Financial Stability and Central Bank Governance · 2014-12-01 · Vol. 10 No. 4 Financial Stability and Central Bank Governance 35 have extended this work and studied the optimal

Financial Stability and Central BankGovernance∗

Michael Koetter,a,b Kasper Roszbach,c,d andGiancarlo Spagnoloe,f,g

aFrankfurt School of Finance and ManagementbHalle Institute for Economic Research

cUniversity of GroningendSveriges Riksbank

eUniversity of Rome Tor VergatafSITE, Stockholm School of Economics

gCEPR

The financial crisis has ignited a debate about the appropri-ate objectives and the governance structure of central banks.We use novel survey data to investigate the relation betweenthese traits and banking system stability, focusing in particu-lar on their role in micro-prudential supervision. We find thatthe separation of powers between single and multiple bank

∗We are grateful to Franklin Allen, Mats Bergman, Mathias Dewatripont,Francesco Giavazzi, Iman van Lelyveld, Enrico Perotti, Marco Pagano, RobertoRocci, two anonymous referees, the editor (Andy Levin), and participants atseminars at Maastricht University, De Nederlandsche Bank, the 2004 Tor VergataConference, the 2006 Bocconi conference on Regulation and Supervision, the 2006Riksbank conference on the Governance of Central Banks, ASSA-AFA 2007, the2007 Central Banking Seminar in Cambridge, the Federal Reserve Bank of SanFrancisco, the 2012 conference on Corporate Governance of Financial Institutionsat De Nederlandsche Bank, and the 2013 Bocconi/Finlawmetrics conference onCentral Banking, Monetary Stability and Systemic Risk for comments and discus-sions on both early and more recent versions of work with these data. Particularthanks go to Lars Frisell, with whom Roszbach and Spagnolo conducted the sur-vey underlying this paper. We are also indebted to the BIS and Tonny Lybek atthe IMF for making their data available to us, and to David Slutsky, Karl Harmen-berg, Harvey Migotti, Mats Levander, and Mattias Quiroz for research assistance.Finally, we thank all central banks that participated in our survey. The viewsexpressed in this paper are solely the responsibility of the authors and shouldnot be interpreted as reflecting the views of the Executive Board of SverigesRiksbank. In an early stage, fragments of this paper circulated under the title“Governing the Governors: A Clinical Study of Central Banks.” Author e-mails:[email protected], [email protected], [email protected].

31

32 International Journal of Central Banking December 2014

supervisors cannot explain credit risk prior to or during thefinancial crisis. Similarly, a large number of central bank gov-ernance traits do not correlate with system fragility. Only theobjective of currency stability exhibits a significant relationwith non-performing loan levels in the run-up to the crisis. Thiseffect is amplified for those countries with most frequent expo-sure to IMF missions in the past. Our results suggest that thecurrent policy discussion on whether to centralize prudentialsupervision under the central bank and the ensuing institu-tional changes some countries are enacting may not producethe improvements authorities are aiming at. Whether otherpotential improvements in prudential supervision due to, forexample, external disciplinary devices, such as IMF conditionallending schemes, are better suited to increase financial stabil-ity requires further research.

JEL codes: G18, G34, G38, E58.

1. Introduction

The recent crisis intensified an already open debate on the appro-priate set of central banks’ objectives and tools, in particular withrespect to the prevention and management of future crises. Rethink-ing the overall financial market regulation, many economists andpolicymakers are asking whether new tools and mandates for pru-dential regulation and systemic financial stability should be givento central banks and, if so, then how to regulate the inevitable con-flicts between monetary policy, macro-systemic stability, and micro-supervisory objectives.

The UK government recently removed the full separationbetween the Bank of England and micro-prudential supervisorytasks allocated to the Financial Services Authority (FSA) abouttwo decades ago. This separation was considered ineffective in lightof the financial crisis, in particular in preventing Northern Rockand the Royal Bank of Scotland from taking too much risk. Part ofthe supervisory tasks will be moved to a subsidiary of the Bank ofEngland called the Prudential Regulation Authority. The previousseparation, however, was introduced in 1997 for the same reason:supervision under the Bank of England was judged ineffective withrespect to preventing excessive risk taking at the Bank of Credit

Vol. 10 No. 4 Financial Stability and Central Bank Governance 33

and Commerce International (BCCI), Barings, etc.1 Moreover, theperception that assigning bank supervision to the central bank is apoor institutional arrangement was so widespread before this crisisthat Norway (1900), Sweden (1907), Denmark (1919), and Canada(1925)2 assigned financial stability tasks to a separate supervisor,and Australia (1998), Japan (1998),3 Korea (1998), Iceland (1999),Austria (2002), and Germany (2002) transferred some responsibili-ties out of the central bank (see Herrero and del Rio 2005).

Notable commentators recently suggested that this widespreadperception was wrong (and so were all the mentioned institutionalreforms), i.e., that it is very important that central banks keepresponsibility for financial supervision (e.g., Davies and Green 2010).However, other at least as prominent observers suggest precisely theopposite, i.e., that an independent bank supervision authority ispreferable and only the new macro-systemic stability tasks and toolsshould be assigned to central banks (e.g., Goodhart 2010, Buiter2011).

The issue is all but obvious, as reasonable theoretical arguments(and models) exist in support of both views. This poses the impor-tant empirical question, which institutional arrangement performsbetter in terms of preventing financial distress, particularly in thewake of a financial crisis: micro supervision to a specialized authorityor to the central bank?

In this paper we try to answer this question by looking forempirical support for one policy or the other. We ask if systematicdifferences across central banks regarding their objectives, supervi-sory mandate, and some other relevant governance features help toexplain banking system fragility prior to and during the crisis. Tothis end, we use a new survey data set providing detailed informationon these central bank traits.

1 Although the FSA in the United Kingdom adopted its name in 1997, it didnot receive its current statutory powers until January 1, 2001.

2For Norway, Sweden, Denmark, and Canada we are referring to the estab-lishment of the legal predecessors of the current FSAs. Each of these countriesmerged its banking supervision authority with one or more other financial sectorsupervisors between 1986 and 1990 and renamed its national authority.

3Until 1998, Japan had a Banking Commission that was an agency of theMinistry of Finance.


We first describe the pre-crisis situation in terms of micro-financial supervision and the other central bank governance vari-ables for a set of forty-seven central banks as collected through ourad hoc survey.4 Then we focus on credit risk and regress mean sharesof non-performing loans (NPL) as a measure of financial fragility ona number of proxies for the governance and institutional traits ofprudential authorities.

Univariate analyses show that the correlation between both thegovernance of supervisory institutions and policy objectives withnon-performing loans is weak.

Multivariate regression analyses highlight that only the objec-tive of currency stability exhibits a consistently negative relation-ship with credit risk prior to the financial crisis. Neither indicatorsof institutional separation (or unification) of prudential and mone-tary policy nor proxies for the quality of governance at regulatorsand central banks influence average credit risk significantly. Over-all, a country’s legal origin, income dispersion, historical levels ofcredit risk, and OECD membership explain most of contemporane-ous credit risk realizations. For the crisis period itself, however, eventhese links break down. We find that countries participating in Inter-national Monetary Fund (IMF) support programs since 1964 mostfrequently are significantly more likely to adhere to explicit foreign-currency objectives. Potentially, the negative relation between cur-rency objectives and lower credit risk in the financial system mighttherefore reflect an improved and more stringent prudential attitudeadopted during longer exposure to disciplining IMF missions.

1.1 Related Literature

A large number of authors, starting with the classic papers by Kyd-land and Prescott (1977), Barro and Gordon (1983), and Rogoff(1985), have investigated how central bank objectives ought to beshaped for the purpose of monetary policy under the risk of politicalcapture by the government and of a consequent political inflationarybias. Persson and Tabellini (1993) and Walsh (1995), among others,

4Some tables, figures, or regressions may be based on fewer observationsbecause data were not available for—or a question did not apply to—a specificcentral bank.


have extended this work and studied the optimal contract for inde-pendent central bankers.5 Their focus, however, remains limited to“the commitment problem,” i.e., to the relevance of independentinstitutions for optimal monetary policy. Surprisingly, the designof the optimal contract for the supervisory tasks, historically theprime activity of many central banks, the possible trade-offs betweenmonetary policy, financial stability, and banking sector efficiency,as well as the link between central bank independence, account-ability, and supervisory performance, have received relatively littleattention.

One exception is Taylor and Quintyn (2002), who argue that“regulatory and supervisory independence” is important for finan-cial stability for the same reasons that central bank independenceis important for monetary stability. They list a number of dimen-sions of regulatory and supervisory independence—for example,budgetary freedom—but also contend that for the agency to befully effective, a supervisor’s independence in general needs to gohand in hand with accountability checks. These authors concludethat improper supervisory arrangements have contributed signifi-cantly to the deepening of several recent systemic banking crises.6

Still, in these papers the concept of “independence” remains focusedon independence from the political process, as a remedy for the riskof capture by the government. On the other hand, the literature onregulatory agencies has emphasized the risk of capture from the reg-ulated firms. Ensuring regulators’ independence from the industrythey regulate—by, for example, limiting revolving doors arrange-ments and post-employment possibilities—may therefore be crucial.Considering the enormous pressure to which central bank governorsare subjected by the financial sector in crisis situations—e.g., to bailout troubled institutions or to cut interest rates—the need to guar-antee this alternative kind of independence appears to have beenexcessively downplayed by the literature on central banks.

5A rich literature has developed out of these seminal contributions, includingrecent work such as Keefer and Stasavage (2001), looking into the importance ofchecks and balances; Eijffinger and Hoeberichts (2002), discussing the need formore transparency in the decision-making process; and Moser (1999), studyingthe factors that empirically determine independence.

6See also Barth, Caprio, and Levine (2004).


The issue of effective central bank governance and regulationbesides monetary policy tasks is not of purely theoretical interest,because it can affect the quality of bank regulation and supervision.Giannone, Lenza, and Reichlin (2011) show that the intensity andquality of bank regulation in a country are crucial for explaining themagnitude of damages caused by the recent financial crisis to thatcountry. Abstracting from most recent events, empirical work byDemirguc-Kunt, Laeven, and Levine (2004), Beck, Demirguc-Kunt,and Levine (2006), and particularly Guiso, Sapienza, and Zingales(2007) suggests that through its effect on the allocation of financialresources, the quality of bank regulation and supervision may havedramatic effects on economic growth.

