Download The Relationship between Structural and Cyclical - ECB

Working Paper Series
Hanno Stremmel and
Balázs Zsámboki
The relationship between
structural and cyclical features
of the EU financial sector
No 1812 / June 2015
Note: This Working Paper should not be reported as representing the views of the European Central Bank (ECB).
The views expressed are those of the authors and do not necessarily reflect those of the ECB
ABSTRACT
In this study, we explore the relationship between certain
structuralfeaturesofthebankingsectorsinEUMemberStates
and the performance of the respective banking sectors over
the financial cycle. Using the financial cycle indicator
developedbyStremmel(2015),weestimatetheimpactofthe
structural features of the banking sector on the amplitude of
thefinancialcycle.Ourresultssuggestthattheconcentration
ofthebankingsector,theshareofforeignbanks,thesizeand
stabilityoffinancialinstitutions,theshareofforeigncurrency
loans and financial interͲlinkages contribute to the amplitude
and hence the variability of financial cycles. This study
provides important insights into the appropriate design of
variousstructuralandcyclicalpolicyinstrumentsaswell.
JELClassification:E44,E61,G18,G21,G28
Keywords:bankingsectorcharacteristics,financialcycle,financialregulation,financialstructure.
ECB Working Paper 1812, June 2015
1
NonͲtechnicalSummary
Theanalysisofsystemicrisksassociatedwithchangesinthecyclicalandstructuralfeatures
offinancialsectorsgainedgrowingimportanceinrecentyears.Atthesametime,theglobal
financial crisis of 2007Ͳ2008 has also triggered a range of policy actions and regulatory
measures that aim to address cyclical and/or structural risks in the financial system. Both
BaselIIIandthenewEuropeanregulatoryframeworkincludeanewsetofmacroͲprudential
tools.
This paper explores the relationship and potential interactions between certain structural
features of the banking sectors in the EU Member States and the performance of the
respective banking sectors over the financial cycle, with the aim of providing guidance to
policymakersontheproperimplementationofcyclicalandstructuralmeasurestoaddress
theassociatedrisks.
In this paper, we follow Stremmel (2015) in creating a financial cycle indicator for 21
Europeancountries.Basedonthisindicatorwederivetwoamplitudemeasurestodescribe
the main characteristics of the financial cycles at the country level. We then relate the
amplitudemeasurestostructuralbankingsectorindicators.Ouranalyticalfindingsprovide
evidencethatcertainstructuralbankingsectorcharacteristics,suchastheconcentrationof
the banking sector, the share of foreign banks as well as the amount and composition of
banksloansandfinancialintegration,areimportantdriversofthefinancialcycleamplitude.
Moreover,thispaperalsoinvestigateswhethermonetarypolicycontributestothefinancial
cycleamplitude.Whileourfindingsare supportiveof the hypothesis thatmonetarypolicy
plays a role in influencing financial cycles, we also find that the banking sector
characteristicstendtooverridetheexplanatorypowerofthemonetarypolicystance.
Our study complements recent literature by providing insights in the longerͲterm
relationship between cyclical and structural features of the banking systems across EU
countries.Thereby,ourpapercontributestotheongoingdiscussionontheimplementation
ofmacroͲprudentialpolicymeasures.Basedontheidentifieddifferencesinfinancialcycles
across EU countries and the impact of certain structural banking characteristics on the
amplitudeofthefinancialcycle,weconcludethattheimplementationofmacroͲprudential
measuresshouldbedifferentiatedacrossEUMemberStates.Thetimingofactivationand
therelativecalibrationofthepolicymeasuresshouldtakeintoconsiderationthedifferences
bothinfinancialcyclesandbankingstructures.
ECB Working Paper 1812, June 2015
2
1
Introduction
Theglobalfinancialcrisisthateruptedin2007hasdrawnparticularattentiontotheanalysis
ofsystemicrisksassociatedwithchangesinthecyclicalandstructuralfeaturesoffinancial
sectorsaroundtheworld.Atthesametime,thecrisishasalsotriggeredarangeofpolicy
actionsandregulatorymeasuresthataimtoaddresscyclicaland/orstructuralrisks.Akey
regulatory initiative in this regard was the development of the new Basel capital and
liquidity framework (Basel III), the implementation of which is accomplished through the
CapitalRequirementsRegulation(CRR)and CapitalRequirementsDirective (CRDIV)inthe
EU.BothBaselIIIandthenewEuropeanregulatoryframeworkincludeanewsetofmacroͲ
prudentialtools,suchasthecapitalconservationbuffer,thecounterͲcyclicalcapitalbuffer,
the capital surcharge for systemically important financial institutions as well as other
instruments,suchasthesystemicriskbufferinEurope.Althoughthecombinedimpactand
possible interactions of these buffers and the underlying risk factors are highly relevant
fromamacroͲprudentialpolicyperspective,theempiricalevidenceoftheseinteractionsis
limited.
Theobjectiveofthispaperistoexploretherelationshipandpotentialinteractionsbetween
certain structural features of the banking sectors in the EU Member States and the
performance of the respective banking sectors over the financial cycle, with the aim of
providingguidancetopolicyͲmakersontheproperimplementationofcyclicalandstructural
measurestoaddresstheassociatedrisks.
Ourinvestigationisrelatedtodifferentstrandsofliterature.Recentliteraturehasrevealed
theimportanceofthefinancialstructureforlendingandeconomicgrowth.Gambacortaet
al. (2014) show that the financial structure is an important driver for output volatility,
notably bankͲbased systems tend to be more resilient than marketͲbased systems in
economic downturns. However, in cases when the economic downturn coincides with a
financial crisis, output losses for bankͲbased systems are higher than for marketͲbased
financial systems. ESRB ASC (2014) finds that bankͲbased systems have a more volatile
creditsupplyandamplifythebusinesscycle.Further,Boltonetal.(2013)elaborateonthe
lending of different types of banks in crisis periods and show that banks involved in
relationshiplendingcontinuetolendinmorefavourabletermsduringfinancialcrises.
Inaddition,thereisanemergingstrandofliteraturefocusingontheanalysisofthefinancial
cycle,tryingtocaptureitsmaincharacteristics(e.g.Aikmanetal.(2010,2014),Claessenset
al. (2011a,b), Drehmann et al. (2012), Stremmel (2015)). Stremmel (2015) provides an
overviewofthevariousapproachesusedintheliteraturetoconstructthefinancialcycle.In
thisstudy,wewillrelyonthefinancialcyclemeasuredevelopedbyStremmel(2015).
OurstudyiscloselyrelatedtoanalyticalworkonthemacroͲprudentialpolicyframeworkas
well. Borio (2013) elaborates on the relevance and implications of understanding the
financialcycleformacroͲprudentialpolicypurposes.Recentliteraturemainlylinkspatterns
ECB Working Paper 1812, June 2015
3
of financial indicators to the implementation of the counterͲcyclical capital buffer (CCB).
Bush et al. (2014), Detken et al. (2014), and Drehmann and Tsatsaronis (2014) provide a
detailedoverviewoftherelevantstudiesandinvestigatetheeffectivenessandadequacyof
cyclical measures, such as the creditͲtoͲGDP gap, for defining and calibrating the counterͲ
cyclicalcapitalbufferrate.Althoughresultsatthecountrylevelaremixed,thesuitabilityof
using the cyclical movements in credit variables as an early warning tool to identify the
buildͲupoffinancialvulnerabilitiesisgenerallynotchallenged(e.g.Detkenetal.(2014)).
OurstudycomplementstheliteraturebyprovidinginsightsinthelongerͲtermrelationship
betweencyclicalandstructuralfeaturesofthebankingsystemsacrossEUcountriesaswell
asbydrawingrelevantpolicyconclusionswithregardtothedesignandimplementationof
cyclicalandstructuralpolicymeasures,suchasthecounterͲcyclicalcapitalbuffer(CCB)and
thesystemicriskbuffer(SRB).
The remainder of the paper is organised as follows: Section 2 elaborates on the financial
cyclemeasureappliedintheanalysis.Section3discussesthemotivationandtheestimation
strategy to investigate the relationship between the financial cycle and the structural
characteristics of the banking sectors. Section 4 describes the data used in the paper,
whereasSection5providestheeconometricanalysisandoffersestimationresults.Section6
providesvariousrobustnesschecks.Section7explorestheimpactofmonetarypolicyonthe
financialcycle.Thelastsectionconcludesandprovidespolicyimplications.
2
FinancialCycles
Fortheanalysisoftheimpactofstructuralfeaturesofthebankingsectoronfinancialcycles,
weneedanindicatorthatappropriatelycapturescyclicalmovementsinthefinancialsector
sincenonaturalmeasureisavailable.Althoughpreviousliteratureprovidedinsightsinthe
developmentoffinancialcycles,itfellshortofdevelopingacommonlyacceptedmediumͲ
term financial cycle measure. Indeed, the literature diverges both as regards the
constructiontechniquesandtheingredientsofthecycle.
In our analysis, we borrow the synthetic financial cycle measure developed by Stremmel
(2015). A synthetic measure allows us to analyse the joint behaviour of different factors
influencingthefinancialcycle.FollowingStremmel(2015)weemployfrequencyͲbasedfilter
techniques to isolate cyclical movements from the trend in each of the underlying time
series.1Weobtainthecyclicalmovementofdifferentpotentialindicators,includingcredit,
assetpriceandbankingsectorindicators,andcombinetheresultingcyclicalmovementsto
construct seven different synthetic financial cycles. Table A1 in the Appendix provides an
1
We use the bandͲpass filter developed by Christiano and Fitzgerald (2003). This is basically a twoͲsided
movingaveragefilterisolatingcertainfrequenciesinthetimeseries.UsingthisbandͲpassmethodology,the
durationofafinancialcyclespansfrom32to120quarters(or8to30years).WealsocrossͲcheckedourresults
usingothersettings.Formoredetails,seeStremmel(2015).
ECB Working Paper 1812, June 2015
4
overviewofthepotentialfinancialcyclemeasures.Stremmel(2015)findsthatthesynthetic
financialcyclemeasurecontainingthecreditͲtoͲGDPratio,housepricesͲtoͲincomeratioand
credit growth offers the best fit. We refer to this measure as the “financial cycle” in our
analysis. However, in the robustness checks we crossͲcheck our results with the other six
potential financial cycle measures considered by Stremmel (2015) which combine various
asset prices and credit aggregates as well as banking sector variables. For the effective
conductofmacroͲprudentialpolicytheunderstandingofthecyclicalbehaviouroffinancial
variables, their main features and drivers, is essential. For this purpose, we compare
financial and business cycles and investigate the synchronicity of the financial cycle over
time.BothapplicationsarereproducedfromStremmel(2015).
