are we riding the waves of a global financial cycle in the
TRANSCRIPT
![Page 1: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/1.jpg)
1NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Are we riding the waves of a global financial cycle in the euro area ?
N. Cordemans J. Tielens Ch. Van Nieuwenhuyze *
Introduction
Followingâ financialâ liberalisation,â deregulationâ andâ innovations,â financialâ marketsâ haveâ becomeâ significantlyâmoreâintegratedâsinceâtheâ1990s.âThisâisâtheâcaseâforâbothâemergingâandâadvancedâeconomies.âVariousâauthorsâ(Rey, 2015â;âMiranda-AgrippinoâandâRey, 2019â;âHabibâandâVenditti, 2019)âhaveâfoundâthatâthisâhasâcontributedâtoâ theâ emergenceâ ofâ aâ âglobalâ financialâ cycleâ.â Theâ conceptâ broadlyâ refersâ toâ theâ ideaâ thatâ fluctuationsâ inâfinancialâmarketsâoccurâonâaâglobalâscale,âconsistingâinâco-movementsâofâcross-borderâcapitalâflows,âassetâprices,âcreditâflowsâandâleverageâacrossâcountries.
Thisâarticleârelatesâtoâtheâburgeoningâliterature 1âonâtheâimportanceâofâtheâglobalâfinancialâcycleâ(GFC)âthatâhasâsoâfarâmainlyâ focusedâonâtheâeffectsâofâ theâGFCâonâcapitalâflowsâofâemergingâmarkets.âWeâcontributeâtoâthisâliteratureâbyâanalysingâtheâimpactâofâtheâGFCâonâdomesticâfinancialâconditionsâinâtheâeuroâareaâcountries.â
Ourâresultsâshowâthatâdomesticâfinancialâconditionsâ inâtheâeuroâareaâare,âonâaverage,âstronglyâcorrelatedâwithâaâmeasureâofâtheâglobalâfinancialâcycle.âFurthermore,âweâlinkâtheâcross-countryâsensitivityâtoâtheâglobalâcycleâtoâvariousâdeterminants,âincludingâtheâsizeâandâtheâcompositionâofâtheâexternalâfinancialâposition.âAâkeyâfindingâisâthatâsensitivityâ toâ theâGFCâdependsâonâtheânetâ internationalâ investmentâposition.âCountriesâwithânetâ liabilitiesâseemâtoâreactâtwiceâasâstronglyâtoâtheâGFCâasâcountriesâthatâhaveânetâassets.
Severalâ policyâ implicationsâ canâ beâ drawnâ fromâ theseâ findings.â First,â theâ strongâ correlationâ betweenâ financialâconditionsâinâtheâeuroâareaâandâtheâglobalâfinancialâcycleâmakesâitâusefulâforâmacroprudentialâpolicyâtoâmonitorâthisâglobalâcycleâandâ/âorâtoâhelpâaddressâextremeâsensitivityâtoâitsâboomâ/âbustâprofile.âSecondlyâââandâimportantlyâinâviewâofâ theâcurrentâdebateâ inâtheâ literatureâââthisâcorrelationâtendsâtoâsuggestâ theâpresenceâofâaââfinancialâdilemmaââinâtheâeuroâarea,âalongâtheâlinesâofâReyâ(2015)âforâemergingâeconomies.âSuchâaâdilemmaâimpliesâthatâwheneverâ theâfinancialâ accountâ isâopen,âmonetaryâandâfinancialâ conditionsâareâ largelyâ inâ theâhandsâofâglobalâfactorsâandâ lessâ inâ thoseâofâanâ independentâmonetaryâpolicy.âWeâshowâthatâ thisâdilemmaâ inâ theâeuroâareaâ isâparticularlyâpresentâwhenâcountriesâhaveâaânegativeânetâexternalâposition.âThisâcallsâforâco-ordinationâbetweenâ
*â WeâwouldâlikeâtoâthankâP.âButzenâandâP.âIlbasâforâusefulâcommentsâandâsuggestions.âWeâareâveryâgratefulâtoâM. HabibâandâF. Venditti forâsharingâtheirâmeasureâofâtheâglobalâfinancialâcycleâandâtoâH.âDewachterâforâdevelopingâanâearlierâversionâofâtheâfinancialâconditionsâindexâusedâinâthisâarticle.
1 Theâconceptâofâtheââglobalâfinancialâcycleââwasâintroducedâinâthe 2015âReyâpaperâandâpresentedâatâthe 2013 KansasâCityâFEDâJacksonâHoleâSymposium.âFollow-upâworkâwasâpresentedâatâthe 2014âIMFâMundell-Flemingâlecture.âTheâpaperâattractedâattentionâandâresponsesâfromâacademicsâandâpolicymakers,âsuchâasâB.âBernankeâatâthe 2015âIMFâMundell-Flemingâlecture.âAâgrowingâliteratureâfollowed,âconcentratingâonâevidenceâinâfavourâofâorâagainstâtheâglobalâfinancialâcycle.
![Page 2: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/2.jpg)
2NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
macroeconomicâ (structural),â macroprudentialâ andâ monetaryâ policyâ inâ theâ euroâ areaâ soâ thatâ theirâ respectiveâobjectives can be better attained.
Theâremainderâofâthisâarticleâisâstructuredâasâfollows.âFirst,âweâreviewâtheâcurrentâliteratureâonâtheâglobalâfinancialâcycleâandâitsâimplications.âInâsection 2 weâconstruct,âforâtheâeuroâareaâcountries,âaâcompositeâmeasureâofâtheirâdomesticâfinancialâconditionsâ(FinancialâConditionsâIndexâââFCI)âandâanalyseâtoâwhatâextentâtheâFCIâisâcorrelatedâwithâtheâglobalâfinancialâcycle.âSection 3 shedsâsomeâ lightâonâcross-countryâheterogeneityâ inâsensitivityâ toâ theâGFCâwhichâweâlinkâtoâvariousâdeterminants,âincludingâtheâsizeâandâcompositionâofâtheâexternalâfinancialâposition.âSection 4 presentsâourâmethodologyâandâempiricalâresults.âGivenâtheseâresults,âweâevaluateârecentâdevelopmentsâinâsection 5.âSection 6 drawsâseveralâpolicyâimplicationsâbeforeâweâconclude.
1. The global financial cycle : evidence, drivers and implications
Followingâfinancialâliberalisation,âderegulationâandâinnovations,âfinancialâmarketsâhaveâbecomeâsignificantlyâmoreâintegratedâsinceâ theâ1990s.âThisâ isâ theâcaseâ forâbothâemergingâandâadvancedâeconomiesâ (seeâboxâ1).âVariousâauthorsâ(Rey, 2015â;âMiranda-AgrippinoâandâRey, 2019âandâHabibâandâVenditti, 2019)âhaveâfoundâthatâthisâhasâcontributedâtoâtheâemergenceâofâaââglobalâfinancialâcycleââ(GFC).âTheâconceptâbroadlyârefersâtoâtheâ ideaâthatâfluctuationsâ inâ financialâmarketsâ occurâ onâ aâ globalâ scale,â consistingâ inâ co-movementsâ ofâ cross-borderâ capitalâflows,âassetâprices,âcreditâflowsâandâleverageâacrossâcountries.â
Inâ theâ literature,â theâglobalâ financialâ cycleâ isâ inâgeneralâ proxiedâbyâ theâ commonâ componentâofâ aâ largeâpanelâofâ assetâ returnsâ (e.g.â 858 assetâ priceâ seriesâ inâ Miranda-Agrippinoâ andâ Rey, 2018â;â stockâ marketâ returnsâ inâ63 economiesâinâHabibâandâVenditti, 2019).âItâisâusuallyâshownâtoâbeârelatedâtoâtwoâmainâdriversâ:âtheâdegreeâofâglobalâriskâaversionâandââcentreââcountryâeconomicâpolicies,âinâparticular,âUSâmonetaryâpolicy 1.âTheâlatterâmightâinfluenceâfinancialâconditionsâandâcapitalâflowsâaroundâtheâworldâthroughâtheâinternationalâroleâofâtheâdollarâinâcreditâmarketsâ(BIS, 2017)âandâtheâleverageâofâglobalâbanksâ(BrunoâandâShin, 2015a,b).â
Theâliteratureâonâtheâeffectsâofâtheâfinancialâcycleâhasâconcentratedâonâtheâimpactâonâcapitalâflowsâandâdomesticâfinancialâconditions.âExamplesâofâtheâformerâincludeâcontributionsâbyâHabibâandâVendittiâ(2019)âandâDavis et al. (2019).â Theseâ contributionsâ confirmâ theâ findingsâ ofâ Forbesâ andâWarnockâ (2012)â stressingâ theâ roleâ ofâ globalâfactors,â suchâ asâ USâ interestâ ratesâ orâ globalâ investorsââ riskâ aversionâ inâ internationalâ grossâ capitalâ flows,â andâepisodesâofâextremeâcapitalâflows.âHabibâandâVendittiâ(2019)âpointâoutâthatââfinancialââshocksâmatterâmoreâthanâUSâmonetaryâpolicy,âwhileâDavis et al.â(2019)âfindâthatâglobalâfactorsâalsoâdetermineânetâcapitalâflows.âAlong theâsameâ lines,â Avdjiev et al.â (2018)â highlightâ theâ importanceâ ofâ distinguishingâ capitalâ flowsâ acrossâ financingâinstrumentsâandâsectors.âMostâofâtheâresearchâfindsâevidenceâofâaâglobalâcycleâinâcapitalâflows,âinâparticularâforâemergingâmarketsâ(Ghosh,âQureshi,âKimâandâZalduendo, 2014).âTheseâfindingsâhaveâbeenâsomehowâchallengedâbyâCerutti,âClaessensâandâRoseâ(2017),âwhoâindicateâthatâglobalâfactorsâdoânotâexplainâmoreâthanâ25 perâcentâofâcapitalâflowâvariationsâacrossâcountries.
