3host reg indi ∗ source reg indi + νij . (4)

Similar to Khwaja and Mian (2008), this identification approach con-trols for any potential credit supply (source banking-system-specific)or demand (host-country-specific) effects simultaneously (Avdjievand Takats 2016). While this formulation identifies the policy inter-action precisely, it also precludes us from being able to observe theimpact of source and host measures in levels. In all estimations wecluster the standard errors across the source (lending) banking sys-tems. For robustness, later we also cluster the standard errors alonghost countries.

5. Estimation Results

5.1 Macroprudential Measures Index: Source BankingSystems vs. Host Countries

We find broad evidence in the full sample that host countries whichhad taken macroprudential measures before the taper tantrum, ascaptured by our Pruc6 index, indeed had more-stable cross-borderbank lending inflows (table 2). However, we do not find evidence ofa significant regulatory impact of source lending systems on cross-border lending outflows.

Vol. 15 No. 1 Can Macroprudential Measures Make Cross-Border 77Tab

le2.

The

Impac

tof

the

Mac

ropru

den

tial

Reg

ula

tion

sIn

dex

(Pru

c6)

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ce0.

147

0.11

10.

291

0.17

5−

0.42

−0.

471

Pru

c6In

dex

(0.7

9)(0

.749

)(1

.392

)(1

.282

)(1

.31)

(1.3

32)

Cum

ulat

ive

Hos

tP

ruc6

1.04

0∗∗∗

1.03

5∗∗∗

0.88

5∗∗

1.02

9∗∗

1.26

3∗∗∗

1.07

9∗∗∗

Inde

x(0

.3)

(0.2

63)

(0.4

34)

(0.3

81)

(0.3

32)

(0.3

38)

Con

stan

t−

21.8

5−

0.55

9∗∗

3.62

6∗∗

39.1

4−

3.56

7∗∗∗

1.47

9−

31.5

6∗0.

750∗

∗∗4.

520∗

∗

(15.

99)

(0.2

19)

(1.4

29)

(30.

99)

(0.4

04)

(2.1

21)

(17.

41)

(0.1

81)

(1.9

72)

Obs

erva

tion

s1,

875

1,87

51,

875

1,59

11,

591

1,59

11,

734

1,73

41,

734

R-s

quar

ed0.

060.

060.

010.

060.

050.

000.

070.

100.

01So

urce

Fix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

0.88

0.66

41.

751.

047

−2.

1−

2.35

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Hos

tC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

8.31

8.28

7.08

8.23

10.1

8.63

Note

s:Tab

le2

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

nd-

ing

bilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

)in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

Pru

c6in

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;

***

p<

0.01

,**

p<

0.05

,*

p<

0.1.


In more detail, in model 1 of table 2 we begin by estimatingequation (1). In this specification we investigate the impact of macro-prudential measures taken in the source (lending) jurisdiction, anduse fixed effects to control for potential credit demand shocks atthe host-country level. Model 1 suggests no statistically significantimpact of source regulatory measures.

In contrast, in model 2 we estimate equation (2) and find sig-nificant effects. Recall that equation (2) examines the impact ofmacroprudential measures taken in the host jurisdiction, and usesfixed effects to control for potential credit supply impacts at thesource level. The significant positive coefficient implies that fromthe perspective of a given source banking system, lending flowsto a host country with an additional pre-crisis prudential measuredeclined by around 1 percentage point less (or accelerated more)than lending to borrowers in those host countries which had appliedless macroprudential stringency.

In model 3 we simultaneously estimate the impact of both sourceand host macroprudential tools through equation (3). The coef-ficient estimate on the host-country regulatory index is similarin magnitude and significance to model 2, and thus confirms thesignificant shock-mitigating effect of host-country macroprudentialrules.

We also examine the economic significance of our findings bycomparing lending inflows into a host country with stricter macro-prudential regulation (at the 90th percentile of the macroprudentialregulation index) with inflows into a host country with less regula-tion (at the 10th percentile). We find that a more strictly regulatedhost country received around 8 percentage points more cross-borderlending inflows than a less-regulated host country. Thus, our resultsare also economically significant.

Next, we delineate our sample by sectors: lending to banks (mod-els 4–6) and to non-banks (models 7–9). The results show roughlythe same picture as the aggregated analysis: macroprudential meas-ures applied in host countries had a significant stabilizing effect inboth target sectors, though somewhat stronger in lending to non-banks. These results hold irrespective of whether we control forcredit supply-side effects (models 5 and 8) or include source rules aswell (models 6 and 9).


5.2 Broader Macroprudential Indexes

We proceed by repeating the previous analysis now using a broadmacroprudential index, the pre-defined overall PruC index, whichalso includes capital requirements in addition to the strictly macro-prudential tools (table 3). The coefficient estimates show the samequalitative picture as those in table 2: the tantrum-shock-mitigatingeffect of host-country macroprudential rules remains significant. Thisis not surprising given the extent of the overlap between our indexexcluding prudential capital regulation (Pruc6) and the pre-definedindex (PruC). These estimations further suggest that the significantresults we find in models 1–3 are mostly driven by lending inflowsinto host countries’ non-bank sectors (models 7–9). We find weakerinterbank results (models 4–6).

In terms of economic significance, a host country with strict pre-tantrum broad macroprudential rules in place saw around 5 per-centage point less reduction in its cross-border lending inflows thana host country with laxer broad macroprudential rules in place. Thisimplied economic impact is smaller than those in table 2—consistentwith the argument that during this period—i.e., the Basel III adop-tion phase—differences in capital requirement measures probablyreflected more variation in the initial capital requirements than dif-ferences in the end-period capital requirements.

5.3 Advanced Economies and Emerging Markets

Next, we divide our sample across the host countries of borrowers,and reestimate our baseline specifications of table 2 for lending toborrowers in advanced economies (table 4) and emerging markets(table 5). These separate estimations reveal three main findings.First, the main pattern we detected in tables 2 and 3, namely thesignificant role of host-country macroprudential measures in stabiliz-ing cross-border lending flows into hosts’ non-bank sectors, is presentboth in lending to advanced economies and in lending to emergingmarkets. Second, similar to table 3, the results on interbank lendingbecome statistically insignificant when we delineate the sample byborrower type. Third, the coefficient estimates on host macropruden-tial rules are larger and more statistically significant for borrowersin advanced economies than in the aggregated sample. As before,


le3.

The

Impac

tof

the

Pre

-defi

ned

Pru

den

tial

Reg

ula

tion

sIn

dex

(Pru

C)

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ceP

ruC

0.00

9−

0.03

20.

129

0.00

7−

0.55

1−

0.54

5In

dex

(0.5

93)

(0.5

66)

(1.1

28)

(1.1

04)

(1.3

44)

(1.3

52)

Cum

ulat

ive

Hos

tP

ruC

0.58

1∗∗

0.57

0∗∗∗

0.48

0.58

5∗0.

693∗

∗∗0.

571∗

∗

Inde

x(0

.23)

(0.1

97)

(0.3

71)

(0.3

15)

(0.2

38)

(0.2

36)

Con

stan

t−

21.7

−0.

691∗

3.57

6∗∗

39.2

−3.

592∗

∗∗1.

34−

30.7

5∗0.

540∗

5.06

5∗

(16.

04)

(0.3

52)

(1.6

75)

(31.

3)(0

.657

)(2

.405

)(1

7.3)

(0.3

08)

(2.7

8)O

bser

vation

s1,

723

1,72

31,

723

1,48

81,

488

1,48

81,

617

1,61

71,

617

R-s

quar

ed0.

060.

060.