Several previous contributions specifically consider whether itis better to allocate financial supervision to a separate, specializedagency rather than to the central bank. One of the earliest studiesthat recognize the trade-off faced by a central bank when attemptingto attain a multiplicity of goals is Schoenmaker (1992). He acknowl-edges that a central bank with supervisory responsibility may chooseto hold down interest rates because of concerns with the bank-ing system, although purely monetary considerations suggest higherrates. Hence, simultaneously striving for price stability and finan-cial stability involves a time-dependent trade-off. In line with this,Haubrich (1996) and Di Noia and Di Giorgio (1999) present evidenceof the inflation rate being higher and more volatile in OECD coun-tries where the central bank also has sole responsibility for bankingsupervision.7

Peek, Rosengren, and Tootell (1999) study how separating bank-ing supervision from monetary policy influences the effectiveness ofthe latter. They find that confidential supervisory information onbank ratings significantly improves forecast accuracy of variablescritical to the conduct of monetary policy. Their findings highlightthe informational synergies often mentioned in support of the argu-ment that central banks should have bank supervision responsibility.However, they do not study how this separation affects the qualityof bank supervision.

7 Whether such a higher rate of inflation is sub-optimal will depend, of course,on the exact weight the government attaches to the bank supervision (financialstability) objective.


Indeed, the main concern against assigning supervisory tasks tocentral banks, stressed once again in the recent theoretical analysisby Boyer and Ponce (2012), is the higher risk of regulatory captureit may generate, and appears to have generated in relation to, forexample, the savings and loans crisis in the United States and theBarings and BCCI cases in the United Kingdom.

Boot and Thakor (1993) provide another reason why supervi-sory tasks should be partially or completely separated from the cen-tral bank. They show that in the presence of uncertainty about thesupervisor’s ability (e.g., in evaluating the quality of banks’ assets),foreclosing a bank may signal poor monitoring ability. Hence, thesupervisor will tend to delay such decisions. Boot and Thakor (1993)suggest that the responsibility for bank closures should thereforebe separated from that for asset-quality monitoring. Naturally, effi-cient separation still requires a mechanism to induce the monitor totruthfully share its information with the regulator.

Kahn and Santos (2005) study the allocation of the lender-of-last-resort function, closure authority (supervision), and deposit insur-ance. They find that in the multi-regulator arrangement, it is bene-ficial to assign the supervision mandate to the deposit insurer. Thechoice between the unified-regulator arrangement and the multi-regulator arrangement involves a trade-off: The multi-regulatorarrangement reduces the forbearance problem at high levels of liquid-ity shortage but may exacerbate it at low levels. To our knowledge,there is not yet any empirical research that assesses the costs ofsupervision in different countries and relates these to the various out-puts of the supervisory system (e.g., official warnings, interventions,bank closures).

Finally, Hasan and Mester (2008) look at the effects of a num-ber of central banks’ organizational features on some performanceindicators in the 1990s, finding little association with the health ofthe commercial banking system as measured by the percentage ofproblem loans. They also find, however, that having financial super-vision as a main task of the central bank is associated with a worseperformance in terms of inflation.

The remainder of this paper is structured as follows. In section2, we explain our methodology and describe the data. Besidesconventional country-specific controls, we present data primarilysourced from our survey on central banks’ objectives, mandates,


and governance structures. Section 3 discusses the results, and weconclude in section 4.

2. Methodology and Data

First, we explain how we measure banking system fragility and con-trol for country traits. Second, we discuss the survey design andpresent in more detail the data relevant to our study.

2.1 Fragility and Country Controls

We use average non-performing loans, a measure of financial sys-tem fragility, as our dependent variable. This very simple measureis part of the financial statistics provided by the World Bank andhas two main advantages. It is comparable across countries and isconsidered by both academics and policymakers a key indicator of(excessive) credit risk taken by the banking system of a country.8 Wefirst short-list variables from the data described below, which exhibitthe highest univariate explanatory power in OLS regressions. Wethen specify multivariate regressions with the mean non-performingloan share relative to total loans over the period 2004 until 2006(pre-crisis NPL) and over the period 2007 until 2009 (crisis NPL).Table 1 provides descriptive statistics and results from univariateOLS for all variables considered, which serve as a selection device ofcovariates in multivariate analyses.

Before turning to the (mostly) survey-based central bank traits,we briefly present the country controls specified in the cross-sectionalregressions. The sample size is at most forty-five, which is determinedby the authorities (countries) responding to the survey. Appendix 1shows all forty-four countries and the European Central Bank (ECB)considered here.

Better enforceability of contracts and protection of creditorsshould generally correlate negatively with the share of NPL. Wemeasure the level of creditor rights suggested initially by La Portaet al. (1998). We obtain data for the years 2001 and 2008 from the

8We also considered capitalization ratios and z-scores (Laeven and Valencia2010), both of which corroborate the absence of results that we report below.Results are available upon request.

Vol. 10 No. 4 Financial Stability and Central Bank Governance 39Tab

le1.

Des

crip

tive

Sta

tist

ics

and

Univ

aria

teR

esults

Univ

aria

teO

LS

onPer

centi

leM

ean

Pre

-Cri

sis

NP

L

Var

iable

Mea

n1s

t99

thN

Coeffi

cien

tSE

R2

Dep

ende

ntVar

iabl

es

Mea

nN

PL

Shar

e(2

004–

6)2.

990.

2010

.00

Mea

nN

PL

Shar

e(2

008–

9)4.

300.

4561

.20

Cou

ntry

and

Cri

sis

His

tory

Con

trol

s

Cre

dit

Rig

hts 2

001

2.15

0.00

4.00

330.

538

[0.3

68]

0.06

5C

redi

tR

ight

s 200

82.

020.

004.

0044

0.48

9[0

.382

]0.

037

Ban

king

Cri

ses 1

970

0.75

0.00

4.00

431.

105

[0.4

53]∗

∗0.

127

Ban

king

Cri

ses 1

990

0.55

0.00

2.00

431.

507

[0.6

27]∗

∗0.

124

Cur

renc

yC

rise

s 197

01.

500.

006.

0043

0.37

7[0

.246

]0.

054

Cur

renc

yC

rise

s 199

00.

550.

003.

0043

0.63

6[0

.541

]0.

033

Deb

tC

rise

s 197

00.

300.

002.

0043

0.53

7[0

.655

]0.

016

Deb

tC

rise

s 199

00.

180.

002.

0043

0.56

9[0

.837

]0.

011

Leg

alO

rigi

nU

K0.

310.

001.

0044

−2.

549

[1.2

20]∗

∗0.

094

Leg

alO

rigi

nFr

ance

0.29

0.00

1.00

44−

0.00

7[0

.892

]0.

000

Leg

alO

rigi

nSo

viet

Uni

on0.

200.

001.

0044

1.84

4[0

.968

]∗0.

080

Leg

alO

rigi

nG

erm

any

0.09

0.00

1.00

44−

0.76

2[1

.410

]0.

007

Leg

alO

rigi

nSc

andi

navi

a0.

110.

001.

0044

0.10

2[0

.891

]0.

000

Gin

iIn

dex

35.9

024

.70

59.3

043

0.07

0[0

.042

]∗0.

065

Seco

ndar

ySc

hool

Enr

ollm

ent 2

000

17.5

03.

5044

.30

37−

0.08

6[0

.040

]∗∗

0.11

5M

ean

GD

P19

96−

2005

7.38

−6.

8411

2.00

440.

003

[0.0

22]

0.00

0A

vera

geTru

stIn

dica

tor

1.67

1.30

1.94

365.

473

[3.0

74]∗

0.08

5H

irsc

hman

-Her

finda

hlIn

dex

0.18

0.03

0.53

37−

4.44

2[3

.757

]0.

038

(con

tinu

ed)


Tab

le1.

(Con

tinued

)

Univ

aria

teO

LS

onPer

centi

leM

ean

Pre

-Cri

sis

NP

L

Var

iable

Mea

n1s

t99

thN

Coeffi

cien

tSE

R2

Cou

ntry

and

Cri

sis

His

tory

Con

trol

s

Pan

zar-

Ros

seH

-Sta

tist

ic0.

56−

0.48

1.52

371.

104

[0.8

89]

0.04

2Fo

reig

n-O

wne

dA

sset

Shar

e0.

310.

000.

9938

0.57

1[1

.350

]0.

005

CR

3R

atio

in20

000.

630.

211.

0043

−2.

062

[2.0

25]

0.02

5N

PL

Shar

ein

2000

7.23

0.50

29.3

042

0.29

6[0

.043

]∗∗∗

0.54

5O

EC

D0.

670.

001.

0044

−2.

409

[0.7

74]∗

∗∗0.

188

Cen

tral

Ban

kO

bjec

tive

s

Obje

ctiv

eM

onet

ary

Pol

icy

0.51

0.00

1.00

44−

0.40

0[0

.811

]0.

006

Obje

ctiv

eFo

reig

nE

xcha

nge

0.18

0.00

1.00

440.

625

[1.0

50]

0.00

8O

bje

ctiv

ePay

men

tSy

stem

0.07

0.00

1.00

44−

1.84

0[1

.589

]0.

031

Mon

etar

yPol

icy:

Yes

0.56

0.00

1.00

44−

1.01

0[0

.802

]0.

036

Mon

etar

yPol

icy:

No

0.13

0.00

1.00

440.

440

[1.1

83]

0.00

3

Supe

rvis

ion

and

Sepa

ration

ofPow

ers

Sole

Supe

rvis

or:C

B0.

510.

001.

0044

1.72

1[0

.769

]∗∗

0.10

7So

leSu

perv

isor

:FSA

0.13

0.00

1.00

44−

2.10

8[1

.140

]∗0.

075

Mul

tipl

eSu

perv

isor

s0.

360.

001.

0044

−0.

658

[0.7

66]

0.01

7

(con

tinu

ed)


le1.

(Con

tinued

)

Univ

aria

teO

LS

onPer

centi

leM

ean

Pre

-Cri

sis

NP

L

Var

iable

Mea

n1s

t99

thN

Coeffi

cien

tSE

R2

Cen

tral

Ban

kG

over

nanc

e

No

Bud

get

App

rova

l0.

770.

001.

0043

−0.

575

[1.0

14]

0.00

8N

oR

estr

icti

ons

onSt

affs’

Inve

stm

ent

0.11

0.00

1.00

44−

1.77

4[1

.252

]0.

046

No

Law

onP

rofit

Allo

cati

onof

CB

0.19

0.00

1.00

411.

934

[1.0

25]∗

0.08

4C

BG

over

nor

Rem

uner

atio

n32

8,64

615

,838

1,28

2,68

533

−0.

000

[0.0

00]

0.02

4Len

gth

ofG

over

nor

Ter

m14

.00

3.00

100.

0044

0.01

0[0

.015

]0.

010

Cou

ntof

Gov

erno

rD

ism

issa

lD

ueto

:1.

840.

006.

0044

−0.

646

[0.3

92]

0.06

1A

ctin

gag

ains

tG

over

nmen

tPol

icy

0.04

0.00

1.00

44−

1.45

9[1

.940

]0.

013

Non

-Ful

fillm

ent

ofR

equi

rem

ents

0.56

0.00

1.00

44−

1.17

2[0

.797

]0.

049

Non

-Ful

fillm

ent

ofD

utie

s0.

400.

001.

