international banking and cross-border effects of ... · pdf fileinternational banking and...

International Banking and Cross-Border Effectsof Regulation: Lessons from Portugal∗

Diana Bonfim and Sonia CostaBanco de Portugal

This paper offers a contribution to understand the cross-border effects of bank regulation using data on Portuguesebanks. We find that the effect of foreign regulation on domes-tic credit growth depends on the type of regulation, on thechannel of transmission, and on the legal form of the bank.Our results show that a tightening in foreign regulation leadsto a decrease in the growth of domestic credit in the caseof concentration ratios and capital requirements and has theopposite effect in the case of sector-specific capital buffers andreserve requirements in foreign currencies. We also find sig-nificant cross-border effects for the loan-to-value limits. Inthis case, cross-border spillovers work in different ways fordomestic banks with international activity and for foreignbanks: after a tightening in this instrument abroad, domesticbanks decrease credit growth in Portugal while foreign banksincrease it. Finally, we show that the cross-border effects ofcapital requirements work differently through branches andsubsidiaries.

JEL Codes: F42, G21, G28.

∗We would like to thank Antonio Antunes, Luısa Farinha, Steven Ongena,Francisca Rebelo, Eric Wong, and the members of the IBRN Polish and Swissteams for relevant comments and suggestions. We would also like to thank Mar-garida Brites, Joao Falcao, Ricardo Martinho, Ana Beatriz Matos, Nuno MoraesSarmento, and Fatima Silva for their help in collecting the data and in the inter-pretation of some results. These are our views and do not necessarily reflect thoseof the Banco de Portugal or the Eurosystem.

341

342 International Journal of Central Banking March 2017

1. Introduction

This paper is part of the joint research project of the InternationalBanking Research Network (IBRN), which offers a unique opportu-nity to study the cross-border effects of banking regulation. Theproject looks at this issue from two angles: how foreign regula-tion affects domestic lending and how destination-country regulationaffects lending of domestic banks abroad. We focus our analysis onthe former, which is labeled the inward transmission channel.

The Portuguese banking system provides an interesting settingto analyze this issue. The international dimension of the Portuguesebanking system is relevant in two dimensions. First, domestic bankshave important activities abroad, and are thus exposed to foreignregulation through their branches and subsidiaries abroad. Second,foreign banks have a meaningful presence in the Portuguese bank-ing system. Recent consolidation developments within the Bank-ing Union suggest that this presence may be reinforced in the nearfuture. Furthermore, Portugal is part of the European MonetaryUnion, and thus does not have a domestically targeted monetarypolicy. Within this setting, macroprudential policy may play a keyrole in the management of country-specific imbalances. It is thusessential to understand the transmission of prudential policy throughbank lending. However, it is not enough to consider domestic pru-dential policy, as foreign regulation may also play an importantrole. The goal of this paper is precisely to gather evidence on thislatter mechanism. This is a very relevant issue for policymakers,most notably when considering the large number of macroprudentialpolicy measures being adopted worldwide.

Foreign banking regulation may have two opposing effects ondomestic credit. On one hand, we could expect that there are cross-border complementary effects arising from regulation: a tighteningin foreign regulation targeted at constraining lending in the homecountry may also lead to less lending in other countries. On the otherhand, there may be cross-border substitution effects: when facing atightening in foreign regulation, banks may actually increase lend-ing in other countries to diversify their exposures and to maximizeprofitability.

To analyze the effects of foreign regulation on domestic credit, weconsider two possible channels. First, we analyze the effect of foreign

Vol. 13 No. S1 Lessons from Portugal 343

regulation on the credit granted in Portugal by Portuguese bankswith activity abroad (specification A). We find that a tightening inforeign regulation yields an increase in the growth of loans granted bydomestic banks in Portugal. This substitution effect works throughsector-specific capital requirements and reserve requirements in for-eign currencies. For the loan-to-value (LTV) ratio, the results showthe opposite sign, thus supporting the cross-border complementaryeffects hypothesis. For this instrument a tightening might imply adecline in the profitability of the affiliates of Portuguese banks, whichcan lead to a reduction in the domestic activity due to the reduc-tion in profits for the banking group as a whole. In this case thereare cross-border complementary effects of foreign regulation, i.e., itseffects on constraining credit go beyond borders. Alternatively, itis also possible that this result is explained by the setting underwhich loan-to-values are usually tightened, which often correspondsto periods of booms in markets in which short-term profitabilitymight be very high despite the tightening of this instrument. If thisis the case, given that banks have limited resources, they may preferto increase credit abroad rather than continue to lend domestically.

Second, we analyze the influence of foreign regulation on thegrowth of credit granted in Portugal by the foreign banks operatingin the country (specification B). In this case we find that a tighten-ing abroad is associated with a decline in credit growth in Portugalin the case of general capital requirements and concentration ratios,while for the loan-to-value ratio we find the opposite effect. In thisspecification, we would expect that a tightening in regulation in thehome country of a given bank should constrain the whole activityof the banking group, including the activity of its affiliates abroad,most notably for instruments that are applied at the consolidatedlevel. This is consistent with the result we obtain for general capi-tal requirements and for concentration ratios. The result we obtainfor the loan-to-value ratio is possibly related to substitution effectsbetween the home and host country. In fact, loan-to-value limits areprobably applied only at the domestic level, thus making the substi-tution effects plausible. Additionally, given that regulators usuallytighten these instruments when home real estate markets are boom-ing and risks are building up, the substitution for credit in the hostcountry might reflect a diversification of exposures internationally.Foreign banks may be worried about the building up of risks in


their home country and thus increase credit abroad to diversify theirexposures and thus mitigate risk.

The results presented above are part of the core analysis com-mon to all country teams participating in the IBRN project andare part of the input for the meta-analysis in Buch and Goldberg(2017). In addition to these results, our paper focuses on one impor-tant additional dimension of analysis: the potentially heterogeneousrole of branches and subsidiaries in the cross-border effects of reg-ulation. More specifically, we use specification B to zoom in on thecross-border transmission of regulation and ask whether the regula-tion implemented in the home countries of foreign banks operatingin Portugal has different effects on the credit granted in Portugalthrough foreign branches and subsidiaries. This distinction is rel-evant if we consider the differences in the legal form of these twotypes of institutions: whereas branches are legally part of the parentforeign bank, subsidiaries are legally independent entities and mightbe allowed to fail on their own. This distinction has important reg-ulatory consequences. For instance, deposits held at subsidiaries areguaranteed by the host country, while those of branches are guaran-teed by the home country. Furthermore, and perhaps more relevantfor the purposes of our study, branches of European Union banksare exempt from capital requirements in the host country. We findthat the negative effects of tighter capital requirements in the homecountry of a foreign bank on credit growth in the host country workonly through branches. In the case of the loan-to-value, the increasein credit growth associated with a tightening operates as expectedthrough both branches and subsidiaries.

This paper is organized as follows. In section 2, we describethe data and present some stylized facts. In section 3 we discussthe empirical methodology and our results. Finally, in section 4 wepresent a few concluding remarks.

2. Data and Stylized Facts for Portugal

2.1 Bank-Level Data

We collect bank-level data from quarterly supervisory reports. Weuse solo-basis data, which allows us to focus the analysis on the


effect of foreign regulation on credit granted in Portugal. If we usedconsolidated data, we would be considering the effects of foreign reg-ulation on all credit granted by Portuguese banks, which includescredit granted by affiliates abroad. Further, all bank controls wouldrefer to this larger perimeter of activity. We considered that thiscould undermine the interpretation of the results.

Our analysis period begins in 2006:Q1 and ends in 2014:Q4.Some of the variables could be computed for earlier periods. How-ever, before 2005, banks used a different accounting system. Usinga longer period would imply important breaks in some series, whichare hard to address without compromising the quality of the data.Furthermore, the quality of analysis could also be compromised ifmany more years were included, as the beginning of that decadewas dominated by a merger wave that substantially changed thelandscape in the Portuguese financial system (for details, see Barroset al. 2014). During the analysis period, the structure of the Por-tuguese banking system was relatively stable. Furthermore, most ofthe changes in foreign regulation affecting Portuguese banks wereimplemented during the sample period.

We collect detailed information on key bank characteristics. Allfinancial institutions are classified as domestic or foreign, depend-ing on their ownership status. Foreign institutions are classified asbranches or subsidiaries and there is information on the country oforigin. Our data set only includes monetary financial institutions(i.e., banks in their classic definition, as these are the only institu-tions authorized to receive deposits from the public). We excludenon-monetary financial institutions from the analysis, as there areimportant differences in their funding models and in their regula-tion that may hamper the interpretation of the results. From apractical point of view, another reason to exclude these institu-tions is that there is no information on their exposures to for-eign countries. In addition, this choice enhances the comparabil-ity of the results with those of other countries participating in theproject.

In order to have data on the international activity of banks,we merge the supervisory bank database with the bank-level dataunderlying the international banking statistics reported to the Bankfor International Settlements (BIS). In particular, we use the BISdata, on a consolidated basis (i.e., excluding intragroup positions)


and on an immediate borrower basis, for the local claims and lia-bilities of the branches and subsidiaries of the Portuguese banks.Additionally, we use bank-level data collected for the constructionof the euro-area monetary financial statistics to obtain informationon assets and liabilities against the banks of the same banking grouplocated abroad. The use of these two alternative data sources impliedthe exclusion of the mutual agricultural credit banks from the sam-ple, as in these sources the data for this type of institution areaggregated at a consolidated level. In any case, given that theseinstitutions are devoted mainly to local activities and have a smallweight on the total credit (around 3.75 percent over the sampleperiod), we believe that their inclusion in the sample would not berelevant for the purpose of this study.1

We also merge the bank database with other data sources com-mon to the project, namely with the IBRN Prudential InstrumentsDatabase, described in section 2.2 (and, in more detail, in Ceruttiet al. 2017) and with economic and financial cycle data (obtained,respectively, from BIS 2014 and Drehmann, Borio, and Tsatsaronis2011). In both databases there is no information for Angola, so wehad to delete from our sample all banks belonging to Angolan bank-ing groups, which have a weight on the domestic credit lower than0.05 percent.