003

0.06

0.05

0.00

10.

070.

090.

004

Sour

ceFix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

CIn

dex

0.54

−0.

190.

780.

05−

2.75

−2.

73

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Hos

tC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

CIn

dex

5.81

5.70

4.32

5.26

6.23

5.14

Note

s:Tab

le3

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

nd-

ing

bilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

)in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

Pru

Cin

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;

***

p<

0.01

,**

p<

0.05

,*

p<

0.1.


le4.

Bor

row

ers

inA

dva

nce

dEco

nom

ies:

The

Impac

tof

the

Mac

ropru

den

tial

Reg

ula

tion

sIn

dex

(Pru

c6)

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ceP

ruc6

0.28

20.

221

0.35

5−

0.00

40.

0153

0.12

3In

dex

(0.9

97)

(0.9

38)

(1.6

99)

(1.5

59)

(1.1

35)

(1.1

17)

Cum

ulat

ive

Hos

tP

ruc6

2.61

3∗∗∗

2.59

0∗∗∗

3.03

32.

895

2.60

8∗∗∗

2.60

8∗∗∗

Inde

x(0

.821

)(0

.899

)(2

.004

)(1

.98)

(0.9

21)

(0.8

82)

Con

stan

t16

.73

−4.

935∗

∗∗2.

659

25.0

8−

12.8

3∗∗∗

−0.

101

9.00

81.

267∗

∗∗3.

850∗

(12.

5)(0

.052

2)(1

.752

)(1

8.43

)(0

.764

)(2

.683

)(7

.592

)(0

.124

)(2

.212

)O

bser

vation

s74

774

774

768

868

868

870

870

870

8R

-squ

ared

0.05

0.11

0.01

0.06

0.08

0.01

0.05

0.13

0.02

Sour

ceFix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

1.41

1.10

72.

13−

0.02

20.

770.

616

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Hos

tC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

13.0

612

.95

15.1

714

.47

13.0

613

.04

Note

s:Tab

le4

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

nd-

ing

bilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

)to

bor

row

ers

inad

vance

dec

onom

ies,

inre

spon

seto

aon

e-unit

chan

gein

the

cum

ula

tive

Pru

c6in

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;**

*p

<0.

01,**

p<

0.05

,*

p<

0.1.


le5.

Bor

row

ers

inEm

ergi

ng

Mar

kets

:T

he

Impac

tof

the

Mac

ropru

den

tial

Reg

ula

tion

sIn

dex

(Pru

c6)

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ceP

ruc6

−0.

414

−0.

349

−0.

297

0.02

99−

1.25

5−

1.45

3In

dex

(1.2

3)(1

.106

)(2

.353

)(2

.303

)(2

.048

)(2

.021

)C

umul

ativ

eH

ost

Pru

c60.

656

0.54

9∗0.

389

0.28

80.

785∗

∗0.

715∗

Inde

x(0

.394

)(0

.318

)(0

.654

)(0

.538

)(0

.381

)(0

.392

)C

onst

ant

−21

.25.

335∗

∗∗6.

112∗

∗39

.99

6.98

0∗∗∗

5.90

9−

30.6

4∗0.

992∗

6.46

5∗∗

(16.

05)

(0.6

14)

(2.2

57)

(31.

31)

(1.0

27)

(4.3

44)

(17.

07)

(0.5

7)(2

.391

)O

bser

vation

s97

697

697

680

080

080

090

990

990

9R

-squ

ared

0.07

0.06

0.00

30.

060.

080.

001

0.09

0.1

0.01

Sour

ceFix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

−2.

49−

2.09

7−

1.78

0.17

9−

7.53

−8.

718

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Hos

tC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

6.56

5.49

3.89

2.88

7.85

7.15

Note

s:Tab

le5

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

nd-

ing

bilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

)to

bor

row

ers

inem

ergi

ng

mar

kets

,in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

Pru

c6in

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;**

*p

<0.

01,**

p<

0.05

,*

p<

0.1.


macroprudential tools applied in source (lending) banking systemsare insignificant for lending flows into both advanced and emerginghost countries.

In more detail, when we examine lending flows into advancedeconomies in table 4, the coefficient estimates on the host-countrymacroprudential tools are larger than in the aggregated sample.An additional macroprudential measure applied at the host-countrylevel mitigated the tantrum shock on lending inflows by around 2.6percentage points (model 2). In terms of economic significance, anadvanced economy host with strict pre-tantrum regulations (at the90th percentile) in place saw a notable 13 percentage point higherlending inflows than a host with laxer macroprudential rules (at the10th percentile). The results for lending to all sectors (models 1–3)are clearly driven by our results for lending to non-banks (models7–9).

Examining the macroprudential effects on lending inflows intoemerging markets in table 5, we find that the impacts are smallerand generally less statistically significant. The coefficient estimateson host macroprudential rules hover in the 0.5–0.8 percentage pointrange, and remain consistently significant for lending to non-banks(models 8 and 9).4 The potential of macroprudential policies to makelending inflows more resilient to the non-bank sector might be partic-ularly relevant for emerging market policymakers because continuedaccess to credit to non-banks is crucial for economic stability andlong-term growth.

Consistently, the mitigating effects of host-country rules are eco-nomically significant for lending to borrowers in emerging marketsas well. The impact amounts to a lending flow difference of around8 percentage points when we compare cross-border inflows into thenon-bank sectors of emerging market hosts at the 90th versus the10th percentile of regulatory stringency.

In additional estimations, we also repeat the analysis foradvanced and emerging markets separately using the pre-definedPruC index (tables 11 and 12). While the advanced-economy results

4Figure 3 suggests that the smaller coefficient estimates may in part be dueto the regularity that emerging market macroprudential rules were on averagestricter during the tantrum episode than those in advanced-economy hosts.


remain statistically significant (with the exception of interbank lend-ing), the tantrum effect-mitigating role of macroprudential regula-tion in emerging host markets loses significance. These additionalregressions suggest that we can observe a similar, but weaker, rela-tionship between the general pre-defined PruC index and lendingstability than the one with our more focused Pruc6 macroprudentialindex.

5.4 Individual Macroprudential Measure Categories

We also check how individual macroprudential tool categoriesaffected cross-border lending during the taper tantrum (table 6).Earlier we found evidence for the impact of macroprudential toolsenacted in host countries. Hence, now we focus on the results for indi-vidual macroprudential tool categories enacted in host countries. Inorder to gain a more representative picture, we look across all coun-try pairs. Similarly, we consider total lending rather than breakingresults into lending to bank and non-bank borrowers. Table 6 showsthe impact of individual macroprudential tool categories one by one.

Three main findings stand out. First, some macroprudential toolgroups seem to be significant even individually. The coefficients aresignificant on concentration limits (model 5, concrat) and on loan-to-value ratio caps (model 7, ltv cap). Notably, the latter suggeststhat the robust role of host-country macroprudential rules in sta-bilizing lending inflows to the non-bank sector may work throughstabilizing non-banks’ credit demand, by increasing the resilience ofthese borrowers. The non-bank private sector might become moreresilient, for instance, if macroprudential tools curb excessive bor-rowing before the shock or, alternatively, they direct funds to morecreditworthy borrowers. In addition, we find a significant role forchanges in local currency reserve requirements (model 9, rr local) inmitigating the tantrum shock.

Second, the insignificance of most individual macroprudentialtools suggests that it is the compounded effects of various toolsthat matters, rather than the application of one “key” tool. Thisis consistent with the observation that macroprudential tools aimto increase the resilience of the financial system or to contain thebuildup of vulnerabilities.


le6.