Figure1:ComparisonofBusinessandFinancialCycles
.1
0
-.01
-.1
-.2
Amplitude Financial Cycle
.2
-.005
0
.005
.01
Amplitude Business Cycle
Sweden
1980q1
1990q1
Financial Cycle
2000q1
2010q1
Crisis
Business Cycle
Source:Stremmel(2015)
Asanexample,Figure1comparesthefinancialandthebusinesscyclesovertimeinSweden.
Other countries are pictured in Stremmel (2015). This figure confirms recent literature
which suggests that the cyclical patterns of both series share certain similarities, but the
durationofthefinancialcyclesislongerwhilebusinesscyclesappeartobemorevolatile.
ECB Working Paper 1812, June 2015
5
Figure2:SynchronicityofCycles
Source:Stremmel(2015)
Following Stremmel (2015), Figure 2 shows the synchronicity of financial cycles across 11
Europeancountries.2ThesynchronicityismeasuredastheoneͲyearcrossͲcountrystandard
deviation of the individual countries’ cycles. This metric can be used to evaluate
convergence(lowerdispersion)anddivergence(higherdispersion)offinancialcycles.
Figure 2 reveals that in periods of common financial stress (darker shaded line) financial
cycle dispersion decreases. In other words, in good times financial cycles are less
synchronized, whereas in stress periods the financial cycles tend to move together. This
increased divergence in boom periods calls for differentiated and wellͲtargeted policy
responses that are properly tailored to individual jurisdictions in order to address specific
emergingrisksinthosecountries.Atthesametime,instressperiodswhencountriesseem
tobeimpactedinasimilarmanner(asreflectedintheincreasedcoͲmovementoffinancial
cycles), a higher level of coordination and harmonisation of policy actions may be
warranted.
Both applications show that the policyͲmakers’ awareness of the characteristics of the
financial cycles across countries is essential for taking adequate macroͲprudential policy
actions.Againstthisfinding,wenowturntotheinvestigationwhetherstructuralfeaturesof
thebankingsectorsinfluencethefinancialcycleinindividualjurisdictionsandwhetherthey
actaspotentialdriversofthefinancialcycleamplitude.
2
Thefollowingcountriesare included intheanalysis:Belgium,Denmark,Finland,France,Germany,Ireland,
Italy,theNetherlands,Spain,SwedenandtheUnitedKingdom.
ECB Working Paper 1812, June 2015
6
3
EstimationApproach
In general, several structural features of the banking sector may have an impact on the
financialcycleanditsamplitude.Bankingsectordepthandsizearepossiblecontributorsto
thevariationsinthefinancialcycle.Otherinfluencingfactorsmayincludetheconcentration
of the banking sector or the structure of the financial system (bankͲ vs. marketͲbased
system).Furthermore,thestabilityoffinancialinstitutions,theactivityofforeignbanksand
theamountofforeigncurrencyloansmayalsobeseenaspotentialfactorsimpactingonthe
amplitudeofthefinancialcycle.Inouranalysisweincorporateseveralpotentialinfluencing
factorstoaccommodateforarangeofpossibleimpactsonthefinancialcycle.
Our estimation strategy to explore the relationship of banking sector features and the
financialcycleinvolvesthefollowingsteps:
i.
ii.
iii.
iv.
v.
WeconstructthefinancialcycleforeachcountrybasedonStremmel(2015).
We identify the tuning points of the financial cycle for each country. The
determinationofthelocalminimaandmaximaofeachcycleallowsustodefinethe
peaks (local maxima) and troughs (local minima) of the financial cycle and to
calculatetheamplitude.
We define the financial cycle phases. An upswing or expansion phase lasts from a
troughtoapeakpoint,whereasadownswingorcontractionphaselastsfromapeak
toatroughpoint.
We calculate the dependent and independent variables for the corresponding
financialcyclephasesaccountingonlyforthedevelopmentsinthespecificfinancial
cyclephase.
Lastly,weemployGeneralizedLinearModel(GLM)estimationtechniquestoanalyse
the relationship between the financial cycle amplitude and banking sector
characteristics.
Inouranalysis,weusetheamplitudeofthefinancialcycleinsteadofacontinuousfinancial
cycle measure as a dependent variable. This has various reasons: A continuous measure
(e.g. variance) requires higher data frequency and hence sufficiently long time series that
arenotavailableformostEuropeancountries.Thefinancialcyclephaseapproachhowever
enablesustoincludemorethan20Europeancountriesinthesample.
Furthermore, given that for many newly joined European Member States available data
starts only in the early 2000s, we are not able to consider complete financial cycle
movements. By using the financial cycle phase measure we can incorporate at least one
phase of the financial cycle and consequently we are able to analyse more countries and
financialcyclemovements.
Another argument for using the phase measures is that structural banking characteristics
arerelativelystableovertimeincomparisontothequicklychangingfinancialcycle.Onthat
ECB Working Paper 1812, June 2015
7
account, a continuous financial cycle measure may misinterpret the influence of the
structural features by overestimating their potential impact on the financial cycle. Against
thisbackground,adiscretemeasureseemstobemoreappropriatetocapturethevariation
inthefinancialcycle.
4
DataandVariables
Inthisstudy,weattempttoincludeasmanyEuropeancountriesintheanalysisaspossible.
Wecreateadatasetfor21Europeancountriesspanningthepotentialperiodof1980Q1to
2012Q4.ThelistofcountriesandthedetailedcoverageofthevariablesarelistedinTable
A2intheAppendix.Thetotalnumberofobservationsacrossthedifferentindicatorsgroups
is266financialcyclephases.
Table1:DescriptionoftheVariables
Variable
ࡲ࢏࢔࡯࢟ࢉ࢒ࡼࢎࢇ࢙ࢋ࡭࢓࢖࡭ ࡲ࢏࢔࡯࢟ࢉ࢒ࡼࢎࢇ࢙ࢋ࡭࢓࢖࡮ Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Bank_Assets/GDP
Market_Cap/GDP
FX_Loans/Loans
Credit/GDP
Foreign_Claims/GDP
Description
LeftͲhandside
NonͲTimeͲAdjustedAmplitudeMeasure
TimeͲAdjustedAmplitudeMeasure
RightͲhandside
AssetsoftheThreeLargestBanksasaShareofTotalBankingAssets(%)
Foreignbanksamongtotalbanks(%)
Bankcredittobankdeposits(%)
BankdepositstoGDP(%)
Depositmoneybanks'assetstoGDP(%)
Stockmarketcapitalizationto GDP(%)
Shareofforeigncurrencyloanstototalloans(%)
Domesticcredittoprivatesector(%ofGDP)
ConsolidatedforeignclaimsofBISreportingbanks(%ofGDP)
Source
AuthorsCalculation
AuthorsCalculation
GFDD/Bankscope
ClaessensandvanHoren(2014)
IMFIFS
IMFIFS
IMFIFS
IMFIFS
ECBSDW
IMFIFS
BISCBS/IMFIFS
Table1providesanoverviewofthevariablesusedintheanalysisaswellastheirunderlying
source. We employ two amplitude measures of the financial cycle as left handͲside
variables. As explanatory variables, we consider nine banking sector characteristics. The
explanatory variables are sourced through different established databases. We use time
series from World Bank Global Financial Development Database (GFDD), International
MonetaryFundInternationalFinancialStatistics(IMFIFS),EuropeanCentralBankStatistical
DataWarehouse(ECBSDW)aswellastheBankforInternationalSettlementsConsolidated
Banking Statistics (BIS CBS). Many variables are used as standard metrics to benchmark
financialsystems(Cihaketal,2013).
FinancialCycleAmplitude
TheleftͲhandsidevariableisdesignedtocapturethemagnitudeofthemovementsinthe
financialcycle.WeobtainthefinancialcyclemeasureconsistingofcreditͲtoͲGDPratio,(ii)
the house pricesͲtoͲincome ratio, and (iii) credit growth for each of the 21 European
countries using the methodology described in Stremmel (2015). The obtained financial
ECB Working Paper 1812, June 2015
8
cycles are illustrated in Figure A1 in the Appendix. In the next step, we identify the time
series’peaksandtroughstodeterminethefinancialcyclephases.3
We employ two different concepts for measuring the amplitude of each financial cycle
phase. The nonͲtimeͲadjusted amplitude measure ሺ‹›…ŽŠƒ•‡’୅ ሻ reflects the
absolutedifferenceofthestartሺ‹›…ŽŠƒ•‡ୗ୘୅ୖ୘ ሻandendvaluesሺ‹›…ŽŠƒ•‡୉୒ୈ )of
thefinancialcyclephase
‹›…ŽŠƒ•‡’୅ ൌ ȁ‹›…ŽŠƒ•‡ୗ୘୅ୖ୘ െ ‹›…ŽŠƒ•‡୉୒ୈ ȁ
Theabsolutedifferencereflectsthemagnitudeofthecyclicalmovementsandquantifiesthe
expansionorcontractionofeachcyclephase.
ThetimeͲadjustedamplitudemeasureሺ‹›…ŽŠƒ•‡’୆ ሻiscalculatedasfollows:
‹›…ŽŠƒ•‡’୆ ൌ
ȁ‹›…ŽŠƒ•‡ୗ୘୅ୖ୘ െ ‹›…ŽŠƒ•‡୉୒ୈ ȁ
‹›…ŽŠƒ•‡ୈ୳୰ୟ୲୧୭୬
The numerator is equivalent to the nonͲtimeͲadjusted amplitude measure. The
denominatorrepresentsthedurationofthefinancialcyclephaseሺ‹›…ŽŠƒ•‡ୈ୳୰ୟ୲୧୭୬ ሻ.In
doing so, the timeͲadjusted amplitude measure accounts for the intensity of changes in
amplitudesand thusincorporatesthetimedimensionintheanalysisofthefinancialcycle
phase.
Toillustratetheintuitionbehindtheamplitudemeasures,weplotSweden’sfinancialcycle
anditsturningpointsinFigure3,comparingtwofinancialcyclephases,theiramplitudesand
correspondingdurations.Thelightredcolouredcyclephasenamesrepresentcontractionor
downswing phase, whereas green coloured phase names correspond to expansions or
upswingcyclephases.