AlthoughâtheâimpactâofâtheâGFCâonâdomesticâfinancialâconditionsâformsâpartâofâtheâoriginalâanalysisâbyâRey (2015),âtheâliteratureâonâthatâsubjectâisâscarcer.âApartâfromâtheâcontributionsâbyâReyâ(2015),âObstfeld et al. (2017)âalsoâlookâ intoâ theâ transmissionâ ofâ globalâ factorsâ toâ domesticâ financialâ andâmacroeconomicâ outcomes.â Again,â theâlargestâeffectsâareâfoundâforâemergingâcountries.
TheâanalysisâofâtheâsensitivityâofâdomesticâfinancialâconditionsâtoâglobalâfactorsâisâcloselyârelatedâtoâtheâdiscussionâregardingâtheâvalidityâofâtheâclassicalâMundell-Flemingââtrilemmaââinâinternationalâeconomics,âwhichâpostulatesâthatâ countriesâ faceâ aâ trade-offâ amongstâ theâ objectivesâ ofâ exchangeâ rateâ stability,â freeâ capitalâ mobilityâ andâ
1â Theâglobalâfinancialâcycleâisâsometimesâalsoâlinkedâtoâconventionalâmeasuresâofâinvestorsââriskâaversion,âsuchâasâtheâVIX.âNoteâthatâthisâmeasureâratherâcapturesâoneâofâtheâdriversâofâtheâglobalâfinancialâcycleâandânotâtheâcycleâasâsuch.
![Page 3: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/3.jpg)
3NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
independentâmonetaryâpolicy 1â(i.e.âinâaâworldâofâfreeâcapitalâmobility,âtoârunâanâindependentâmonetaryâpolicyâisâfeasibleâifâandâonlyâifâtheâexchangeârateâisâfloating).âAccordingâtoâReyâ(2015),âtheâexistenceâofâaâglobalâfinancialâcycleâ transformsâ thisâ âtrilemmaââ intoâ aâ âdilemmaââ:â runningâ anâ independentâ monetaryâ policyâ orâ allowingâcapitalâ toâflowâ freely.âThus,âwhileâ itâ remainsâ trueâ thatâfixedâexchangeâ ratesâdoânotâallowâ forâanâ independentâmonetaryâpolicy,âcross-borderâcapitalâflowsâwouldâtransmitâtheâmonetaryâpolicyâstanceâofâtheââcentreââeconomyâworldwide,âevenâtoâeconomiesâwithâfloatingâexchange-rateâregimes.âThisâboilsâdownâtoâspill-overâeffectsâofâUSâmonetaryâ policyâ (withâ theâUSâ beingâ theâ âcentreâ)â onâmonetaryâ andâ financialâ conditionsâ inâ otherâ economiesâandâ thusâ limitingâmonetaryâ independence 2â inâ thoseâ countries.â Reyâ (2019)â thereforeâ characterisesâ theâ FEDâ asâaââhegemonâ,âessentiallyâdescribingâtheâFEDâasâtheâdeâfactoâcentralâbankerâofâtheâworld.âOnâtheâotherâhand,âseveralâauthorsâprovideâevidenceâ inâfavourâofâtheâtrilemma,âbasedâonâtheâfindingâthatâfloatingâexchangeâratesâinsulateâ economiesââ monetaryâ andâ domesticâ financialâ conditionsâ fromâ globalâ factorsâ (Shambaugh, 2004â;âObstfeld,â Shambaughâ andâ Taylor, 2005â;â Kleinâ andâ Shambaugh, 2015â;â Obstfeld,â Ostryâ andâ Qureshi, 2017).âSo far,âmostâofâtheâ literatureâhasâ lookedâintoâtheâevidenceâforâEMEs,âasâEMEsâareâ inâprincipleâmoreâsubjectâtoâtheâswingsâofâtheâglobalâcycleâgivenâtheirâdependenceâonâdollarâborrowings.â
Ourâarticleâcontributesâtoâthisâburgeoningâliteratureâinâseveralâways.âWeâaimâtoâfillâinâaâgapâbyâconcentratingâonâtheâeffectsâofâtheâGFCâonâfinancialâconditionsâinâtheâeuroâareaâcountries.âGivenâtheâspecificâfeaturesâofâtheâeuroâarea,âi.e.âtheâsingleâcurrency,âandâtheâhighâdegreeâofâfinancialâ integration,âweâlinkâtheâcross-countryâsensitivityâtoâ theâ globalâ cycleâ toâ variousâ determinants,â includingâ theâ sizeâ andâ theâ compositionâ ofâ cross-borderâ financialâholdings.âFurthermore,âweâanalyseâwhetherâ theâevidenceâ favoursâaâfinancialâ trilemmaâorâdilemmaâ inâ theâeuroâarea.âFinally,âweâdrawâconclusionsâforâtheâvariousâeconomicâpolicyâdomainsâinâtheâeuroâarea.
1â Economicâsystemâconfigurationsâhaveâbeenâdesignedâinâlineâwithâtheââtrilemmaââthroughoutâhistoryâ:âduringâtheâgoldâstandardâ(approximatelyâfromâtheâ1870sâtoâtheâ1930s),âexchangeârateâstabilityâandâfreeâcapitalâmobilityâwereâassured,âatâtheâexpenseâofâmonetaryâautonomy.âByâcontrast,âtheâBrettonâWoodsâeraâ(inâtheâaftermathâofâWWII)âwasâcharacterisedâbyâmonetaryâindependenceâandâexchangeârateâstability,âwhileâcapitalâmobilityâwasârestricted.âTheâperiodâthereafterâ(sinceâ1973)âhasâseenâanâincreaseâinâeconomiesâwithâfreeâcapitalâmobility,âmonetaryâautonomyâandâexchangeârateâflexibility.
2â Inâtheâcontextâofâtheâtrilemmaâ/âdilemmaâdiscussion,âmonetaryâindependenceâgoesâfurtherâthanâtheâsettingâofâtheâshort-termâpolicyârate,âandâalsoâincludesâtheâfactâthatâmonetaryâpolicymakersâcanâsteerâtheâbroaderâdomesticâfinancialâconditions.