The

Impac

tof

Indiv

idual

Mac

ropru

den

tial

Reg

ula

tion

inH

ost

Cou

ntr

ies

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Borr

ower

Sec

tor:

Tota

lTota

lTota

lTota

lTota

lTota

lTota

lTota

lTota

lV

aria

ble

s

Cum

ulat

ive

Hos

tss

cbIn

dex

1.87

8on

Res

iden

tial

Len

ding

(1.8

85)

Cum

ulat

ive

Hos

tss

cbIn

dex

0.98

4on

Con

sum

erLen

ding

(3.0

27)

Cum

ulat

ive

Hos

tss

cbIn

dex

0.39

5on

Oth

erLen

ding

(3.9

76)

Cum

ulat

ive

Hos

tca

pre

q5.

318

(4.6

15)

Cum

ulat

ive

Hos

tco

ncra

t2.

195∗

Inde

x(1

.206

)C

umul

ativ

eH

ost

ibex

Inde

x0.

996

(2.1

37)

Cum

ulat

ive

Hos

tltv

cap

1.66

6∗∗

(0.7

89)

Cum

ulat

ive

Hos

trr

fore

ign

−0.

343

(0.5

58)

Cum

ulat

ive

Hos

trr

loca

l0.

779∗

(0.4

05)

Con

stan

t−

0.27

30.

133

0.11

5−

4.41

−0.

826

−2.

456

−4.

587∗

∗∗0.

341

1.53

2∗∗

(0.4

75)

(0.2

08)

(0.8

68)

(3.8

65)

(0.9

27)

(2.6

58)

(0.7

43)

(0.2

28)

(0.6

93)

Obs

erva

tion

s1,

723

1,72

31,

723

1,56

41,

007

603

890

1,72

31,

723

R-s

quar

ed0.

056

0.05

60.

056

0.06

20.

072

0.18

40.

090.

056

0.05

7So

urce

Fix

edE

ffec

tsYes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Hos

tFix

edE

ffec

tsN

oN

oN

oN

oN

oN

oN

oN

oN

o

Note

s:Tab

le6

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

ndin

gbilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

),in

resp

onse

toa

one-

unit

chan

gein

the

regu

lato

rysu

b-index

assh

own.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;

***

p<

0.01

,**

p<

0.05

,*

p<

0.1.


Third, the capital requirement prudential tool (model 4, cap req)is also insignificant. This further confirms our reasoning in subsec-tion 5.2, namely that excluding capital requirements from the mainanalysis is reasonable not only based on the definition of macropru-dential tools but also from an empirical perspective.

The interpretation of these tool-specific results needs to be rathercautious. Importantly, we do not see our results as ranking macro-prudential tools by effectiveness for cross-border bank lending stabi-lization. Overall, our main body of evidence suggests that macropru-dential tools work together to increase the resilience of the financialsystem. Therefore, we take the results of table 6 to generally suggestthat the significant impact of some of the individual tools, as well asthe marked differences across tools, shall serve as solid motivationfor future research on comparing and contrasting the effectivenessof individual tools.

5.5 Interaction of Source and Host Regulations

In studying the stabilizing roles of source and host macroprudentialrules simultaneously, the question which naturally arises is: how dothese macroprudential policies interact? We address this question intable 7. Models 1–3 examine all borrower countries, while models4–6 and models 7–9 delineate the sample by advanced-economy andemerging market borrowers, respectively. For each sample, we exam-ine the roles of the levels of source (models 1, 4, and 7) and host(models 2, 5, and 8) regulations as well as their interactions, employ-ing one-sided fixed effects to control for unobservable credit shocks.The application of fixed effects follows and extends the Khwaja andMian (2008) type of identification applied on both sides in bilat-eral lending relationships, as in Avdjiev and Takats (2016). We theninvestigate the interaction of source and host regulatory rules by esti-mating equation (4)—applying fixed effects both across source (lend-ing) banking systems and across host countries (models 3, 6, and 9).By doing so, we isolate the interaction of source and host macropru-dential measures from any unobservable host-country-specific creditdemand or source (lending) banking-system-specific credit supplyshocks.

Our estimations show no evidence of interactions between sourceand host regulatory stringency, including those specifications with


Tab

le7.

Inte

ract

ion

ofSou

rce

and

Hos

tR

egula

tion

s:T

he

Impac

tof

Inte

ract

ion

bet

wee

nSou

rce-

and

Hos

t-C

ountr

yM

acro

pru

den

tial

Reg

ula

tion

son

Chan

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Host

Countr

y:

Full

Sam

ple

Adva

nce

dEco

nom

ies

Em

ergin

gM

arket

s

Len

din

gFlo

ws

by

Borr

ower

Sec

tor:

Tota

lTota

lTota

lTota

lTota

lTota

lTota

lTota

lTota

lV

aria

ble

s

Cum

ulat

ive

Sour

ceP

ruc6

0.24

0.34

5−

0.38

2In

dex

(0.9

22)

(0.9

78)

(1.7

45)

Sour

ceP

ruc6

Inde

x*H

ost

−0.

0527

−0.

0852

−0.

0443

−0.

218

−0.

29−

0.29

1−

0.00

772

0.12

60.

143

Pru

c6In

dex

(0.1

37)

(0.1

89)

(0.1

96)

(0.3

69)

(0.3

51)

(0.3

55)

(0.1

88)

(0.2

86)

(0.3

16)

Cum

ulat

ive

Hos

tP

ruc6

1.14

2∗∗∗

2.85

5∗∗∗

0.51

1In

dex

(0.2

79)

(0.8

19)

(0.3

41)

Con

stan

t−

22.2

7−

0.75

8∗−

26.9

17.1

9−

4.98

8∗∗∗

11.9

−21

.28

5.95

4∗∗∗

−22

.67

(16.

12)

(0.3

92)

(16.

82)

(12.

36)

(0.0

754)

(12.

6)(1

6.42

)(1

.165

)(1

8.95

)O

bser

vation

s1,

875

1,87

51,

875

747

747

747

976

976

976

R-s

quar

ed0.

056

0.06

30.

115

0.04

80.

110.

150.

072

0.06

20.

133

Sour

ceFix

edE

ffec

tsN

oYes

Yes

No

Yes

Yes

No

Yes

Yes

Hos

tFix

edE

ffec

tsYes

No

Yes

Yes

No

Yes

Yes

No

Yes

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

0.98

5−

0.73

8−

0.35

1.39

7−

0.44

−1.

31−

2.46

62.

853

1.61

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Hos

tC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

c6In

dex

−0.

428.

45−

0.38

4−

0.98

12.9

6−

0.44

1−

0.09

6.53

3.24

3

Note

s:Tab

le7

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

nd-

ing

bilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

),in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

Pru

c6in

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;

***

p<

0.01

,**

p<

0.05

,*

p<

0.1.


fixed effects on both the source and host sides. Estimating equation(4) for all countries (model 3), for borrowers in advanced-economyhosts (model 6), and in emerging market hosts (model 9) all yield sta-tistically insignificant coefficients. The results remain insignificantwhen we delineate the sample by target sector (bank and non-banklending, results available upon request). This suggests that macro-prudential tools applied in source (lending) banking systems and inhost countries did not significantly strengthen or weaken each other’slending shock-mitigating impact during the taper tantrum.

However, the results for the regressions which include the lev-els of regulatory stringency together with the interaction term con-firm our earlier results. The significant shock-mitigating impact ofhost-country macroprudential regulations is evident even after theinclusion of the interaction term, while the role of source regulationsremains insignificant (see models 1, 2, 4, 5, 7, and 8). In other words,our main results are robust to controlling for potential source–hostregulatory interactions.