3
Further,wemakeanassumptionontheendpointofthefinalfinancialcyclephase.Ifthelastphaseisnot
completed,weconsiderthelastobservationofthefinalfinancialcycletobeaturningpointsothatweare
able to complete the corresponding financial cycle phase. Of course, this final turning point may not be an
accurate estimation as the phase might last longer. However, this assumption allows us to incorporate the
structuralbankingsectorcharacteristicsafterthe2007GlobalFinancialCrisis.
ECB Working Paper 1812, June 2015
9
Figure3:FinancialCyclePhases
Figure3revealsthatfinancialcyclephasesmaydifferinamplitude,durationandadjustment
speed.ToillustratethedifferentmetricswefocusonupwardsPhases2and4.Bothphases
are quite similar regarding their duration (22 and 19 quarters, respectively), but their
amplitudesaremarkedlydifferent(0.175vs.0.072increaseinthefinancialcyclemeasures,
respectively).Webelievethattherelationshipbetweenthedurationandtheamplitudeof
thefinancialcyclealsohasimportantimplicationsforfinancialstabilityassessmentandthe
design of macroͲprudential policy action. In particular, a rapid increase may be more of a
financial stability concern than a longͲterm gradual buildͲup of the cycle as such a rapid
increase,possiblysupportedbylooserlendingstandards,mayswiftlyrevealvulnerabilities
in the financial sector, narrowing the scope and shortening the available time for policy
action.Nonetheless,weuseboththestandardandthetimeͲadjustedamplitudemeasures
toverifyourresults.
StructuralFeaturesoftheBankingSector
The explanatory variables (RHS) are based on a set of structural banking sector features
which are expected to have an influence on the financial cycle and in particular on its
amplitude. Overall, we employ nine structural banking sector variables grouped into six
categories. To be in line with the LHS variable, the RHS variables also reflect the
developments of structural banking sector features in the corresponding financial cycle
phase.Thereisawidevarietyofpotentialstatisticalmethodstomodelthesedevelopments.
We opt to use two simple approaches to capture these developments of the structural
features. On the one side, for rather sluggish variables we obtain the medians across all
observationsinthecorrespondingcycleͲphase.Weapplythismediancalculationforthetwo
marketsharemeasures.Weexpectthatthemarketsharesoflargeinstitutionsandforeign
bankschangeonlygraduallyoverthecyclephase.Fortheremainingindicators,wecalculate
theabsolutedifferencesineachphase(i.e.thedifferencebetweenthestartandendvalues
ECB Working Paper 1812, June 2015
10
ofeachcyclephase).Theinterpretationisstraightforward,becausetheunitsoftheabsolute
differencesareexpressedinpercentagepoints.
TheRHSvariablesaregroupedintosixcategories:(i)concentrationofthebankingsystem,
(ii) market share of foreign banks, (iii) institution sizeand stability, (iv)financial depth, (v)
bank loans, and (vi) financial integration. Each variable is obtained at the country level.
Table A2 in the Appendix gives an overview of the coverage of the indicators regarding
financialcyclephasesforeachvariablegroupatthecountrylevel.
The first two categories only contain a single variable. First, we approximate the
concentration of the banking sector, Concentration, by calculating the assets of the three
largestbanksasafractionofthetotalbankingassets.Theanalysisincludes60cyclephase
observations. The empirical evidence is inconclusive on the effects of banking sector
concentration on financial stability (Berger et al, 2009). Recent papers suggest an inverse
relationshipbetweenmarketconcentrationandfinancialstability(e.g.BoydandDeNicolo
(2005), Boyd et al. (2006), De Nicolo and Loukoianova (2007), Schaeck et al. (2009)).4
Accordingtothislineofarguments,weexpecthigherbankingsectorconcentrationtohave
apositiveinfluencetotheamplitudeofthefinancialcycle.
In the second category, Foreign_Banks, we consider the activity of foreign banks in the
domesticmarket.Weincorporateameasurethatrelatesthenumberofforeignbankstothe
totalnumberofbanksineachcountry.ThismeasureisbasedonthedatabasebyClaessens
andvanHoren(2014),whereasabankisdefinedasforeignif50%ofitssharesareholdby
nonͲresidentshareholders.Welookatthenumberofinstitutionsinsteadofforeignbanks'
shareintotalassetsduetolongeravailabletimeseries.Thiswayweareabletoinclude26
cyclephaseobservationsforthiscategory.
Therecentglobalfinancialcrisishighlightedthepotentialrisksassociatedwiththeactivityof
foreignbanksandcrossͲborderlending.DeHaasandvanLelyveld(2014)showthatforeign
banks are not a source of credit provision in times of credit tightening. Instead, foreign
banksandsubsidiariesadjusttheirlendingevenstrongerthandomesticcreditinstitutionsin
response to shocks (e.g. Aiyar (2012), Popov and Udell (2012), De Haas et al. (2013), and
Fungácováetal.(2013)).Therefore,weexpectthatahighershareofforeignbanksamplifies
thefinancialcycle.
The remaining four categories contain two explanatory variables each. The third group of
variables“Institutionsizeandstability”aimsatcapturingthesizeandthefundingstabilityof
thebankingsystem.WefollowtheapproachbyCihaketal.(2013)tomeasurethesizeof
the banking sector. The first variable, the bank depositsͲtoͲGDP ratio (%), Deposits/GDP,
4
In addition, the sameconclusions canbe derived from economic theory.Bergeretal.(2004), Beck(2008),
Uhde and Heimsehoff (2009), and Degryse et al. (2013) provide literature reviews on the theoretical and
empiricalapplications.
ECB Working Paper 1812, June 2015
11
indicates the amount of deposit resources available to the financial sector for its lending
activitiesinrelationtotherealeconomyatthecountrylevel.Thesecondvariable,thebank
creditͲtoͲbank deposits ratio (%), Credit/Deposits, measures the banking sector’s funding
stability.Thisratioincreasesifcreditcreationishigherthandepositgrowthanddecreases
whendepositgrowthexceedscreditgrowth.Weexpectthatanincreasingindicatorofthe
bankingsystem'ssizeaswellasanincreasingcreditͲtoͲdepositratiocontributepositivelyto
theamplitudeofthefinancialcycle.Weincludeintheanalysis62cyclephaseobservations
forthisgroup.
The fourth category, “Financial depth”, accounts for the importance of various financial
markets for financing the economy. The depth of the banking system is traditionally
measured by the deposit money banks' assetsͲtoͲGDP ratio (%), Bank_Assets/GDP. The
depth of the stock market, Market_Cap/GDP, is captured by using the stock market
capitalizationͲtoͲGDP ratio (%). Recent literature argues that a certain level of financial
depth is needed to sustain longͲterm economic growth. However, there is also evidence
thatatoodeepfinancialsystemcanalsobeaccompaniedbyundesiredeffectsonfinancial
stability and economic growth (e.g. Arcand et al. (2012), Cecchetti and Kharroubi (2012),
ESRBASC(2014)).Weexpectthatadeeperfinancialsystemamplifiesthefinancialcycle.In
addition, these indicators could also be used to investigate whether a financial system is
morebankͲ,ormarketͲbasedandwhetherthesecharacteristicshavedifferentimpactson
financial cycles. In total, we are able to include 51 cycle phase observations for this
category.
Thefifthcategory,“Bankloans”,dealswiththeamountandcompositionofbankloans.We
measure bank lending in the economy by using the domestic private sector creditͲtoͲGDP
ratio(%),Credit/GDP.5Additionally,weincorporatethecurrencycompositionofloansusing
theforeigncurrencyloansͲtoͲtotalloansratio(%),FX_Loans/Loans.Foreigncurrencyloans
are an instrument that allows financial intermediaries to provide additional credit to their
clients even in cases when credit origination in domestic currency could be constrained.6
Foreigncurrencyloansmayalsoimposeadditionalrisksoncreditorsanddebtorsalike.We
expectbothindicatorstoincreasetheamplitudeofthefinancialcycle.Intotal,weareable
toinclude34cyclephaseobservationsforthisgroup.
5
This ratio is similar to one of the components of the financial cycle measure, but the intuition for
incorporatingthismeasureasaRHSvariableisdifferent.Inthissectionwearenotinterestedindetermining
the cyclical movement of the creditͲtoͲGDP ratio, but the indicator is rather used to capture the overall
amount of credit provided by financial intermediaries relative to the level of economic development.
Therefore, we look at the levels and not the filtered series. The correlation of the two series is rather low
(below 0.3), therefore we are confident that employing this measure as a RHS variable is appropriate.
Moreover,weusetheIMFIFScreditdatatodefinetheRHSvariableinsteadoftheBIScreditdatausedforthe
LHSvariable.
6
The volume of foreign currency credit can either be driven by the demand or the supply side. For more
informationonforeigncurrencyloansseeLucaandPetrova(2008),Brownetal.(2010),Bassoetal.(2010)and
BrownandDeHaas(2012).
ECB Working Paper 1812, June 2015
12
Thelastcategory,“Financialintegration”,accountsforinternationalfinanciallinkagesacross
countries.InadditiontoFX_Loans/Loansusedinthepreviousspecification,wealsoinclude
the ratio of BIS reporting banks’ consolidated foreign claimsͲtoͲGDP, Foreign_Claims/GDP,
to approximate the international financial linkages. In line with the argument on the
presence of foreign banks and the impact of more financial development, we expect that
crossͲbroader claims have a positive influence on the amplitude of the financial cycle. In
total,weareabletoinclude34cyclephaseobservationsforthiscategory.
DescriptiveStatistics
Table2revealsthedescriptivestatisticsofthefinancialcyclephasesforthe21EUMember
Statesincludedintheanalysis.Thegreyshadedrowsexhibitcountriesforwhichboththe
standard and the timeͲadjusted amplitude measures are higher than their respective
medians.Thestandardamplitudemeasure(AmplitudeA)seemstobemoremarkedinup
than in down phases, whereas with the alternative timeͲadjusted amplitude measure
(Amplitude B) the distinction is less pronounced. For all countries, the duration of the
financial cycle phase is similar with around 20 quarters per cycle phase. Nevertheless, on
average the up phases tend to last longer than the down phases (25 and 14 quarters,
respectively).