0044

0.39

6[0

.833

]0.

005

wit

hout

Pre

cise

Rea

soni

ng0.

130.

001.

0044

−0.

943

[1.1

76]

0.01

5Poo

rPer

form

ance

0.07

0.00

1.00

44−

1.75

6[1

.934

]0.

019

Mis

cond

uct/

Cri

min

alO

ffens

e0.

600.

001.

0044

−0.

370

[0.8

33]

0.00

5A

pptm

nt.C

ond.

onG

ood

Con

duct

0.04

0.00

1.00

44−

2.75

1[1

.906

]0.

047

Sepa

rate

Supe

rvis

ory

Boa

rd0.

490.

001.

0044

−0.

484

[0.8

11]

0.00

8E

xter

nalSu

perv

isor

yB

oard

Mem

bers

0.80

0.00

1.00

29−

0.06

3[1

.799

]0.

000

Num

ber

ofM

eeti

ngs

per

Yea

r12

.80

4.00

52.0

019

0.01

8[0

.047

]0.

009

CB

Foun

dati

onY

ear

1908

1668

1998

440.

011

[0.0

05]∗

∗0.

106

Note

s:D

ata

are

from

the

Wor

ldB

ank’

sW

orld

Dev

elop

men

tIn

dica

tors

(WD

I)an

dG

loba

lD

evel

opm

ent

Fin

ance

(GD

F)

data

base

,da

taba

nk.w

orld

bank

.org

/ddp

/hom

e.do

.


website of Andrei Shleifer.The variable is an index that ranges from0 to 4 in our sample, where higher values indicate better creditorrights. Univariate results in table 1 indicate that creditor rightsdo not correlate with NPL shares. We disregard the variable inmultivariate analysis.

A country’s history of crises may increase risk adversity of agentsand improve prudential monitoring procedures subsequently. But itmight also create path dependency if it takes long-lasting effortsto restore trust in financial markets. We therefore include a countvariable for three types of crisis sourced from Laeven and Valencia(2010): banking crises, currency crises, and debt crises. We spec-ify two counts. The first starts in 1970 and the second starts in1990. Only banking crises correlate significantly and positively withpre-crisis NPL shares, which indicates that bad debt is persistent.

Next, we control for the differences in legal origins across coun-tries (also obtained from the home page of Andrei Shleifer). Giventhe limited degrees of freedom due to the number of countries stud-ied here, we specify in the multivariate analyses below only onedummy that is significant in univariate analysis, namely for coun-tries with a legal UK heritage. Relative to any other legal heritage,UK-type legal systems exhibit lower levels of NPL shares in oursample.

To control for the differences in income and economic activityacross countries, we tested a number of conventional macro indi-cators provided in the World Development Indicators of the WorldBank, such as gross domestic product (GDP), GDP per capita, GDPgrowth across various time episodes, etc. We show in table 1 thedescriptive statistics for the specified measure of income dispersionacross countries, the Gini coefficient, which is weakly positively cor-related with NPL shares. Higher values indicate a larger dispersionof income within a country. We also show the share of studentsenrolled in secondary education in the total population as a measureof human capital intensity as well as mean GDP between 1996 and2005. Both variables are excluded below due to high correlation withthe Gini coefficient and limited degrees of freedom in multivariateanalysis.

Trust is another potential determinant of deciding whetherto extend a loan, as well as the risk of the latter turningnon-performing. We use a measure from the World Value Survey


questionnaire also used in, for instance, Guiso, Sapienza, and Zin-gales (2008). It gauges how much people generally think others canbe trusted. Higher values indicate larger skepticism. Despite univari-ate significance, we excluded the covariate from multivariate analysisbelow because of high correlation with other country controls andfairly low coverage of sampled countries.

Many studies discuss the relation between banking market struc-ture, competition, and risk (see, e.g., Keeley 1990, Koetter andPoghosyan 2009, and Martınez Miera and Repullo 2010). We there-fore test simple measures of concentration (Hirschman-Herfindahlindices and concentration ratios) and competition measures (Panzar-Rosse H-statistics), as well as the share of foreign-owned banks’assets. None of these correlates significantly with mean NPL shares.We nonetheless include one proxy in the multivariate regressionsbelow given the ample indications in previous literature that com-petition and market structure are important in explaining the risktaking of banks.

Bad credit is persistent and we expect that historically high lev-els of NPL shares are unlikely to vanish quickly. Therefore, we alsospecify the NPL share in the year 2000, which exhibits a signifi-cantly positive univariate effect on mean NPL shares in the threeyears prior to the crisis. Finally, we account for OECD member-ship in this fairly heterogeneous sample of countries by including anaccording indicator variable. OECD membership is also statisticallysignificant and correlates negatively with mean NPL shares.

2.2 Survey Data on Central Banks

In spring 2004 we formulated a list with nineteen composite ques-tions regarding the governance characteristics of a central bank. Werefer to appendix 1 for the complete list. The questions broadlyfocused on five areas: (i) the objectives and tasks of the bank, andthe exclusivity of its responsibilities; (ii) ownership structure, bud-getary freedom, and restrictions on the allocation of profits; (iii) theappointment, remuneration, tenure, and dismissal of governors; (iv)the appointment, tenure, and dismissal of supervisory board mem-bers; and (v) limitations on investments by bank staff and governors.In this paper, we use data from areas (i) and (ii).


To the extent that central banks provided information on theabove-mentioned areas, we collected the answers to these questionsfrom the websites of forty-seven central banks between June andAugust 2004. Between June and August 2005, we checked the web-sites once more to verify if more data was available.

The starting point for our sample consisted of all thirty-onemembers and two candidate members of the OECD as of 2005 andall but one G20 members that are not OECD members (nine). Tocome closer to a sample size of about fifty countries, we added fourdeveloped or emerging Asian economies as well as an EU candidatemember.

On September 16, 2005, we followed up on the first step of thedata collection by sending out a written questionnaire to all forty-six central banks with a request for relevant information that wasnot provided on the banks’ websites.9 Surveys were individualizedin the sense that questions to which we already had answers weredeleted from the “gross” list. In case a central bank had a researchdepartment at that point in time, we sent the survey to the direc-tor of research. For the remaining central banks, we directed theform to the respective heads of the economics department or to whatappeared to be the nearest alternative. Addresses, departments, andcontact persons were taken from the Bank for International Settle-ment’s (BIS’s) International Directory of central banks, which isupdated monthly by the BIS. Each questionnaire was accompaniedby a letter (appendix 2). Central banks were asked to return theform by October 6, 2005.

By October 6, 2005, we had received replies from four of the forty-seven central banks, although by October 13, this number had risento fourteen. On November 1, 2005, we sent out a reminder letter byregular mail to the twenty-six central banks that had not yet replied.If we were able to find the e-mail address of the addressee on theInternet, we also sent a copy of the reminder by e-mail. Within thefour weeks after this reminder, we received another nine replies, mak-ing the total, including Sweden, twenty-nine. Another four centralbanks had confirmed receipt of the questionnaire, without sendinga reply, one of which replied they would not be able to answer thequestions.

9No survey was sent to the Riksbank.


Between December 2005 and June 2006, one more bank con-firmed receipt of the survey without replying. On June 19, 2006, wesent a second reminder by e-mail to the remaining sixteen banksand followed up with a phone call within a week. This resulted inanother eleven responses. A personal letter we sent to the governoror a deputy governor of three central banks on July 5, 2006 led toanother two replies. By September 4, 2006, we had the replies fromforty-three of the forty-seven central banks and concluded our datacollection effort.

For eight central banks, the answer to at least one sub-questionwas missing or incomplete in the reply. In those cases, we contactedthe respondent with a request for a completion and attempted oncemore to locate the information on the website of that bank. Forsome variables and central banks, omitted replies indicated that aquestion did not apply. For some others it indicated unwillingness toreply and/or an inability by us to obtain an answer—for example,because we did not get the complete answer after a reminder. In thefinal data set, for example, nine countries did not fill out an answerto the question about governor wages. In sections 2.2.1 through 2.2.3below, we briefly discuss the survey results we use in this study. Formore information, we refer to a predecessor of this paper—Frisell,Roszbach, and Spagnolo (2008)—where the full survey data set isdiscussed.

2.2.1 Central Bank Objectives

We define three groups of objectives from these responses. First, wecode an indicator variable for monetary policy objectives if author-ities replied that they pursue price stability or monetary stability.Second, we specify a dummy for foreign exchange objectives if centralbanks respond that they aim to preserve either the external valueor the stability of the purchasing power of the currency. Third, wespecify payment system stability as the objective if central banksindicate to ensure an efficient functioning of payment systems orguide sound banking operations.

As figure 1 shows, these three objectives jointly apply to 70.2percent of the sampled countries. Roughly 50 percent of the banksindicate that price or monetary stability is their main objective,just over 8 percent that they have an efficient payments system or


Figure 1. Objectives of Central Banks

Notes: Replies from questionnaire; see appendix 1 for details. Multiple repliesper central bank are possible (N = 47).

banking system stability as their objective, and 18 percent that theyhave a forex objective. The remaining central banks indicate morecomplex objectives. The data in table 1 demonstrate that none ofthese objectives exhibits significant univariate explanatory power.We test for potential multivariate effects below. We also test whetherthe simple indicator “monetary policy is stipulated in the centralbank law as the prime objective” can explain mean NPL shares andfind this indicator is also uncorrelated.

2.2.2 Separation of Powers

We specify three categories to test for different arrangements regard-ing the separation of power. The first dummy indicates that acountry has only one supervisor, namely the central bank. Of allcountries, 51 percent belong to this category. The second dummyequals 1 for countries with just one supervisory authority but wherean institution other than the central bank, such as a financial super-visory agency, fulfills this function. Thirteen percent are in thiscategory. A third dummy variable captures jurisdictions with mul-tiple supervisors. This regime applies to more than one out of threejurisdictions.


Univariate regressions indicate that centralized supervision cor-relates with NPL shares. If central banks act as the sole prudentialsupervisor, mean NPL shares prior to the crisis are higher. Soleresponsibilities resting with an independent authority different thanthe central bank correlates negatively with NPL shares.

In contrast to objectives as such, differences in mandates thusseem to bear explanatory power for NPL shares. Below, we willexplore this possible link in more detail using multivariate analysis.

2.2.3 Governance and Independence

Commonly used measures of central bank independence take intoaccount to what extent a bank can conduct monetary policy with-out a formal possibility for parliament or the government to influ-ence decision making. In our survey and in this paper, we insteadattempt to capture to what extent central banks can be consideredfinancially independent from the political sphere. In particular, whenstudying the impact of the institutional structure of financial super-vision on financial system stability, such measures better capturethe relevant governance features of a central bank and the leeway ithas to conduct stabilizing operations in financial markets. Our sur-vey therefore focused on a bank’s freedom to allocate and distributeprofits, as well as on its freedom to set its own budget. Such rightsare likely to influence the ability of a central bank to adjust thesize and composition of its balance sheet when conducting financialstability policy. Figure 2 contains data on the freedom that centralbanks have to allocate profits. In our survey 11 percent report thatthere are no laws governing the allocation of profits, while an equalshare reports that the allocation of profits is governed by law. Thevast majority reports a more complex solution—for example, withinformal arrangements or through regulations other than by law.