Chart 1
Mechanism of the global financial cycle and our contribution 1
USmonetary
policy
Domesticfinancial
conditions
Riskaversion
Capitalflows
Globalfinancialcycle
US dollar
1 BasedâonâHabibâandâVendittiâ(2018).âTheârelationsâonâwhichâweâfocusâinâthisâarticleâareâindicatedâinâgreen.
![Page 4: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/4.jpg)
4NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Financial globalisation : a state of play
Theâthreeâdecadesâprecedingâtheâglobalâfinancialâcrisisâof 2008-2009 wereâmarkedâbyâaâmassiveâincreaseâinâgrossâcapitalâflowsâworldwide.âThisâwasâtheâresultâofâcapitalâcontrolsâbeingâtakenâdown,âaâdecreaseâinâbothâfinancialâregulationâandâtransactionâcosts,âandâtheâemergenceâofâfinancialâinnovationsâ(Gourinchasâandâ Rey, 2014â &â BIS, 2017).â Consequently,â cross-borderâ holdingsâ ofâ financialâ assetsâ andâ liabilitiesâ(expressedâasâaâ ratioâofâGDP)âââwhichâ canâbeâ referredâ toâasâaâmeasureâofââfinancialâglobalisationââorâinternationalâ financialâ integrationâ (Laneâ andâ Milesi-Ferretti, 2001)â ââ underwentâ aâ remarkableâ surge.âInâ Europeâ inâ particular,â financialâ opennessâ acceleratedâmoreâmarkedlyâ fromâ theâ lateâ 1990s,â afterâ theâintroductionâofâtheâeuroâhelpedâboostâcross-borderâtransactions.â
Thus,âbetweenâ1980 and 2007,âtheâsumâofâcross-borderâfinancialâclaimsâandâliabilities,âscaledâbyâannualâGDP,âroseâfromâaroundâ60 %âtoâalmostâ400 %âforâadvancedâeconomiesâ(G7 average),âandâfromâroughlyâ25 %âtoâmoreâthanâ110 %âforâemergingâmarketâeconomiesâ(BRICSâaverage).â
BOX 1
Real and financial globalisation
1970
1975
1980
1985
1990
1995
2000
2005
2010
2015
0
50
100
150
200
250
300
350
400
450
500
1980
1985
1990
1995
2000
2005
2010
2015
0
100
200
300
400
500
600
700
800
900
1000
1100
90
100
110
120
130
140
150
160
170
180
190
200
Euro areaJapanUS
Brics
G7Financial, euro areaInternational trade (right-hand scale)
Financial, world
Financial globalisation 1, 2, 3
(selected countries, total external assets and liabilities, in % GDP)
Real 4 and financial 5 globalisation(total exports and imports & external assets and liabilities, in % GDP, index 1982 = 100)
Sourcesâ:âLaneâandâMilesi-Ferretti,âWorldâBank,âNBB.1 TheâG7 countriesâareâCanada,âFrance,âGermany,âItaly,âJapan,âtheâUKâandâtheâUS.â2 TheâBricsâcountriesâareâBrazil,âRussia,âIndia,âChinaâandâSouthâAfrica.â3â Theâeuroâareaâfiguresârelateâtoâtheâeuroâareaâasâaâwholeâandâdoânotâincludeâintra-euroâareaâassetsâorâliabilities.â4â Totalâexportsâandâimports,âinâ%âofâGDP.5â Totalâexternalâassetsâandâliabilities,âinâ%âofâGDP.â
u
![Page 5: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/5.jpg)
5NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Financialâ globalisationâ isâ inâ partâ relatedâ toâ realâ globalisationâ sinceâ internationalâ tradeâ bothâ dependsâonâandâgeneratesâfinancialâ linkages.âTradeâneedsâ toâbeâfinancedâandâ itâ thereforeâ inducesâ cross-borderâpaymentâflows.âItâmayâalsoârequireâhedging,âwhenâdenominatedâinâforeignâcurrencyâorâwhenâconductedâinâ aâ riskyâ environment.â Finally,â itâ canâ boostâ foreignâ directâ investments,â forâ instanceâwhenâ companiesâdecideâtoâestablishâglobalâvalueâchainsâtoâoptimiseâproductionâcosts.âTradeâthusâinducesâtheâaccumulationâofâinternationalâassetsâandâliabilitiesâand,âusually,âcountriesâthatâareâmoreâinvolvedâinâtradeâareâalsoâmoreâfinanciallyâopen.â
Nevertheless,âfinancialâglobalisationâisâalsoâcharacterisedâbyâintricateâfinancialâlinksâestablishedâsolelyâforâfinancialâpurposesâ(BIS, 2017).âAsâtheâdemandâfor,âandâsupplyâof,âfinancialâproductsâandâservicesâincreasesâwithâtheâwealthâofâbusinessesâandâhouseholds,âfinancialâopennessâtendsâtoâincreaseâwithâtheâincomeâlevel.âItâ isâ thereforeânoâ surpriseâ thatâfinancialâglobalisationâhasâgrownâmuchâmoreâ rapidlyâ thanâ internationalâtradeâ sinceâ theâ1980s.âHowever,â inâ someâcountries,âpartâofâ theâfinancialâ integrationâmightâ containâanââartificialââ componentâ relatedâ toâ tax-optimisationâ strategiesâ whichâ inflateâ assetsâ andâ liabilitiesâ toâ aâsimilarâextentâ(e.g.âthroughâcross-borderâintragroupâloans,âseeâalsoâsection 3).âSinceâtheâglobalâfinancialâcrisisâof 2008-2009,âtheâgrowthâinâcross-borderâassetâpositionsâinârelationâtoâGDPâ(i.e.âcapitalâflows)âhasâslowedâdownâsignificantly 1.âThreeâfactorsâmayâbeâputâforwardâtoâexplainâthisâdevelopment.âTheâfirstâ isâpreciselyâaâdecelerationâofâinternationalâtradeâandâaâdemand-inducedâweaknessâinâtrade-intensiveâphysicalâinvestments.âTheâsecondâisâaâdeclineâinâcross-borderâactivityâbyâbanks,âconcentratedâinâbankâloans,âandâlargelyâconfinedâtoâEuropeanâbanksâ(BIS, 2017).âAndâtheâthirdâisâsimplyâanâincreaseâinâtheârelativeâweightâofâemergingâmarketâeconomiesâinâglobalâGDPâwhile,âatâtheâsameâtime,âtheseâeconomiesâtendâtoâbeâlessâfinanciallyâintegratedâ(i.e.âholdâlowerâexternalâassetsâandâliabilities)âcomparedâtoâadvancedâeconomies.â
Nonetheless,âtheâoutstandingâexternalâassetsâandâliabilitiesâofâbothâemergingâandâdevelopedâeconomiesâremainâcloseâtoâtheirâhighestâlevel.âLikeâReyâ(2015),âweâtakeâthisâasâaâstartingâpointâtoâanalyseâwhetherâthisâhasâimplicationsâforâtheâevidenceâinâfavourâofâtheâglobalâfinancialâcycleâandâitsâtransmission.
1â Forâaâdetailedâdescriptionâofâtheâevolutionâofâfinancialâglobalisationâsinceâtheâglobalâfinancialâcrisis,âseeâLaneâandâMilesi-Ferrettiâ(2017).
2. Financial cycles in the euro area
2.1 Financial conditions index
Theâpreviousâsectionâshowedâthatâtheâ literatureâfindsâevidenceâofâaâglobalâfinancialâcycleâ inâbothâcapitalâflowsâandâfinancialâconditions.âOurâworkâmainlyârelatesâtoâthisâsecondâbranchâofâliteratureâ(e.g.âimpactâofâtheâGFCâonâassetâpricesâandâcreditâgrowth,âasâ inâObstfeld et al., 2017).âSinceâweâwantâ toâbroadenâourâscopeâasâmuchâasâpossible,âthisâalsoâraisesâtheâquestionâconcerningâwhichâfinancialâconditionsâweâshouldâconsiderâ;âthatâquestionâisâcloselyâlinkedâtoâtheâdiscussionâonâexactlyâwhatâaâfinancialâcycleâis.