ECB Working Paper 1812, June 2015
13
ECB Working Paper 1812, June 2015
14
Phases
Amplitude
A
Country
0.043
Austria
0.032
Belgium
0.095
Denmark
0.083
Finland
0.076
France
0.052
Germany
0.056
Greece
0.133
Hungary
0.085
Ireland
0.116
Italy
0.364
Latvia
0.393
Lithuania
0.107
Luxembourg
0.063
Malta
0.052
Netherlands
0.175
Poland
0.042
Portugal
0.072
Slovakia
0.145
Spain
0.103
Sweden
0.158
UnitedKingdom
0.080
Median/Sum
AllPhases
Amplitude Duration
B
(quarter)
0.0020
21.5
0.0017
16.8
0.0047
19.2
0.0037
18.3
0.0030
24.0
0.0014
29.7
0.0027
20.0
0.0070
19.5
0.0037
25.0
0.0051
23.2
0.0199
19.0
0.0197
20.0
0.0043
29.0
0.0036
17.0
0.0021
25.8
0.0099
14.5
0.0022
16.5
0.0057
11.5
0.0059
25.8
0.0052
17.5
0.0059
28.3
0.0035
20.0
Amplitude
A
0.036
0.033
0.106
0.100
0.096
0.048
0.088
0.176
0.106
0.150
0.412
0.587
0.167
0.085
0.062
0.322
0.055
0.111
0.163
0.125
0.191
0.092
#Obs.
2
6
6
6
5
3
2
2
4
5
2
2
2
2
4
2
4
2
4
6
3
74
UpswingPhases
Amplitude Duration
B
(quarter)
0.0020
18.0
0.0019
15.8
0.0047
21.7
0.0042
23.3
0.0033
31.0
0.0012
29.0
0.0028
31.0
0.0065
27.0
0.0030
34.5
0.0054
32.0
0.0172
24.0
0.0196
30.0
0.0034
49.0
0.0038
22.0
0.0025
24.0
0.0129
25.0
0.0025
21.5
0.0074
15.0
0.0051
31.5
0.0058
21.0
0.0042
45.0
0.0057
25.0
37
1
4
3
3
2
2
1
1
2
2
1
1
1
1
2
1
2
1
2
3
1
#Obs.
Table2:FinancialCyclePhasesacrosstheCountries7
0.048
0.050
0.029
0.084
0.066
0.062
0.059
0.023
0.091
0.065
0.094
0.316
0.199
0.046
0.041
0.042
0.027
0.029
0.033
0.128
0.082
0.141
Amplitude
A
0.0057
0.0020
0.0015
0.0048
0.0033
0.0029
0.0019
0.0025
0.0075
0.0043
0.0049
0.0226
0.0199
0.0051
0.0034
0.0017
0.0068
0.0019
0.0041
0.0067
0.0046
0.0067
14.0
25.0
19.0
16.7
13.3
19.3
31.0
9.0
12.0
15.5
17.3
14.0
10.0
9.0
12.0
27.5
4.0
11.5
8.0
20.0
14.0
20.0
DownswingPhases
Amplitude Duration
B
(quarter)
37
1
2
3
3
3
1
1
1
2
3
1
1
1
1
2
1
2
1
2
3
2
#Obs.
AmplitudeAreferstothenonͲtimeͲadjustedamplitudemeasure(‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஺ )andAmplitudeBreferstothetimeͲadjustedamplitudemeasure(‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஻ ).
7
5
EstimationResults
The empirical model employs two independent variables (‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஺ and
‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஻ ) and six groups of explanatory measures (Concentration, Foreign
banks,Institutionsizeandstability,Financialdepth,Bankloans,Financialintegration).Due
tothelownumberofoverlappingobservationsamongthegroups,wehavetoanalysethe
influenceofeachvariablegroupseparately.
To enhance the credibility and plausibility of our regressions, in Section 6 we supplement
theanalysisbyotheramplitudeindicatorsandestimationtechniquesforrobustnesschecks.
We employ different estimations techniques: Ordinary Least Squares (OLS) with robust
standards errors and General Linear Model (GLM) with either robust standards errors or
clustered standard errors. We believe that different estimation techniques and varying
standard errors are capable of accommodating the required demands of this setting.
Overall, we obtain six regressions per estimation technique and per amplitude measure.
Moreover, we reͲestimate our model for all potential financial cycle measures defined by
Stremmel(2015).
For the sake of convenience, we show results only for our preferred timeͲadjusted phase
amplitude metric (‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஻ ) and the General Linear Model with robust
standards errors. The results for other metrics and estimation techniques are very similar
andwillbeconsideredinSection6.
Table 3 shows the regression results for each indicator groups. The six columns (1) to (6)
correspond to the six groups of structural banking characteristics defined in the previous
section.Confirmingtheimpressionsfromthecorrelationtable(TableA2intheAppendix),
each variable features a significant influence on the amplitude measure. The obtained
marginal values can be interpreted in semiͲelastic terms. The results exhibit only the
combinedcontributionofeachvariablegrouptotheexplanatorypoweroftheregressions.
ThecontributionofindividualvariableswillbecoveredindetailinSection6.
ECB Working Paper 1812, June 2015
15
Table3:RegressionsofFinancialCycleMeasure
‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஻ Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Market_Cap/GDP
Bank_Assets/GDP
FX_Loans/Loans
Credit/GDP
FX_Loans/Loans
Foreign_Claims/GDP
(1)
0.0074***
(2)
(3)
(4)
(5)
0.0093***
0.0046*
0.0067**
0.0021*
0.0071***
No.ofObservation
Constant
AdjustedR2(fromOLS)
BIC
0.0925***
0.0098***
60
yes
0.48
Ͳ464.00
(6)
26
yes
0.33
Ͳ201.71
62
yes
0.22
Ͳ480.98
51
yes
0.23
Ͳ391.34
34
yes
0.55
Ͳ271.45
0.1229***
0.0058***
33
yes
0.51
Ͳ261.76
*p<.1,**p<.05,***p<.01
ThevariablesConcentrationandforeignbanks(Model(1)and(2))seemtohavethehighest
positiveimpactontheamplitudeofthefinancialcycle.Bothmodelsalsoofferconsiderable
explanatorypowerintermsofahighadjustedR2measure.Inaddition,Model(5)suggests
thatForeignCurrencyLoansalsocontributesignificantlytotheamplificationofthefinancial
cycle.Model(6)exhibitsthatFinancialLinkagesarealsoimportantdriversoftheamplitude.
In contrast, the impacts of Financial depth, Model (4), and the explanatory power of this
specification tend to be limited in terms of low adjusted R2 measure. Nevertheless, the
componentsofthismeasureneedtobedifferentiated.Incomparisontothedepthofstock
market,therelativesizeofthebankingsectorseemstobethemaindriverofthefinancial
cycle. Finally, the Institution size and stability specification – Model (3) – is also able to
explain a notable part of the variation of the amplitude measures, although its total
explanatorypowerislowerincomparisontoothergroups.
Overall, our regression results suggest that structural features of national banking sectors
haveasignificantimpactontheamplitudeofthefinancialcycle.Althoughallbankingsector
indicator groups have some explanatory power, the magnitude of the impact varies
significantly across the indicator groups. In particular, banking concentration, the share of
foreignbanks,banksloansandfinanciallinkagesofferconsiderablyhighexplanatorypower.
6
RobustnessChecks
Todemonstratetherobustnessofourfindings,weperformthreerobustnesschecks.First,
we successively add variables to the individual model specification to investigate model
stability. Second, we explore whether the influence of banking sector characteristics on
ECB Working Paper 1812, June 2015
16
financial cycles diverges in upͲ and downswing phases. Lastly, we estimate our model
specificationforotherleftͲhandsidevariablemeasures.
Inthefirststep,weexplorethestabilityoftheindividualparametersofthebankingsector
characteristicsandtheircontributiontotheexplanatorypowerofvariablegroupsbasedon
the results in Section 5. For each category, we gradually extend the specification by
sequentiallyintroducingthevariables.ThefirsttwocolumnsinTable3areidenticalwiththe
firsttwocolumnsinTable4duetothesinglebankingsectorvariableinthosespecifications.
For the remaining characteristic categories we employ three model specifications in each
case. The first two specifications reflect the individual category components, whereas the
thirdspecificationofeachgrouprepresentsthecombinedinfluenceofbothcomponents.
All employed banking sector characteristics indicators in Table 4 are significant at least at
the10%confidencelevel.Themarginalvalueoftheindividualindicatorsremainsstableby
adding additional components. Further, the sum of the individual explanatory powers in
termsoftheadjustedR2valuesaddupquitecloselytoaggregatemeasures.Thissuggests
thateachoftheusedindicatorsoffersadditionalandcomplementaryexplanatorypower.8
However,theaddedexplanatorypowerisdifferentamongtheindicators.Variablessuchas
Market_CAP/GDP or Deposits/GDP only increase the explanatory power marginally. In
comparison, other variables are better placed to explain large parts of the variation (e.g.
Concentration, Foreign currency loans). All in all, the selected indicators seem to be well
determined and remain robust individually and in combination, but their individual
contributiontotheexplanatorypowerofeachspecificationdiffersmarkedly.
In a second robustness check, we investigate whether the influence of the structural
banking sector variables on the amplitude of the financial cycle varies across different
phasesofthecycle.WesplitthesampleintoupͲanddownͲphasesofthefinancialcycle.The
regressionresultsarepresentedinTable5.Importantly,duetothelownumberoffinancial
cyclephasesincludedintheanalysis,theresultshavetobeinterpretedwithcaution.
8
ThesamplesofthelattertwobankingsectorcharacteristicsgroupsͲModel(10)to(14)Ͳarenotidenticaland
thereforetheircomparabilityislimited.Althoughthenumberoftheobservationsissimilar,thecoverageof
thecountriesisdifferent(TableT1intheAppendix),hencethemarginalvaluesoftheidenticalvariablesvary
betweenthetwogroups.