Of course, this is not the final word on possible interactions asso-ciated with the application of macroprudential tools. We should becautious before dismissing the potential cross-border interaction ofmacroprudential tools for three main reasons. First, our method-ology of two-dimensional fixed effects in a single cross-section isvery demanding of our data. Second, we focus on the taper tantrumepisode alone—such regulatory interaction effects might be observedduring other stress episodes or during “normal” times. Third, ourregressions measure direct interactions, while macroprudential poli-cies may affect lending behavior indirectly.

6. Robustness Checks

We undertake a number of robustness tests to ensure that our resultshold under various specifications. Most importantly, we add a num-ber of control variables to confirm that our main results do notstem from omitted-variables bias (table 8). Specifically, in panel Aof table 8 we first add the pre-tantrum credit-to-GDP gap of sourceand host countries to our table 2 specifications. We then includeaverage credit growth over 2009–12 in panel B; average GDP growthover 2009–12 in panel C; and the average current-account-to-GDPratio over 2009–12 in panel D. In panel E, we add a measure of


le8.

Rob

ust

nes

sC

hec

ks

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

A.C

redi

tG

ap

Cum

ulat

ive

Sour

ceP

ruC

6−

0.91

9−

0.87

20.

0152

−0.

421

−1.

484

−1.

417

Inde

x(0

.975

)(0

.988

)(1

.557

)(1

.368

)(1

.423

)(1

.382

)So

urce

Cre

dit

Gap

0.40

2∗∗

0.38

5∗∗

0.04

480.

047

0.44

9∗∗

0.49

0∗∗

(0.1

47)

(0.1

41)

(0.2

56)

(0.2

13)

(0.2

05)

(0.1

83)

Cum

ulat

ive

Hos

tP

ruC

60.

749∗

0.68

9∗0.

759

0.61

11.

039∗

∗∗0.

967∗

∗

Inde

x(0

.395

)(0

.384

)(0

.612

)(0

.539

)(0

.367

)(0

.385

)H

ost

Cre

dit

Gap

0.24

10.

271

0.22

10.

254

0.17

60.

198

(0.1

63)

(0.1

63)

(0.3

16)

(0.3

14)

(0.1

21)

(0.1

22)

Obs

erva

tion

s1,

380

1,67

21,

210

1,24

41,

440

1,10

81,

272

1,56

61,

136

R-s

quar

ed0.

080.

070.

020.

050.

060.

000.

110.

080.

03

B.C

redi

tG

rowth

Cum

ulat

ive

Sour

ceP

ruC

6−

0.72

2−

0.73

6−

1.42

2−

1.89

9−

0.39

8−

0.18

8In

dex

(1.0

69)

(1.1

25)

(1.8

56)

(1.4

10)

(1.5

75)

(1.6

33)

Sour

ceC

redi

tG

row

th0.

897

0.87

61.

834

1.86

4−

0.08

11−

0.15

4(0

.860

)(0

.876

)(1

.345

)(1

.154

)(0

.720

)(0

.758

)C

umul

ativ

eH

ost

Pru

C6

1.32

7∗∗∗

1.27

3∗∗∗

1.60

71.

386

1.33

4∗∗

1.40

6∗∗∗

Inde

x(0

.446

)(0

.454

)(1

.163

)(1

.169

)(0

.513

)(0

.503

)H

ost

Cre

dit

Gro

wth

−0.

229

−0.

275

−0.

548

−0.

472

−0.

0288

−0.

237

(0.3

55)

(0.4

00)

(1.1

79)

(1.2

11)

(0.3

80)

(0.4

15)

Obs

erva

tion

s1,

380

1,67

21,

210

1,24

41,

440

1,10

81,

272

1,56

61,

136

R-s

quar

ed0.

080.

060.

010.

050.

060.

010.

110.

070.

01

(con

tinu

ed)


le8.

(Con

tinued

)

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

C.G

DP

Gro

wth

Cum

ulat

ive

Sour

ceP

ruC

60.

581

0.39

10.

631

0.20

3−

0.40

8−

0.45

9In

dex

(0.8

10)

(0.7

43)

(1.5

56)

(1.3

95)

(1.5

17)

(1.5

12)

Sour

ceG

DP

Gro

wth

−0.

75−

0.47

7−

0.61

7−

0.20

4−

0.06

010.

0121

(0.8

86)

(0.8

23)

(1.6

62)

(1.5

49)

(0.9

48)

(0.9

43)

Cum

ulat

ive

Hos

tP

ruC

61.

017∗

1.08

6∗∗

0.29

70.

525

1.41

6∗∗

1.32

3∗∗

Inde

x(0

.533

)(0

.508

)(1

.236

)(1

.195

)(0

.543

)(0

.538

)H

ost

GD

PG

row

th0.

0466

−0.

082

1.13

60.

96−

0.32

4−

0.52

1(0

.743

)(0

.742

)(1

.965

)(1

.943

)(0

.737

)(0

.740

)O

bser

vation

s1,

876

1,85

11,

851

1,59

21,

576

1,57

61,

734

1,71

81,

718

R-s

quar

ed0.

060.

060.

010.

050.

060.

000.

100.

070.

01

D.C

urre

ntAcc

ount

Bal

ance

Cum

ulat

ive

Sour

ceP

ruC

6−

0.15

5−

0.21

30.

850.

688

−1.

35−

1.42

4In

dex

(0.9

07)

(0.8

78)

(1.3

13)

(1.2

86)

(1.4

12)

(1.3

97)

Sour

ceC

urre

ntA

ccou

nt0.

362

0.37

6−

0.78

5−

0.73

61.

160∗

∗∗1.

160∗

∗∗

Bal

ance

(0.3

34)

(0.3

37)

(0.5

09)

(0.4

72)

(0.2

93)

(0.2

87)

Cum

ulat

ive

Hos

tP

ruC

61.

010∗

∗∗1.

002∗

∗∗0.

761

0.91

8∗1.

199∗

∗∗1.

044∗

∗∗

Inde

x(0

.322

)(0

.289

)(0

.482

)(0

.453

)(0

.340

)(0

.346

)H

ost

Cur

rent

Acc

ount

0.06

550.

075

0.26

50.

233

0.18

40.

225

Bal

ance

(0.3

42)

(0.3

38)

(0.5

84)

(0.5

82)

(0.2

06)

(0.2

06)

Obs

erva

tion

s1,

855

1,85

11,

830

1,58

41,

576

1,56

81,

713

1,71

81,

697

R-s

quar

ed0.

060.

060.

010.

050.

060.

010.

100.

090.

03So

urce

FE

Yes

No

No

Yes

No

No

Yes

No

No

Hos

tFE

No

Yes

No

No

Yes

No

No

Yes

No

(con

tinu

ed)


le8.

(Con

tinued

)

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

E.C

ontrol

ling

for

Inst

itut

iona

lQ

ualit

y

Cum

ulat

ive

Sour

ceP

ruC

60.

0585

0.15

20.

126

0.13

6−

0.41

4−

0.39

4In

dex

(0.7

25)

(0.6

54)

(1.3

50)

(1.2

21)

(1.2

98)

(1.2

95)

Sour

ceD

TF—

Eas

eof

−0.

506∗

∗−

0.47

3∗∗

−0.

491

−0.

479

−0.

33−

0.31

Doi

ngB

usin

ess

(0.2

19)

(0.2

20)

(0.3

24)

(0.3

36)

(0.2

34)

(0.2

37)

Cum

ulat

ive

Hos

tP

ruC

60.

971∗

∗∗1.

023∗

∗∗0.