Althoughâ thereâ isâ currentlyâ noâ generallyâ acceptedâ definitionâ ofâ theâ financialâ cycle,â itâ isâ oftenâ describedâ asâaâ cyclicalâ movementâ commonâ toâ multipleâ financialâ sectorâ segments,â suchâ asâ creditâ andâ realâ estateâ marketsâ(see e.g. Borio, 2012 andâDrehmann et al., 2012).âToâoperationaliseâthisâdefinition,âcompositeâ indicatorsâareâaâusefulâ toolâ forâextendingâtheâstandardâunivariateâapproachâ(e.g.âcredit-to-GDPâgapâasâfinancialâcycleâmeasure)â
![Page 6: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/6.jpg)
6NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
toâmoreâ holisticâ approachesâ whereâ theâ financialâ cycleâ isâ extractedâ fromâ aâ largeâ rangeâ ofâ relevantâ data.â Theâmethodologyâbehindâourâcompositeâindicator,âtheâfinancialâconditionsâindexâ(FCI),âisâdescribedâinâboxâ2.
Theâ FCIâoffersâ aâ viewâonâ theâpropertiesâofâ theâfinancialâ cyclesâ inâ theâeuroâareaâ countries.âUnderstandingâ theâdevelopmentâofâ theâfinancialâcycleâ isâkeyâ forâmacroprudentialâpolicy.âTheâ literatureâsuggestsâ thatâ theâfinancialâcycleâisâsubjectâtoâaâboomâ/âbustâprofile.âDuringâtheâboomâphase,âsystemicârisksâareâbuildingâupâandâtheâpeaksâofâtheâcycleâcanâserveâasâearlyâwarningâsignalsâforâfinancialâcrises.â
Notwithstandingâitsâimportance,âempiricalâanalysisâregardingâtheâfeaturesâofâtheâfinancialâcycleâinâEuropeâisâscarce.âAâlimitingâfactorâisâtheâlackâofâaâconsensusâdefinitionâforâtheâfinancialâcycle,âregardingâbothâitsâcompositionâandâitsâmethodology.âTheâdifficultyâofâobtainingâharmonisedâlong-termâseriesâinâEuropeâalsoâplaysâaârole.âMerler (2015)âandâSchĂźler et al.â (2015)âwereâamongâtheâfirstâtoâcharacteriseâtheâfinancialâcycleâ inâEurope.âBothâauthorsâfindâââ asâ âstylisedâ factsââ ââ thatâ financialâ cyclesâ areâ inâ generalâ longerâ thanâ theâ traditionalâ businessâ cycle,â therebyâconfirmingâtheâfindingsâofâBorio et al.â(2012).âBothâauthorsâpointâtoâtheâexistenceâofâaâfinancialâcycleâinâtheâeuroâareaâwithâaâclearâboomâ/âbustâprofileâaroundâfinancialâcrises,â illustratingâitsâearlyâwarningâcapabilities.âHowever,âfinancialâcyclesâshowâstrongâheterogeneityâ/âdivergenceâacrossâeuroâareaâcountries,âwithâvaryingâamplitudesâandâdifferentâcyclicalâpositions 1.
Asâ shownâ inâChart 2,â theâ FCIâ largelyâ confirmsâ theseâ findings.âNoteâ thatâ ourâ financialâ cycleâmeasureâ isâmoreâbroadlyâdefinedâthanâtheâconceptsâutilisedâinâMerlerâ(creditâandâhouseâprices)âandâSchĂźlerâ(credit,âhouse,âequityâandâbondâprices).âTheâaverageâFCIâinâtheâeuroâareaâ(Figureâ2 ââleftâpanel)âshowsâevidenceâofâaâboom-bustâprofileâandâreachesâitsâhighestâpeakâbeforeâtheâglobalâfinancialâcrisisâof 2008.âOnâaverage,âtheâFCIâresultsâinâpersistentâcyclesâthatâoperateâatâ lowerâfrequenciesâthanâtheâclassicâbusinessâcycle.âFigureâ2 (rightâpanel)âdepictsâhowâtheâ
1â Germanyâsââsafe-havenââstatusâisâlikelyâtoâcontributeâtoâitsâdivergingâfinancialâcycle,âresultingâinâhigherâdemandâforâGermanâgovernmentâbondsâwhenâglobalâriskâaversionâincreases.âFurthermore,âinâtheâfirstâpartâofâtheâsample,âGermanâhouseâpricesâdeviatedâfromâtheâgeneralârisingâtrendâdueâtoâtheâoversupplyâcausedâbyâhouse-buildingâincentivesâafterâGermanâreunification.âTheseâelementsâmightâexplainâtheââatypicalââbehaviourâofâtheâGermanâFCI.
Chart 2
Financial conditions index as a measure of the financial cycle
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
Euro area cross-country distribution
Loose
Tight
FCI, euro area average
25 %-75 %
Profile around systemic financial crisis 1, 2
FCI, euro area average
25 %-75 %
â24 â20 â16 â12 â8 â4 0 4 8 12 16 20 240
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
Start of a financial crisis
Sourcesâ:âECB,âNBB.1 Numberâofâquartersâbeforeâ(â)âorâafterâ(+)âtheâstartâofâaâfinancialâcrisis.2 CrisisâeventsâasâinâLoâDucaâet al.â(2017)âandâdefinedâasâallâsystemicâcrisesâwithâatâleastâpartlyâdomesticâoriginâandâconsideredâbyâEuropeanânationalâauthoritiesâasârelevantâfromâaâmacroprudentialâperspective.
![Page 7: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/7.jpg)
7NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
FCIâ startsâ toâ increaseâwellâaheadâofâ systemicâcrisesâandâ reachesâ itsâpeakâaroundâ2 yearsâbeforeâaâcrisisâ starts.âItâcanâbeâshownâthatâtheâFCIâhasâgoodâearlyâwarningâpropertiesâ(AUROC 1âaboveâ0.85),âthatâoutperformâthoseâofâunivariateâfinancialâcycleâmeasuresâsuchâasâ theâcredit-to-GDPâgap 2.âTheseâpropertiesâholdâ forâaâmajorityâofâcountries,âalthoughâtheâFCIâshowsâsomeâcross-countryâheterogeneityâ (inâparticularâ inâtheâbuild-upâphase).âTheâfollowingâsectionâanalysesâthisâinâmoreâdetail.
1â AreaâUnderâtheâReceiverâOperatingâCharacteristics.âThisâmeasureâroughlyâcapturesâtheâprobabilityâofâcorrectâprediction,âwithâ1 correspondingâtoâperfectâpredictionâandâ0.5 toânoâpredictiveâpowerâ(equivalentâtoâtossingâaâcoin).
2â ForâmoreâdetailsâregardingâtheâearlyâwarningâperformanceâofâtheâFCIârelativeâtoâotherâmethodsâmeasuringâcyclicalâsystemicârisk,âseeââCyclicalâsystemicâriskâmeasurementââ(2019),âECBâOccasionalâWorkingâPaper,âforthcoming.
Financial conditions index (FCI)
TheâFCIâisâaâbroad-basedâcompositeâindicatorâofâdomesticâfinancialâconditions,âaggregatingâfiveâfinancialâriskâ dimensions 1â (creditâ developments,â realâ estate,â privateâ sectorâ debt,â bankingâ sectorâ andâ financialâmarketâ conditions)â intoâ anâ overallâ indicatorâ usingâ time-varyingâweightsâ basedâ onâ theâ dataâ correlationâstructure.âTheâcurrentâversionâofâtheâindicatorâcontainsâ17 variables.
Inâaâfirstâstep,âtheâvariablesâareâtransformedâbyâmeansâofâorderâstatistics 2âsuchâthatâhigherâvaluesâindicateâlooserâfinancialâconditionsâandâlowerâvaluesâcorrespondâtoâtighterâfinancialâconditions.âTheâorderâstatisticâofâvariableâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
âatâtimeâtâisâdenotedâbyâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
,âandâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
âdenotesâtheâk-thâvalueâinâtheâ(ascending)âseriesâofâtheâvariableâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
.âTheseâorderâstatisticsâareâcalculatedâonâtheâbasisâofâ theâempiricalâdistributionâ functionâandâtakeâaâvalueâbetweenâ0 (bust)âandâ1 (boom).