ECB Working Paper 1812, June 2015
17
ECB Working Paper 1812, June 2015
18
60
yes
0.48
Ͳ464.00
0.0074***
(1)
Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Market_Cap/GDP
Bank_Assets/GDP
FX_Loans/Loans
Credit/GDP
FX_Loans/Loans
Foreign_Claims/GDP
No.ofObservation
AdjustedR2(fromOLS)
BIC
(2)
(3)
(4)
51
yes
0.01
Ͳ381.27
0.0023**
(6)
51
yes
0.22
Ͳ393.58
0.0071***
(7)
51
yes
0.23
Ͳ391.34
0.0021*
0.0071***
(8)
0.0955**
34
yes
0.30
Ͳ259.22
(9)
33
0.61
Ͳ269.15
0.0118***
0.0025
(6)
(7)
0.0074***
0.0014*
0.0096*** 0.0362
0.0074***
Ͳ0.0043
0.0066*** 24
12
11
32
0.36
0.80
0.61
0.46
Ͳ180.52
Ͳ118.27
Ͳ102.18
Ͳ246.87
UpswingFinancialCyclePhase
(3)
(4)
(5)
(10)
7
0.35
Ͳ70.05
0.0062**
(8)
34
yes
0.55
Ͳ271.45
0.0925**
0.0098***
(11)
33
yes
0.45
Ͳ260.43
0.1243***
(12)
29
Ͳ0.05
Ͳ229.90
0.0004
0.0038
27
0.02
Ͳ212.89
Ͳ0.0003
0.0036**
(12)
*p<.1,**p<.05,***p<.01
0.1495***
Ͳ0.0034
22
22
0.44
0.61
Ͳ162.57
Ͳ170.73
0.1100**
0.0087
0.1229***
0.0058***
33
yes
0.51
Ͳ261.76
(14)
*p<.1,**p<.05,***p<.01
0.006***
33
yes
0.06
Ͳ242.28
(13)
DownswingFinancialCyclePhase
(9)
(10)
(11)
34
yes
0.25
Ͳ256.82
0.0102***
Table5:RobustnessCheckforFinancialCyclePhase
62
yes
0.22
Ͳ480.98
62
Yes
0.07
Ͳ474.51
0.0046*
0.0067**
(5)
Table4:RobustnessCheckforVariablesGroups
0.0077***
0.0119***
19
0.50
Ͳ141.86
(2)
62
yes
0.15
Ͳ479.11
0.0074***
28
0.49
Ͳ218.58
(1)
26
yes
0.33
Ͳ201.71
0.0093*** 0.0049*
‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஻ No.ofObservation
Constant
AdjustedR2(fromOLS)
BIC
Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Market_Cap/GDP
Bank_Assets/GDP
FX_Loans/Loans
Credit/GDP
FX_Loans/Loans
Foreign_Claims/GDP
‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஻ In Table 5, the first six columns (Model 1Ͳ6) reflect the regression results for the upswing
financial cyclephases,whereasthelattersixcolumns(Model7Ͳ12)reflect theregressions
forthedownswingfinancialcyclephases.Theregressionresultssuggestthattheinfluence
ofthestructuralbankingfeaturesvariesacrossthefinancialcyclephases.TheConcentration
ofthebankingsectorseemstobeanimportantdriverinbothtypesoffinancialcyclephase.
AlthoughthemarketshareofForeignbankshasaninfluenceinbothphases,itsinfluence
tendstobebiggerinupswingsphases.TheInstitutionsizeandstabilitygrouptendstohave
an influence on the cycle phases in the upswing. The Financial depth group occurs to be
importantinbothphases,albeititsimpactishigherintheupswingphase.TheShareofFX
Loansseemstobeamajordriverinthefinancialcycleamplitudeduringdownswingphases,
indicating that countries with high levels of foreign currency loans are exposed to more
severe contractions in these periods. In addition, Financial integration
(Foreign_Claims/GDP) seems to be important for the amplitude in upswing phases of the
financial cycle. However, due to the limited number of financial cycle phases, we restrain
fromdrawingfarͲreachingconclusionsfromtheseresults.Obviously,thisrobustnesscheck
leaves room for improvement by adding further phases and investigating the impact of
structuralfeaturesinmoredetail.
As a third step we investigate the robustness of the banking sector characteristics by
enlarging the econometric analysis to other financial cycle measures. Stremmel (2015)
investigated a number of potential indicators to obtain the best performing measure to
portraythefinancialcycle.Inthisroundweemployallpotentialsyntheticcyclemeasuresto
explore the robustness of our findings. Table A1 in the Appendix provides an overview of
thedifferentfinancialcyclemeasures.
WereͲestimatetheregressionsforallfinancialcyclemeasures.Asinthebaselinemodel,we
have to determine the financial cycle phases and calculate their amplitude as well as the
corresponding banking sector characteristics for each financial cycle indicator. In addition,
wealsoruntheregressionjustfortheelevencountriesusedinStremmel(2015)tocontrol
whethertheinfluenceofvariableshaschanged.Weemploythreeestimationtechniquesfor
the seven financial cycle measures with a potential maximum of 42 specifications per
banking sector characteristicscategory.9 To manage the number of regressions, we opt to
visualize the condensed overview of results in Table 6. We only exhibit the proportion of
wellͲdeterminedRHSvariablesthataresignificantat10%levelacrossthemodels.
9
The number of 42 regressions is based on the following procedure. We estimate seven regressions per
financialcycleamplitude–foreachcategoryoneregression.Weusetwodifferentamplitudephasemeasures
– ‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஺ and ‫݌݉ܣ݁ݏ݄݈ܽܲܿݕܥ݊݅ܨ‬஻ – and we employ three different estimation techniques –
OrdinaryLeastSquares(OLS)withrobuststandardserrorsandGeneralLinearModel(GLM)witheitherrobust
standards errors or clustered standard errors. The theoretical maximum would be 252 regressions (42
multiplied by 6 indicators groups). Due to the requirements of having an appropriate number of 20
observationsperestimation,weonlyinclude198specificationsintherobustnesschecks.
ECB Working Paper 1812, June 2015
19
Table6:ProportionofSignificantandWellͲDeterminedRHSVariables
LHSMeasure
Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Market_Cap/GDP
Bank_Assets/GDP
FX_Loans/Loans
Credit/GDP
FX_Loans/Loans
Foreign_Claims/GDP
Overall**
#Regression
ࡲ࢏࢔࡯࢟ࢉ࢒ࡼࢎࢇ࢙ࢋ࡭࢓࢖࡭ *
ࡲ࢏࢔࡯࢟ࢉ࢒ࡼࢎࢇ࢙ࢋ࡭࢓࢖࡮ *
Total
42
24
100.0%
100.0%
100.0%
42.9%
100.0%
100.0%
75.0%
100.0%
83.3%
91.7%
100.0%
100.0%
100.0%
42.9%
100.0%
100.0%
91.7%
100.0%
91.7%
83.3%
100.0%
100.0%
100.0%
42.9%
100.0%
100.0%
83.3%
100.0%
87.5%
87.5%
88.8%
88.3%
42
42
24
24
198
87.8%
*Asacriticalsignificantthresholdwechoosethe10%confidencelevelforeachvariable.
**Theoverallaverageiscalculatedastheaccuracyoftheunderlyingmodelweightedbythenumberofobservation.
Table 6 demonstrates that the overall accuracy of different financial cycle measures is
remarkably high. In 88% of all regressions the coefficients of the banking sector
characteristicsarecorrectlyspecifiedandsignificant.Thisconfirmsthatourfindingsarenot
specifictothechosenfinancialcyclemeasurebutgenerallyapplicabletoallfinancialcycle
measures. Nonetheless, there are differences across banking sector characteristics. While,
thelargemajorityoftheindicatorshavetheexpectedsignandaresignificantineverysingle
specification, there are some exceptions. Regarding Foreign_Claims/GDP and
FX_Loans/Loansintwodifferentspecifications,theshareofcorrectlyspecifiedindicatorsis
high.Incontrast,Deposits/GDPisonlysignificantin4outof10cases.Thisfindingisinline
with the previous robustness check that revealed that the explanatory power of this
componentisratherlimited.Somewhatsurprisingly,bothmeasuresofFinancialdepthare
significantthroughoutallmodels,althoughtheirexplanatorypowertendstoberatherlow.
Basedonallrobustnesschecks, weareconfidentthatourfindingsfromSection5arenot
conditional on the choice of a specific financial cycle measure but have validity in more
generalterms.
As ageneral caveat it has to be pointed out that ourstudy is only able to includea small
numberofobservationsreflectingtheavailabilityoftheunderlyingdata.Therestricteddata
availability for the LHS variableencouraged usto use financialcycle phases instead of full
financial cycles. The constraints of the RHS variables led us to employ the banking
characteristicindicatorsindividuallyorpairwise.Inourview,thesearesensiblewaystodeal
withdataconstraints.Anothercaveatisthatthespecificationsdivergeinboththenumber
of observations and country coverage. An obvious solution would be to find a common
denominator of phase coverage through all models but this would reduce the number of
observationsdramatically.Nonetheless,weareconvincedthatourapproachappropriately
accounts for the data limitations and provide consistent, robust and insightful regression
results.
ECB Working Paper 1812, June 2015
20
7
ImpactsfromMonetaryPolicy
Relatedtothequestionofwhetherstructuralbankingfeaturesinfluencethefinancialcycle
is the question whether monetary policy also contributes to the development of the
financial cycle. We explore this relationship by extending our structural banking sector
specificationsbyincorporatingameasureofthemonetarypolicystance.
RecentliteraturearguesthatmonetarypolicycontributestothebuildͲupoffinancialcycles
by extending banks' balance sheets, triggering additional bank riskͲtaking and boosting
credit supply (e.g. Adrian and Shin (2008), Altunbas et al. (2010), Maddaloni and Peydró
(2011),HoubenandKakes(2013),Dell’Aricciaetal.(2013),andBorio(2014)).10
To measure the impacts of monetary policy on the financial cycle, we calculate the
differencebetweentheactualpolicyrateandtheimpliedpolicyrateusingtheTaylorrule.
The literature generally considers the Taylor rule as an accurate approximation of the
monetary policy rate decisions in modern times (Hofmann and Bogdanova, 2012). It
mechanically links policy rates to the deviations in the inflation rate and the output gap.
Therefore, the implied Taylor rate is often used as a yardstick to gauge the stance of the
monetarypolicy.Wefollowtheclassicalsimpleformulationofthe Taylor(1993)rule(e.g.
Orphanides(2007)):
‫ݎ‬௜ǡ௧ ൌ ʹ ൅ ߨ௜ǡ௧ ൅ ͲǤͷ൫ߨ௜ǡ௧ െ ʹ൯ ൅ ͲǤͷ‫ݕ‬௜ǡ௧ ,
where”୧ǡ୲ istheimpliedkeypolicyratebytheTaylorruleforcountryiinyeart,Ɏ୧ǡ୲ isthe
inflationrateandthecurrentoutputgapisrepresentedby›୧ǡ୲ .