With regards to their budgets, central banks have more free-dom. As displayed in figure 3, three out of four report that they donot need approval from parliament or the government, the typicalprincipals of central banks.

We tested a large number of governance indicators from the sur-vey, shown in the bottom panel of table 1, on a relation with twobanking system stability measures. With the exception of two indi-cators, none correlates with observed NPL shares. We discuss here


Figure 2. Are Any Laws Set upon Central Banks thatGovern How Profits Should Be Allocated?

Notes: Replies from questionnaire; see appendix 1 for details. (N = 47).

Figure 3. Do Central Banks Need Approval fromParliament or the Government before Setting Their

Budget?

Notes: Replies from questionnaire; see appendix 1 for details. (N = 47).

only the three measures we include in the multivariate regressionsbelow.

First, we include a dummy indicating that the government or leg-islature does not have to approve central banks’ expenditure budget.


Budget autonomy indicates higher central bank independence andshould thus indicate that it is better safeguarded from political influ-ence. Second, we specify an indicator equal to 1 if no law exists thatgoverns the allocation of central bank profits. This variable indicateshigher financial independence and correlates weakly, but positively,with mean NPL shares. Third, we find that a more recent founda-tion data of central banks is positively related to NPL shares prior tothe crisis. Younger central banks may correlate positively with NPLshares if a shorter history of the institution gauges also less experi-ence and skill. In contrast, younger central banks might have fewervested interests—for instance, resulting from entanglement of for-mer supervisors with the commercial banking sector—and thereforecould reduce NPL shares after controlling for other factors. Thus,the effect of central banks’ foundation years on NPL shares in amultivariate context remains a priori ambiguous.

Other measures of central bank governance—such as the levelof pay, the existence and composition of supervisory boards, themeeting cycle, terms lengths, or differently motivated dismissals ofgovernors—show no significant univariate correlation with NPL forour sample.

2.3 Auxiliary Data

In addition to the data from our questionnaire and the country con-trols described above, we used data from three other sources: a listof regulatory authorities and supervisory agencies maintained by theBIS (BIS 2005), central bank governance data collected by the IMF(Lybek and Morris 2004; Berger, Nitsch, and Lybek 2006), and theElectronic Compendium on Central Bank Governance maintainedby the BIS Central Bank Governance Forum (BIS 2004). In addi-tion, we collected the founding year of each central bank from itsrespective website.

We used the BIS list of regulatory authorities and supervisoryagencies to cross-check and supplement information provided by cen-tral banks on the importance and exclusivity of their supervisorytasks. For countries we did not have first-hand knowledge on, weverified the data on the BIS website to make sure the list was up todate. All three supervision and separation of power variables (“SoleSupervisor: CB,” “Sole Supervisor: FSA,” “Multiple Supervisors”)


were constructed using this information. Because we only used theother BIS and IMF data in exploratory analyses, but not in any ofthe results we present in the paper, it suffices here to refer to theabove articles for a description of these data.

3. Results

As discussed in the literature review, there are arguments in favor ofand against separating financial supervision from other tasks typicalof central banks, but most theoretical analyses appear to suggestthat the benefits from separation are likely higher than the costs.Many recent policy papers and statements from central banks andthe current UK government argue instead in the opposite direction.Empirical evidence is lacking, as the study closest to ours, Hasanand Mester (2008), finds no significant effect of this feature for thepre-crisis period. Here, we test whether the consideration of the cri-sis period paired with the specification of our novel survey responsedata reported directly by central banks themselves sheds a differentlight on the issue.

3.1 Pre-Crisis Credit Risk

Table 2 shows results from OLS regressions of non-performing loan(NPL) shares averaged over the period 2004 until 2006 for eachcountry.10 Column 1 presents results for a specification includingall covariates that are significantly different from 0 in univariatetests and a number of test variables added at our discretion.11 Thefirst panel illustrates that neither more frequent crises in the past,nor banking market structure, influence mean NPL shares as such.Higher income dispersion and high historical levels of NPL explain

10We also specified annual estimations and various different time intervals.Results are qualitatively unaffected. Our preference for the period 2004–6 resultsfrom the timing of the questionnaire (mostly around 2005), a sufficiently long timelag to historical NPL levels (2000 in our case), and the aim to have a symmetriccrisis time window (2007–9; see next sub-section). All alternative specificationsare available upon request.

11We tested a battery of different combinations, using stepwise regression toexclude or include alternative covariates for each of the four “blocks” in theregressions.


le2.

Mea

nN

on-P

erfo

rmin

gLoa

ns

and

Cen

tral

Ban

kTra

its

Pri

orto

Cri

sis

(200

4–6)

Spec

ifica

tion:

(1)

(2)

(3)

(4)

(5)

(6)

Vari

able

sFull

Model

Obje

ctiv

esO

bje

ctiv

esand

Mandate

Cou

ntr

yan

dCri

sis

His

tory

Con

trol

s

oecd

−4.

002

−1.

990

−2.

444

−2.

448

−2.

443

−2.

469

[1.0

85]∗

∗∗

[0.7

28]∗

∗∗

[0.8

49]∗

∗∗

[0.8

80]∗

∗∗

[0.8

67]∗

∗∗

[0.8

29]∗

∗∗

lego

ruk

−1.

855

−1.

262

−1.

334

−1.

338

−1.

345

−1.

380

[0.9

74]∗

[0.6

58]∗

[0.7

03]∗

[0.7

38]∗

[0.7

25]∗

[0.6

61]∗

∗

giniin

dex

0.16

50.

127

0.12

90.

129

0.13

00.

129

[0.0

54]∗

∗∗

[0.0

41]∗

∗∗

[0.0

41]∗

∗∗

[0.0

42]∗

∗∗

[0.0

41]∗

∗∗

[0.0

40]∗

∗∗

npl20

00w

b0.

319

0.26

80.

281

0.28

10.

280

0.28

1[0

.063

]∗∗

∗[0

.041

]∗∗

∗[0

.042

]∗∗

∗[0

.044

]∗∗

∗[0

.043

]∗∗

∗[0

.042

]∗∗

∗

bc

1970

lv−

0.55

7[0

.485

]cc

1970

lv−

0.15

0[0

.269

]cr

320

00−

0.30

5[1

.573

]

Cen

tral

Ban

kO

bjec

tive

s

obje

ctiv

esfx

−3.

745

−3.

675

−3.

956

−3.

961

−3.

948

−3.

979

[1.3

00]∗

∗∗

[1.0

83]∗

∗∗

[1.1

28]∗

∗∗

[1.1

70]∗

∗∗

[1.1

51]∗

∗∗

[1.1

07]∗

∗∗

obje

ctiv

esps

−2.

580

−1.

358

−1.

513

−1.

516

−1.

499

−1.

514

[1.3

94]∗

[1.0

31]

[1.0

48]

[1.0

76]

[1.0

55]

[1.0

33]

obje

ctiv

esm

p0.

620

[0.7

47]

Supe

rvis

ion

and

Sep

arat

ion

ofPow

ers

mult

iple

super

viso

r−

0.93

2−

0.01

90.

084

[1.0

71]

[0.8

92]

[0.6

68]

(con

tinu

ed)


Tab

le2.

(Con

tinued

)

Spec

ifica

tion:

(1)

(2)

(3)

(4)

(5)

(6)

Vari

able

sFull

Model

Obje

ctiv

esO

bje

ctiv

esand

Mandate

Supe

rvis

ion

and

Sep

arat

ion

ofPow

ers

cbso

lesu

per

viso

r−

1.93

7−

0.35

9−

0.38

9[1

.210

][1

.638

][2

.189

]fs

apru

dsu

p0.

367

0.35

20.

651

0.72

0[1

.725

][1

.875

][0

.824

][0

.606

]

Cen

tral

Ban

kG

over

nan

ce

cbfo

und

year

−0.

008

[0.0

05]

budap

no

0.58

0[0

.817

]pro

fitn

olaw

−0.

995

[1.4

63]

Con

stan

t15

.441

−0.

903

−0.

704

−0.

667

−1.

088

−1.

057

[10.

651]

[1.6

34]

[2.3

07]

[2.9

16]

[1.6

73]

[1.6

29]

Obse

rvat

ions

3941

4141

4141

R-S

quar

ed0.

818

0.74

20.

753

0.75

30.

753

0.75

3N

3941

4141

4141

Note

s:T

he

dep

enden

tva

riable

isth

eav

erage

ofth

enon-p

erfo

rmin

glo

an

share

for

the

yea

rs2004–6.Vari

able

sare

defi

ned

as

follow

s:bc

1970

lv:num

ber

ofbankin

gcr

ises

since

1970

acc

ord

ing

toLaev

enand

Vale

nci

a(2

008);

cc1970

lv:num

ber

ofcu

rren

cycr

ises

since

1970

acc

ord

ing

toLaev

enand

Vale

nci

a(2

010);

cbfo

und

yea

r:se

lf-r

eport

edfo

undati

on

yea

rofth

ece

ntr

albank;le

gor

uk:in

dic

ato

rva

riable

equalto

1if

the

countr

yhas

ale

galsy

stem

wit

hU

Kori

gin

.T

he

refe

rence

gro

up

isall

oth

erle

galher

itages

as

des

crib

edin

Shle

ifer

and

Vis

hny

(1998).

gin

iin

dex

:in

com

edis

per

sion

mea

sure

;cr

32000:th

em

ark

etsh

are

of

asy

stem

’sla

rges

tth

ree

banks

inte

rms

of

gro

ssto

talass

ets

in2000

(Worl

dB

ank);

npl2000

wb:non-p

erfo

rmin

glo

an

share

inth

eyea

r2000

(Worl

dB

ank);

OEC

D:in

dic

ato

rva

riable

equalto

1if

the

countr

yis

am

ember

ofth

eO

EC

D;obje

ctiv

esm

p:in

dic

ato

req

ualto

1if

the

centr

albank

answ

ered

topurs

ue

eith

erpri

ceor

money

stability

as

pri

me

obje

ctiv

e;obje

ctiv

esfx

:in

dic

ato

req

ualto

1if

the

centr

albank

answ

ered

topurs

ue

eith

ercu

rren

cyor

purc

hasi

ng

pow

erof

curr

ency

stability

as

pri

me

obje

ctiv

e;obje

ctiv

esps:

indic

ato

req

ualto

1if

the

centr

albank

answ

ered

topurs

ue

eith

eran

effici

ent

pay

men

tsy

stem

or

stable

bankin

goper

ati

ons

as

pri

me

obje

ctiv

e;cb

sole

super

vis

or:

indic

ato

req

ualto

1if

the

centr

albank

isin

volv

edin

at

least

som

esu

per

vis

ion

and

ifth

ere

isonly

one

pru

den

tialsu

per

vis

or;

mult

iple

super

vis

or:

indic

ato

req

ualto

1if

mult

iple

super

vis

ors

;fs

apru

dsu

p:eq

ual

to1

ifth

ere

exis

tsa

financi

alsu

per

vis

ion

auth

ori

tynex

tto

the

centr

albank;cb

pru

dsu

p:in

dic

ato

req

ualto

1if

the

centr

albank

isin

volv

edin

at

least

som

esu

per

vis

ion;budapno:in

dic

ato

req

ualto

1if

no

law

inpla

cere

quir

ing

appro

valofce

ntr

albank

budget

;pro

fitn

ola

w:in

dic

ato

req

ualto

1if

no

law

inpla

cere

quir

ing

dis

trib

uti

on

ofce

ntr

albank

pro

fits

.Sta

ndard

erro

rsare

inbra

cket

s.∗

∗∗p

<0.0

1,

∗∗p

<0.0

5,

∗p

<0.1

.


a large part of the cross-country variation and correlate positivelywith pre-crisis NPL shares. Membership in the OECD and a legalsystem with a UK heritage reduces NPL shares.