Inâ aâ secondâ step,â theâ sub-indicesâ areâ compiledâ onâ theâ basisâ ofâ anâ unweightedâ averageâ ofâ theâ orderâstatisticsâ ofâ allâ theâ variablesâ assignedâ toâ theâ specificâ sub-indices,â whereâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
â denotesâ theâ numberâ ofâvariablesâassignedâtoâtheâsub-indexâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
and
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
âdenotesâtheâorderâstatisticâforâeachâofâtheseâvariables.
1 TheâselectionâofâriskâdimensionsâisâbasedâonâtheâcategoriesâsuggestedâforâmonitoringâcyclicalâsystemicâriskâinâESRBârecommendationâESRBâ/â2014â/â1,âwithâtheâexceptionâofâtheâriskâcategoryââexternalâimbalancesâ.âTheâexclusionâofâthisâcategoryâbenefitsâtheâanalysisâinâtheârestâofâthisâarticle,âasâweâavoidâendogeneityâissuesâbetweenâourâmeasureâofâdomesticâfinancialâconditionsâandâinternationalâcapitalâflows.
2 Theâuseâofâorderâstatisticsâisârelevantâasâitâmakesâtheâresultingâstatistic(s)âlessâsensitiveâtoâextremeârealisationsâofâtheâvariable (seeâHollĂłâet al., 2012).
BOX 2
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
u
![Page 8: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/8.jpg)
8NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Inâaâthirdâstep,âtheâsub-indicesâareâaggregatedâintoâanâoverallâindicatorâofâtheâfinancialâcycleâbyâapplyingâbothâanâindex-specificâandâaâtime-varyingâweightingâfunction,âfollowingâHollĂłâet al.â(2012).âDenotingâtheâvectorâofâ index-specificâweightsâbyâwâandâ theâvectorâofâ theâvalueâ forâ theâ sub-indicesâatâ timeâ tâbyâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
,âthe financialâ cycleâ indicator,â
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
,â canâ beâ constructedâ asâ theâ weightedâ quadraticâ formâ ofâthe sub-indicesâ.â
with °â theâ Hadamard-productâ andâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
â aâ time-varyingâ weightingâ matrixâ reflectingâ theâ time-varyingâ(positive)â bilateralâ co-movementâ betweenâ theâ respectiveâ variables.â Theâ latterâ isâ constructedâ byâ takingââ element-wiseâââtheâmaximumâbetweenâzeroâandâtheâtime-varyingâpair-wiseâcorrelationsâbetweenâeachâcombinationâofâsub-indices.âTheâpair-wiseâcorrelationsâareâconstructedâusingâanâexponentiallyâweightedâmovingâ averageâ (EWMA)â filterâ forâ theâ variance-covarianceâ matrix.â Inâ operationalisingâ thisâ statistic,âthe index-specificâweightsâareâequalâtoâ0.2 (orâ1â/ânumberâofâ
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
).â
10
Box 2 - Financial conditions index (FCI)
The FCI is a broad-based composite indicator of domestic financial conditions, aggregating five financial risk dimensions(1) (credit developments, real estate, private sector debt, banking sector and financial market conditions) into an overall indicator using time-varying weights based on the data correlation structure. The current version of the indicator contains 17 variables.
In a first step, the variables are transformed by means of order statistics(2) such that higher values indicate looser financial conditions and lower values correspond to tighter financial conditions. The order statistic of variable đĽđĽ!,! at time t is denoted by đ§đ§!,!, and đĽđĽ!,[!] denotes the k-th value in the (ascending) series of the variable đĽđĽ!. These order statistics are calculated on the basis of the empirical distribution function and take a value between 0 (bust) and 1 (boom).
đ§đ§!,! = đšđš đĽđĽ!,! âś=
đđđđ
đđđđđđ đĽđĽ!,[!] ⤠đĽđĽ!,! < đĽđĽ!,[!!!]
1 đđđđđđ đĽđĽ!,! ⼠đĽđĽ!,[!]
In a second step, the sub-indices are compiled on the basis of an unweighted average of the order statistics of all the variables assigned to the specific sub-indices, where đđ! denotes the number of variables assigned to the sub-index đđ! and đ§đ§!,! denotes the order statistic for each of these variables.
đđ!,! =1đđ!
đ§đ§!,! , đđ = 1,⌠, 5
!!
!!!
In a third step, the sub-indices are aggregated into an overall indicator of the financial cycle by applying both an index-specific and a time-varying weighting function, following HollĂł et al. (2012). Denoting the vector of index-specific weights by w and the vector of the value for the sub-indices at time t by đşđş!, the financial cycle indicator, đšđšđšđšđšđš!, can be constructed as the weighted quadratic form of the sub-indices:
đšđšđšđšđšđš! = đđ â đşđş! â˛đ¸đ¸! đđ â đşđş!
with â the Hadamard-product and đ¸đ¸! a time-varying weighting matrix reflecting the time-varying (positive) bilateral co-movement between the respective variables. The latter is constructed by taking - element-wise - the maximum between zero and the time-varying pair-wise correlations between each combination of sub-indices. The pair-wise correlations are constructed using an exponentially weighted moving average (EWMA) filter for the variance-covariance matrix. In operationalising this statistic, the index-specific weights are equal to 0.2 (or 1/number of đşđş!).
Next, we use a weight of 0.98 in the EWMA for the variance-covariance matrix which assigns a significantly larger weight to more recent observations. In any given period, the FCI maximum (minimum) value of 1 (0) can be attained only if each of the sub-indices reaches the maximum (minimum) value at a time where the cycles are also perfectly coincident.
Input and output
As an input, 17 variables are used which are presumed to be relevant for shaping the financial cycle. The selection is based on the empirical literature and availability over a longer time period. Our sample contains the euro area countries. The data go back as far as 1970Q1, but the length of the time series varies across series and countries(3). The data set is mixed in terms of frequency (monthly and quarterly), nominal and real variables, levels, data in differences and gap measures (using a recursive HP-filter). The indicators with quarterly frequency are transformed to
FCI composition: 17 data series
Series (17) Transformation Sample (max)
Sub-index 1 : Credit developments (3 series)
Bank credit gap NFPS gap, % points 1970 Q4
HH bank loan growth y-o-y % 1998 Sep
NFC bank loan growth y-o-y % 1998 Sep
Sub-index 2 : Real estate (5 series)
Price-to-income ratio, level level 1970 Q1
Price-to-income ratio, gap gap,â% points 1971 Q1
Affordability (1), level level 1996 Q1
Affordability (1), gap gap,â% points 1997 Q1
Nominal house prices, gap gap,â% points 1970 Q1
Sub-index 3 : Private debt (3 series)
Debt-to-GDP ratio NFPS y-o-y difference 1971 Q4
Debt service ratio HH y-o-y difference 1981 Q4
Debt service ratio NFC y-o-y difference 1981 Q4
Sub-index 4 : Banking sector (4 series)
Financial sector assets y-o-y % 2000 Q1
Bank lending margin level (-) 2003 Q1
Credit spread HH loans (vs 10Y sovereign) level (-) 2003 Jan
Credit spread NFC loans (vs 10Y sovereign) level (-) 2003 Jan
Sub-index 5 : Financial markets (2 series)
Real equity prices y-o-y % 1981 Q1
Bond yield : 10Y sovereign level (-) 1970 Q1
Sources : ECB, NBB.Note : Gap measures calculated using a Hodrick-Prescott filter consistent with the Basel credit gap (lambda = 400 000).1 Estimates of the over / undervaluation of residential property prices : average of different valuation measures for all types of property.
u
![Page 9: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/9.jpg)
9NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Next,â weâ useâ aâ weightâ ofâ 0.98 inâ theâ EWMAâ forâ theâ variance-covarianceâ matrixâ whichâ assignsâ aâsignificantlyâlargerâweightâtoâmoreârecentâobservations.âInâanyâgivenâperiod,âtheâFCIâmaximumâ(minimum)âvalueâofâ1 (0)âcanâbeâattainedâonlyâifâeachâofâtheâsub-indicesâreachesâtheâmaximumâ(minimum)âvalueâatâaâtimeâwhereâtheâcyclesâareâalsoâperfectlyâcoincident.