ߝ௜ǡ௧ ൌ ݅௜ǡ௧ െ ‫ݎ‬௜ǡ௧ Thedifferencebetweentheactualpolicyrate‹୧ǡ୲ andtheimpliedpolicyratebytheTaylor
rule – ୧ǡ୲ are the Taylor rule residuals ɂ୧ǡ୲ . In general, negative (positive) deviations are
associated with looser (tighter) monetary policy in a given jurisdiction (Hofmann and
Bogdanova,2012).11
10
Smets(2014)providesadetailedoverviewoftheliteratureonthistopic.
WeconstructcountryͲspecificshadowpolicyratesimpliedbytheTaylorruleforeachquarter.Therefore,we
calculatealsocountryͲspecificTaylorratesforthecountriesthathaveadoptedtheeuro.
11
ECB Working Paper 1812, June 2015
21
ECB Working Paper 1812, June 2015
22
16
Yes
0.14
Ͳ141.36
34
yes
0.65
Ͳ308.80
29
yes
0.31
Ͳ247.65
0.0008*
0.0100***
0.0057***
No.ofObservation
Constant
AdjustedR2(fromOLS)
BIC
(3)
0.0267**
0.2962***
29
yes
0.45
Ͳ66.26
(3)
(2)
TIMEͲADJUSTED
AMPLITUDEMEASURE
Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Market_Cap/GDP
Bank_Assets/GDP
FX_Loans/Loans
Credit/GDP
FX_Loans/Loans
Foreign_Claims/GDP
TaylorRuleResiduals
0.1592**
16
Yes
0.12
Ͳ32.37
(2)
(1)
0.0048***
(1)
0.1071***
34
yes
0.50
Ͳ81.64
NONͲTIMEͲADJUSTED
AMPLITUDEMEASURE
Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Market_Cap/GDP
Bank_Assets/GDP
FX_Loans/Loans
Credit/GDP
FX_Loans/Loans
Foreign_Claims/GDP
TaylorRuleResiduals
No.ofObservation
Constant
AdjustedR2(fromOLS)
BIC
30
yes
0.37
Ͳ257.42
21
yes
0.57
Ͳ184.61
Ͳ0.0005
34
yes
0.01
Ͳ276.31
0.0904***
0.0068***
20
yes
0.30
Ͳ167.2
0.0604*
0.0085***
(7)
0.0023**
0.0052***
(6)
20
yes
0.49
Ͳ54.08
Ͳ0.0188*
34
Yes
0.07
Ͳ60.69
(5)
21
yes
0.75
Ͳ65.94
(7)
(4)
30
yes
0.41
Ͳ65.53
(6)
1.2246**
0.1843***
0.3247*
0.2188***
0.0561**
0.1381***
(5)
(4)
34
yes
0.64
Ͳ305.27
(8)
0.0048***
Ͳ0.0002
(8)
0.1071***
Ͳ0.0113
34
yes
0.51
Ͳ79.63
Ͳ0.0003
16
yes
0.08
Ͳ138.76
0.0057***
(9)
Ͳ0.0144
16
yes
0.11
Ͳ30.67
0.1656*
(9)
Table7:ImpactofMonetaryPolicyontheFinancialCycle
Ͳ0.0002
29
yes
0.31
Ͳ245.18
0.0009*
0.0090***
(10)
Ͳ0.0154
29
yes
0.5
Ͳ66.68
0.0307***
0.2482***
(10)
Ͳ0.0003
30
yes
0.36
Ͳ254.35
0.0023**
0.0050***
(11)
Ͳ0.0107
30
yes
0.43
Ͳ63.99
0.0541**
0.1309***
(11)
1.4420***
0.1843***
Ͳ0.0089
20
yes
0.49
Ͳ52.25
(13)
0.1042***
0.0068***
Ͳ0.0001
20
yes
0.3
Ͳ165.55
(13)
*p<.1,**p<.05,***p<.01
0.0695*
0.0082***
Ͳ0.0006
21
yes
0.56
Ͳ182.22
(12)
*p<.1,**p<.05,***p<.01
Ͳ0.0041
21
yes
0.74
Ͳ63.27
0.4418*
0.2151***
(12)
WeobtaintheTaylorruleresidualsߝ௜ǡ௧ for11Europeancountries,startingfrom1990.12The
inflationrateandtheoutputgapdataaresourcedthroughIMFWEOdatabaseandactual
keypolicyratesareobtainedviaHaverAnalytics.AftercalculatingtheTaylorruleresiduals,
we incorporate them into the model. We calculate the median value of the Taylor rule
residualsinthecorrespondingfinancialcyclephase.13WeemploytheGeneralLinearModel
with robust standards errors with both the nonͲadjusted (‹›…ŽŠƒ•‡’୅ ) and the
timeͲadjustedphaseamplitudemetric(‹›…ŽŠƒ•‡’୆ )forthefinancialcyclemeasure.
TheupperpanelofTable7showstheresultsforthenonͲtimeadjustedamplitudemetricof
thefinancialcyclephases(‹›…ŽŠƒ•‡’୅ ),whilethelowerpanelincludestheresults
for the timeͲadjusted amplitude metric (‹›…ŽŠƒ•‡’୆ ). The regressions of both
panels provide similar conclusions. In each panel, the first six columns represent the
baselineresultsforthestructuralbankingfeatures.Thesespecificationsremainrobustfor
theselectedsubͲperiod(Columns1Ͳ6).Model7representstheinfluenceoftheTaylorrule
residuals without accounting for structural banking sector characteristics. The monetary
policyindicatorturnsouttobesignificantforthenonͲtimeadjustedmeasures(upperpanel
of Table 7). As suggested by the literature, the Taylor rule residuals have a negative sign,
implyingthatloosermonetarypolicyinflatestheamplitudeofthefinancialcycle.However,
bytakingintoaccountthespeedoftheadjustmentofthefinancialcycleamplitude(lower
panel),theimpactofmonetarypolicybecomesinsignificant.
Thelattersixcolumnsofeachpanel(Columns8Ͳ13)containmodelscoveringtheinfluence
of both the banking characteristic group specifications and the Taylor rule residuals. By
combining both impacts, only the banking sector characteristics are robust across the
specifications. Although the marginal values of the Taylor rule residuals also remain at
comparablelevels,noneoftheseindicatorsturnouttobesignificant.Theseresultssuggest
that banking sector characteristics tend to override the explanatory power of Taylor rule
residuals in explaining the financial cycle amplitude. This finding also indicates that
structuralbankingfeaturesmattermorethanthemonetarypolicystanceforbuildingupof
thefinancialcyclephaseamplitudeoverthemediumterm.
8
Conclusion
In this paper we explore the relationship and potential interactions between certain
structuralfeaturesofthebankingsectorsintheEUMemberStatesandtheperformanceof
thebankingsectorsoverthefinancialcycle.Overall,ouranalyticalfindingsprovideevidence
12
We need to restrict the investigation to this subͲperiod due to data constraints. In detail, we include the
following countries: Belgium, Denmark, Finland, France, Germany, Ireland, Italy, the Netherlands, Spain,
SwedenandtheUnitedKingdom.ThesecountriesarealsousedinStremmel(2015)todeterminethefinancial
cycle.
13
We also tried to shift the calculation window of the median value of the Taylor residual to account for
potentialtimelagsofmonetarypolicy(6to8quarters).Theresultsremainsimilartothoseprovidedinthis
section.
ECB Working Paper 1812, June 2015
23
that structural banking sector characteristics do influence the amplitude of the financial
cycle. We find robust results across the variable groups using both different estimation
techniques and different financial cycle measures. The robustness checks as well as the
specificationtestsconfirmthechoiceofourvariables.Thestructuralcharacteristicsofthe
banking sector, such as the concentration of the banking sector and the share of foreign
banksaswellastheamountandcompositionofbanksloansandfinancialintegrationseem
tobeimportantdriversofthefinancialcycleamplitude.
Besides these influencing factors, the depth of financial intermediation and the size and
stabilityoffinancialinstitutionsshowweakerimpactsontheamplitude.Wealsofindthat
monetary policy contributes to the financial cycle amplitude, but the banking sector
characteristicstendtooverridetheexplanatorypowerofmonetarypolicystance.
We believe that our findings also contribute to the onͲgoing discussion on the
implementation of macroͲprudential policy measures, in particular as regards certain
structural and cyclical policy instruments. Based on the identified differences in financial
cyclesacrossEUcountriesaswellastheimpactofcertainstructuralbankingcharacteristics
ontheamplitudeofthefinancialcycle,wecanconcludethattheimplementationofmacroͲ
prudential measures should be differentiated across EU Member States. The timing of
activationandtherelativecalibrationofthepolicymeasuresshouldtakeintoconsideration
thedifferencesbothinfinancialcyclesandbankingstructures.
In particular, our results suggest that the activation and calibration of structural policy
measures,suchasthesystemicriskbuffer(SRB),shouldbemindfulofthecyclicalpositionof
thebankingsystem.Ontheonehand,ifastructuralmeasureisactivatedandphasedͲinina
boomperiodtoaddressstructuralrisks,itmay,atthesametime,alsomitigatetheupward
swingsinthefinancialcycle,inparticularifitcoincideswiththeimplementationofcounterͲ
cyclical measures, such as the counterͲcyclical buffer (CCB). On the other hand, if a
structural measure is activated in a recessionary phase, it may counteract other cyclical
policymeasures,suchasthereleaseoftheCCB.
The regression results also confirm the intuition that the activation and calibration of
counterͲcyclicalpolicymeasures(e.g.CCB)shouldnotonlydependonthecyclicalsituation
ofthebankingsector,butitshouldalsotakeintoconsiderationthestructuralcharacteristics
ofthebankingsystemsinindividualMemberStates.Concretely,intheabsenceofstructural
measures in place, in countries where the banking sector is more concentrated, more
integratedand/ordominatedbyforeignbanksandforeigncurrencylending,thecalibration
oftheCCBmayneedtobemorestringent,giventhatthosebankingsystemsarefoundtobe
more exposed to cyclical swings. However, if systemic risk buffers or other structural
measuresareinplace,thesemeasuresmayalsocontributetoreducingtheamplitudeofthe
cycle,providedthattheunderlyingstructuralrisksareaddressedeffectively.
ECB Working Paper 1812, June 2015
24
Nonetheless, further analyses are needed to achieve a better understanding of the
combined impact of cyclical and structural policy measures that may ultimately have an
impactontheirrelativecalibrationandthepropertimingoftheiractivation.