The final data set includes fifty-seven banks (twenty-five domes-tic and thirty-two foreign), which account on average over the sampleperiod for 96 percent of the credit granted by banks in Portugal.

2.1.1 Dependent Variables

In both specifications A and B, our dependent variable is ΔYb,t,which is defined as the quarterly change in credit granted by bank bto non-financial residents in Portugal in quarter t, measured in logpercentage points.

2.1.2 Balance Sheet Characteristics

To ensure the consistency in the IBRN project, it is of utmost impor-tance to guarantee that the explanatory variables used are as close

1All the bank-level data are subject to confidentiality rules.


as possible in the papers of each country team and described inBuch and Goldberg (2017). The variables considered in our spec-ifications are the percentage of a bank’s portfolio of assets that isilliquid (IlliquidAssetsRatiob,t−1), the percentage of the bank’s bal-ance sheet financed with core deposits (CoreDepositsRatiob,t−1), abank’s capital-to-asset ratio (CapitalRatiob,t−1), the percentage ofthe bank’s net external intragroup funding relative to its total lia-bilities

(NetIntragroupFundingb,t−1

), the log of total assets (Log-

TotalAssetsb,t−1), and the percentage of the assets plus liabilities ofa bank’s affiliates abroad relative to total assets plus total liabilities(InternationalActiv ityb,t−1). All the variables are defined in detailin table 6 in the appendix. Table 1 summarizes these indicators forthe full sample of banks operating in Portugal, as well as for domes-tic and foreign banks separately. Domestic banks are larger, bettercapitalized, less illiquid, and rely more on core deposits and less onnet external intragroup funding than foreign banks.

2.2 Data on Prudential Instruments

We use the IBRN Prudential Instruments Database, which isdescribed in Cerutti et al. (2017). The database includes quarterlyinformation on the timing of tightening or loosening of a number ofprudential tools in sixty-four countries over the period 2000–14.

The prudential tools considered are capital requirements,sectoral-specific capital buffers (which include an aggregate indexas well as indexes for real estate, consumption, and other loans),loan-to-value limits, foreign and local reserve requirements, inter-bank exposure limits, and concentration ratios. The database alsoincludes some summary measures of all the above tools. For each pru-dential tool, the database includes one index for its change, wherea negative value corresponds to a loosening, a positive value corre-sponds to a tightening, and zero signals that no change has occurredin the quarter.

In specification A we want to evaluate the impact of the pruden-tial regulation implemented in the countries where the Portuguesebanks have branches and subsidiaries. Thus, in line with the har-monized methodology for the IBRN project, we construct for eachPortuguese bank and prudential instrument an index (ExpPb,t) forthe change of the host countries’ regulation (HostPi,t), weighted


Tab

le1.

Sum

mar

ySta

tist

ics

onB

ank

Cre

dit

and

Char

acte

rist

ics

All

Ban

ks

Por

tugu

ese

Ban

ks

For

eign

Ban

ks

(N=

57)

(N=

25)

(N=

32)

Var

iable

Mea

nM

edia

nSD

Mea

nM

edia

nSD

Mea

nM

edia

nSD

Bal

ance

Shee

tD

ata

(for

Eac

hBan

kian

dQ

uart

ert)

Dep

ende

ntV

aria

bles

:D

omes

tic

Cre

dit

(Ln

Cha

nge)

(%)

0.31

8−

0.16

915

.34

0.38

0−

0.29

314

.13

0.26

6−

0.07

2016

.30

Inde

pend

ent

Var

iabl

es:

Log

Ass

ets

7.27

87.

088

1.95

27.

805

7.53

82.

090

6.83

16.

881

1.70

5C

apit

alR

atio

(%)

6.45

95.

116

12.7

78.

580

6.51

715

.30

4.66

03.

436

9.79

9Illiq

uid

Ass

ets

Rat

io(%

)79

.95

89.8

824

.13

78.6

188

.16

24.0

481

.09

92.5

724

.17

Inte

rnat

iona

lA

ctiv

ity

(%)

——

—2.

429

04.

075

——

—N

etIn

trag

roup

Fund

ing

(%)

25.3

64.

763

42.4

81.

297

09.

798

45.7

756

.42

48.8

5C

ore

Dep

osit

sR

atio

(%)

16.2

210

.34

18.3

025

.29

22.5

920

.72

8.52

22.

386

11.2

2

Note

s:T

heta

ble

prov

ides

sum

mar

yst

atis

tics

for

bank

bala

nce

shee

tan

dcr

edit

data

.D

ata

are

obse

rved

quar

terl

yfr

om20

06:Q

1to

2014

:Q4.

Ban

king

data

are

repor

ted

atth

eso

lole

vel.

All

vari

able

sar

ede

fined

inta

ble

6in

the

appen

dix.


by the bank foreign exposures to the host countries (∅b,i,t−1). Inthe calculation we used the weights data on the previous fourquarters.

ExpP b,t =∑

i

HostP i,t∅b,i,t−1

∅b,i,t−1 =∑t−1

t=t−4 exposureb,i,t∑

i

∑t−1t=t−4 exposureb,i,t

The exposure of the domestic bank b to country i is measured by theclaims plus liabilities of the branches and subsidiaries of that bankon country i, denominated in local currency (i.e., in the currency ofcountry i) and on an immediate borrower basis.

In the construction of these exposure-weighted prudential policyindexes, only exposures to countries with data available in the pru-dential database could be considered. In our sample, this means weare taking into account 87 percent of the total foreign exposures ofthe Portuguese banks, through their affiliates abroad.

With specification B we are interested in evaluating the impact ofthe regulation adopted in the home country of each foreign bank withbranches and subsidiaries in Portugal. Thus, in this case the regu-lation variables used in the regressions correspond to the indexes ofthe prudential database for the change in the prudential instrumentsin the countries of the parent banks (HomeP j,t).

Table 2 reports some descriptive statistics on the prudential pol-icy variables. Around 18 percent of the observations in the sample ofPortuguese banks (used in specification A) and around 14 percentin the sample of foreign banks (used in specification B) are asso-ciated with some change in the prudential variables.2 We excludefrom the analysis the indexes referring to the decomposition of thesectoral-specific capital buffer as well as other regulatory measures

2The sample used in specification B includes both domestic and foreign banks,but the statistics for the incidence of regulation were calculated using only for-eign banks. In fact, since we are interested in estimating the impact of foreignregulation, the regulation variable was set to zero for Portuguese banks in theregressions of specification B. This means regulation in Portugal is not explicitlyincluded in the regressions, although its effects are embedded in the time fixedeffects.


Tab

le2.

Sum

mar

ySta

tist

ics

onC

han

ges

inP

ruden

tial

Inst

rum

ents

Inw

ard:Spec

ifica

tion

A

Exposu

re-

Wei

ghte

dB

ase

Dat

a(B

efore

Aggre

gat

ing

toExposu

re-W

eighte

dM

easu

res)

Obse

rvat

ions

No.of

No.of

No.of

Countr

y-

Countr

y-

No.of

Countr

y-

Tim

eT

ime

Ban

k-

Pro

port

ion

Pro

port

ion

Tim

eC

han

ges

Chan

ges

Tim

eB

ase-

MP

PExpP

t

Inst

rum

ent

Chan

ges

(Tig

hte

nin

g)

(Loose

nin

g)

Chan

ges

Non-z

ero

Non-z

ero

Pru

dent

ialIn

dex

107

8027

209

0.01

40.

175

Gen

eral

Cap

ital

Req

uire

men

ts30

300

550.

003

0.03

5Se

ctor

-Spec

ific

Cap

ital

Buff

er17

152

360.

002

0.05

2Se

ctor

-Spec

ific

Cap

ital

Buff

er(R

ealE

stat

e)10

100

190.

001

0.03

1

Sect

or-S

pec

ific

Cap

ital

Buff

er(C

onsu

mpt

ion)

65

115

0.00

10.

019

Sect

or-S

pec

ific

Cap

ital

Buff

er(O

ther

)4

31

90.

001

0.01

2

Loa

n-to

-Val

ueR

atio

Lim

its

1811

736

0.00

20.

049

Res

erve

Req

uire

men

ts:Fo

reig

n19

109

310.

002

0.03

6R

eser

veR

equi

rem

ents

:Loc

al32

1319

600.

004

0.07

0In

terb

ank

Exp

osur

eLim

it11

110

160.

001

0.01

4C

once

ntra

tion

Rat

ios

87

117

0.00

10.

023

(con

tinu

ed)

Vol. 13 No. S1 Lessons from Portugal 351Tab

le2.

(Con

tinued

)

Inw

ard:Spec

ifica

tion

B

No.of

No.of

No.of

Countr

y-

Countr

y-

No.of

Countr

y-

Tim

eT

ime

Ban

k-

Pro

port

ion

Tim

eC

han

ges

Chan

ges

Tim

eH

om

ePt

Inst

rum

ent

Chan

ges

(Tig

hte

nin

g)

(Loose

nin

g)

Chan

ges

Non-z

ero

Pru

dent

ialIn

dex

4130

1113

60.

142

Gen

eral

Cap

ital

Req

uire

men

ts15

150

480.

050

Sect

or-S

pec

ific

Cap

ital

Buff

er10

82

210.

022

Sect

or-S

pec

ific

Cap

ital

Buff

er(R

ealE

stat

e)5

50

160.

017

Sect

or-S

pec

ific

Cap

ital

Buff

er(C

onsu

mpt

ion)

32

13

0.00

3

Sect

or-S

pec

ific

Cap

ital

Buff

er(O

ther

)2

11

20.

002

Loa

n-to

-Val

ueR

atio

Lim

its

30

323

0.02

4R

eser

veR

equi

rem

ents

:Fo

reig

n4

22

40.

004

Res

erve

Req

uire

men

ts:Loc

al14

59

290.

030

Inte

rban

kE

xpos

ure

Lim

it6

60

120.

013

Con

cent

ration

Rat

io3

30

190.