Input and output
Asâ anâ input,â 17 variablesâ areâ usedâwhichâ areâ presumedâ toâ beâ relevantâ forâ shapingâ theâ financialâ cycle.âThe selectionâ isâbasedâonâ theâempiricalâ literatureâandâavailabilityâoverâ aâ longerâ timeâperiod.âOurâ sampleâcontainsâtheâeuroâareaâcountries.âTheâdataâgoâbackâasâfarâasâ1970Q1,âbutâtheâlengthâofâtheâtimeâseriesâvariesâacrossâseriesâandâcountries 1.âTheâdataâsetâisâmixedâinâtermsâofâfrequencyâ(monthlyâandâquarterly),ânominalâandârealâvariables,âlevels,âdataâinâdifferencesâandâgapâmeasuresâ(usingâaârecursiveâHP-filter).âTheâindicatorsâwithâquarterlyâfrequencyâareâtransformedâtoâaâmonthlyâfrequencyâusingâstandardâlinearâinterpolation 2.
1 Providedâtheâfinancialâcycleâcanâtakeâmoreâthanâ20 years,âpreferenceâwasâgivenâtoâlong-termâseries.âInâtheâcaseâofâmissingâvariables,âtheâsub-indicatorsâtakeâtheâaverageâoverâtheâotherâvariables.âIfâdataâareâmissingâatâtheâlevelâofâtheâsub-indicators,âweightsâareâadjustedâ(1â/ânumberâofâsub-indicators).âTheâuseâofâorderâstatisticsâandâweightedâaveragesâlimitsâtheâimpactâofâthisâchangingâcompositionâonâtheâaggregateâindex.
2 Theâindicatorâisâcalculatedâusingâaâbalancedâsampleâatâtheâend.âToâcaterâforâdifferentâpublicationâlags,âmissingâobservationsâareâreplacedâbyâtheâlatestâobservation.
Financial conditions index (FCI) for Belgium and sub-components(1980Q1-2019Q2)
1980
1982
1984
1986
1989
1991
1993
1995
1998
2000
2002
2004
2007
2009
2011
2013
2016
2018
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Correlation component
Financial markets
Real estate
FCIBanking sector
Private debt
Credit developments
Sourceâ:âNBB
u
![Page 10: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/10.jpg)
10NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
2.2 Synchronisation and correlation of the FCI with the global financial cycle
BasedâonâtheâcorrelationâbetweenâtheâindividualâcountriesââFCIâandâtheâaverageâeuroâareaâFCI,âsynchronisationâofâ financialâ cyclesâ isâ ââ onâ averageâ ââ relativelyâ highâ (averageâ bilateralâ correlationâ ofâ 0.74).â However,â thereâ isâsubstantialâ cross-countryâ heterogeneity,âwithâweakerâ correlationsâ forâ someâ countriesâ (0.18 forâGermany)â andâstrongerâcorrelationsâforâothersâ(0.94 forâFrance).âNoteâthatâ inâcontrastâtoâtheâbusinessâcycle,â largeâeconomiesâmayâdeviateâmarkedlyâfromâtheâaverageâeuroâareaâfinancialâcycle.
Theâkeyâquestionâweâraiseâinâthisâarticleâconcernsâtheâdegreeâofâsynchronisationâbetweenâtheâfinancialâcycleâinâtheâeuroâareaâandâtheâglobalâfinancialâcycle.âAsâaâstartingâpoint,âweâthereforeâcalculateâtheâcorrelationâbetweenâtheâaverageâeuroâareaâFCI 1âandâaâmeasureâofâtheâglobalâcycle.âForâtheâlatterâweâuseâtheââGlobalâStockâMarketâFactorââ ofâ Habibâ andâ Vendittiâ (2019).â Thisâ factorâ isâ extractedâ fromâ aâ globalâ panelâ ofâ stockâmarketâ returns.âAlternativeâ measuresâ includeâ theâ Miranda-Agrippinoâ andâ Reyâ factorâ (2019)â whichâ capturesâ theâ commonâcomponentâinâ858 assetâpriceâseries.âSinceâtheâvariousâmeasuresâofâtheâGFCâtendâtoâbeâhighlyâcorrelatedâ(HabibâandâVenditti, 2019),âtheâresultsâareâinâgeneralârobustâtoâtheâchoiceâofâGFCâmeasure.â
ItâturnsâoutâthatâtheâaverageâeuroâareaâFCIâandâtheâglobalâfinancialâcycleâmeasureâareâhighlyâcorrelatedâ(0.89).âTheâ highâ correlationâ isâ remarkable,â givenâ thatâ theâ twoâmeasuresâ haveâ differentâ purposesâ (domesticâ financialâconditionsâ versusâ globalâ financialâ cycle),â areâ derivedâ fromâ completelyâ differentâ datasetsâ (broadâ spectrumâ ofâmacrofinancialâseriesâversusâstockâmarketâreturns)âandâareâbasedâonâdifferentâmethodologiesâ(compositeâ indexâversusâfactorâanalysis).
Theâstrongâcorrelationâwithâtheâglobalâfinancialâcycleâalsoâholdsâatâtheâlevelâofâtheâindividualâcountries,âalbeitâtoâvaryingâdegrees.âTheâcorrelationârangesâfromâ0.27 (Germany)âtoâ0.86 (Luxembourg)âandâ isâ largelyâ inâ lineâwithâtheâsynchronicityâofâeachâcountryâsâcycleâwithinâtheâeuroâarea.â
1â ThroughoutâthisâpaperâweâuseâtheâaverageâFCIâasârepresentingâtheâeuroâareaâfinancialâcycle.âAlternatively,âoneâcouldâapplyâaâprincipalâcomponentâanalysis.âTheâvarianceâofâourâeuroâareaâaverageâlargelyâcorrespondsâwithâtheâresultâofâaâprincipalâcomponentâanalysisâ(selectingâtwoâfactors)âandâhasâtheâadvantageâofâbeingâsimple.
TheâmainâoutputâisâtheâcompositeâFCIâindicatorâonâaâmonthlyâorâquarterlyâbasisâ(transformedâbyâtakingâaverages).âTheâchartâbelowâillustratesâtheâFCIâandâitsâsub-componentsâforâBelgium.â
Byâconstruction,âtheâFCIâoffersâanâabsoluteâinterpretationâforâfinancialâconditions,âasâtheâvariableâisâcontainedâbetweenâ0 andâ1.âInâanyâgivenâperiod,âtheâmaximumâ(minimum)âvalueâofâ1 (0)âcanâbeâattainedâonlyâifâeachâofâtheâsub-indicesâreachesâtheâcountry-specificâmaximumâ(minimum)âvalueâatâaâtimeâwhereâtheâcyclesâareâalsoâperfectlyâcoincident.âMoreover,âincreasesâinâtheâFCIâcanâariseâbecauseâofâeitherâanâincreaseâinâ(someâof)âtheâindividualâsub-indicesâ(riskâdimensions)âorâbecauseâofâanâincreaseâinâtheâco-incidenceâinâtheâcyclesâofâtheârespectiveâsub-indices.âTheâfinancialâconditionsâ indicatorâ thusâexplicitlyâ takesâ intoâaccountâ theâcorrelationâbetweenâtheâfinancialâvariables.âThisâcorrelationâorâco-movementâisâanâessentialâfeatureâofâfinancialâcyclesâandâtendsâtoâbeâstrongâaroundâfinancialâcrises.âAsâanâintermediateâproduct,âtheâ(unweighted)âsub-indicesâââwhichâ averageâ theâ orderâ statisticsâ ofâ theâ variablesâ ââ canâ beâ usedâ toâmonitorâ tensionsâ inâ theâ specificârisk categories.â
![Page 11: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/11.jpg)
11NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
3. How sensitive are euro area countries to the global financial cycle ?
Soâ far,â weâ haveâ shownâ thatâ domesticâ financialâ conditionsâ inâ theâ euroâ areaâ areâ closelyâ linkedâ toâ theâ globalâfinancialâ cycle.â Atâ theâ sameâ time,â theâ correlationâwithâ theâGFCâ differsâ acrossâ countries,â suggestingâ thatâ theâcountriesâ co-movementâwithâtheâGFCâisâinfluencedâbyâcountry-specificâfactors.âWhichâfeaturesâcanâmagnifyâorâattenuateâcountriesââsensitivityâtoâtheâGFCâ?âTheâmostânaturalâcandidatesâareâtheâpolicyâvariablesâofâtheâfinancialâtrilemma,âi.e.âfinancialâaccountâopennessâandâtheâexchangeârate.