ECB Working Paper 1812, June 2015
25
References
ADRIAN, T. and H.S. SHIN (2008): “Financial Intermediaries, Financial Stability, and Monetary Policy”, in
MaintainingStabilityinaChangingFinancialSystem,ProceedingsoftheFederalReserveBankofKansas
CityEconomicSymposium,JacksonHole,Wyoming,August21–23.
AIKMAN,D.,A.G.HALDANEandB.NELSON(2010):“CurbingtheCreditCycle”,SpeechgivenattheColumbia
UniversityCenteronCapitalismandSocietyAnnualConference,NewYork,November.
AIKMAN, D., A.G. HALDANE and B. NELSON (2014): “Curbing the Credit Cycle”, The Economic Journal,
forthcoming.
AIYAR,S.(2012):“FromFinancialCrisistoGreatRecession:TheRoleofGlobalizedBanks”,AmericanEconomic
Review,102,225–30.
AIZENMAN,J.,B.PINTOandS.VLADYSLAV(2013):"FinancialSectorUpsandDownsandtheRealSectorinthe
OpenEconomy:UpbytheStairs,DownbytheParachute",EmergingMarketsReview,16(C),1–30.
ALTUNBAS, Y., L. GAMBACORTA and D. MARQUÉZͲIBAÑEZ (2010): “Does Monetary Policy Affect Bank RiskͲ
Taking?”,ECBWorkingPaper,No1166.
ARCAND,J.–L.,E.BERKESandU.PANIZZA(2012):“TooMuchFinance?”,IMFWorkingPaper,NoWP/12/161.
BASSO,H.,O.CALVO–GONZALEZandM.JURGILAS(2010):“FinancialDollarization:TheRoleofForeign–Owned
BanksandInterestRates“,JournalofBankingandFinance,35,794–806.
BECK,T.(2008):“BankCompetitionandFinancialStability:FriendsorFoes?”,WorldPolicyResearchWorking
Paper,No4656.
BERGER, A. N., A. DEMIRGÜC–KUNT, R. LEVINE and J. G. HAUBRICH (2004): “Bank Concentration and
Competition:AnEvolutionintheMaking”,JournalofMoney,CreditandBanking,36,434–50.
BERGER,A.N.,L.F.KLAPPERandR.TURK–ARISS(2009):“BankCompetitionandFinancialStability”,Journalof
FinancialServicesResearch,35,99–118.
BOLTON,P.,X.FREIXAS,L.GAMBACORTAandP.MISTRULLI(2013):"RelationshipandTransactionLendingOver
theBusinessCycle",BISWorkingPaper,No417.
BORIO,C.(2013):“MacroprudentialPolicyandtheFinancialCycle:SomeStylizedFactsandPolicySuggestions”,
Speech given at the “Rethinking Macro Policy II: First Steps and Early Lessons” hosted by the IMF in
Washington,DC.
BORIO,C.(2014):"MonetaryPolicyandFinancialStability:WhatRoleinPreventionandRecovery?",BISWorking
Paper,No440.
BOYD,J.andG.DENICOLO(2005):"TheTheoryofBankRiskTakingRevisited",JournalofFinance,60(3),1329–
43.
BOYD,J.,G.DENICOLOandA.M.JALAL(2006):"BankRiskTakingandCompetitionRevisited:NewTheoryand
Evidence",IMFWorkingPaper,NoWP/06/297.
BRACKE,P.(2013):"HowLongDoHousingCyclesLast?ADurationAnalysisfor19OECDCountries",Journalof
HousingEconomics,22,213–30.
BROWN, M. and R. DE HAAS (2012): “Foreign Banks and Foreign Currency Lending in Emerging Europe”,
EconomicPolicy,27(69),57–98.
BROWN,M.,K.KIRSCHENMANNandS.ONGENA(2010):“ForeignCurrencyLoans:DemandorSupplyDriven?“,
CEPRDiscussionPaper,No7952.
BUSH,O.,R.GUIMARAESandH.STREMMEL(2014):“BeyondtheCreditGap:QuantityandPriceofRiskIndicators
forMacro–PrudentialPolicy”,BankofEngland,Manuscript.
CECCHETTI,S.G.andE.KHARROUBI(2012):“ReassessingtheImpactofFinanceonGrowth“,BISWorkingPaper,
No381.
CHRISTIANO,L.andT.FITZERALD(2003):“TheBand–PassFilter”,InternationalEconomicReview,44(2),435–65.
CIHAK,M.,A.DEMIRGÜÇ–KUNT,E.FEYEN,andR.LEVINE(2013):“BenchmarkingFinancialSystemsaroundthe
World”,NBERWorkingPaper,No18946.
CLAESSENS,S.,M.KOSEandM.TERRONES(2011a):“FinancialCycles:What?How?When?”,IMFWorkingPaper,
NoWP/11/76.
CLAESSENS, S., M. KOSE and M. TERRONES (2011b): “How Do Business and Financial Cycles Interact?”, IMF
WorkingPaper,NoWP/11/88.
CLAESSENS,S.,andN.VANHOREN(2014):“ForeignBanks:TrendsandImpact”,JournalofMoney,Creditand
Banking,46(s1),295–326.
DEHAAS,R.andN.VANHOREN(2013):“RunningfortheExit?InternationalBankLendingduringaFinancial
Crisis”,ReviewofFinancialStudies,26,244–85.
ECB Working Paper 1812, June 2015
26
DEHAAS,R.andN.VANLELYVELD(2014):“MultinationalBanksandtheGlobalFinancialCrisis:Weatheringthe
PerfectStorm?”,JournalofMoney,CreditandBanking,46(2),333–64.
DENICOLO,G.andE.LOUKOIANOVA(2007):“BankOwnership,MarketStructureandRisk”,IMFWorkingPaper,
NoWP/07/215.
DEGRYSE, H., M. A. ELAHI, and M. F. PENAS (2013): “Determinants of Banking System Fragility: A Regional
Perspective”,ECBWorkingPaper,No1567.
DELL’ARICCIA,G.,L.LAEVEN,andG.SUAREZ(2013):“BankLeverageandMonetaryPolicy’sRiskͲTakingChannel:
EvidencefromtheUnitedStates”,IMFWorkingPaper,NoWp/13/143.
DETKEN,C.,O.WEEKEN,L.ALESSI,D.BONFIM,M.M.BOUCINHA,C.CASTRO,S.FRONTCZAK,G.GIORDANA,J.
GIESE,N.JAHN,J.KAKES,B.KLAUS,J.H.LANG,N.PUZANOVAandP.WELZ(2014):“Operationalizingthe
Countercyclical Capital Buffer: Indicator Selection, Threshold Identification and Calibration Options”,
ESRBOccasionalPaper,5.
DREHMANN,M.,C.BORIOandK.TSATSARONIS(2012):“CharacterisingtheFinancialCycle:Don'tLoseSightof
TheMediumTerm!”,BISWorkingPaper,No380.
DREHMANN, M. and K. TSATSARONIS (2014): “The Credit–to–GDP Gap and Countercyclical Capital Buffers:
QuestionsandAnswers”,BISQuarterlyReview,March2014.
EUROPEANSYSTEMICRISKBOARDAdvisoryScientificCommittee(ESRBASC)(2014),“IsEuropeOverbanked?”,
ASCReport,No4,June2014.
FUNGÁCOVÁ,Z.,R.HERRALA,andL.WEILL(2013):“TheInfluenceofBankOwnershiponCreditSupply:Evidence
fromtheRecentFinancialCrisis”,EmergingMarketsReview,15,136–47.
GAMBACORTAL.,J.YANGandK.TSATSARONIS(2014):“FinancialStructureandGrowth”,BISQuarterlyReview,
March2014.
GERDESMEIER,D.,H.–E.REIMERSandB.ROFFIA(2010):"AssetPriceMisalignmentsandtheRoleofMoneyand
Credit",InternationalFinance,13,377–407.
HOFMANN,B.andB.BOGDANOVA(2012):“Taylorrulesandmonetarypolicy:aglobalgreatdeviation?”,BIS
QuarterlyReview,September(2012),pp.37–49
HOUBEN,A.andJ.KAKES(2013):“FinancialImbalancesandMacroprudentialPolicyinaCurrencyUnion”,DNB
OccasionalStudies,11,5.
LUCA,A.andI.PETROVA(2008):“WhatDrivesCreditDollarizationinTransitionEconomies?“,JournalofBanking
&Finance,32,858–69.
MADDALONI,A.andJ.ͲL.PEYDRÓ(2011):“BankRiskͲtaking,Securitization,Supervision,andLowInterestRates:
EvidencefromEuroͲareaandU.S.LendingStandards”,ReviewofFinancialStudies,24(6),2121–65.
ORPHANIDES,A(2007):“TaylorRules”,FinanceandEconomicsDiscussionSeries2007Ͳ18,BoardofGovernors
oftheFederalReserveSystem(U.S.).
POPOV,A.andG.F.UDELL:(2012):“Cross–BorderBanking,CreditAccess,andtheFinancialCrisis”,Journalof
InternationalEconomics,87,147–61.
SCHAECK, K., M. CIHAK, and S. WOLFE (2009): “Are Competitive Banking Systems More Stable?”, Journal of
Money,Credit,andBanking,41(4),711–34.
SMETS,F.(2014):“FinancialStabilityandMonetaryPolicy:HowCloselyInterlinked?”,InternationalJournalof
CentralBanking,10(2),263Ͳ300.
STREMMEL,H.(2015):“CapturingtheFinancialCycleinEurope”,ECBWorkingPaperSeries,No1811.
TAYLOR,J.B.(1993):“DiscretionversusPolicyRulesinPractice”,CarnegieͲRochesterConferenceSeriesonPublic
Policy39,December,195Ͳ214.
UHDE, A. and U. HEIMESHOFF (2009): "Consolidation in Banking and Financial Stability in Europe: Empirical
Evidence",JournalofBanking&Finance,33(7),1299–1311.