The second panel shows that objectives do matter for credit riskas well. However, in contrast to univariate analysis, the objectiveof monetary policy exhibits no significant relation. In fact, pursu-ing external stability as a prime objective correlates most signifi-cantly, and negatively, with pre-crisis NPL shares. The third panelfurther shows that differences in the prudential supervisory architec-ture across sampled economies have no significant influence on NPLshares. Likewise, models with interactions terms between objectivesand supervisory architecture differences (not reported) yield no sig-nificant differential effect of certain combinations between the twosets of variables.

Hence, much of the ongoing debate on whether to central-ize prudential supervision, the potential role of central banks ver-sus alternative authorities, and the potential interplay betweenprice and financial stability objectives appears less relevant thanexpected.

Regarding central bank governance proxies shown in the bottompanel of table 1, we also fail to detect significant relations with NPLshares. Overall, macroeconomic differences across countries are mostrelevant to explain credit risk.

The lack of significant relations before and during the crisis isconsistent with the negative findings of Hasan and Mester (2008) forthe non-crisis period in the 1990s. However, the absence of resultsmight merely reflect the fairly low degrees of freedom in our sam-ple. Therefore, we test in columns 2–6 whether our findings arecorroborated when excluding insignificant covariates. Consider firstspecification (2), which scrutinizes the results for central bank objec-tives and drops any prudential supervision, governance, or insignifi-cant macro covariate. The negative relation between foreign stabilityobjectives and NPL shares is corroborated, but the previously weaksignificance regarding payment system stability vanishes. We inves-tigate below whether this NPL-reducing effect of foreign stabilityobjectives potentially reflects past experiences with IMF missions.IMF involvement might have caused generally higher standards ofpolicy procedures at central banks that eventually also affected NPLlevels positively.


Let us first consider a number of permutations in columns 3–6 that augment the reduced specification with different combina-tions of prudential mandate indicators. We focus on these vari-ables because of the intense policy debate (and actions) reflectedby, for instance, the enactment of the Dodd-Frank bill in theUnited States and the inception of the European Systemic RiskBoard under the auspices of the ECB that intend to establishautonomous and centralized prudential supervisory institutions. Inshort, none of the specifications indicate any significant correlationbetween differences in prudential mandates across sample coun-tries and NPL shares. Hence, this finding confirms that the cur-rent policy debate might be misguided to the extent that alterna-tive arrangements did not affect pre-crisis NPL shares in the firstplace.

3.2 Crisis Credit Risk

A possible explanation for the absence of any such effect mightbe that NPL shares in “normal” times just prior to the crisiswere monitored equally (in)effectively by either prudential architec-ture. Therefore, table 3 shows identical specifications as table 2—using, however, the mean share of NPL during the crisis, specificallybetween 2007 and 2009.

The ability of the previously chosen covariates to explain NPLshares vanishes almost entirely. Only the number of currency crisessince 1970, higher banking market concentration, and more recentfoundation of the central bank exhibit statistically weak evidence infavor of lower NPL ratios. This result vividly illustrates that duringthe crisis the institutional arrangements are of even lesser impor-tance. Presumably, the statement of objectives of central banksmight have meant little in times of crisis when authorities around theglobe took unusual measures to support crippled financial systems(see, e.g., Stolz and Wedow 2010 for a survey of “unconventional”measures during the crisis). Likewise, the existence of single or multi-ple supervisors is likely to mean little in times of international ad hoccollaboration among representatives from all authorities, includingtop-level government representatives, to stabilize financial systemsin concerted efforts.


le3.

Mea

nN

on-P

erfo

rmin

gLoa

ns

and

Cen

tral

Ban

kTra

its

duri

ng

the

Cri

sis

(200

7–9)

Spec

ifica

tion:

(1)

(2)

(3)

(4)

(5)

(6)

Var

iable

sFull

Model

Obje

ctiv

esO

bje

ctiv

esan

dM

andat

e

Cou

ntry

and

Cri

sis

His

tory

Con

trol

s

oecd

−0.

119

−0.

893

−1.

980

−1.

628

−1.

634

−2.

144

[6.6

86]

[4.8

48]

[5.8

16]

[6.0

27]

[5.9

26]

[5.6

77]

lego

ruk

2.81

2−

4.00

1−

3.95

2−

3.58

8−

3.57

9−

4.25

0[6

.133

][4

.400

][4

.776

][5

.007

][4

.917

][4

.491

]gi

niin

dex

−0.

117

0.15

30.

158

0.16

80.

167

0.16

1[0

.333

][0

.271

][0

.279

][0

.286

][0

.279

][0

.275

]np

l20

00w

b−

0.19

1−

0.07

2−

0.03

6−

0.04

9−

0.04

8−

0.03

3[0

.391

][0

.272

][0

.294

][0

.301

][0

.296

][0

.289

]bc

1970

lv−

1.40

0[2

.985

]cc

1970

lv4.

586

[1.6

72]∗

∗

cr3

2000

16.9

31[9

.695

]∗

Cen

tral

Ban

kO

bjec

tive

s

obje

ctiv

esfx

−6.

680

−4.

886

−5.

521

−5.

052

−5.

067

−5.

675

[8.0

04]

[7.2

11]

[7.6

69]

[7.9

50]

[7.8

07]

[7.5

19]

obje

ctiv

esps

1.07

6−

1.72

9−

2.07

9−

1.77

8−

1.79

9−

2.08

5[8

.647

][6

.873

][7

.114

][7

.295

][7

.148

][7

.006

]ob

ject

ives

mp

2.87

6[4

.601

]

Supe

rvis

ion

and

Sepa

ration

ofPow

ers

cbso

lesu

per

viso

r1.

603

−2.

328

0.47

2[7

.535

][1

1.11

6][1

4.83

7]

(con

tinu

ed)


Tab

le3.

(Con

tinued

)

Spec

ifica

tion:

(1)

(2)

(3)

(4)

(5)

(6)

Var

iable

sFull

Model

Obje

ctiv

esO

bje

ctiv

esan

dM

andat

e

Supe

rvis

ion

and

Sepa

ration

ofPow

ers

mul

tipl

esu

per

viso

r3.

580

1.75

71.

632

[6.5

93]

[6.0

45]

[4.5

23]

fsa

prud

sup

−0.

399

0.91

10.

549

1.89

3[1

1.78

7][1

2.78

7][5

.744

][4

.315

]

Cen

tral

Ban

kG

over

nanc

e

cbfo

und

year

0.05

4[0

.030

]∗

buda

pno

−1.

095

[5.1

57]

profi

tnol

aw7.

080

[9.0

09]

Con

stan

t−

112.

002

2.09

93.

862

0.44

50.

956

1.57

5[6

6.19

0][1

0.91

0][1

5.70

2][1

9.80

6][1

1.39

4][1

1.11

0]

Obs

erva

tion

s38

4040

4040

40R

-Squ

ared

0.45

10.

036

0.04

40.

046

0.04

60.

042

N38

4040

4040

40

Note

s:T

he

dep

enden

tva

riable

isth

eav

erage

ofth

enon-p

erfo

rmin

glo

an

share

for

the

yea

rs2007–9.Vari

able

sare

defi

ned

as

follow

s:bc

1970

lv:num

ber

ofbankin

gcr

ises

since

1970

acc

ord

ing

toLaev

enand

Vale

nci

a(2

010);

cc1970

lv:num

ber

ofcu

rren

cycr

ises

since

1970

acc

ord

ing

toLaev

enand

Vale

nci

a(2

008);

cbfo

und

yea

r:se

lf-r

eport

edfo

undati

on

yea

rofth

ece

ntr

albank;le

gor

uk:in

dic

ato

rva

riable

equalto

1if

the

countr

yhas

ale

galsy

stem

wit

hU

Kori

gin

.T

he

refe

rence

gro

up

isall

oth

erle

galher

itages

as

des

crib

edin

Shle

ifer

and

Vis

hny

(1998).

gin

iin

dex

:in

com

edis

per

sion

mea

sure

;cr

32000:th

em

ark

etsh

are

of

asy

stem

’sla

rges

tth

ree

banks

inte

rms

of

gro

ssto

talass

ets

in2000

(Worl

dB

ank);

npl2000

wb:non-p

erfo

rmin

glo

an

share

inth

eyea

r2000

(Worl

dB

ank);

oec

d:in

dic

ato

rva

riable

equalto

1if

the

countr

yis

am

ember

ofth

eO

EC

D;obje

ctiv

esm

p:in

dic

ato

req

ualto

1if

the

centr

albank

answ

ered

topurs

ue

eith

erpri

ceor

money

stability

as

pri

me

obje

ctiv

e;obje

ctiv

esfx

:in

dic

ato

req

ualto

1if

the

centr

albank

answ

ered

topurs

ue

eith

ercu

rren

cyor

purc

hasi

ng

pow

erof

curr

ency

stability

as

pri

me

obje

ctiv

e;obje

ctiv

esps:

indic

ato

req

ualto

1if

the

centr

albank

answ

ered

topurs

ue

eith

eran

effici

ent

pay

men

tsy

stem

or

stable

bankin

goper

ati

ons

as

pri

me

obje

ctiv

e;cb

sole

super

vis

or:

indic

ato

req

ualto

1if

the

centr

albank

isin

volv

edin

at

least

som

esu

per

vis

ion

and

ifth

ere

isonly

one

pru

den

tialsu

per

vis

or;

mult

iple

super

vis

or:

indic

ato

req

ualto

1if

mult

iple

super

vis

ors

;fs

apru

dsu

p:eq

ual

to1

ifth

ere

exis

tsa

financi

alsu

per

vis

ion

auth

ori

tynex

tto

the

centr

albank;cb

pru

dsu

p:in

dic

ato

req

ualto

1if

the

centr

albank

isin

volv

edin

at

least

som

esu

per

vis

ion;budapno:in

dic

ato

req

ualto

1if

no

law

inpla

cere

quir

ing

appro

valofce

ntr

albank

budget

;pro

fitn

ola

w:in

dic

ato

req

ualto

1if

no

law

inpla

cere

quir

ing

dis

trib

uti

on

ofce

ntr

albank

pro

fits

.Sta

ndard

erro

rsare

inbra

cket

s.∗

∗∗p

<0.0

1,

∗∗p

<0.0

5,

∗p

<0.1

.