020

Note

s:T

his

table

show

ssu

mm

ary

stat

isti

cson

the

chan

ges

inth

ere

gula

tion

onpru

den

tial

inst

rum

ents

use

din

our

sam

ple

.In

the

pan

elfo

rsp

ecifi

cati

onA

,th

edat

are

fer

toch

ange

sin

regu

lati

onin

the

count

ries

wher

eth

ebra

nch

esan

dsu

bsi

dia

ries

ofth

ePor

tugu

ese

ban

ksar

elo

cate

dov

erth

eper

iod

2005

:Q4–

2014

:Q4.

Inth

epan

elfo

rsp

ecifi

cati

onB

,th

edat

are

fer

toch

ange

sin

regu

lati

onin

the

hom

eco

unt

ries

ofth

efo

reig

nban

ksop

erat

ing

inPor

tuga

lov

erth

eper

iod

2005

:Q4–

2014

:Q4.

Dat

aon

the

inst

rum

ents

com

efr

omth

eIB

RN

Pru

den

tial

Inst

ru-

men

tsD

atab

ase

des

crib

edin

Cer

utt

iet

al.(2

017)

and

are

onth

equ

arte

rle

vel.

The

num

ber

ofch

ange

sin

pru

den

tial

inst

rum

ents

isre

por

ted

onse

vera

ldim

ensi

ons,

i.e.

,on

the

count

ry-t

ime

leve

lan

don

the

ban

k-ti

me

leve

l.T

he

table

also

show

sth

esh

are

ofpru

den

tial

chan

ges

toto

talob

serv

atio

ns

(i.e

.,th

esh

are

ofnon

-zer

oob

serv

atio

ns)

.In

the

top

pan

el,th

eco

lum

n“E

xpos

ure

-Wei

ghte

dO

bse

rvat

ions”

isbas

edon

the

under

lyin

gdat

aon

pru

den

tial

chan

ges

info

reig

nco

unt

ries

(col

um

ns

under

the

“Bas

eD

ata”

hea

din

g).

The

repor

ted

dat

aar

ebas

edon

the

regr

essi

onsa

mple

.


(the interbank exposure limit and, in the case of approach B, alsothe reserve requirements in foreign currencies) with a sample varia-tion less than 2 percent, given that for these measures we were notable to obtain robust results.

In the case of capital requirements, as explained in Cerutti et al.(2017), all the changes correspond to tightening movements, sincethey refer to the implementation of Basel. For the sectoral-specificcapital buffer and the concentration limits, most of the changesin our sample also correspond to tightening movements. By con-trast, for the reserve requirements in both specifications, and for theloan-to-value limits in specification A, both tightening and looseningmovements occurred during the sample period.

2.3 Stylized Facts

In the period under analysis, credit granted in Portugal witnessedstrong movements. While in the mid-2000s credit was expandingquickly, it started to decelerate in 2008–09 during the global finan-cial crisis and has been declining since the beginning of the euro-areasovereign debt crisis and the Economic and Financial Assistance Pro-gramme to Portugal. Figure 1 shows that the evolution of credit inour sample is broadly consistent with the aggregate data of the mon-etary financial statistics. In this period, the behavior of domestic andforeign banks operating in Portugal has not always been alike (seefigure 2). In particular, while in the years 2010–11 domestic insti-tutions faced with the increase in funding difficulties and the needto deleverage started to reduce credit, foreign banks continued toexpand the credit granted in Portugal (Costa and Farinha 2011).In the most recent years, foreign banks have also cut their activityin Portugal. Nevertheless, their market share in the credit marketremained around 25 percent, which is slightly higher than what wasobserved before the crisis.

The Portuguese banking system is highly concentrated. The fivelargest banking groups accounted for around 75 percent of bankcredit to non-financial residents in Portugal in the last quarter of2014. One of these five groups is part of a large foreign bankinggroup. The rest of the Portuguese banking system comprises manysmall and medium-sized banks. Most of these banks are small-scaleuniversal banks, competing directly with the five largest banking


Figure 1. Credit Granted by Banks in Portugal

Notes: The figure depicts the year-on-year growth rate of credit granted bydomestic and foreign banks operating in Portugal. The solid line refers to dataused in this paper, which were compiled from supervisory reports, while thedashed line refers to data from the monetary and financial statistics publishedby Banco de Portugal.

Figure 2. Credit Granted by Domestic andForeign Banks in Portugal

Note: The figure depicts the year-on-year growth rate of credit granted by domes-tic and foreign banks operating in Portugal in solid and dashed lines, respectively,for the banks in the sample used in this paper.


groups. A few of them have specialized business models, offeringonly specific products such as consumer loans or asset managementservices.

By ownership nationality, Spanish banks dominate the marketwith a weight on the total credit granted by foreign banks of morethan 65 percent over the period under analysis. The other countrieswith a non-negligible presence in the Portuguese credit market arethe United Kingdom, Germany, and France.

Spain has also a dominant weight in the international activ-ity of Portuguese banks, accounting for around 30 percent of thetotal foreign exposure through affiliates over the period 2006–14.Additionally, domestic banks were, during our sample period, signif-icantly exposed to Poland and to a lesser extent to Greece, France,and some emerging market economies, such as Brazil, Angola, andMozambique.

3. Empirical Method and Regression Results

In this section we discuss the results of our empirical estimations,trying to understand how foreign regulation affected the evolu-tion of credit granted in Portugal. In section 3.1 we present theresults of the baseline specifications, which are common to all coun-try teams analyzing the inward transmission mechanism. In section3.2 we discuss the results of an extension to the baseline analy-sis, where we explore in depth the results from specification B totry to understand whether the transmission of foreign regulationthrough foreign banks operating in Portugal is different for branchesand subsidiaries. Finally, in section 3.3 we describe some of therobustness analyses and minor extensions done on these baselinespecifications.

3.1 Baseline Analysis of Inward Transmission ofPrudential Policies

3.1.1 Empirical Approach

The empirical approach we use to analyze the inward transmission offoreign regulation on loans granted by banks in Portugal is described


in detail in Buch and Goldberg (2017) and includes two differentspecifications.3

In specification A, the objective is to understand how foreignregulation affects the evolution of credit granted by domestic banks.The channel in focus in this specification comes from the exposuresthat domestic banks have abroad. To capture this, the regressions areestimated only for domestic banks. In this specification, the followingregression is estimated:

Specification A: Exposure-weighted inward transmission of regu-lation (see table 3).

ΔYb,t =2∑

k=0

αk+1ExpPb,t−k + α4Xb,t−1

+2∑

k=0

βk+1ExpPb,t−kXb,t−k + fb + ft + εb,t, (1)

where ΔYb,t is the quarterly log change in domestic credit of bank bat time t (measured in percentage). The prudential policy changesare captured by ExpPb,t−k, which measures exposure-weighted pru-dential policy outside Portugal. Xb,t−1 is the vector of bank con-trol variables. Its interaction with ExpPb,t−k captures the degree towhich a bank is exposed to changes in regulation through ex ante bal-ance sheet composition and market access. These regressions includebank and time fixed effects.

In specification B, the goal is to understand how foreign regu-lation affects the growth of credit granted in Portugal by branchesand subsidiaries of foreign branches. In this second approach, thefollowing specification is estimated:

3The IBRN project considers two mechanisms for the cross-border transmis-sion of prudential policies: inward and outward. We chose not to analyze theoutward transmission channel because the regulation of the Portuguese bankingsystem did not change significantly during the sample period and an importantpart of the international activity of Portuguese banks relates to emerging marketeconomies not covered in the prudential database.


Specification B: Inward transmission of home prudential policyvia affiliates (see table 4).

ΔYb,j,t = α0 +2∑

k=0

αk+1HomePj,t−k + α4Xb,j,t−1 + α5Zj,t

+2∑

k=0

βk+1HomePj,t−kXb,j,t−k + fb + ft + εb,j,t, (2)

where ΔYb,j,t is the quarterly change in log loans extended by bankb, from country j, to residents in Portugal at time t (in percentage).The prudential policy changes are captured by HomePj,t−k pruden-tial policy in the home country j of the parent bank. Xb,j,t−k is thevector of bank control variables. Its interaction with HomePj,t−k

captures the degree to which a bank b is exposed to changes in reg-ulation of country j through ex ante balance sheet composition andmarket access. Zj,t represents the economic and credit cycle vari-ables for country j. These regressions include bank and time fixedeffects. Standard errors are clustered by country.

Besides controlling for time fixed effects, as in equation (1), theseregressions control for macroeconomic and financial conditions in thehome country of foreign banks. The regressions are estimated for thefull sample, including domestic banks. However, for this latter group,the regulation variables and the financial and business cycle variablesare set to zero. This allows all the identification on the regulationand cycle variables to come from foreign banks. Domestic banksenter the regressions to provide more strength on the conclusionsregarding the effect of bank characteristics on credit growth.

3.1.2 Main Results

Table 3 presents the results of the estimation of equation (1). Weconsider contemporaneous effects and two lags for the regulationvariable. In the first lines of the table we report the results for thesethree terms and in the bottom of the table the results for the sum ofthe three coefficients. Given space constraints, for the interactionsof regulation with the bank control variables we report only thejoint economic and statistical significance of these three coefficients,


le3.

Inw

ard

Tra

nsm

issi

onof

Pol

icy

thro

ugh

Inte

rnat

ional

Expos

ure

sof

Dom

estic

Ban

ks

ExpP

=ExpP

=ExpP

=ExpP

=Sec

tor-

Res

erve

Res

erve

ExpP

=ExpP

=C

apit

alSpec

ific

ExpP

=R

equir

e-R

equir

e-C

once

n-

Pru

den

tial

Req

uir

e-C

apit

alLT

Vm

ents

:m

ents

:tr

atio

nIn

dex

Cm

ents

Buffer

Rat

ioFore

ign

Loca

lR

atio

s(1

)(2

)(3

)(4

)(5

)(6

)(7

)

Exp

Pt

41.5

7∗∗∗

−58

.08∗

∗∗13

.91

29.5

91.

467∗

∗0.

644

−1.

444∗

∗∗

(9.9

94)

(19.

08)

(22.

83)

(25.

49)

(533

.8)

(24.

57)

(382

.0)

Exp

Pt−

110

.82

41.5

837

.61

−15

6.3∗

390.

615

.04

−57

.00

(21.

03)

(39.

24)

(40.

59)

(79.