ECB Working Paper 1812, June 2015
27
Appendix
FigureA1:FinancialCyclePhases
Belgium
Denmark
.1
-.05
-.1
-.03
-.04
-.02
-.02
-.01
0
0
0
.05
.02
.01
.02
.04
Austria
1980q1
1990q1
2000q1
Synthetic Financial Cycle
Turning point (peak)
2010q1
1980q1
Turning point (trough)
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
1980q1
Turning point (trough)
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
Turning point (trough)
France
Germany
1980q1
1990q1
2000q1
Synthetic Financial Cycle (FC3)
Turning Point (Peak)
2010q1
1980q1
Turning Point (Trough)
-.05
-.1
-.05
-.05
0
0
0
.05
.05
.05
.1
.1
Finland
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
1980q1
Turning point (trough)
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
Turning point (trough)
Hungary
Ireland
1980q1
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
1980q1
Turning point (trough)
.1
.05
0
-.05
-.1
-.05
-.02
0
0
.02
.05
.04
.1
.06
.15
Greece
1990q1
Synthetic Financial Cycle
2010q1
1980q1
Turning point (trough)
1990q1
Synthetic Financial Cycle
2010q1
Turning point (trough)
Lithuania
.3
-.2
-.4
-.2
-.1
-.1
-.05
0
0
0
.1
.2
.2
.05
2000q1
Turning point (peak)
.4
Latvia
.1
Italy
2000q1
Turning point (peak)
1980q1
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
1980q1
Turning point (trough)
1990q1
Synthetic Financial Cycle
2010q1
1980q1
Turning point (trough)
1990q1
Synthetic Financial Cycle
2010q1
Turning point (trough)
Netherlands
.04
1980q1
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
Turning point (trough)
ECB Working Paper 1812, June 2015
1980q1
-.05
-.1
-.04
-.02
-.05
0
0
.05
.02
.05
0
2000q1
Turning point (peak)
.1
Malta
.1
Luxembourg
2000q1
Turning point (peak)
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
Turning point (trough)
1980q1
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
Turning point (trough)
28
Slovakia
1980q1
1990q1
2000q1
Synthetic Financial Cycle
Turning point (peak)
2010q1
.05
-.05
-.2
-.04
-.1
-.02
0
0
0
.1
.02
.04
.1
Portugal
.2
Poland
1980q1
Turning point (trough)
1990q1
2000q1
Synthetic Financial Cycle
2010q1
Turning point (peak)
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
Turning point (trough)
United Kingdom
-.1
-.1
-.1
-.05
-.05
-.05
0
0
0
.05
.05
.05
.1
.1
.1
Sweden
.15
Spain
1980q1
Turning point (trough)
1980q1
1990q1
2000q1
Synthetic Financial Cycle
Turning point (peak)
2010q1
1980q1
Turning point (trough)
1990q1
2000q1
Synthetic Financial Cycle
2010q1
Turning point (peak)
1980q1
Turning point (trough)
1990q1
Synthetic Financial Cycle
2000q1
Turning point (peak)
2010q1
Turning point (trough)
Source:Stremmel(2015)
Each of the 21 country panels reflects the financial cycle with the identified turning points over time. The financial cycle measure is
borrowedfromStremmel(2015)).Turningpointsaretheresultofavisualinspectionofthefinancialcycletimeseriesforeachcountry.The
determination of the local minima and maxima of each cycle allows us to define the peaks and troughs of the financial cycle and to
calculatetheamplitude.Thefinancialcyclephaselastsfromthelastfromturningpointtothenextoneandcorrespondstoanexpansionor
contractionphaseofthefinancialcycle.Therefore,anupswingperiod(expansionphase)ofthefinancialcyclemeasurewillendurefroma
trough to peakpoint and, vice versa, a downswing period (contractionphase)lasts from a peak to a trough pointof the financial cycle
measure.Unfortunately,forsomecountries(e.g.Greece,LatviaorSlovakia)wefacedataconstraintsandthereforewemaynotbeableto
captureafullfinancialcycle.Thisfactalsoprovidesargumentsforusingfinancialcyclephasesinsteadoffullfinancialcycles.Foradetailed
interpretationofthesecountrypanelspleaseseeSection4inthepaper.
TableA1:FinancialCycleMeasures
FinancialCycle
FC1
FC2
FC3
FC4
FC5
FC6
FC7
Ingredients
CreditͲtoͲGDPratio
CreditͲtoͲGDPratio,Housepricestoincomeratio
CreditͲtoͲGDPratio,Housepricestoincomeratio,Creditgrowth
CreditͲtoͲGDPratio,Housepricestoincomeratio,Creditgrowth,Housepricegrowth
CreditͲtoͲGDPratio,Housepricestoincomeratio,Creditgrowth,Bankfundingratio
CreditͲtoͲGDPratio,Housepricestoincomeratio,Creditgrowth,Banknetincometototalassets
CreditͲtoͲGDPratio,Housepricestoincomeratio,Creditgrowth,Loanstototalassets
Source:Stremmel(2015)
ThistableexhibitsvarioussyntheticfinancialcyclemeasureconsideredinStremmel(2015)todeterminethefinancialcyclemeasure.Fora
detailed description and review of the underlying components in the financial cycle see Stremmel (2015). The ingredients are obtained
usingfrequencyͲbasedfiltertechniquestoisolatecyclicalmovementsfromthetrendineachoftheunderlyingtimeseries.Thefinancial
measuresrepresentthecombinationofindividualcyclicalingredients.
ECB Working Paper 1812, June 2015
29
Bankloans
(Start:1997Q3)
2
5
3
5
3
2
2
1
1
3
1
1
2
2
3
1
2
1
1
3
1
4
2
4
4
4
2
3
3
3
2
3
2
3
2
2
2
1
1
2
1
1
1
1
1
1
3
26
62
51
2
5
4
5
3
3
2
2
2
3
2
2
2
2
3
2
4
2
3
4
3
1
3
2
3
1
1
1
1
єSample
60
1
1
1
1
SumofObservations
Financialdepth
(Start:1989Q4)
2
6
5
6
5
2
2
1
4
5
1
1
2
2
4
Austria
Belgium
Denmark
Finland
France
Germany
Greece
Hungary
Ireland
Italy
Latvia
Lithuania
Luxembourg
Malta
Netherlands
Poland
Portugal
Slovakia
Spain
Sweden
UnitedKingdom
Financialintegration
(Start:1998Q3)
Institutionsizeandstability
(Start:1981Q1)
Marketshareofforeignbanks
(Start:1996Q1)
Concentrationofthebankingsestor
(Start:1988Q4)
Country
TableA2:CountryͲlevelDataAvailabilityforEachPhaseandIndicatorGroup
2
3
2
3
2
2
2
1
1
2
1
1
2
2
1
1
1
1
3
1
1
2
1
11
25
18
25
16
12
11
7
9
16
7
7
8
8
12
5
20
8
14
18
9
34
33
266
Thistableprovidesanoverviewoftheavailabilityoftheindicatorsintermsoffinancialcyclephasesforeachvariablegroupatthecountry
level.ForSample2,weareabletoinclude266phaseͲobservations.ItisobviousthatfornewEUmemberstatessuchasHungary,Latviaor
Lithuaniathedatahistoryisratherlimited.ThecategoriesthatprovidethebestcoveragewiththelongesttimehorizonsareInstitutionsize
andstabilityandConcentration.Inaddition,itisalsoinsightfultonotethatforsomecountriessuchasIreland,Luxembourg,Malta,the
Netherlands,Poland,andtheUnitedKingdomcertainexplanatoryvariablesarenotavailable.Thisismainlyduetothelackofdataonthe
Shareofforeignbanks.AfurtherremarkconcernsthecategoriesofBankloansandFinancialintegration.Bothdataseriesstartonlyatthe
endofthe1990sandthereforetheoverallnumberofobservationsisrathersmall.
ECB Working Paper 1812, June 2015
30
TableA3:CorrelationMatrixofVariablestoFinancialCyclePhases
Correlation
Concentration
Foreign_Banks
Credit/Deposits
Deposits/GDP
Market_Cap/GDP
Bank_Assets/GDP
FX_Loans/Loans
Credit/GDP
Foreign_Claims/GDP
ࡲ࢏࢔࡯࢟ࢉ࢒ࡼࢎࢇ࢙ࢋ࡭࢓࢖࡭ 0.69
0.58
0.41
0.41
0.20
0.41
0.44
0.50
0.52
ࡲ࢏࢔࡯࢟ࢉ࢒ࡼࢎࢇ࢙ࢋ࡭࢓࢖࡮ 0.70
0.60
0.40
0.31
0.17
0.46
0.57
0.45
0.43
ThistableexhibitsthePearsoncorrelationsofbankingsectorcharacteristicsandbothfinancialcyclephaseamplitudemeasures.Thetable
showsthatConcentrationandtheShareofforeigncurrencyloanshavehighcorrelations.Otherindicators,suchasthevariablesofFinancial
depth are associated with a lower correlation to the financial cycles. All measures, except the Market_Cap/GDP ratio, are statistically
significantlycorrelatedatthe5%confidencelevel.
ECB Working Paper 1812, June 2015
31
Acknowledgements
The authors are very grateful to Sophia Chen, Luis Gutierrez de Rozas, Benjamin Klaus, Tuomas Peltonen, Willem Schudel, Peter
Welz, Bent Vale as well as Benjamin Nelson and Jonathan Frost for their useful comments on an earlier version of this paper. We also
thank seminar participants at the European Central Bank (2014), 6th IFABS conference (2014), 46th MMF conference (2014), 11th
International Conference Western Economic Association International (2015) as well as an anonymous referee for valuable comments
and suggestions. The views expressed in this paper are those of the authors and do not necessarily reflect the views of the ECB or the
Eurosystem.
Hanno Stremmel
WHU – Otto Beisheim School of Management;
e-mail: [email protected]
Balázs Zsámboki
European Central Bank;
e-mail: [email protected]
© European Central Bank, 2015
Postal address
Telephone
Internet
60640 Frankfurt am Main, Germany
+49 69 1344 0
www.ecb.europa.eu
All rights reserved. Any reproduction, publication and reprint in the form of a different publication, whether printed or produced
electronically, in whole or in part, is permitted only with the explicit written authorisation of the ECB or the authors.
This paper can be downloaded without charge from www.ecb.europa.eu, from the Social Science Research Network electronic library at
http://ssrn.com or from RePEc: Research Papers in Economics at https://ideas.repec.org/s/ecb/ecbwps.html.
Information on all of the papers published in the ECB Working Paper Series can be found on the ECB’s website,
http://www.ecb.europa.eu/pub/scientific/wps/date/html/index.en.html.
ISSN
ISBN
DOI
EU catalogue number
1725-2806 (online)
978-92-899-1625-7
10.2866/176942
QB-AR-15-052-EN-N