784∗

1.00

5∗∗

1.14

9∗∗∗

1.00

6∗∗∗

Inde

x(0

.280

)(0

.248

)(0

.444

)(0

.409

)(0

.325

)(0

.323

)H

ost

DT

F—

Eas

eof

Doi

ng0.

0797

0.12

30.

0226

0.02

370.

161

0.23

8∗

Bus

ines

s(0

.141

)(0

.154

)(0

.352

)(0

.349

)(0

.110

)(0

.121

)O

bser

vation

s1,

695

1,72

41,

695

1,46

11,

489

1,46

11,

589

1,61

71,

589

R-s

quar

ed0.

060.

070.

010.

050.

060.

000.

100.

070.

01

F.C

umul

atin

gPru

C6

Inde

xSt

arting

2010

Cum

ulat

ive

Sour

ceP

ruC

61.

211

1.05

70.

0637

−0.

549

2.22

62.

319

Inde

x(1

.330

)(1

.298

)(2

.071

)(1

.940

)(1

.535

)(1

.476

)C

umul

ativ

eH

ost

Pru

C6

1.90

7∗∗

1.77

0∗∗

1.63

1.78

41.

456

0.96

3In

dex

(0.8

36)

(0.8

29)

(1.5

00)

(1.4

97)

(0.9

26)

(0.9

91)

Obs

erva

tion

s1,

723

1,72

31,

723

1,48

81,

488

1,48

81,

617

1,61

71,

617

R-s

quar

ed0.

060.

060.

010.

050.

060.

000.

090.

070.

01So

urce

FE

Yes

No

No

Yes

No

No

Yes

No

No

Hos

tFE

No

Yes

No

No

Yes

No

No

Yes

No

(con

tinu

ed)


le8.

(Con

tinued

)

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

G.Exc

ludi

ngLoc

alRes

erve

Req

uire

men

ts

Cum

ulat

ive

Sour

ceP

ruC

6−

0.03

61−

0.15

2−

0.31

6−

0.35

7−

0.20

6−

0.28

3In

dex

(0.7

51)

(0.7

13)

(1.2

51)

(1.1

62)

(1.3

00)

(1.3

00)

Cum

ulat

ive

Hos

tP

ruC

61.

192∗

∗∗1.

188∗

∗∗1.

117∗

1.26

8∗∗

1.39

3∗∗∗

1.20

6∗∗

Inde

x(0

.366

)(0

.328

)(0

.575

)(0

.508

)(0

.469

)(0

.478

)O

bser

vation

s1,

875

1,87

51,

875

1,59

11,

591

1,59

11,

734

1,73

41,

734

R-s

quar

ed0.

060.

060.

010.

050.

060.

000.

100.

070.

01

H.Foc

usin

gon

USD

-Den

omin

ated

Flo

ws

Cum

ulat

ive

Sour

ceP

ruC

61.

434

1.20

14.

088

3.87

20.

55−

0.06

16In

dex

(3.4

20)

(3.3

08)

(3.7

37)

(3.5

97)

(3.6

90)

(3.6

35)

Cum

ulat

ive

Hos

tP

ruC

61.

549∗

∗∗1.

301∗

∗1.

366

1.08

82.

151∗

∗∗1.

734∗

∗

Inde

x(0

.535

)(0

.508

)(0

.844

)(0

.906

)(0

.692

)(0

.788

)O

bser

vation

s1,

493

1,49

31,

493

1,12

01,

120

1,12

01,

319

1,31

91,

319

R-s

quar

ed0.

150.

080.

000.

090.

120.

000.

130.

080.

01So

urce

FE

Yes

No

No

Yes

No

No

Yes

No

No

Hos

tFE

No

Yes

No

No

Yes

No

No

Yes

No

Note

s:Tab

le8

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

t-st

andin

gbilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

),in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

mac

ropru

den

tial

regu

lato

ryin

dex

.T

he

resu

lts

inpan

els

Aan

dB

cont

rol

for

sourc

e-or

hos

t-co

unt

rycr

edit

-to-

GD

Pga

pan

dcr

edit

grow

th,

resp

ecti

vely

.In

pan

els

Can

dD

,th

ead

ded

cont

rols

are

GD

Pgr

owth

and

curr

ent

acco

unt

bal

ance

ofso

urc

eor

hos

tco

unt

ries

,re

spec

tive

ly.

The

resu

lts

inpan

elE

cont

rolfo

rin

stit

uti

onal

qual

ity

ofth

eso

urc

ean

dhos

tco

unt

ries

.T

he

resu

lts

inpan

els

Fan

dG

use

alte

rnat

ive

Pru

Cin

dex

es:

one

cum

ula

ting

star

ting

in20

10an

don

eex

cludin

glo

calcu

rren

cyre

serv

ere

quir

emen

ts,re

spec

tive

ly.Las

tly,

pan

elH

show

sre

sult

sfo

res

tim

atio

ns

usi

ng

only

USD

-den

omin

ated

lendin

gflow

sin

the

calc

ula

tion

ofth

edep

enden

tva

riab

le.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;**

*p

<0.

01,**

p<

0.05

,*

p<

0.1.


host-country institutional quality: the World Bank’s ease of doingbusiness index. Among these controls, we find that source credit-to-GDP gap is positively related to interbank lending flows during thetaper tantrum, as is the source current account balance in lending tonon-banks. We also find that source institutional quality amplified,while host institutional quality somewhat mitigated the tantrumshock to cross-border lending flows. In general, the estimates confirmour benchmark results from table 2: host-country macroprudentialmeasures seem to stabilize cross-border bank lending inflows, andthis effect is stronger for lending to non-banks than for interbanklending.

A second potential concern relates to the robustness of our resultsto the specific method of how we constructed our main PruC6 regu-latory index. We show that using the pre-defined PruC index (whichincludes capital requirements as well) does not materially affect ourbenchmark results (tables 3, 11, and 12). Furthermore, in order toexamine the robustness of our results to variations to the IBRNdatabase’s particular index construction method, we create a newindex similar to our Pruc6 index with the same eight sub-indexes—but without constraining the quarterly index value on the {–1, 0,1} spectrum. Again, the results remain robust and only marginallyweaker (detailed results are available upon request).

Third, we address concerns that the long period (2000–12) overwhich our regulatory index accumulates may drive our results. Thislong accumulation period might lead to reverse causality (due tomore sustained booms leading to larger accumulation of macropru-dential tools, for instance) or to increased noise (due to early-2000spolicies becoming less relevant by the time of the taper tantrum, forinstance). To address such concerns related to a long accumulationperiod, we define a new PruC7 index which cumulates macropruden-tial policies only after the financial crisis (starting in 2010:Q1), anduse this measure to reestimate our benchmark estimations. Our mainresults remain robust, but again with interbank lending coefficientslosing significance (table 8, panel F).

A fourth concern is that the inclusion of local currency require-ments among the macroprudential policies in the construction of ourPruC6 index may bias our results. This may be the case since theselocal currency reserve requirements may in some instances be used asa monetary policy tool (in lieu of policy interest rates, particularly


in emerging markets). When we reconstruct our macroprudentialindex without local currency reserve requirements and reestimateour benchmark specifications (table 8, panel G), we observe coef-ficient estimates that are closely comparable in significance to ourbenchmark table 2 results. This suggests that local currency reserverequirements are not a key driver of our results—and also furtherconfirms our benchmark specifications.