69)

(372

.3)

(28.

38)

(266

.2)

Exp

Pt−

228

.97

1.33

158

.39∗

∗−

112.

5∗∗

−40

8.2

19.9

7−

166.

5(1

7.73

)(2

1.62

)(2

7.30

)(4

7.77

)(3

04.7

)(4

3.10

)(1

47.2

)Log

Tot

alA

sset

s t−

12.

109

1.89

51.

062

3.00

21.

083

1.43

91.

331

(2.6

33)

(2.6

53)

(2.5

13)

(2.9

60)

(2.3

29)

(2.3

98)

(2.3

64)

Cap

ital

Rat

iot−

10.

0671

0.05

390.

0768

0.06

760.

106∗

0.08

230.

0886

∗

(0.0

692)

(0.5

334)

(0.0

585)

(0.0

625)

(0.0

539)

(0.0

509)

(0.0

478)

Illiq

uid

Ass

ets

Rat

iot−

10.

0337

0.04

190.

0543

0.02

560.

0451

0.03

860.

0346

(0.1

17)

(0.1

09)

(0.1

07)

(0.1

18)

(0.1

09)

(0.1

09)

(0.1

09)

Inte

rnat

iona

lA

ctiv

ity t

−1

0.14

10.

828∗

∗0.

347

0.76

3∗∗

0.56

70.

452

0.50

7(0

.245

)(0

.310

)(0

.282

)(0

.326

)(0

.352

)(0

.423

)(0

.337

)N

etIn

trag

roup

Fund

ing t

−1

0.09

150.

0955

0.05

760.

137∗

∗0.

0659

0.08

410.

0846

(0.0

672)

(0.0

648)

(0.0

683)

(0.0

663)

(0.0

655)

(0.0

621)

(0.0

627)

Cor

eD

epos

its

Rat

iot−

10.

101

0.07

680.

108

0.06

800.

0673

0.07

700.

0892

(0.1

35)

(0.1

25)

(0.1

30)

(0.1

36)

(0.1

24)

(0.1

29)

(0.1

29)

Log

Tot

alA

sset

s*E

xpP

−6.

19∗∗

∗4.

45∗∗

∗−

2.41

40.1

3∗∗∗

−10

6.59

∗∗∗

−6.

32−

70.9

0(8

.651

1)(1

1.29

28)

(0.6

192)

(5.0

245)

(9.4

879)

(1.5

684)

(0.8

783)

Cap

ital

Rat

io*E

xpP

0.04

2.35

∗∗∗

−0.

022.

201.

7∗∗∗

0.32

−4.

33∗

(0.2

869)

(10.

7245

)(0

.096

1)(1

.992

1)(3

3.61

21)

(0.2

125)

(2.7

86)

Illiq

uid

Ass

ets

Rat

io*E

xpP

−0.

34−

0.48

−0.

95∗∗

−0.

75−

7.35

∗∗∗

0.10

16.3

9∗∗∗

(0.5

509)

(1.7

82)

(3.4

517)

(1.1

176)

(9.4

271)

(0.5

166)

(6.0

442)

(con

tinu

ed)

358 International Journal of Central Banking March 2017Tab

le3.

(Con

tinued

)

ExpP

=ExpP

=ExpP

=ExpP

=Sec

tor-

Res

erve

Res

erve

ExpP

=ExpP

=C

apit

alSpec

ific

ExpP

=R

equir

e-R

equir

e-C

once

n-

Pru

den

tial

Req

uir

e-C

apit

alLT

Vm

ents

:m

ents

:tr

atio

nIn

dex

Cm

ents

Buffer

Rat

ioFore

ign

Loca

lR

atio

s(1

)(2

)(3

)(4

)(5

)(6

)(7

)

Inte

rnat

iona

l2.

74∗

1.41

∗∗0.

72−

0.92

35.5

5∗1.

65∗

34.2

9∗A

ctiv

ity*

Exp

P(2

.900

3)(3

.966

3)(1

.994

3)(1

.175

1)(2

.824

8)(2

.901

1)(2

.548

8)N

etIn

trag

roup

0.34

0.48

∗∗0.

82−

3.44

∗∗0.

1∗0.

033.

98Fu

ndin

g*E

xpP

(1.6

361)

(4.4

14)

(1.9

257)

(3.1

076)

(2.8

478)

(0.2

281)

(1.4

908)

Cor

eD

epos

its

−0.

08−

0.68

∗∗∗

−0.

27∗∗

∗−

4.12

∗∗3.

080.

5519

.7∗∗

Rat

io*E

xpP

(1.1

85)

(7.7

25)

(9.2

166)

(3.3

567)

(0.7

987)

(1.0

793)

(4.0

381)

Exp

P(E

xpP

t+E

xpP

t−1

81.3

531∗

∗−

15.1

710

9.90

69∗∗

∗−

239.

2609

∗14

49.2

935

.65

−13

67.1

461∗

∗

+E

xpP

t−2)

F-s

tatist

ics

(4.9

682)

(0.0

839)

(8.8

012)

(4.2

453)

(2.3

015)

(0.4

48)

(4.8

28)

p-va

lues

0.04

0.77

0.01

0.05

0.14

0.51

0.04

Ave

rage

Mar

gina

lE

ffec

tsof

Exp

P:

For

All

Ban

ks11

.32

−12

.32

11.9

7∗−

71.6

3∗∗

196.

99∗∗

13.3

6−

93.3

2Fo

rB

anks

/Per

iods

10.8

7∗∗∗

0.95

8.04

∗−

10.0

467

.85∗

∗∗7.

85−

4.55

with

Cha

nges

inE

xpP

Obs

erva

tion

s70

370

370

370

370

370

370

3A

djus

ted

R2

0.04

0.04

0.02

0.03

0.03

0.01

0.01

No.

ofB

anks

2525

2525

2525

25

Note

s:T

his

table

repor

tsth

eeff

ects

ofch

ange

sin

regu

lati

onan

dfirm

char

acte

rist

ics

and

thei

rin

tera

ctio

ns

onlo

gch

ange

sin

dom

esti

clo

ans.

The

dat

aar

equ

arte

rly

from

2006

:Q1

to20

14:Q

4fo

ra

pan

elof

dom

esti

cban

ks.For

eign

-exp

osure

-wei

ghte

dre

gula

tion

Exp

Pis

calc

ula

ted

asth

ew

eigh

ted

aver

age

ofch

ange

sin

fore

ign

regu

lati

onw

her

eth

ew

eigh

tsar

eas

sets

and

liab

ilit

ies

ofth

eban

kaffi

liat

esin

the

resp

ecti

vefo

reig

nco

unt

ry.For

Exp

Pin

tera

ctio

neff

ects

,th

ere

por

ted

coeffi

cien

tis

the

sum

ofth

eco

ntem

por

aneo

us

term

and

two

lags

,w

ith

the

corr

espon

din

gF-s

tati

stic

sfo

rjo

int

sign

ifica

nce

inpar

enth

eses

.For

mor

edet

ails

onth

eva

riab

les,

see

table

6in

the

appen

dix

.Eac

hco

lum

ngi

ves

the

resu

ltfo

rth

ere

gula

tory

mea

sure

spec

ified

inth

eco

lum

nhea

dline.

All

spec

ifica

tion

sin

clude

tim

ean

dban

kfixe

deff

ects

.Sta

ndar

der

rors

are

not

clust

ered

.**

*,**

,an

d*

indic

ate

sign

ifica

nce

atth

e1

per

cent

,5

per

cent

,an

d10

per

cent

leve

l,re

spec

tive

ly.


i.e., the results for the sum of the interactions with the contem-poraneous and lagged regulation. In order to have an idea of theimpact of regulation when both the direct effect and the interac-tions effects are taken into account, at the bottom of the table wealso include the average marginal effects of changes in regulation.The marginal effects are calculated both for all banks and for onlythe banks/periods subject to changes in regulation in our sample.

The first column reports the results for the aggregate pruden-tial index (PrudentialIndexC ) and the remaining columns showthe results for each prudential tool individually. By examiningthe lines of the table with the marginal effects, we can con-clude that foreign regulation affects the evolution of loans granteddomestically through the international exposures of domestic banks.This effect is statistically significant for the aggregate prudentialindex. This aggregate effect seems to work through specific instru-ments, for which we obtain statistically significant marginal effects:sector-specific capital requirements, loan-to-value ratios, and reserverequirements on foreign currencies. For the remaining instruments(general capital requirements, reserve requirements on local curren-cies, and concentration ratio), the effects of foreign regulation on thegrowth of credit granted by Portuguese banks are not statisticallysignificant.

Analyzing the statistical significance of the marginal effectsallows us to establish that there are cross-border spillovers of reg-ulation. However, it is also very important to understand in whichdirection these spillovers go. Does a tightening in regulation abroadlead to more or less credit at home? In aggregate terms, we find thata tightening in foreign regulation yields an increase in the growthof loans granted by domestic banks in Portugal. This result sug-gests that Portuguese banks operating internationally divert theirresources to internal markets when they face tougher regulationabroad. This aggregate effect is coming from the sector-specific capi-tal requirements and the reserve requirements on foreign currencies.For the loan-to-value ratio the effect is the opposite: a tighteningof this instrument abroad decreases credit growth domestically. Forthese instruments a tightening might imply a decline in the prof-itability of the affiliates, which can lead to a reduction in the domes-tic activity. It is also possible to argue that despite tighter loan-to-value limits, banks still find it profitable to lend abroad, given


that this instrument is usually tightened when credit and real estatemarkets are booming and hence (short-term) profitability might bevery high. Assuming that resources are limited, this might imply aconstraint in domestic credit. Cerutti et al. (2017) find that thereis a positive correlation between credit growth and the increase ofloan-to-value limits, thus supporting this hypothesis.

The magnitude of the marginal effects reflects the average impact(in percentage points) on the growth rate of credit of a simulta-neous tightening in regulation in all countries where Portuguesebanks have affiliates. Thus, we have computed the economic effectsof these changes by rescaling the marginal effects for the aver-age value of the exposure-weighted prudential measure observed inour sample in the periods of regulatory changes. After a tighten-ing in the sector-specific capital requirements, a tightening in thereserve requirements on foreign currencies, and a loosening in theloan-to-value ratio, on average, the quarterly loan growth rate forthe banks/periods exposed to the regulatory changes in our sampleincreased 3.1, 12.2, and 3.6 percentage points, which is around 35percent, 144 percent, and 44 percent of the mean absolute change ofcredit for these banks/periods. These large effects should be inter-preted with caution given the small number of regulatory changesanalyzed.