Mostâ ofâ theâ literatureâ analysingâ theâ sensitivityâ ofâ financialâ conditionsâ toâ theâGFCâhasâ beenâ concentratingâ onâtheseâvariables,âandâ inâparticularâonâ theâexchangeâ rateâ regimeâ (Rey, 2015â;âObstfeld et al., 2017).â In general,âthe evidenceâ isâ mixed,â resultingâ inâ varyingâ conclusionsâ regardingâ theâ existenceâ ofâ aâ financialâ trilemmaâ(the exchangeârateâmatters)âorâdilemmaâ(theâexchangeârateâisâirrelevant).âSinceâeuroâareaâcountriesâshareâtheâeuroâasâsingleâcurrency,âtheâexchangeârateâcannotâexplainâtheâdifferencesâinâtheâimpactâofâtheâGFCâacrossâcountries.âThe onlyâremainingâvariableâisâtheâfinancialâaccountâopennessâ(i.e.âtheâdegreeâofâfinancialâintegration).â
Inâ fact,â theâ externalâ assetsâ andâ liabilitiesâ formâ aâ keyâ channelâ throughâ whichâ globalâ financialâ conditionsâ areâtransmittedâ toâ anâ economy.â Apartâ fromâ financialâ openness,â whichâ weâ canâ quantifyâ byâ meansâ ofâ theâ grossâpositionâ definedâ asâ theâ sumâ ofâ externalâ assetsâ andâ liabilitiesâ scaledâ byâ GDP,â weâ addâ otherâ dimensionsâ ofâthe countriesââexternalâfundingâasâpotentialâdeterminantsâforâtheirâsensitivityâtoâtheâGFC.
Theseâotherâdimensionsâincludeâtheâcompositionâofâtheâexternalâfundingâinâtermsâofâinstrumentsâ(direct,âportfolioâandâotherâ investment).âAlso,â apartâ fromâ theâgrossâ position,âweâ analyseâ theâpossibleâ roleâ ofâ theânetâ position,âwhichâequalsâtheâdifferenceâbetweenâtheâexternalâassetsâandâliabilitiesâscaledâbyâGDPâ(i.e.âtheânetâinternationalâinvestmentâpositionâââNIIP).âAâlastâdimension,âasâaâcomplementâtoâtheâstocks,âcomprisesâtheâgrossâandânetâcapitalâflows,âincludingâtheirâbreakdownâbyâinstrument.
Chart 3
Are financial conditions in the euro area correlated with the global financial cycle ?
Average FCI, euro area 1
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
â2
â1.5
â1
â0.5
0
0.5
1
1.5
2
2.5
3
Corr. = 0.89
Global financial cycle (ECB measure) 2
LU FR AT ES NL IE IT GR FI MT LT CY PT BE SI LV EE SK DE0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Correlation with global cycle
Correlation with average FCI, euro area
FCI euro area and global financial cycle(standardized data, 1990-2017)
Synchronization of FCI in the euro area(2000-2017)
Sourcesâ:âECB,âNBB.1 Financialâconditionsâindex,âeuroâareaâcross-countryâaverage.2 GlobalâfinancialâcycleâmeasureâfromâHabibâandâVendittiâ(2019).
![Page 12: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/12.jpg)
12NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Theâexternalâ fundingâ isâaânaturalâ candidateâ toâ influenceâ theâ impactâofâ theâGFC,ânotâonlyâbecauseâaâ strongârelationshipâbetweenâdomesticâcreditâgrowthâandâinternationalâcapitalâflowsâisâanâestablishedâfactâ(LaneâandâMcQuade, 2013)âbutâalsoâbecauseâ theâ literatureâhasâ shownâ thatâ theâglobalâfinancialâ cycleâhasâaâ significantâinfluenceâ onâ capitalâ flowsâ themselvesâ (Forbesâ andâ Warnock, 2012),â beâ itâ inâ grossâ orâ netâ termsâ (Davis et al., 2019)âorâbyâtypeâofâcapitalâflowâ(Avdjiev et al., 2018).âGlobalâfactors,âsuchâasâUSâinterestâratesâorâglobalâriskâ aversionâ actâ asâ âgatekeepersââ forâ capitalâ in-â andâ outflowsâ toâ andâ fromâ emergingâ economiesâ (Ghosh,âQureshi,âKimâandâZalduendo, 2014).âHabibâandâVendittiâ (2019)âprovideâevidenceâofâaââglobalâ capitalâflowsâcycleâ.âMoreover,âitâhasâbeenâshownâthat,âduringâfinancialâcrises,âsomeâcapitalâflowsâtendâtoâbeâmoreâvolatileâthanâothersâ(Bussière, 2016).âAsâsuch,âweâexpectâthatâtheâsizeâandâcompositionâofâtheâexternalâfundingâplaysâanâessentialâ roleâ inâdeterminingâcountriesââ sensitivityâ toâ theâGFC,âparticularlyâ inâ theâeuroâarea,âwhereâ thereâareâwideâcross-countryâvariationsâinâtheâsizeâandâcompositionâofâtheâexternalâfunding,âwhereasâtheâexchangeârateâisâtheâsameâforâallâcountries.
Figureâ 4 showsâ theâ cross-countryâ variationâ inâ theâ grossâ andâ netâ externalâ position.â Asâ advancedâ economies,âthe euroâareaâcountriesâshowâaâhighâdegreeâofâfinancialâintegration.âInâallâeconomiesâtheâstockâofâexternalâassetsâandâ liabilitiesâexceedsâGDP.âAsâexplainedâ inâboxâ1,âfinancialâ integrationâhasâ increasedâmarkedly,âparticularlyâ inâtheâeuroâareaâwhereâtheâeuroâactedâasâaâcatalystâforâcross-borderâfinancialâflowsâsinceâtheâcreationâofâtheâEMU.âAlthoughâ thatâprocessâhasâ comeâ toâaâhaltâ sinceâ theâfinancialâ crisis,âwithâ lowerâ capitalâflows,â theâoutstandingâstocksâareâstillâcloseâtoâtheirâhighestâlevels.âAsâmentionedâinâboxâ1,âapartâfromâtheâmacroeconomicâfundamentals,âtheâfiscalâ regimeâandâpresenceâofâ largeâmultinationalsâ inâsomeâcountriesâcontributesâ toââaccounting-inspiredââflowsâ thatâ inflateâ assetsâ andâ liabilitiesâ toâ aâ similarâ extentâ (e.g.â cross-borderâ intragroupâ loans),âmakingâpartâofâtheâintegrationâartificialâandâvolatile.âInâtheseââfinancialâcentresâ,âtheâgrossâpositionâtakesâextremeâvaluesâ(aboveâ1000%âGDP).
Chart 4
Gross and net financial position
LT SI SK LV EE IT EL ES AT PT DE FR FI BE NL CY IE MT LU0
400
800
5 00010 00015 00020 00025 00030 00035 00040 000
Gross financial position 1
(in % GDP)
2018Q4
1999Q1
DE BE AT FI EA IT FR SI EE LT LV SK ES PT CY EL IEâ200
â150
â100
â50
0
50
100
Net financial position 2 by instrument(in % GDP, 2018Q4)
Reserve assets
Other investment and financial derivatives
Portfolio investment
Direct investment
Total economy (i.e. net external position)
Sourcesâ:âECB,âIMF,âNBB.1 Sumâofâtheâexternalâassetsâandâliabilities.2 Differenceâbetweenâtheâoutstandingâassetsâandâliabilities,âbyâfinancingâinstrument.