A fifth robustness check concerns the currency denomination ofcross-border bank lending flows. More specifically, we examine thepossibility that macroprudential policies had potentially strongereffects of USD-denominated lending during the taper tantrum. Avd-jiev and Takats (2016) show that cross-border bank lending declinedmore in bilateral lending relationships with higher USD shares. Fur-thermore, evidence from Takats and Temesvary (2017) points tocurrency denomination to be a channel of monetary policy transmis-sion. These results may imply that the taper tantrum, a U.S.-relatedfinancial shock, might affect USD-denominated lending dispropor-tionately. We redefine our dependent variable to include only USD-denominated lending flows, and reestimate our benchmark specifica-tions (table 8, panel H). We find that host-country macroprudentialpolicies indeed had a larger stabilizing impact on USD-denominatedlending inflows to the non-bank sector than on flows measured acrossall currencies. The implication is that host-country macroprudentialmeasures in place are particularly potent in stabilizing lending flowsin the currency whose issuing country the financial shock originatesfrom—and this is particularly so for lending to non-banks, the sec-tor whose uninterrupted credit access is particularly important forrobust real economic activity. This result is thus important from apolicy perspective as well.

Furthermore, the main thrust of the results remains robust evenwhen we specify a longer crisis window: we reestimate our benchmarkspecifications comparing changes in lending flows from two quartersbefore with two quarters after the taper tantrum (as opposed to theone quarter pre- and post-tantrum formulation in the benchmarkspecifications). The results are again comparable and we continueto see a significant stabilizing impact of macroprudential rules onlending to non-banks, with coefficient estimates similar in magni-tude to our benchmark results. This finding confirms that the timewindow specification is a not a key driver of our results.


In addition, our benchmark table 2 results remain generallyrobust to the following variations on our baseline specifications(detailed results available upon request):

• clustering standard errors along host countries,• no clustering of standard errors,• dropping source (lending) banking systems one by one from

the sample (to ensure that outliers do not drive the results),• dropping the host countries of borrowers one by one from the

sample (for the same reason as above).

Another feature of our analysis that we examine more closely is theuse of nationality-based banking statistics. In order to provide moreinsight into the role of this choice, we reestimate our benchmarkspecifications on residence-based data (which links lending countriesto borrowing countries; table 13 in the appendix). We find evidencethat host-country macroprudential tools stabilized total cross-borderlending inflows during the taper tantrum, though the sector-specificresults are weaker. The result is not fully surprising given the sim-ilarity between the residence- and nationality-based data sets inmost jurisdictions in our sample (apart from financial centers). Weattribute the somewhat weaker results in the residence-based datamainly to the fact that by construction in the residence-based dataour fixed effects are generally less potent in capturing unobserv-able differences across lending banking systems. This might in turnprovide for a less clear identification.

Lastly, in order to control for the potential confounding effectsof cross-country differences in monetary policy in our results, weinvestigate how the impact of macroprudential policies worked inan area with uniform monetary policy: within the euro zone. Werun estimations in which we constrain our sample to intra-euro-arealending, i.e., including only country pairs where both source (lend-ing) banking systems and host countries of borrowers are located inthe euro area. These intra-euro-area results (table 14 in the appen-dix) are generally in line with our earlier findings. Though the coef-ficients remain significant in the overall sample, statistical signifi-cance declines in the sector-specific delineations, perhaps reflectingthe smaller number of observations and less cross-country variationin macroprudential rules.5

5Our benchmark results also remain significant in the ex-euro-area sample.


7. Caveats

There are some caveats which deserve consideration in interpretingour results. First, we must consider the possibility that changes inmacroprudential tools are endogenous, i.e., they also signal macro-economic or financial vulnerabilities of either the source (lending)banking system or the host country of borrowers. This would implythat one could observe a negative association between the applica-tion of macroprudential tools and the resilience of cross-border banklending. However, this vulnerability signaling effect, to the degree itexists in the data, would work against the significance of our results,as it would imply weaker or outright negative coefficient estimates.Furthermore, the potential existence of such a signaling effect sug-gests that our results should be seen as a lower bound on the impactof macroprudential tools.

Second, the IBRN macroprudential database does not containinformation on the stance of individual macroprudential policy tools,which would allow cross-sectional comparison at a given point intime. Furthermore, we have no objective measure on how large afinancial shock the taper tantrum constituted. Combined, these fac-tors make it more difficult to precisely interpret the economic impactwe show. While it is clear that we identify an economically signifi-cant lending stabilization in the context of the taper tantrum shock,it is less clear how much stabilization a marginal macroprudentialpolicy action would yield during other economic shocks.

Third, there are potentially relevant non-linearities in the rela-tionship between macroprudential stringency and cross-border lend-ing stabilization. As we examine the circumstances of a particularidentified shock—the taper tantrum—our results do not make pre-cise predictions about the potential lending effects of other, differentglobal funding shocks. Therefore, our estimates should be read assuggesting that macroprudential tools can have a sizable stabilizingimpact during a stress scenario—not as a precise calibration.

Fourth, a potential complication is that the timing of the effectiveimplementation of tighter macroprudential measures is unknown,and thereby the effects can be confounded by expectations. Regula-tory actions are often expected beforehand, and the same is true formacroprudential measures. Our methodology, choosing a long win-dow prior to the tantrum episode, largely addresses this concern.


While our results hold for shorter windows as well, the potential roleof these expectations implies that applying even shorter time win-dows biases the analysis against finding significant results, becausemacroprudential actions in any given short window might have beenexpected beforehand.

Finally, we find heterogeneity in the lending-stabilization impactof macroprudential tools, across host countries (borrowers inadvanced economies versus in emerging markets), counterparties(bank versus non-bank borrowers), and macroprudential tools.Hence, our results should be read at most as providing broad sup-port for the role of macroprudential tools in stabilizing cross-borderbank lending, rather than as assessing specific tools or their effecton specific countries or sectors.

8. Conclusion

We show that macroprudential tools applied prior to the U.S. tapertantrum episode helped to stabilize lending during the tantrum.Most importantly, macroprudential tools enacted in the host coun-tries of borrowers stabilized cross-border bank lending significantly.The impacts are stronger for borrowers in advanced economies thanfor borrowers in emerging markets, and are also economically sig-nificant. Although the effect is present in lending to both bank andnon-bank borrowers, the statistical significance of our results forlending to non-banks is less robust.

The results are relevant for policymakers: they demonstrate thatmacroprudential tools not only stabilize the domestic financial sys-tem directly but also make cross-border bank lending inflows intohost economies more stable. The results which show strong and sig-nificant effects for macroprudential tools applied in host countriessuggest that, though international coordination can play a usefulrole, it is paramount that countries “keep their house in order.”While we do not find significant evidence for direct policy inter-actions, this does not imply that externalities do not exist. Morefinancial stability in one country could indirectly enhance financialstability elsewhere.

We hope that our results will pave the way for future research—which, together with our findings, will help us better understand theinternational ramifications of macroprudential policies.


Appendix

Table 9. Characterization of the BIS IBS Stage 1Enhanced Data

Currency Residence of Nationality ofBreakdown of Data Composition Borrower Lending Bank

Consolidated Data No Yes NoLocational Data

By Residence Yes Yes NoBy Nationality Yes No YesStage 1 Data Yes Yes Yes

Table 10. Types of Prudential Indexes

Nine Categories

sscb res Change in sector-specific capital buffer: Real estate credit.Requires banks to finance a larger fraction of theseexposures with capital.

sscb cons Change in sector-specific capital buffer: Consumer credit.Requires banks to finance a larger fraction of theseexposures with capital.

sscb oth Change in sector-specific capital buffer: Other sectors.Requires banks to finance a larger fraction of theseexposures with capital.

cap req Change in capital requirements. Implementation of Baselcapital agreements.

concrat Change in concentration limit. Limits banks’ exposures tospecific borrowers or sectors.

ibex Change in interbank exposure limit. Limits banks’ exposuresto other banks.

ltv cap Change in the loan-to-value ratio cap. Limits on loans toresidential borrowers.

rr foreign Change in reserve requirements onforeign-currency-denominated accounts.

rr local Change in reserve requirements onlocal-currency-denominated accounts.


le11

.B

orro

wer

sin

Adva

nce

dEco

nom

ies:

The

Impac

tof

Pru

den

tial

Reg

ula

tion

(Pru

C)

Inte

ract

ion

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ceP

ruC

0.02

07−

0.02

81−

0.19

7−

0.51

1−

0.01

390.