Though the signal of the effects of foreign regulation on theevolution of domestic credit is of primary interest, it is also rele-vant to understand exactly through which mechanisms these effectsare transmitted across borders. Our specification allows us to dothat through the analysis of the interaction terms. The substitu-tion effects of foreign regulation leading to an increase in domesticcredit growth, which work through sectoral capital buffers and for-eign reserve requirements, are stronger for smaller banks, as well asfor banks with more liquid assets and with more intense internationalactivity. In turn, the complementary effects arising from a tighten-ing in the loan-to-value ratio are reinforced for smaller banks andfor banks with more net external intragroup funding and a highercore deposits ratio. Banks’ size and liquidity thus seem to play animportant role in how foreign regulation affects domestic credit.

Table 4 presents the results of the estimation of equation (2),i.e., specification B. In this case, the goal is to understand how for-eign regulation affects credit granted in Portugal by branches and


le4.

Inw

ard

Tra

nsm

issi

onof

Pol

icy

via

Affi

liat

esof

For

eign

-Ow

ned

Ban

ks

Hom

eP=

Hom

eP=

Hom

eP=

Sec

tor-

Res

erve

Hom

eP=

Hom

eP=

Cap

ital

Spec

ific

Hom

eP=

Req

uir

e-C

once

n-

Pru

den

tial

Req

uir

e-C

apit

alLT

Vm

ents

:tr

atio

nIn

dex

Cm

ents

Buffer

Rat

ioLoca

lR

atio

s(1

)(2

)(3

)(4

)(5

)(6

)

Hom

ePt

1.15

6−

10.1

913

.78

82.4

7∗∗∗

−3.

993

8.65

7(6

.126

)(1

2.15

)(1

0.62

)(2

2.66

)(1

1.82

)(7

.127

)H

omeP

t−1

22.9

2∗∗

24.6

1∗34

.93∗

∗18

.43∗

∗∗−

0.67

3−

33.3

1∗∗∗

(7.8

32)

(11.

28)

(13.

15)

(3.5

01)

(10.

82)

(8.1

97)

Hom

ePt−

2−

2.69

1−

12.6

6∗∗

31.4

6∗18

.15∗

∗10

.08

−46

.09∗

∗∗

(5.4

56)

(5.5

55)

(15.

82)

(6.5

53)

(16.

26)

(7.9

25)

Log

Tot

alA

sset

s t−

1−

1.81

4−

1.30

0−

1.70

9−

1.51

4−

1.46

5−

1.43

2(1

.586

)(1

.557

)(1

.581

)(1

.558

)(1

.580

)(1

.532

)C

apital

Rat

iot−

10.

0713

∗0.

0835

∗0.

0798

∗0.

0753

∗∗0.

0790

∗0.

0785

∗

(0.0

341)

(0.0

374)

(0.0

396)

(0.0

329)

(0.0

368)

(0.0

366)

Illiq

uid

Ass

ets

Rat

iot−

1−

0.04

16−

0.05

77−

0.04

36−

0.07

39−

0.05

78−

0.05

31(0

.075

5)(0

.007

4)(0

.074

4)(0

.077

5)(0

.075

5)(0

.073

2)N

etIn

trag

roup

Fund

ing t

−1

−0.

0427

−0.

0408

−0.

0578

−0.

0376

−0.

0546

−0.

0525

(0.0

488)

(0.0

448)

(0.0

484)

(0.0

410)

(0.0

472)

(0.0

467)

Cor

eD

epos

its

Rat

iot−

10.

106∗

0.08

420.

0922

0.09

73∗

0.08

030.

0780

(0.0

557)

(0.0

651)

(0.0

643)

(0.0

492)

(0.0

656)

(0.0

662)

Fin

anci

alC

ycle

(Hom

eC

ount

ry)

−0.

0332

∗−

0.04

05∗

−0.

0438

−0.

0350

−0.

0473

−0.

0537

∗

(0.0

155)

(0.0

209)

(0.0

256)

(0.0

209)

(0.0

276)

(0.0

246)

Bus

ines

sC

ycle

(Hom

eC

ount

ry)

1.21

4∗∗

1.37

5∗∗

1.48

9∗∗

1.24

6∗∗

1.45

6∗∗

1.57

7∗∗∗

(0.4

43)

(0.4

70)

(0.5

06)

(0.4

95)

(0.4

96)

(0.4

76)

Log

Tot

alA

sset

s*H

omeP

0.51

0.23

1.51

1.10

1.10

−0.

90(0

.274

4)(0

.012

4)(0

.798

1)(2

.115

1)(0

.127

2)(0

.244

3)C

apital

Rat

io*H

omeP

0.03

−0.

54−

0.91

∗∗2.

44∗∗

∗0.

581.

6∗∗

(0.0

056)

(1.2

924)

(7.3

717)

(28.

9896

)(1

.035

8)(6

.287

2)Illiq

uid

Ass

ets

Rat

io*H

omeP

−0.

150.

16−

1.09

∗∗∗

−1.

51∗∗

∗−

0.18

0.49

∗∗∗

(1.2

312)

(0.7

227)

(24.

3021

)(1

7.74

)(1

.627

9)(1

2.26

58)

Net

Intr

agro

upFu

ndin

g*H

omeP

−0.

16−

0.37

∗0.

24∗

0.31

∗∗∗

0.08

0.18

∗∗∗

(1.6

166)

(3.9

579)

(4.8

046)

(14.

68)

(1.1

721)

(14.

1518

)C

ore

Dep

osits

Rat

io*H

omeP

−0.

55∗

−0.

41−

0.43

−0.

54∗∗

∗−

0.48

3.04

∗∗∗

(3.3

943)

(1.6

546)

(0.6

513)

(87.

823)

(1.8

278)

(56.

9368

)

(con

tinu

ed)


Tab

le4.

(Con

tinued

)

Hom

eP=

Hom

eP=

Hom

eP=

Sec

tor-

Res

erve

Hom

eP=

Hom

eP=

Cap

ital

Spec

ific

Hom

eP=

Req

uir

e-C

once

n-

Pru

den

tial

Req

uir

e-C

apit

alLT

Vm

ents

:tr

atio

nIn

dex

Cm

ents

Buffer

Rat

ioLoca

lR

atio

s(1

)(2

)(3

)(4

)(5

)(6

)

Hom

eP(H

omeP

t+H

omeP

t−1+

Hom

ePt−

2)

21.3

8∗1.

7580

.17∗

∗∗11

9.05

∗∗∗

5.41

−70

.75∗

∗∗

F-s

tatist

ics

(4.6

434)

(0.0

171)

(43.

432)

(20.

3492

)(0

.058

2)(1

6.47

92)

P-v

alue

s0.

060.

900.

000.

000.

810.

00

Ave

rage

Mar

gina

lE

ffec

tsof

Hom

eP:

For

Fore

ign

Ban

ks1.

00−

7.1∗

4.87

24.9

1∗∗∗

0.99

5.49

For

Fore

ign

Ban

ks/P

erio

ds1.

08−

6.73

∗3.

6911

.1∗∗

∗2.

36−

5.21

∗∗

with

Cha

nges

inH

omeP

Obs

erva

tion

s1,

619

1,61

91,

619

1,61

91,

619

1,61

9A

djus

ted

R2

0.05

00.

046

0.04

60.

052

0.03

80.

034

No.

ofB

anks

5757

5757

5757

Note

s:T

his

table

repor

tsth

eeff

ects

ofch

ange

sin

regu

lati

onan

dfirm

char

acte

rist

ics

and

thei

rin

tera

ctio

ns

onlo

gch

ange

sin

dom

esti

clo

ans.

The

dat

aar

equ

arte

rly

from

2006

:Q1

to20

14:Q

4.H

omeP

refe

rsto

the

chan

ges

inre

gula

tion

inth

ehom

e(i

.e.,

par

ent-

ban

k)co

unt

ryof

fore

ign

affiliat

es.For

Hom

ePin

tera

ctio

neff

ects

,th

ere

por

ted

coeffi

cien

tis

the

sum

ofth

eco

ntem

por

aneo

us

term

and

two

lags

,w

ith

the

corr

espon

din

gF-s

tati

stic

sfo

rjo

intsi

gnifi

cance

inpar

enth

eses

.For

the

Por

tugu

ese

ban

ksth

ere

gula

tion

vari

able

san

dth

efinan

cial

and

busi

nes

scy

cle

vari

able

sar

eze

ro.For

mor

edet

ails

onth

eva

riab

les,

see

table

6in

the

appen

dix

.Eac

hco

lum

ngi

ves

the

resu

ltfo

rth

ere

gula

tory

mea

sure

spec

ified

inth

eco

lum

nhea

dline.

All

spec

ifica

tion

sin

clude

tim

ean

dban

kfixe

deff

ects

.Sta

ndar

der

rors

are

clust

ered

byco

unt

ry.**

*,**

,an

d*

indic

ate

sign

ifica

nce

atth

e1

per

cent

,5

per

cent

,an

d10

per

cent

leve

l,re

spec

tive

ly.


subsidiaries of foreign banks. As shown in equation (2), we considercontemporaneous effects and two lags for the foreign regulation vari-able. As in the previous table, the reported coefficients for interactioneffects are the sum of the contemporaneous term and two lags. Forthe direct effects we report both the coefficients of the three HomePterms (in the first lines of the table) and their sum (at the bottomof the table). The table also includes the average marginal effects ofchanges in regulation and their significance, calculated for all the for-eign banks and for only the foreign banks/periods subject to changesin regulation in our sample.