3.3 Why Currency Stability Objectives?

The result that pre-crisis NPL shares are basically only correlatedwith the existence of explicit objectives of central banks to maintaina stable currency warrants some further discussion.

A first potential reason could be that governments in countriesthat pursue fixed exchange rate regimes are more prone, and pow-erful, to prick (perceived) bubbles as they emerge. Such differencesin intrusive policy propensity and ability might partly be capturedby indices of economic freedom. Table 4 shows according regres-sions where we specify each of the five individual freedom indexcomponents as well as a summary index collected since 1970 by theFraser Institute for a comprehensive sample of countries (Gwart-ney, Lawson, and Gartzke 2005).12 Each pair of columns refers toNPL prior to the financial crisis (2004–6) and during the financialcrisis (2007–9) as dependent variable, respectively. The result thatpre-crisis credit risk is negatively associated with a foreign exchangeobjective of the central bank remains intact. Likewise, NPL levelsduring the crisis continue to remain not significantly affected bystated objectives of central bankers. Hence, foreign exchange objec-tives are unlikely to merely gauge potentially omitted differences ingovernment’s hesitance to regulate banks.

Another potential reason why the objective to maintain a stablecurrency coincides with lower levels of NPL shares could result froma larger exposure to external disciplining devices. One such devicecould, in our view, be the involvement of the IMF once it extendsa loan to a country. IMF lending is usually conditioned on com-plying with procedures to select and monitor investment projectsfunded with IMF support. We also assume that IMF missions atcentral banks of countries having received support occur more oftenin order to ensure a proper management of IMF funds indepen-dent of political interests, e.g., by the ministry of finance. IMF staffpresence might either directly or indirectly improve governance andprocesses at central banks—for instance, through training and skilltransfers or simply by setting an example on acting prudently and

12The sub-components of the index are listed in table 6. Data pertain to thesurvey year 2005 to match the survey year of central banks. Data have beenobtained from http://www.freetheworld.com/.

58 International Journal of Central Banking December 2014Tab

le4.

NP

LP

rior

toan

dD

uri

ng

the

Cri

sis

and

Eco

nom

icFre

edom

Indic

ator

s

Specifi

cati

on:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

Baseline

Siz

eofG

overnm

ent

LegalSyste

mSound

Money

Freedom

toTrade

Regula

tion

Sum

mary

Index

Varia

ble

spre

durin

gpre

durin

gpre

durin

gpre

durin

gpre

durin

gpre

durin

gpre

durin

g

Cou

ntr

yan

dCri

sis

His

tory

Con

trol

s

oecd

−4

.121

1.5

84

−3

.936

−1

.344

−4

.160

1.2

09

−3

.972

13

.530

−4

.199

9.8

91

−4

.376

−5

.758

−4

.053

4.7

96

[1.1

45]∗

∗∗

[7.8

88]

[1.2

28]∗

∗∗

[8.2

92]

[1.1

98]∗

∗∗

[8.2

71]

[1.3

93]∗

∗∗

[8.3

92]

[1.2

44]∗

∗∗

[6.8

72]

[1.3

23]∗

∗∗

[8.6

09]

[1.5

20]∗

∗[1

0.4

59]

legor

uk

−1

.448

0.9

65

−1

.290

−1

.328

−1

.506

0.3

58

−1

.368

7.4

68

−1

.570

14

.289

−1

.725

−7

.366

−1

.381

4.2

01

[0.9

31]

[6.5

08]

[1.0

04]

[6.8

06]

[1.0

28]

[7.3

01]

[1.0

33]

[6.3

19]

[1.1

61]

[6.5

35]∗

∗[1

.166]

[7.8

02]

[1.3

52]

[9.4

62]

gin

iin

dex

0.1

34

0.3

67

0.1

38

0.3

12

0.1

36

0.3

82

0.1

31

0.1

05

0.1

36

0.1

92

0.1

41

0.5

41

0.1

33

0.3

19

[0.0

46]∗

∗∗

[0.3

14]

[0.0

47]∗

∗∗

[0.3

17]

[0.0

48]∗

∗∗

[0.3

29]

[0.0

49]∗

∗[0

.298]

[0.0

47]∗

∗∗

[0.2

62]

[0.0

49]∗

∗∗

[0.3

16]

[0.0

49]∗

∗[0

.335]

cr3

2000

0.0

28

16

.348

0.1

24

14

.883

−0

.084

15

.266

0.1

89

29

.266

−0

.102

30

.209

−0

.237

8.6

23

0.1

12

20

.289

[1.5

69]

[10

.805]

[1.6

07]

[10

.838]

[1.7

68]

[12

.277]

[1.7

97]

[10

.836]∗

∗[1

.750]

[9.6

78]∗

∗∗

[1.7

24]

[11

.218]

[1.9

94]

[13

.733]

npl2000

wb

0.2

79

−0

.209

0.2

75

−0

.126

0.2

84

−0

.165

0.2

76

−0

.424

0.2

83

−0

.607

0.2

81

−0

.175

0.2

77

−0

.315

[0.0

55]∗

∗∗

[0.3

81]

[0.0

56]∗

∗∗

[0.3

87]

[0.0

65]∗

∗∗

[0.4

48]

[0.0

57]∗

∗∗

[0.3

50]

[0.0

60]∗

∗∗

[0.3

32]∗

[0.0

56]∗

∗∗

[0.3

65]

[0.0

65]∗

∗∗

[0.4

47]

Cen

tral

Ban

kO

bjec

tive

s

obje

cti

ves

fx−

3.6

98

−2

.448

−3

.583

−4

.290

−3

.684

−2

.313

−3

.634

2.6

17

−3

.802

8.5

91

−3

.740

−3

.631

−3

.667

−1

.019

[1.2

79]∗

∗∗

[8.8

06]

[1.3

23]∗

∗[8

.925]

[1.3

09]∗

∗∗

[9.0

21]

[1.3

44]∗

∗[8

.097]

[1.4

23]∗

∗[7

.858]

[1.3

06]∗

∗∗

[8.4

50]

[1.3

76]∗

∗[9

.442]

obje

cti

ves

ps

−2

.835

5.4

19

−2

.776

4.7

52

−2

.835

5.3

49

−2

.749

12

.453

−2

.878

10

.396

−2

.990

0.4

95

−2

.804

6.9

61

[1.3

92]∗

[9.6

84]

[1.4

21]∗

[9.6

59]

[1.4

22]∗

[9.8

99]

[1.4

87]∗

[9.0

54]

[1.4

41]∗

[8.0

48]

[1.4

67]∗

[9.6

71]

[1.4

90]∗

[10

.364]

obje

cti

ves

mp

0.9

12

4.4

31

0.7

76

6.5

89

0.9

24

4.5

41

0.9

38

6.4

78

0.8

30

13

.196

0.9

59

5.7

63

0.9

09

4.2

79

[0.7

58]

[5.2

17]

[0.8

22]

[5.5

52]

[0.7

78]

[5.3

58]

[0.7

84]

[4.7

23]

[0.8

94]

[4.9

43]∗

∗[0

.780]

[5.0

47]

[0.7

75]

[5.3

16]

Supe

rvis

ion

and

Sep

arat

ion

ofPow

ers

cb

sole

superv

isor

−0

.202

2.6

67

−0

.301

4.2

19

−0

.285

1.8

85

−0

.225

0.8

68

−0

.127

−5

.334

−0

.133

4.6

91

−0

.211

2.2

76

[2.2

65]

[15

.599]

[2.3

12]

[15

.593]

[2.3

78]

[16

.403]

[2.3

15]

[13

.944]

[2.3

48]

[12

.963]

[2.3

12]

[14

.965]

[2.3

17]

[15

.888]

mult

iple

superv

isor

−0

.499

7.1

42

−0

.441

6.3

06

−0

.568

6.4

63

−0

.477

8.8

64

−0

.514

8.9

26

−0

.517

6.4

79

−0

.474

8.3

09

[1.0

36]

[7.1

45]

[1.0

60]

[7.1

53]

[1.1

55]

[8.0

42]

[1.0

63]

[6.4

13]

[1.0

61]

[5.8

67]

[1.0

55]

[6.8

45]

[1.1

13]

[7.6

66]

fsa

pru

dsu

p1

.363

−1

.113

1.1

56

2.4

23

1.3

14

−1

.641

1.3

25

−4

.034

1.4

83

−13

.469

1.4

46

0.8

46

1.3

30

−2

.570

[1.9

79]

[13

.683]

[2.0

58]

[13

.996]

[2.0

47]

[14

.221]

[2.0

29]

[12

.267]

[2.1

23]

[11

.732]

[2.0

24]

[13

.135]

[2.0

73]

[14

.247]

Cen

tral

Ban

kG

over

nan

ce

cb

found

year

−0

.007

0.0

58

−0

.006

0.0

46

−0

.007

0.0

57

−0

.007

0.0

77

−0

.007

0.0

79

−0

.007

0.0

45

−0

.007

0.0

66

[0.0

05]

[0.0

33]∗

[0.0

05]

[0.0

34]

[0.0

05]

[0.0

34]

[0.0

05]

[0.0

30]∗

∗[0

.005]

[0.0

28]∗

∗∗

[0.0

05]

[0.0

32]

[0.0

05]

[0.0

37]∗

Budapno

0.7

80

−0

.552

0.6

81

0.7

63

0.7

71

−0

.581

0.7

29

−4

.795

0.8

11

−4

.224

0.8

86

2.9

51

0.7

57

−1

.724

[0.8

16]

[5.7

73]

[0.8

55]

[5.8

70]

[0.8

36]

[5.8

99]

[0.8

72]

[5.4

03]

[0.8

49]

[4.8

37]

[0.8

70]

[5.8

66]

[0.8

97]

[6.3

62]

Pro

fitn

ola

w−

0.6

64

−3

.490

−0

.575

−4

.959

−0

.716

−3

.970

−0

.544

6.1

28

−0

.762

6.8

43

−0

.887

−9

.808

−0

.605

−0

.714

[1.3

42]

[9.2

41]

[1.3

77]

[9.2

96]

[1.4

13]

[9.7

38]

[1.4

99]

[9.0

32]

[1.4

70]

[8.1

11]

[1.4

71]

[9.5

32]

[1.6

09]

[11

.044]

(con

tinu

ed)


le4.