109

Inde

x(0

.756

)(0

.728

)(1

.38)

(1.2

96)

(1.1

)(1

.087

)C

umul

ativ

eH

ost

Pru

C2.

367∗

∗2.

409∗

∗2.

939

2.82

12.

302∗

∗2.

350∗

∗∗

Inde

x(1

.099

)(1

.129

)(2

.38)

(2.3

51)

(0.8

63)

(0.8

24)

Con

stan

t17

.04

−7.

311∗

∗∗0.

811

25.9

9−

15.8

9∗∗∗

−1.

599.

05−

1.04

1.72

1(1

2.52

)(1

.18)

(2.2

08)

(18.

7)(3

.416

)(3

.781

)(7

.434

)(0

.733

)(3

.005

)O

bser

vation

s74

774

774

768

868

868

870

870

870

8R

-squ

ared

0.05

0.11

0.01

0.06

0.08

0.01

0.05

0.13

0.01

Sour

ceFix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

CIn

dex

0.1

−0.

14−

0.98

−2.

55−

0.07

0.54

3

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Adv

ance

dH

ostEco

nom

ies

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

CIn

dex

11.8

312

.05

14.6

914

.111

.51

11.7

5

Note

s:Tab

le11

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

nd-

ing

bilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

)to

bor

row

ers

inad

vance

dec

onom

ies,

inre

spon

seto

aon

e-unit

chan

gein

the

cum

ula

tive

Pru

Cin

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;**

*p

<0.

01,**

p<

0.05

,*

p<

0.1.


le12

.B

orro

wer

sin

Em

ergi

ng

Mar

kets

:T

he

Impac

tof

Pru

den

tial

Reg

ula

tion

(Pru

C)

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ceP

ruC

−0.

0081

3−

0.06

060.

633

0.74

5−

1.31

7−

1.42

1In

dex

(0.9

59)

(0.8

53)

(1.8

65)

(1.8

79)

(1.8

78)

(1.8

47)

Cum

ulat

ive

Hos

tP

ruC

0.21

60.

164

0.03

70.

005

0.26

70.

248

Inde

x(0

.294

)(0

.236

)(0

.521

)(0

.415

)(0

.238

)(0

.248

)C

onst

ant

−21

.66

5.90

1∗∗∗

7.02

2∗∗∗

37.7

97.

512∗

∗∗5.

481

−28

.98∗

1.65

5∗∗∗

9.08

2∗∗∗

(16.

4)(0

.62)

(2.5

59)

(32.

24)

(1.1

25)

(4.3

23)

(16.

79)

(0.4

57)

(3.1

27)

Obs

erva

tion

s97

697

697

680

080

080

090

990

990

9R

-squ

ared

0.07

0.06

0.00

10.

060.

080.

001

0.1

0.09

0.01

Sour

ceFix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

CIn

dex

−0.

05−

0.36

3.80

4.47

−7.

90−

8.52

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Em

ergi

ngH

ostM

arke

tsat

the

90th

vers

usth

e10

thPer

cent

ileof

Pru

CIn

dex

3.03

2.30

0.52

0.07

3.74

3.47

Note

s:Tab

le12

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

t-st

andin

gbilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

)to

bor

row

ers

inem

ergi

ng

mar

kets

,in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

Pru

Cin

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;**

*p

<0.

01,**

p<

0.05

,*

p<

0.1.


Tab

le13

.R

esid

ence

-Bas

edD

ata:

The

Impac

tof

Pru

den

tial

Reg

ula

tion

(Pru

C6)

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ceP

ruC

6−

0.33

9−

0.26

1−

0.40

3−

1.03

40.

343

0.39

7In

dex

(0.8

66)

(0.8

30)

(1.0

44)

(1.1

61)

(0.8

89)

(0.8

37)

Cum

ulat

ive

Hos

tP

ruC

61.

268∗

∗∗1.

169∗

∗∗2.

196

2.38

4∗0.

240.

194

Inde

x(0

.275

)(0

.248

)(1

.337

)(1

.310

)(0

.323

)(0

.333

)O

bser

vation

s1,

094

1,09

41,

094

924

924

924

1,02

61,

026

1,02

6R

-squ

ared

0.04

0.09

0.01

0.09

0.09

0.01

0.05

0.04

0.00

Sour

ceFix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

C6

Inde

x

−1.

7−

1.3

−2.

01−

5.17

1.37

1.59

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Hos

tC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

C6

Inde

x

8.87

8.19

13.1

714

.31

1.68

1.36

Note

s:Tab

le13

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

resi

den

ce-b

ased

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

tsta

ndin

gre

siden

ce-b

ased

bilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

),in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

Pru

C6

index

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;**

*p

<0.

01,**

p<

0.05

,*

p<

0.1.


le14

.In

tra-

Euro

-Are

aLen

din

g:T

he

Impac

tof

Mac

ropru

den

tial

Reg

ula

tion

(Pru

c6)

onC

han

ges

inLen

din

gFlo

ws

from

Bef

ore

toA

fter

the

Tap

erTan

trum

Model

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

Len

din

gFlo

ws

by

Non-

Non-

Non-

Borr

ower

Sec

tor:

Tota

lTota

lTota

lB

anks

Ban

ks

Ban

ks

ban

ks

ban

ks

ban

ks

Var

iable

s

Cum

ulat

ive

Sour

ceP

ruc6

−0.

209

−0.

149

−0.

711

−0.

939

0.94

60.

971

Inde

x(1

.211

)(1

.163

)(2

.26)

(2.1

09)

(2.1

28)

(1.9

95)

Cum

ulat

ive

Hos

tP

ruc6

2.17

9∗∗

2.10

3∗∗

3.44

3.07

51.

877

1.98

0∗

Inde

x(0

.98)

(0.8

8)(2

.196

)(2

.049

)(1

.072

)(1

.027

)C

onst

ant

−14

.34

1.36

42.

734

−18

.52

−1.

103

2.49

1−

7.31

10.

816

2.43

6(8

.365

)(0

.803

)(3

.072

)(1

1.5)

(1.4

42)

(4.5

71)

(7.2

23)

(0.9

05)

(3.6

32)

Obs

erva

tion

s21

121

121

119

119

119

121

121

121

1R

-squ

ared

0.09

0.11

0.02

0.13

0.1

0.01

0.09

0.15

0.03

Sour

ceFix

edE

ffec

tsN

oYes

No

No

Yes

No

No

Yes

No

Hos

tFix

edE

ffec

tsYes

No

No

Yes

No

No

Yes

No

No

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Out

flow

sfrom

Ban

ksin

Sour

ceBan

king

Syst

ems

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

CIn

dex

−0.

836

−0.

596

−2.

84−

3.76

4.73

4.85

Diff

eren

cein

the

Tan

trum

Effec

ton

Bila

tera

lLen

ding

Inflow

sto

Bor

rower

sin

Hos

tC

ount

ries

atth

e90

thve

rsus

the

10th

Per

cent

ileof

Pru

CIn

dex

10.8

910

.52

17.2

15.3

79.