As in table 3, the first column reports the results for the aggre-gate prudential index and the remaining columns show the resultsfor each prudential tool individually. At the aggregate level, changesin foreign regulation do not affect credit granted in Portugal in thisspecification. This is possibly because of the mixed directions ofeffects coming from different prudential tools. While for the loan-to-value ratio a tightening abroad is associated with more creditgrowth in Portugal, for the general capital requirements and theconcentration ratios we find the opposite. After a tightening in thecapital requirements, a tightening in the concentration ratio and aloosening in the loan-to-value limit, on average, the quarterly loangrowth rate for the banks/periods with changes in these regulatorymeasures in our sample declined 6.7, 5.2, and 11.1 percentage points,which is around 75 percent, 81 percent, and 138 percent of the meanabsolute change for these subsamples. As before, the magnitude ofthese effects should be interpreted with caution.

To better understand these results, it is important to discuss ourexpectations of this transmission channel. When regulation is tight-ened in the home country of a given bank, this might affect the wholeactivity of the banking group, including its affiliates abroad if theregulation is applied at the consolidated level. So, while in the pre-vious specification domestic banks could to some extent substitutebetween foreign and domestic credit when regulation was tightenedor loosened abroad, in this specification this substitution might bemore likely to occur in the case of regulations that are not appliedat the consolidated level. The results we obtain are in line withthis reasoning. In fact, both capital requirements and concentrationratios are usually applied at the consolidated level, while limits tothe loan-to-value ratio are most often applied at the local level, whenspecific risks are building up in the home country of the bank, where


most of its activity is usually concentrated. To be more effective,these instruments are typically targeted to the vulnerabilities theywant to address and thus do not cover the international activity ofbanks.

As before, our empirical strategy allows us to understand throughwhich channels these mechanisms are working by exploring theinteraction terms in the regressions. The negative effect of tightercapital requirements on credit growth in Portugal by foreign banksis mitigated when banks have less intragroup external net debt.Other indicators of banks’ financial strength and business modelsare not statistically significant. For concentration ratios the nega-tive effect on credit growth is mitigated by higher capital ratios,more illiquid assets, more net intragroup external debt, and morecore deposits, thus not providing a very clear picture about how thefinancial health of a banks’ affiliate affects this cross-border effect.Looking at the positive effect of a tightening in the loan-to-valueratio, we find that this effect is stronger when the affiliate becomesbetter capitalized and more liquid. This suggests that foreign bankswith better financial standing substitute some of the credit grantedabroad by domestic loans when lending requirements become tighterat home. Additionally, the substitution effect is stronger for the affil-iates that rely more on intragroup funding and less on deposits fromresidents in the host country.

In sum, the results suggest that the cross-border effects of regu-lation depend on the prudential tool considered as well as on thechannel of transmission. A tightening in foreign regulation leadsto a decrease in domestic credit growth in the case of concentra-tion ratios and capital requirements. These effects operate throughforeign banks located in Portugal. By contrast, in the cases of sector-specific capital buffers and foreign reserve requirements, a tighten-ing in foreign regulation leads to an increase in credit growth inPortugal. These effects operate through the domestic banks withinternational activity. We also find significant cross-border effectsfor the loan-to-value limits. In this case, it is interesting to notethat the cross-border spillovers work in different ways for domes-tic banks with international activity and for foreign banks—after atightening in this instrument abroad, domestic banks decrease creditgrowth in Portugal while foreign banks increase it. Since the tighten-ing of loan-to-value limits generally occurs when real estate markets


are booming, one possible explanation for these different effects isthat Portuguese banks might constrain their domestic credit growthto be able to increase credit abroad, while foreign banks might bemore worried with the building up of risks in the home country(where most of their activity is concentrated) and thus increasecredit growth abroad.

3.2 Further Exploring Cross-Border Spillovers ofPrudential Policies

In this section we extend our previous analysis in several directionswith two purposes: to gain further insight on some issues and to testthe robustness of the results to different specifications.

The most important extension is related to an effort to under-stand how the cross-border transmission of prudential policy worksthrough different types of foreign banks. More specifically, we lookseparately at the transmission through foreign branches and sub-sidiaries located in Portugal, as their legal form has implications forthe way regulation is applied. In this analysis we will focus on theprudential tools for which we find evidence of transmission throughforeign banks to the domestic economy and for which we have enoughvariation in our data: capital requirements and loan-to-value limits.

Afterwards, we summarize the results of the extensive battery ofrobustness tests we conducted on the baseline results.

3.2.1 Cross-Border Spillovers through Branches andSubsidiaries

A bank might be present in a foreign country through two differ-ent legal forms: a branch or a subsidiary. A branch is not a legallyautonomous entity and belongs directly to the parent bank. In turn,a subsidiary is a legally independent institution in the host coun-try. In legal terms, it works in a very similar way to the domesticbanks operating in that country, with the main difference being thatits capital is held by a foreign bank. For an uninformed customerthe differences between a branch and a subsidiary would not beperceptible, as the management of their operations and their rela-tionships with customers have no reason to differ. However, impor-tant differences apply in regulatory terms due to the legal nature


of each institution. For instance, deposits held by customers in abranch are guaranteed by the deposit guarantee scheme of the homecountry, while for the subsidiary the responsibility lies entirely withthe host country. More importantly for the purposes of our study,some prudential instruments are applied differently for branchesand subsidiaries. Cerutti, Dell’Ariccia, and Martınez Perıa (2007),Dell’Ariccia and Marquez (2010), Focarelli and Pozzolo (2005), andGoldberg and Saunders (1981) discuss in more detail some of thedifferences between branches and subsidiaries and the way bankschoose to expand internationally, while Peek and Rosengren (1997,2000) analyze the implications on the transmission of shocks.

The most relevant example in the European Union is perhaps thecase of capital requirements: branches of EU banks are exempt fromfulfilling capital requirements in the host country, but are directlysubject to capital requirements in the home country. In this setting,the cross-border implications of regulations may be differentiated.While both branches and subsidiaries are affected by the capitalrequirements implemented in the home country, only subsidiariesare affected by changes in capital requirements in the host country.In contrast, loan-to-value ratios are usually applied directly to expo-sures in markets in which there are concerns regarding the buildupof risks in real estate markets. Thus, if the regulator applies thismeasure in the home country, the loans granted by home-countryaffiliates abroad should not be directly affected.

Given these important differences, the cross-border effects of reg-ulation may depend on the legal form of foreign banks. To analyzethis, we adapt equation (2) and estimate the following regression:

Specification B1: Inward transmission of home prudential policyvia branches and subsidiaries (see table 5).

ΔYb,j,t = α0 +2∑

k=0

αk+1HomePj,t−kBranchb,t

+2∑

k=0

αk+4HomePj,t−kSubsidiaryb,t + α7Xb,j,t−1 + α8Zj,t

+2∑

k=0

βk+1HomePj,t−kXb,j,t−kBranchb,t


+2∑

k=0

βk+4HomePj,t−kXb,j,t−kSubsidiaryb,t

+ fb + ft + εb,j,t (3)

All the variables and estimation restrictions are the same as inequation (2). The only difference is that the prudential variable isinteracted with a categorical variable for branches and subsidiaries.The omitted category is the one referring to domestic banks. Theseregressions include bank and time fixed effects. Standard errors areclustered by country.

The results of this approach are presented in table 5.4 The resultsin table 4 show that tighter capital requirements in the home countryof a foreign bank are associated with less credit growth in the hostcountry. By looking at the marginal effects in table 5, we are able tofind that this cross-border spillover of regulation works only throughbranches. As discussed above, the impact of foreign regulation shouldin theory affect both types of foreign banks. One possible explana-tion for this difference might be the different way branches and sub-sidiaries are affected by capital regulation. Branches are only affectedby their home-country regulation, and so it makes sense to find thisstatistically significant spillover. In turn, subsidiaries are simultane-ously affected by home and host regulation. Capital requirementswere higher in Portugal than in most other European countries dur-ing a large part of the sample period. These measures were taken tostrengthen the resilience of the Portuguese banking system amidstan environment of erosion of trust. Given this backdrop, when cap-ital requirements were tightened in the home countries, their effecton subsidiaries was possibly not felt, as they were already subjectto more demanding capital requirements due to host regulation.

Regarding the loan-to-value ratio, in table 4 we reported that atightening in the home country implies more credit growth in thehost country through foreign banks. In table 5, we report positivemarginal effects both for branches and subsidiaries, which supportsour hypothesis that this instrument should affect in the same waythe two types of institutions.

4Given space constraints, we do not report the coefficients of the direct effectsof bank control variables.


Tab

le5.

Inw

ard

Tra

nsm

issi

onof

Pol

icy

via

Affi

liat

esof

For

eign

-Ow

ned

Ban

ks—

Bra

nch

esvs.

Subsi

dia

ries

Hom

eP=

Hom

eP=

Cap

ital

Pru

den

tial

Index

CR

equir

emen

tsH

om

eP=

LT

VR

atio

(1)

(2)

(3)

Hom

ePt*

Subs

idia

ries

−12

.01

−54

.73

−43

.82∗

∗∗

(6.6

77)

(39.

71)

(4.8

02)

Hom

ePt*

Bra

nche

s37

.28∗

11.6

625

.66∗

∗∗

(17.

57)

(26.

87)

(7.1

22)

Hom

ePt−

1*S

ubsi

diar

ies

4.11

4−

81.4

3∗26

.55∗

∗∗

(11.

40)

(44.

36)

(4.5

60)

Hom

ePt−

1*B

ranc

hes

30.2

5∗∗

28.4

134

.83∗

∗∗

(9.9

70)

(16.

16)

(8.1

06)

Hom

ePt−

2*S

ubsi

diar

ies

−6.

396

−81

.46∗

∗∗−

14.5

7(2

5.72

)(2

2.70

)(1

0.21

)H

omeP

t−2*B

ranc

hes

−13

.75

−8.

625

46.3

5∗∗∗

(9.4

95)

(7.3

57)

(8.0

19)

Fin

anci

alC

ycle

(Hom

eC

ount

ry)

−0.

0251

−0.

0348

−0.

0286

(0.0

164)

(0.0

232)

(0.0

234)

Bus

ines

sC

ycle

(Hom

eC

ount

ry)

1.23

1∗∗

1.39

8∗∗

1.23

5∗∗

(0.4

41)

(0.4

92)

(0.5

08)

Log

Tot

alA

sset

s*H

omeP

*Sub

sidi

arie

s3.