(Con

tinued

)

Specifi

cati

on:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

Baseline

Siz

eofG

overnm

ent

LegalSyste

mSound

Money

Freedom

toTrade

Regula

tion

Sum

mary

Index

Varia

ble

spre

durin

gpre

durin

gpre

durin

gpre

durin

gpre

durin

gpre

durin

gpre

durin

g

Eco

nom

icFre

edom

Indi

cato

rs

Siz

eofG

overn

ment

−0

.140

2.2

29

[0.2

96]

[2.0

28]

LegalSyst

em

0.0

59

0.5

65

[0.3

97]

[2.8

10]

Sound

Money

−0

.090

−7

.235

[0.4

58]

[2.7

64]∗

∗

Fre

edom

toTra

de

0.0

97

−10

.341

[0.5

30]

[2.9

41]∗

∗∗

Regula

tion

0.1

93

5.5

46

[0.4

74]

[3.1

31]∗

Sum

mary

Index

−0

.059

−2

.789

[0.8

44]

[5.8

27]

Const

ant

11

.854

−133.9

88

11

.098

−124.1

10

11

.698

−134.9

54

12

.144

−111.8

03

11

.487

−98.2

11

11

.229

−148.1

38

11

.942

−130.4

15

[10

.185]

[70

.949]∗

[10

.475]

[71

.198]∗

[10

.451]

[72

.635]∗

[10

.499]

[63

.910]∗

[10

.586]

[58

.928]

[10

.479]

[68

.334]∗

∗[1

0.4

79]

[72

.553]∗

Obse

rvati

ons

39

38

39

38

39

38

39

38

39

38

39

38

39

38

R-S

quare

d0

.804

0.2

62

0.8

06

0.3

00

0.8

04

0.2

63

0.8

05

0.4

37

0.8

05

0.5

27

0.8

06

0.3

54

0.8

04

0.2

69

Note

s:

The

dependent

vari

able

isth

eavera

ge

of

the

non-p

erf

orm

ing

loan

share

for

the

years

2004–7

(pre

)and

2007–9

(duri

ng).

Vari

able

sare

defined

as

follows:

bc

1970

lv:num

ber

ofbankin

gcri

ses

since

1970

accord

ing

toLaeven

and

Vale

ncia

(2010);

cc

1970

lv:num

ber

ofcurr

ency

cri

ses

since

1970

accord

ing

toLaeven

and

Vale

ncia

(2008);

cb

found

year:

self-r

eport

ed

foundati

on

year

ofth

ecentr

albank;le

gor

uk:in

dic

ato

rvari

able

equalto

1if

the

countr

yhas

ale

galsy

stem

wit

hU

Kori

gin

.T

he

refe

rence

gro

up

isall

oth

er

legalheri

tages

as

desc

ribed

inShle

ifer

and

Vis

hny

(1998).

gin

iin

dex:in

com

edis

pers

ion

measu

re;cr3

2000:th

em

ark

et

share

ofa

syst

em

’sla

rgest

thre

ebanks

inte

rms

ofgro

ssto

talass

ets

in2000

(Worl

dBank);

npl2000

wb:non-p

erf

orm

ing

loan

share

inth

eyear

2000

(Worl

dBank);

oecd:

indic

ato

rvari

able

equal

to1

ifth

ecountr

yis

am

em

ber

of

the

OEC

D;

obje

cti

ves

mp:

indic

ato

requal

to1

ifth

ecentr

al

bank

answ

ere

dto

purs

ue

eit

her

pri

ce

or

money

stability

as

pri

me

obje

cti

ve;obje

cti

ves

fx:in

dic

ato

requalto

1if

the

centr

albank

answ

ere

dto

purs

ue

eit

her

curr

ency

or

purc

hasi

ng

power

ofcurr

ency

stability

as

pri

me

obje

cti

ve;

obje

cti

ves

ps:

indic

ato

requal

to1

ifth

ecentr

al

bank

answ

ere

dto

purs

ue

eit

her

an

effic

ient

paym

ent

syst

em

or

stable

bankin

gopera

tions

as

pri

me

obje

cti

ve;cb

sole

superv

isor:

indic

ato

requalto

1if

the

centr

albank

isin

volv

ed

inat

least

som

esu

perv

isio

nand

ifth

ere

isonly

one

pru

denti

alsu

perv

isor;

mult

iple

superv

isor:

indic

ato

requal

to1

ifm

ult

iple

superv

isors

;fs

apru

dsu

p:

equal

to1

ifth

ere

exis

tsa

financia

lsu

perv

isio

nauth

ori

tynext

toth

ecentr

al

bank;

cb

pru

dsu

p:in

dic

ato

requalto

1if

the

centr

albank

isin

volv

ed

inat

least

som

esu

perv

isio

n;budapno:in

dic

ato

requalto

1if

no

law

inpla

ce

requir

ing

appro

valof

centr

albank

budget;

pro

fitn

ola

w:in

dic

ato

requalto

1if

no

law

inpla

ce

requir

ing

dis

trib

uti

on

ofcentr

albank

pro

fits

.Econom

icfr

eedom

data

isfu

rther

desc

ribed

inta

ble

6.Sta

ndard

err

ors

are

inbra

ckets

.∗

∗∗p

<0.0

1,

∗∗p

<0.0

5,

∗p

<0.1

.


Table 5. Correspondence of Currency Objectives andPast IMF Exposure 1963–2005

FX Objective

IMF NO YES Total

NO 28 1 29YES 6 5 11Total 34 6 40

Notes: The table shows tabulation of IMF and FX-objective indicators. IMF is equalto 1 if the share of years with IMF support measures for a country is larger thanthe 75th percentile of the distribution of the share of years with support of all yearssince 1964. This share percentile is equal to 28 percent. FX objective is equal to 1 ifthe central bank answered to pursue stability of the currency or purchasing power ofthe currency. The area under the receiver operating characteristics curve is 0.71. Thestandard error is 0.081. Thus, this area is different from either 0.5 (non-determinacy)or 1 (perfectly deterministic).

independently. The result may be positive spillovers to local author-ities, which eventually helps them in supervising their own financialsystems according to mimicked standards and procedures.

Obviously, a full-fledged statistical test of this narrative is beyondthe scope of this paper. It would require detailed information onactual IMF missions as well as more direct data proxying for poten-tial skill transfers and, ultimately, adoption of better procedures,governance, and policymaking with domestic authorities.

But as an exploration of this potential reason why foreign sta-bility objectives correlate significantly with lower shares of NPL, wemanually collected from IMF annual reports the outstanding vol-umes of support funding to any of our sampled countries between1964 and 2005, i.e., just up to the first period included in the pre-crisis mean NPL share. On average, countries in our sample had in19 percent of the years since 1964 an exposure to the IMF. Inter-estingly, despite the large share of OECD countries in our sample,only seven out of forty-four countries never had an exposure to theIMF. We define intensive IMF exposure countries as those that havean exposure above the 75th percentile of the frequency distribution,i.e., in more than 28 percent of the years between 1964 and 2005.

Table 5 shows the tabulation between high (low) IMF exposurecountries and the existence (absence) of foreign stability objectives


le6.

Eco

nom

icFre

edom

Index

Cat

egor

ies

1.Siz

eof

2.Leg

alSyst

eman

d4.Fre

edom

toTra

de

Gov

ernm

ent

Pro

per

tyR

ights

3.Sound

Money

Inte

rnat

ional

ly5.R

egula

tion

1AG

over

nmen

tco

nsum

ptio

n2A

Judi

cial

inde

pen

denc

e3A

Mon

eygr

owth

4ATar

iffs

5AC

redi

tm

arke

tre

gula

tion

s1B

Tra

nsfe

rsan

dsu

bsid

ies

2BIm

part

ialco

urts

3BSt

anda

rdde

viat

ion

ofin

flation

4AiR

even

uefr

omtr

ade

taxe

s(%

oftr

ade

sect

or)

5AiO

wne

rshi

pof

bank

s

1CG

over

nmen

ten

terp

rise

san

din

vest

men

t

2CP

rote

ctio

nof

prop

erty

righ

ts3C

Infla

tion

:M

ost

rece

ntye

ar4A

iiM

ean

tari

ffra

te5A

iiP

riva

tese

ctor

cred

it

1DTop

mar

gina

lta

xra

te2D

Mili

tary

inte

rfer

ence

inru

leof

law

and

pol

itic

s

3DFr

eedo

mto

own

fore

ign-

curr

ency

acco

unts

4Aiii

Stan

dard

devi

atio

nof

tari

ffra

tes

5Aiii

Inte

rest

rate

cont

rols

1DiTop

mar

gina

lin

com

eta

xra

te

2EIn

tegr

ity

ofth

ele

galsy

stem

4BR

egul

ator

ytr

ade

barr

iers

5BLab

orm

arke

tre

gula

tion

s

1Dii

Top

mar

gina

lin

com

ean

dpa

yrol

lta

xra

te

2FLeg

alen

forc

emen

tof

cont

ract

s4B

iN

on-t

ariff

trad

eba

rrie

rs5B

iH

irin

gre

gu-

lation

san

dm

inim

umw

age

2GR

egul

ator

yre

stri

ctio

nson

the

sale

ofre

alpr

oper

ty

4Bii

Com

plia

nce

cost

sof

impor

ting

and

expor

ting

5Bii

Hir

ing

and

firin

gre

gula

tion

s

2HR

elia

bilit

yof

pol

ice

4CB

lack

mar

ket

exch

ange

rate

s5B

iiiC

entr

aliz

edco

llect

ive

barg

aini

ng2I

Bus

ines

sco

sts

ofcr

ime

4DC

ontr

ols

ofth

em

ovem

ent

ofca

pita

lan

dpeo

ple

5Biv

Hou

rsre

gula

tion

s

4DiFo

reig

now

ners

hip/

inve

stm

ent

rest

rict

ions

5Bv

Man

date

dco

stof

wor

ker

dism

issa

l

(con

tinu

ed)


Tab

le6.