399.

9

Note

s:Tab

le14

show

sth

eper

cent

age

poi

ntch

ange

inth

eta

per

tant

rum

effec

ton

bilat

eral

lendin

gflow

s(m

easu

red

asth

ech

ange

inou

t-st

andin

gbilat

eral

ban

kcl

aim

sfr

ombef

ore

toaf

ter

the

taper

tant

rum

)w

ithin

the

euro

area

,in

resp

onse

toa

one-

unit

chan

gein

the

cum

ula

tive

Pru

Cin

dex

.Sum

mar

yst

atis

tics

and

vari

able

defi

nit

ions

are

show

nin

table

1.R

obust

stan

dar

der

rors

are

show

nin

par

enth

eses

bel

owth

eco

effici

ent

esti

mat

es;**

*p

<0.

01,**

p<

0.05

,*

p<

0.1.


References

Agenor, P.-R., E. Kharroubi, L. Gambacorta, G. Lombardo, andL. a Pereira da Silva. 2017. “The International Dimensions ofMacroprudential Policies.” BIS Working Paper No. 643.

Aiyar, S., C. Calomiris, J. Hooley, G. Korniyenko, and T. Wieladek.2014. “The International Transmission of Bank Capital Require-ments: Evidence from the UK.” Journal of Financial Economics113 (3): 368–82.

Alper, K., F. Altunok, T. Capacioglu, and S. Ongena. 2016. “TheEffect of US Unconventional Monetary Policy on Cross-BorderBank Loans: Evidence from an Emerging Market.” WorkingPaper.

Avdjiev, S., C. Koch, P. McGuire, and G. von Peter. 2017. “Inter-national Prudential Policy Spillovers: A Global Perspective.”International Journal of Central Banking 13 (S1): 5–33.

Avdjiev, S., R. N. McCauley, and H. S. Shin. 2015. “Breaking Free ofthe Triple Coincidence in International Finance.” BIS WorkingPaper No. 524.

Avdjiev, S., P. McGuire, and P. Wooldridge. 2015. “Enhancementsto the BIS International Banking Statistics.” IFC Bulletin (Bankfor International Settlements) 39 (April).

Avdjiev, S., A. Subelyte, and E. Takats. 2016. “The ECB’s QEand Euro Cross-Border Bank Lending.” BIS Quarterly Review(September): 99–113.

Avdjiev, S., and E. Takats. 2014. “Cross-Border Bank Lending dur-ing the Taper Tantrum: The Role of Emerging Market Funda-mentals.” BIS Quarterly Review (September): 49–60.

———. 2016. “Monetary Policy Spillovers and Currency Networks inCross-border Bank Lending.” Forthcoming in Review of Finance.

Berrospide, J., R. Correa, L. Goldberg, and F. Niepmann. 2017.“International Banks and Cross-Border Effects of Regulation:Lessons from the United States.” International Journal of Cen-tral Banking 13 (S1, March): 435–76.

Borio, C. 2003. “Towards a Macroprudential Framework for Finan-cial Supervision and Regulation?” BIS Working Paper No. 128.

Buch, C., and L. Goldberg. 2016. “Cross-Border Prudential Pol-icy Spillovers: How Much? How Important? Evidence from the


International Banking Research Network.” Staff Report No. 801,Federal Reserve Bank of New York (November).

Cecchetti, S., I. Fender, and P. McGuire. 2010. “Towards a GlobalRisk Map.” BIS Working Paper No. 309.

Cerutti, E., S. Claessens, and L. Laeven. 2017. “The Use and Effec-tiveness of Macroprudential Policies: New Evidence.” Journal ofFinancial Stability 28 (February): 203–24.

Cerutti, E., R. Correa, E. Fiorentino, and E. Segalla. 2016. “Changesin Prudential Policy Instruments—A New Cross-Country Data-base.” IMF Working Paper No. 16/110.

Cetorelli, N., and L. S. Goldberg. 2012. “Banking Globalization andMonetary Transmission.” Journal of Finance 67 (5): 1811–43.

Committee on the Global Financial System. 2011. “Global Liquid-ity — Concept, Measurement and Policy Implications.” CGFSPapers No. 45.

Crockett, A. 2000. “Marrying the Micro- and MacroprudentialDimensions of Financial Stability.” Remarks before the EleventhInternational Conference of Banking Supervisors, Basel, Switzer-land, September 20–21.

Fender, I., and P. McGuire. 2010. “Bank Structure, Funding Riskand the Transmission of Shocks across Countries: Concepts andMeasurement.” BIS Quarterly Review (September): 63–79.

Forbes, K., D. Reinhardt, and T. Wieladek. 2017. “The Spillovers,Interactions, and (Un)intended Consequences of Monetary andRegulatory Policies.” Journal of Monetary Economics 85 (C):1–22.

Forbes, K. J., and F. E. Warnock. 2012. “Debt and Equity-LedCapital Flow Episodes.” NBER Working Paper No. 18329.

Galati, G., and R. Moessner. 2011. “Macroprudential Policy — ALiterature Review.” BIS Working Paper No. 337.

Houston J., C. Lin, and Y. Ma. 2012. “Regulatory Arbitrage andInternational Bank Flows.” Journal of Finance 67 (5): 1845–95.

IMF–FSB–BIS (International Monetary Fund–Financial StabilityBoard–Bank for International Settlements). 2016. “Elements ofEffective Macroprudential Policies — Lessons from InternationalExperience.” (August).

Karolyi, G. A., J. Sedunov, and A. G. Taboada. 2017. “Cross-BorderBank Flows and Systemic Risk.” Working Paper.


Khwaja, A. I., and A. Mian. 2008. “Tracing the Impact of BankLiquidity Shocks: Evidence from an Emerging Market.” Ameri-can Economic Review 98 (4): 1413–42.

Laeven, L., L. Ratnovski, and H. Tong. 2014. “Bank Size and Sys-temic Risk.” IMF Staff Discussion Note 14/04.

Miranda-Agrippino, S., and H. Rey. 2015. “U.S. Monetary Policyand the Global Financial Cycle.” NBER Working Paper No.21722.

Ocampo, J. A. 2011. “Reforming the International Monetary Sys-tem.” WIDER Annual Lecture No. 14, United Nations Univer-sity World Institute for Development Economics Research (UNU-WIDER).

Ongena, S., A. Popov, and G. F. Udell. 2013. “When the Cat’sAway the Mice Will Play: Does Regulation at Home Affect BankRisk-Taking Abroad?” Journal of Financial Economics 108 (3):727–50.

Ongena, S., I. Schindele, and D. Vonnak. 2014. “In Lands of For-eign Currency Credit, Bank Lending Channels Run Through?The Effects of Monetary Policy at Home and Abroad on theCurrency Denomination of the Supply of Credit.” CFS WorkingPaper No. 474.

Takats, E., and J. Temesvary. 2017. “The Currency Dimension ofthe Bank Lending Channel in International Monetary Transmis-sion.” Finance and Economics Discussion Series No. 2017-001,Board of Governors of the Federal Reserve System.

Temesvary, J. 2018. “The Role of Regulatory Arbitrage in U.S.Banks’ International Flows: Bank-Level Evidence.” EconomicInquiry 56 (4): 2077–98.

Can Macroprudential Measures Make Cross-Border Lending ...Vol. 15 No. 1 Can Macroprudential Measures...

Documents

Transcript of Can Macroprudential Measures Make Cross-Border Lending ...Vol. 15 No. 1 Can Macroprudential Measures...