100

24.7

664∗

∗∗2.

768

(1.2

878)

(39.

8632

)(1

.438

2)Log

Tot

alA

sset

s*H

omeP

*Bra

nche

s−

3.05

92∗

−2.

801

13.9

214∗

∗∗

(4.4

34)

(1.5

096)

(61.

3848

)C

apital

Rat

io*H

omeP

*Sub

sidi

arie

s0.

497

4.45

52∗∗

∗6.

6972

∗∗∗

(0.7

014)

(10.

9829

)(2

8.81

27)

Cap

ital

Rat

io*H

omeP

*Bra

nche

s−

0.30

8−

0.78

7−

0.77

52∗∗

(0.1

381)

(0.9

531)

(7.3

751)

Illiq

uid

Ass

ets

Rat

io*H

omeP

*Sub

sidi

arie

s−

0.05

6−

0.12

5−

0.12

0(0

.050

3)(0

.044

3)(0

.079

9)Illiq

uid

Ass

ets

Rat

io*H

omeP

*Bra

nche

s−

0.27

50.

125

−2.

2814

∗∗∗

(1.2

421)

(0.2

071)

(89.

1302

)N

etIn

trag

roup

Fund

ing*

Hom

eP*S

ubsi

diar

ies

0.04

6−

0.04

50.

111

(0.0

798)

(0.8

222)

(0.2

571)

Net

Intr

agro

upFu

ndin

g*H

omeP

*Bra

nche

s−

0.15

8−

0.41

85∗

0.28

52∗

(1.4

909)

(4.6

719)

(4.9

366) (con

tinu

ed)


Tab

le5.

(Con

tinued

)

Hom

eP=

Hom

eP=

Pru

den

tial

Cap

ital

Hom

eP=

Index

CR

equir

emen

tsLT

VR

atio

(1)

(2)

(3)

Cor

eD

epos

its

Rat

io*H

omeP

*Sub

sidi

arie

s−

0.50

17∗

−0.

6817

∗−

0.69

82∗∗

∗

(4.8

538)

(3.3

879)

(20.

8261

)C

ore

Dep

osits

Rat

io*H

omeP

*Bra

nche

s−

0.63

6−

1.22

60.

364

(1.0

467)

(3.2

624)

(2.0

952)

Hom

eP(H

omeP

t+H

omeP

t−1+

Hom

ePt−

2)*

Subs

idia

ries

−14

.297

−21

7.62

86∗∗

∗−

31.8

47(0

.214

9)(1

5.24

4)(3

.088

)0.

654

0.00

40.

113

Hom

eP(H

omeP

t+H

omeP

t−1+

Hom

ePt−

2)*

Bra

nche

s53

.775

4∗∗

31.4

4810

6.84

09∗∗

∗

(6.4

632)

(1.2

814)

(34.

985)

0.03

20.

287

0.00

0

Ave

rage

Mar

gina

lE

ffec

tsof

Hom

ePfo

rFo

reig

nB

anks

:Fo

rSu

bsid

iari

es5.

4672

∗1.

038

44.2

201∗

∗∗

For

Bra

nche

s−

3.11

2−

12.2

22∗∗

∗27

.876

8∗∗∗

Ave

rage

Mar

gina

lE

ffec

tsof

Hom

ePFo

reig

nB

anks

/Per

iods

with

Cha

nges

inH

omeP

:Fo

rSu

bsid

iari

es6.

0252

∗∗−

2.26

913

.214

∗∗∗

For

Bra

nche

s−

2.60

5−

8.55

57∗∗

∗8.

1721

∗∗∗

Obs

erva

tion

s1,

619

1,61

91,

619

Adj

uste

dR

20.

056

0.04

70.

055

No.

ofB

anks

5757

57

Note

s:T

his

table

repor

tsth

eeff

ects

ofch

ange

sin

regu

lati

onan

dfirm

char

acte

rist

ics

and

thei

rin

tera

ctio

ns

onlo

gch

ange

sin

dom

esti

clo

ans.

The

dat

aar

equ

arte

rly

from

2006

:Q1

to20

14:Q

4.H

omeP

refe

rsto

the

chan

ges

inre

gula

tion

inth

ehom

e(i

.e.,

par

ent-

ban

k)co

unt

ryof

fore

ign

affiliat

es.For

Hom

ePin

tera

ctio

neff

ects

wit

hban

kch

arac

teri

stic

s,th

ere

por

ted

coeffi

cien

tis

the

sum

ofth

eco

ntem

por

aneo

us

term

and

two

lags

,w

ith

the

corr

espon

din

gF-s

tati

stic

sfo

rjo

int

sign

ifica

nce

inpar

enth

eses

.For

the

Por

tugu

ese

ban

ksth

ere

gula

tion

vari

able

san

dth

efinan

cial

and

busi

nes

scy

cle

vari

able

sar

eze

ro.For

mor

edet

ails

onth

eva

riab

les,

see

table

6in

the

appen

dix

.Eac

hco

lum

ngi

ves

the

resu

ltfo

rth

ere

gula

tory

mea

sure

spec

ified

inth

eco

lum

nhea

dline.

All

spec

ifica

tion

sin

clude

tim

ean

dban

kfixe

deff

ects

.Sta

ndar

der

rors

are

clust

ered

byco

unt

ry.**

*,**

,an

d*

indic

ate

sign

ifica

nce

atth

e1

per

cent

,5

per

cent

,an

d10

per

cent

leve

l,re

spec

tive

ly.


3.2.2 Further Extensions and Robustness Tests

To be sure of the validity of the results presented above, we didseveral further extensions to the analysis.5

On the construction of the data set, we estimated the baselineregressions using a sample with all explanatory variables winsorizedat the 1st and 99th percentile. The results are broadly consistent.Further, we tested the implications of excluding the smallest banksin the sample from the analysis. When we exclude banks with amarket share smaller than 0.05 percent in the loan market, thereare some changes in the results. One possible reason for this mightbe that when we exclude these small banks, there is much less vari-ation in foreign regulation, thereby affecting the precision of theestimation of the cross-border effects of regulation. This happensbecause even though these banks are very small, they represent animportant share of the number of observations (the total number ofbanks in the sample decreases from fifty-seven to thirty-one).

Since we decided to use solo data from supervisory reportsinstead of consolidated data, we considered that a relevant robust-ness test would be to estimate the regressions including bankinggroup fixed effects. The results are consistent in qualitative terms.

Another issue that could affect the results is the treatment ofmissing observations in the IBRN Prudential Instruments Database.In the baseline specifications, the missing observations are treated aszeros. If we keep them as missing, thereby losing some observations,the results are consistent.

One issue in which we differ significantly from the approach usedin other countries participating in the IBRN project is the definitionof the capital ratio. We use an unweighted accounting capital ratioinstead of regulatory tier 1 risk-based capital. Using tier 1 capitalwould eliminate from the sample all branches exempt from capitalrequirements in Portugal. Nevertheless, given the important differ-ences between the two variables, we also estimated whenever possiblethe regressions with tier 1 capital for a subsample, excluding foreignbranches for which there is no data on regulatory capital. The dif-ferences we find in the results are attributable to the change in thesample and not to change in the definition of the capital ratio.

5The results are not reported, but are available upon request.


In specification B we chose to include all banks, domestic and for-eign. All the observations concerning the home cycle and the homeregulation were set to zero for Portuguese banks. For robustnesspurposes, we estimate the regressions in specification B only for for-eign banks. We find some differences in the results, including thelack of significance of the negative marginal effect associated with atightening of capital requirements. However, we would like to notethat the estimations with foreign banks rely on a much smaller sam-ple of banks, most of which are relatively small. Furthermore, thereis a lot of heterogeneity in the business models of these banks, withsome being more targeted to consumer loans, others to asset manage-ment and investment banking, and others to local retail activities,for instance. The volatility generated by this smaller sample is themain reason to justify our inclusion of domestic banks in the baselineregressions.

Finally, as discussed before, there is a strong integration betweenthe Spanish and the Portuguese banking systems. Recent consolida-tion trends within the European Union will possibly reinforce thisintegration. As such, it might be interesting to focus in more depthon the transmission of regulation implemented in Spain. We reesti-mate specification B including only changes in Spanish regulation,and our previous results are much stronger: we find significant cross-border effects of regulation for all the instruments considered. Forthe aggregate prudential index, we find that tighter regulation inSpain is associated with more credit growth in Portugal, thus show-ing the existence of non-negligible substitution effects between thesetwo highly integrated banking markets. These effects work mainlythrough loan-to-value ratios and local reserve requirements. In con-trast, a tightening in general and sectoral capital requirements inSpain leads to a decrease in the growth of credit granted in Portugal.

Still focusing in Spain, there is one prudential instrumentthat deserves further analysis: dynamic provisions. As discussed inJimenez et al. (2015), this is one of the few time-varying regula-tory tools in the world. This tool was introduced in Spain in July2000 to improve the regulatory coverage of credit risk. The previ-ously existing provisioning system was highly procyclical, increasingin bad times, and one of the main goals of the new tool was toreduce that procyclicality (Trucharte and Saurina 2013). Given theprominent role that the countercyclical capital buffer plays in theBasel III framework and its similarities with the spirit of dynamic


provisions, this analysis offers a key input by providing for thefirst time evidence on the cross-border effects of a cyclical regu-latory tool. We do not find a significant effect of changes in thedynamic provisioning system on Portuguese banks working throughtheir exposures in Spain (specification A), but when we consider thecredit behavior of the affiliates of Spanish banks in Portugal, we findthat when dynamic provisions are loosened in Spain, the growth ofcredit granted by these affiliates increases in Portugal, thereby show-ing that this measure has significant cross-border spillovers (spec-ification B). However, it should be noted that our sample periodincludes only two episodes in which the regime of dynamic provi-sions was loosened, both during the global financial crisis. As such,these results should be interpreted with some reservations.

4. Concluding Remarks

The IBRN offers a unique opportunity to explore a common researchquestion with a common methodology across different countries,using high-quality data available at central banks worldwide. In thispaper we offer a contribution to the IBRN project on the cross-border impacts of prudential regulation. This contribution entailsthe estimation of the baseline specifications common to the projecton the inward transmission of foreign prudential regulation. Thisis the basis of the meta-analysis conducted in Buch and Goldberg(2017).