(Con

tinued

)

1.Siz

eof

2.Leg

alSyst

eman

d4.Fre

edom

toTra

de

Gov

ernm

ent

Pro

per

tyR

ights

3.Sound

Money

Inte

rnat

ional

ly5.R

egula

tion

4Dii

Cap

ital

cont

rols

5Bvi

Con

scri

ptio

n4D

iiiFr

eedo

mof

fore

igne

rsto

visi

t

5CB

usin

ess

regu

la-

tion

s

5CiA

dmin

istr

ativ

ere

quir

emen

ts5C

iiB

urea

ucra

cyco

sts

5Ciii

Star

ting

abu

sine

ss5C

ivE

xtra

pay-

men

ts/

brib

es/

favo

ritism

5Cv

Lic

ensi

ngre

stri

ctio

ns5C

viTax

com

plia

nce

Note

s:T

he

econ

omic

free

dom

index

isob

tain

edfr

omG

war

tney

,Law

son,an

dG

artz

ke(2

005)

.D

ownlo

aded

dat

aco

rres

pon

dto

the

2005

surv

eyre

sult

san

dar

ebas

edon

the

late

stve

rsio

nof

the

Eco

nom

icFre

edom

ofth

eW

orld

Annu

alR

epor

t,av

aila

ble

atht

tp:/

/ww

w.fre

ethew

orld

.com

/.H

igher

scor

esin

dic

ate

mor

efr

eedom

.


for the countries included in specifications (2) through (6) in pre-ceding multivariate analyses. Around half of the countries with themost substantial IMF exposure also indicated the pursuit of for-eign exchange stability as an objective. In turn, 83 percent of allcountries indicated foreign exchange stability objectives were alsoexposed most frequently to the IMF. The area under the ROC curveis 0.71—according to Hosmer and Lemeshow (2005), an indication ofa good discriminatory power of these two measures. The five coun-tries in the lower right cell of the tabulation are Argentina, Brazil,Bulgaria, Chile, and Mexico. Many of these countries actually expe-rienced currency crises in the past, on average two since 1990 andaround four since 1970 according to the crisis definitions of Laevenand Levine (2010).

Consequently, past experiences of crisis that brought in someexternal disciplining device in the form of IMF missions might have amore relevant long-run effect on improving prudential attitudes com-pared with the currently debated issues on whether or not to central-ize supervision and, if so, with whom the prudential mandate shouldrest, central banks or financial stability authorities. Note, however,that we are not able to detect a significant relationship of an IMFexposure indicator variable on the NPL share prior to, or during thecrisis for this fairly confined, cross-sectional sample. Therefore, fur-ther research into the effects of skill transfer and externally imposeddiscipline might constitute an important, but so far neglected, routefor further research on the relation between prudential supervisionand financial stability.

4. Conclusion

The crisis in the sub-prime mortgage market and the subsequentfinancial crisis ignited a debate about the past and future role of cen-tral banks and financial supervisory agencies. In several countries,new laws have been or are being adopted that alter the objectivesof central banks, change the allocation of prudential supervisoryauthority, or change the governance structure of central banks.

In this paper, we use newly collected survey data on the gover-nance structure of central banks in forty-seven developed and emerg-ing jurisdictions, to shed more light on whether the formal objective,


the mandate, and a large number of governance features of a cen-tral bank can help explain a country’s banking system fragility, bothprior to and during the 2007–9 financial crisis. We focus in partic-ular on empirical evidence of whether micro-prudential supervisionshould be with the central banks or with a separate independentagency, as the issue is at the core of the current policy debate.

We find that neither having formal responsibility for bankingsupervision or banking system stability nor other central bank gov-ernance features explain cross-country variation in financial fragility,both prior to and during the crisis. Instead we find a country’slegal origin, income dispersion, historical levels of credit risk, andOECD membership explain most of the variation in pre-crisis finan-cial fragility. During the crisis period, however, these links breakdown.

The only governance characteristic that exhibits a significantrelation with non-performing loan levels is whether central bankspursue a currency stability objective. This effect is amplified forcountries with more frequent exposures to IMF missions in the past.

These results suggest that the current focus of policy discussionson whether prudential supervision should be shared or concentratedat the central bank may be somewhat misguided. Our results do notlend support in favor of institutional reforms in this direction.

Instead, one way to interpret our results is that countries thatare experiencing a big crisis and are exposed to IMF crisis packagestighten up their financial supervision and reduce financial fragility,irrespective of their central bank’s governance structure. Unfortu-nately, our data lack time-series variation and do not allow us to fur-ther investigate this matter. Additional work on the subject, startingfrom the data-collection stage, is needed to support policymakersand guide the design of any institutional reforms.

References

Bank for International Settlements. 2004. Electronic Compendiumon Central Bank Governance. BIS Central Bank GovernanceForum. Available at www.bis.org/hub/cbgf.htm (downloadedNovember 25, 2004).

———. 2005. “Regulatory Authorities and Supervisory Agencies.”BIS central bank hub, www.bis.org/regauth.htm.


Barro, R. J., and D. B. Gordon. 1983. “Rules, Discretion and Rep-utation in a Model of Monetary Policy.” Journal of MonetaryEconomics 12 (1): 101–21.

Barth, J. R., G. Caprio, Jr., and R. Levine. 2004. “Bank Regula-tion and Supervision: What Works Best?” Journal of FinancialIntermediation 13 (2): 205–48.

Beck, T., A. Demirguc-Kunt, and R. Levine. 2006. “Bank Supervi-sion and Corruption in Lending.” Journal of Monetary Econom-ics 53 (8): 2131–63.

Berger, H., V. Nitsch, and T. Lybek. 2006. “Central Bank Boardsaround the World: Why Does Membership Size Differ?” IMFWorking Paper No. 06/281.

Boot, A. W. A., and A. V. Thakor. 1993. “Self-Interested BankRegulation.” American Economic Review 83 (2): 206–12.

Boyer, P., and J. Ponce. 2012. “Regulatory Capture and Bank-ing Supervision Reform.” Journal of Financial Stability 8 (3):206–17.

Buiter, W. H. 2011. “Accountability of the Bank of England.” Sup-plementary written material submitted to the UK Parliament.

Davies, H., and D. Green. 2010. Banking on the Future: The Fall andRise of Central Banking. Princeton, NJ: Princeton UniversityPress.

Demirguc-Kunt, A., L. Laeven, and R. Levine. 2004. “Regulations,Market Structure, Institutions, and the Cost of Financial Inter-mediation.” Journal of Money, Credit and Banking 36 (3): 593–622.

Di Noia, C., and G. Di Giorgio. 1999. “Should Banking Supervi-sion and Monetary Policy Tasks Be Given to Different Agen-cies?” International Finance 2 (3): 361–78.

Eijffinger, S. C. W., and M. Hoeberichts. 2002. “Central BankAccountability and Transparency: Theory and Some Evi-dence.” International Finance 5 (1): 73–96.

Frisell, L., K. Roszbach, and G. Spagnolo. 2008. “Governing the Gov-ernors: A Clinical Study of Central Banks.” CEPR DiscussionPaper No. 6888.

Giannone, D., M. Lenza, and L. Reichlin. 2011. “Market Freedomand the Global Recession.” IMF Economic Review 59 (1): 111–35.


Goodhart, C. 2010. “The Changing Role of Central Banks.” BISWorking Paper No. 326 (November).

Guiso, L., P. Sapienza, and L. Zingales. 2007. “The Cost of BankingRegulation.” EUI Working Paper No. ECO 2007/43.

———. 2008. “Trusting the Stock Market.” Journal of Finance 63(6): 2557–2600.

Gwartney, J., R. Lawson, and E. Gartzke. 2005. Economic Freedomof the World: 2005 Annual Report. Fraser Institute.

Hasan, I., and L. J. Mester. 2008. “Central Bank Institutional Struc-ture and Effective Central Banking: Cross-Country EmpiricalEvidence.” Comparative Economic Studies 50 (4): 620–45.

Haubrich, J. G. 1996. “Combining Bank Supervision and Mone-tary Policy.” Economic Commentary (Federal Reserve Bank ofCleveland) (November).

Herrero, A. G., and P. del Rio. 2005. “Central Banks as Mone-tary Authorities and Financial Stability.” In Handbook of CentralBanking and Financial Authorities in Europe: New Architecturesin the Supervision of Financial Markets, ed. D. Masciandaro,chapter 1. Edward Elgar Publishing.

Hosmer, D. W., and S. Lemeshow. 2005. Applied Logistic Regression,Second ed. John Wiley & Sons, Inc.

Kahn, C. M., and J. A. C. Santos. 2005. “Allocating Bank Reg-ulatory Powers: Lender of Last Resort, Deposit Insurance andSupervision.” European Economic Review 49 (8): 2107–36.

Keefer, P., and D. Stasavage. 2001. Checks and Balances, Pri-vate Information, and the Credibility of Monetary Commitments.World Bank Publications.

Keeley, M. 1990. “Deposit Insurance, Risk, and Market Power inBanking.” American Economic Review 80 (5): 1183–1200.

Koetter, M., and T. Poghosyan. 2009. “The Identification of Tech-nology Regimes in Banking: Implications for the Market Power-Fragility Nexus.” Journal of Banking and Finance 33 (8): 1413–22.

Kydland, F. E., and E. C. Prescott. 1977. “Rules Rather than Discre-tion: The Inconsistency of Optimal Plans.” Journal of PoliticalEconomy 85 (3): 473–91.

Laeven, L., and F. Valencia. 2010. “Resolution of Banking Crises:The Good, the Bad, and the Ugly.” IMF Working Paper No.10/146.


La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny.1998. “Law and Finance.” Journal of Political Economy 106 (6):1113–55.

Lybek, T., and J. Morris. 2004. “Central Bank Governance: ASurvey of Boards and Management.” IMF Working Paper No.04/226.

Martınez Miera, D., and R. Repullo. 2010. “Does CompetitionReduce the Risk of Bank Failure?” Review of Financial Studies23 (10): 3638–64.

Moser, P. 1999. “Checks and Balances, and the Supply of Cen-tral Bank Independence.” European Economic Review 43 (8):1569–93.

Peek, J., E. Rosengren, and M. B. Tootell. 1999. “Is Bank Super-vision Central to Central Banking?” Quarterly Journal of Eco-nomics 114 (2): 629–53.

Persson, T., and G. Tabellini. 1993. “Designing Institutions for Mon-etary Stability.” Carnegie-Rochester Conference Series on PublicPolicy 39 (1): 53–84.

Rogoff, K. 1985. “The Optimal Degree of Commitment to an Inter-mediate Monetary Target.” Quarterly Journal of Economics 100(4): 1169–89.

Schoenmaker, D. 1992. “Institutional Separation between Supervi-sory and Monetary Agencies.” Special Paper No. 52, FinancialMarkets Group.

Stolz, S. M., and M. Wedow. 2010. “Extraordinary Measures inExtraordinary Times. Public Measures in Support of the Finan-cial Sector in the EU and the United States.” ECB OccasionalPaper No. 117.

Taylor, M., and M Quintyn. 2002. “Regulatory and SupervisoryIndependence and Financial Stability.” IMF Working Paper No.02/46.

Walsh, C. E. 1995. “Optimal Contracts for Central Bankers.” Amer-ican Economic Review 85 (1): 150–67.

World Bank, World Development Indicators (WDI) and GlobalDevelopment Finance (GDF) database. Available at http://databank.worldbank.org/data/home.aspx.

Documents

Financial Stability and Central Bank Governance · 2014-12-01 · Vol. 10 No. 4 Financial Stability and Central Bank Governance 35 have extended this work and studied the optimal