We find that a tightening in foreign regulation yields an increasein the growth of credit granted by domestic banks in Portugal,which suggests the presence of substitution effects. This effect worksthrough the sector-specific capital requirements and the reserverequirements on foreign currencies (and only through the foreignexposures of domestic banks). For the loan-to-value ratio, we obtainthe opposite sign, thus suggesting the existence of complementaryeffects. Indeed, a tightening of the loan-to-value ratio abroad is asso-ciated with a decrease in the growth of domestic loans granted byPortuguese banks. This result might stem from the reduction inprofits for the banking group as a whole. Alternately it might reflectthe conditions under which this instrument is usually applied, i.e.,periods of booms in real estate markets. Having limited resources,banks may prefer to limit domestic lending to continue to lendabroad if this market still yields high profitability despite the tighterregulation.


When we analyze the influence of foreign regulation on thegrowth of credit granted in Portugal by the foreign banks operatingin the country, the effects are mixed. A tightening of general capitalrequirements and concentration ratios is associated with less creditgrowth in Portugal, while a tightening in loan-to-value ratios has theopposite effect, reflecting possible substitution effects. These resultsare in line with what could be expected given that when regulation istightened in the home country of a given bank, substitution effectsare more likely to occur if regulation is applied at the local levelthan if it is applied at the consolidated level. It is interesting to notethat for the loan-to-value ratio the cross-border spillovers work indifferent ways for domestic banks with international activity and forforeign banks.

Our contribution also tries to understand whether the transmis-sion of foreign prudential policy through foreign banks operating in agiven country works differently through branches or subsidiaries. Wefind, as expected, that in the case of the loan-to-value ratio the posi-tive effect works both through branches and through subsidiaries. Bycontrast, the negative effect of tighter capital requirements, in thehome country of a foreign bank, on credit in the host country worksonly through branches. One possible explanation for this differencemight be the fact that when capital requirements were tightenedin the home countries, their effect on subsidiaries was possibly notfelt, as these banks were already subject to more demanding capi-tal requirements due to Portuguese regulation. These results showthat the legal form of credit institutions plays an important role inthe cross-border transmission of prudential regulation, most notablydue to differences in the scope and perimeter of application of theinstruments.

With increasingly harmonized regulation across the world, thisproject contributes to an understanding of how changes in pruden-tial tools in one country might affect the evolution of credit grantedin another country. This is relevant to thinking about intended andunintended international spillovers when designing regulation. Withincreased pressure for international reciprocity between regulators(as set out, for instance, in the countercyclical capital buffer frame-work), having at hand empirical evidence on the way regulationaffects lending in other countries will certainly be highly valuablefor policymakers.


Appen

dix

Tab

le6.

Con

stru

ctio

nof

Bal

ance

Shee

tIn

dep

enden

tV

aria

ble

s

Var

iable

Nam

eD

escr

ipti

onC

omm

ents

Dat

aSou

rce

Illiq

uid

Ass

ets

Rat

io(1

–(C

ash

and

Cla

ims

onC

entr

alB

anks

and

Cre

dit

Inst

itut

ions

/Tot

alA

sset

s))

(in

%)

The

defin

itio

nis

diffe

rent

from

that

used

inot

her

IBR

Nco

untr

yte

ams

due

todi

ffere

nces

inth

esu

perv

isor

yre

port

ing

tem

plat

es.T

heir

corr

elat

ion

shou

ldno

neth

eles

sbe

high

.

Supe

rvis

ory

Dat

a(B

anco

dePor

tuga

l)

Log

Ass

ets

Ln

(Tot

alA

sset

s/G

DP

Defl

ator

2012

)Su

perv

isor

yD

ata

(Ban

code

Por

tuga

l)an

dN

atio

nal

Acc

ount

s(S

tati

stic

sPor

tuga

l)C

ore

Dep

osit

sR

atio

(Tim

eD

epos

its

from

Res

iden

ts+

Dep

osit

sR

edee

mab

leat

Not

ice

from

Res

iden

ts+

Savi

ngs

Dep

osit

sfr

omR

esid

ents

)/Tot

alA

sset

s(i

n%

)

Supe

rvis

ory

Dat

a(B

anco

dePor

tuga

l)

Cap

ital

Rat

ioE

quity

Cap

ital

/Tot

alA

sset

s(i

n%

)T

his

vari

able

diffe

rssi

gnifi

cant

lyfr

omth

ere

gula

tory

tier

1ri

sk-b

ased

capi

talus

edby

othe

rIB

RN

coun

try

team

s.W

ede

cide

tous

eac

coun

ting

capi

talm

ainl

ybe

caus

ew

edo

not

wan

tto

excl

ude

from

the

sam

ple

fore

ign

bank

sex

empt

from

fulfi

lling

capi

talre

quir

emen

tsin

Por

tuga

l(s

uch

asth

ebr

anch

esfr

omE

Uco

untr

ies)

,as

thes

ein

stit

utio

nspl

ayan

impo

rtan

tro

lefo

rth

ean

alys

isof

the

cros

s-bo

rder

tran

smis

sion

ofba

nkre

gula

tion

.T

here

gula

tory

capi

talra

tio

also

has

the

disa

dvan

tage

ofbe

ing

affec

ted

bych

ange

sin

defin

itio

ndu

ring

our

sam

ple

peri

od.

Supe

rvis

ory

Dat

a(B

anco

dePor

tuga

l)

(con

tinu

ed)


Tab

le6.

(Con

tinued

)

Var

iable

Nam

eD

escr

ipti

onC

omm

ents

Dat

aSou

rce

Net In

trag

roup

Fund

ing

(Dep

osit

sof

Ban

ksof

the

Sam

eB

anki

ngG

roup

Loc

ated

Abr

oad

–C

redi

t,D

ebt

Secu

riti

esSh

ares

,an

dO

ther

Equ

ity

toB

anks

ofth

eSa

me

Ban

king

Gro

upLoc

ated

Abr

oad)

/Tot

alLia

bilit

ies

(in

%)

Mon

thly

Bal

ance

Shee

tSt

atis

tics

and

Supe

rvis

ory

Dat

a(B

anco

dePor

tuga

l)

Inte

rnat

iona

lA

ctiv

ity

Loc

alC

laim

s+

Loc

alLia

bilit

ies

(Den

omin

ated

inLoc

alC

urre

ncy)

ofth

eB

ranc

hes

and

Subs

idia

ries

(ofth

ePor

tugu

ese

Ban

ks)

Loc

ated

Out

side

Por

tuga

l/(T

otal

Ass

ets

and

Tot

alLia

bilit

ies

ofth

ePar

ent

Ban

k+

Loc

alC

laim

san

dLia

bilit

ies

ofth

eB

ranc

hes

and

Subs

idia

ries

Loc

ated

Out

side

Por

tuga

l)(i

n%

)

Ban

k-Lev

elD

ata

ona

Con

solid

ated

Bas

isU

nder

lyin

gth

eR

epor

tto

the

Inte

rnat

iona

lB

anki

ngSt

atis

tics

ofth

eB

ISan

dSu

perv

isor

yD

ata

(Ban

code

Por

tuga

l)


References

Bank for International Settlements. 2014. “Debt and the Finan-cial Cycle: Domestic and Global.” In 84th Annual Report, 65–84(chapter IV) Basel, Switzerland: Bank for International Settle-ments.

Barros, P. B., D. Bonfim, M. Kim, and N. Martins. 2014. “Counter-factual Analysis of Bank Mergers.” Empirical Economics 46 (1):361–91.

Buch, C., and L. Goldberg. 2017. “Cross-Border Prudential Pol-icy Spillovers: How Much? How Important? Evidence from theInternational Banking Research Network.” International Journalof Central Banking 13 (S1).

Cerutti, E., R. Correa, E. Fiorentino, and E. Segalla. 2017. “Changesin Prudential Policy Instruments: A New Cross-Country Data-base.” International Journal of Central Banking 13 (S1).

Cerutti, E., G. Dell’Ariccia, and M. S. Martınez Perıa. 2007. “HowBanks Go Abroad: Branches or Subsidiaries?” Journal of Bank-ing and Finance 31 (6): 1669–92.

Costa, S., and L. Farinha. 2011. “The Behaviour of Domestic andNon-domestic Banks in the Housing Credit Market: An AnalysisBased on Microeconomic Data.” In Financial Stability Report,November, 167–83. Banco de Portugal.

Dell’Ariccia, G., and R. Marquez. 2010. “Risk and the CorporateStructure of Banks.” Journal of Finance 65 (3): 1075–96.

Drehmann, M., C. Borio, and K. Tsatsaronis. 2011. “AnchoringCountercyclical Capital Buffers: The Role of Credit Aggregates.”International Journal of Central Banking 7 (4): 189–240.

Focarelli, D., and A. Pozzolo. 2005. “Where Do Banks ExpandAbroad? An Empirical Analysis.” Journal of Business 78 (6):2435–64.

Goldberg, L., and A. Saunders. 1981. “The Growth of Organiza-tional Forms of Foreign Banks in the U.S.: Note.” Journal ofMoney, Credit and Banking 13 (3): 365–74.

Jimenez, G., S. Ongena, J. L. Peydro, and J. Saurina. 2017. “Macro-prudential Policy, Countercyclical Bank Capital Buffers, andCredit Supply: Evidence from the Spanish Dynamic ProvisioningExperiments.” Forthcoming in Journal of Political Economy.


Peek, J., and E. Rosengren. 1997. “The International Transmissionof Financial Shocks: The Case of Japan.” American EconomicReview 87 (4): 495–505.

———. 2000. “Collateral Damage: Effects of the Japanese Bank Cri-sis on Real Activity in the United States.” American EconomicReview 90 (1): 30–45.

Trucharte, C., and J. Saurina. 2013. “Spanish Dynamic Provisions:Main Numerical Features.” Estabilidad Financiera (Banco deEspana) 25 (November): 11–47.

international banking and cross-border effects of ... · pdf fileinternational banking and...

Documents