![Page 13: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/13.jpg)
13NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Theânetâpositionâ isâunaffectedâbyâ theseââartificialââflows,â insofarâ asâ theyâdriveâupâassetsâ andâ liabilitiesâ toâ theâsameâextent.âNonetheless,âtheâNIIPâalsoâshowsâsubstantialâcross-countryâdifferences,ârangingâatâtheâendâof 2018âfrom â143 %â(Ireland)âtoâ61 %âGDPâ(Germany).âMostâofâtheâcountriesâinâtheâeuroâareaâareânetâdebtorsâ(liabilitiesâexceedâ assets).â Theâ NIIPâ isâ theâ aggregateâ netâwealthâ ofâ theâ domesticâ sectors,â andâ largeâ negativeâ valuesâ areâconsideredâunsustainable.â TheâNIIPâ isâmonitoredâ closelyâwithinâ theâenhancedâEuropeanâeconomicâgovernanceâframeworkâ (Europeanâ Semester)â sinceâ itâ isâ oneâ ofâ theâ indicatorsâ includedâ inâ theâMacroeconomicâ ImbalanceâProcedureâ (MIP).â Valuesâ belowâ â35 %âGDPâ canâ beâ consideredâ asâ anâ excessiveâ imbalance.â Theâ chartâ furtherâdecomposesâ theâNIIPâ accordingâ toâ theâ typeâofâ funding.âAmongâ theânetâ debtors,â aâ largeâpartâ ofâ theâ fundingâconsistsâofâotherâinvestment,âwhichâisâmainlyâbank-relatedâfunding.â
Inâ theâ nextâ section,âweâ analyseâwhetherâ theâ sizeâ andâ compositionâ ofâ theâ externalâ positionâ ofâ theâ euroâ areaâcountriesâ canâexplainâ theâdifferenceâ inâ sensitivityâ toâ theâGFC.â Forâ thisâ purpose,âweâ constructedâaâdatasetâ forâtheâ19 euroâareaâcountriesâonâaâquarterlyâbasisâsinceâ1990,ârelyingâonâEurostatâ/âECBâandâonâtheâIMFâBalanceâofâPaymentsâStatisticsâ forâhistoricalâdata.âTheâdatasetâcontainsâbothâexternalâassetsâandâ liabilities,âcapitalâ in-âandâoutflowsâandâaâbreakdownâbyâmainââfunctionalââcategoriesâ(direct,âportfolioâandâotherâinvestment).âSeriesâthatâshowedâ aâ breakâ betweenâ theâ twoâ sourcesâ wereâ retropolated.âWhereâ necessaryâ weâ interpolatedâ theâ annualâobservationsâlinearlyâtoâobtainâquarterlyâdata.
4. Empirical results on sensitivity to the global financial cycle
Thisâsectionâprovidesâempiricalâevidenceâonâ(i)âwhetherâdomesticâfinancialâconditionsâinâtheâeuroâareaâcountriesâmoveâinâlineâwithâtheâglobalâfinancialâcycleâandâ(ii)âtoâwhatâextentâthisâco-movementâisâmagnifiedâorâattenuatedâbyâfeaturesâofâtheirâexternalâfunding.âTherefore,âweâletâvariousâvariablesââinteractââwithâtheâGFC,âsuchâasâtheâgrossâandânetâexternalâposition,âasâwellâasâtheâgrossâandânetâcapitalâflows.âToâthatâend,âweâestimateâtheâfollowingâpanelâregressionâspecificationâ:
in which
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
âdenotesâ theâdomesticâfinancialâ conditionsâ indexâofâ countryâ
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
and
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
â isâ theâglobalâfinancialâcycleâ(takenâfromâHabibâandâVenditti, 2019).âToâgainâinsightâintoâwhatâdrivesâcountriesââsensitivityâtoâtheâGFC,â
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
âcapturesâtheâvariousâfeaturesâofâtheirâexternalâfunding,âwhichâweâletâ interact,âone-by-one,âwithâ
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
E.g. if
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
,â definedâ asâ assetsâminusâ liabilitiesâ scaledâ byâGDP,â
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
â indicatesâ theâ degreeâ toâwhichâ countriesââsensitivityâ toâ theâGFCâdependsâonâ theirâ netâ internationalâ investmentâposition.âWeâmakeâaâ similarâ assessmentâinâtheâotherâregressionsâthatâtestâforâtheârelevanceâofâtheâgrossâexternalâpositionâ
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
,âsumâofâexternalâassetsâ andâ liabilitiesâ scaledâbyâGDP),â grossâflowsâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
,â averageâofâ in-â andâoutflowsâ scaledâbyâGDP)âandânetâflowsâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
,âdifferenceâbetweenâout-âandâinflowsâscaledâbyâGDP).â
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
âcapturesâcountryâfixedâeffectsâandâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
âisâaâvectorâofâlaggedâmacroeconomicâcontrolâvariablesâtakenâfromâtheâliteratureâ(domesticâandâglobalâinflation,âdomesticâandâglobalârealâgrowth,ârealâandâfinancialâopennessâofâtheâcountryâ;âseeâRey, 2015â;âObstfeld et al., 2017â;âDavis et al., 2019â;âHabibâ&âVenditti, 2019âwhoâuseâaâsimilarâframeworkâtoâanalyseâtheâimpactâofâtheâGFC).
Allâ modelsâ areâ estimatedâ usingâ ordinaryâ leastâ squaresâ onâ aâ sampleâ ofâ euroâ areaâ countries.â Standardâ errorsâareâclusteredâatâ theâcountryâ level.âAllâ variablesâareâatâ theâquarterlyâ frequencyâ runningâ fromâ1990Q1-2017Q4.âStationarityâisâverifiedâalongâtheâlinesâofâIm,âPesaranâ&âShinâ(2003).âWeâdropâtheâcountriesâidentifiedâasââfinancialâcentresââ fromâ theâ analysisâ asâ theirâ grossâ capitalâ flowsâmateriallyâ exceedâGDPâ andâ areâ typicallyâ veryâ volatile. 1
1â WeâconsiderâLuxemburg,âMalta,âCyprus,âIrelandâandâtheâNetherlandsâasâfinancialâcentresâsinceâtheâsumâofâtheirâexternalâassetâandâliabilitiesâexceedsâ1000â%âGDP.âNoteâthatâBelgiumâisâaâborderlineâcaseâwithâtheâgrossâpositionâamountingâtoâ826â%âGDPâatâtheâendâof 2018â(dueâtoâaâlargeâshareâofâintragroupâloans).âAlso,âSloveniaâisâexcludedâfromâourâanalysisâdueâtoâaâlackâofâsufficientlyâlongâseries.
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
![Page 14: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/14.jpg)
14NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Allâmodelsâ includeâquarterlyâdummiesâandâaâ linearâtimeâtrend.â Inâunreportedâresults,âweâdocumentâtheâresultsâpresentedâbelowâtoâbeârobustâtoâvariousâmodificationsâtoâaforementionedâset-up. 1
Tableâ1 (inâannex)âsummarisesâtheâmainâresults.âColumnâ(1)âshowsâthatâdomesticâfinancialâconditionsâinâtheâeuroâareaâareâpositivelyârelatedâtoâtheâglobalâfinancialâcycle.âQuantitatively,âaâ1.0 standardâdeviationâ(s.d.)âincreaseâinâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
âisâassociatedâwithâaâ0.27 s.d.âincreaseâinâtheâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
.âItâisâworthâemphasisingâatâthisâpointâthatâtheâregressionâcoefficientâ indicatesâ correlationâ ââ notâ causalityâ (seeâ alsoâ Rey, 2015â forâ aâ discussion).â Importantly,â columnâ (2)ârevealsâ thatâ thisâ co-movementâ isâ strongerâ forâ countriesâ thatâ haveâ aâ negativeâ netâ externalâ positionâ (negativeâcoefficient).â Inâ orderâ toâ betterâ appreciateâ theâquantitativeâ significanceâofâ thisâ result,â specificationâ (3)â replacesâthe interactionâtermâwithâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
,âwhere
Theâ interactionâ termâ inâ columnâ (3)â indicatesâ thatâ thisâ increasedâ co-movementâ isâ sizeableâ:â inâ absoluteâ terms,âaâ1.0 s.d.â increaseâ inâ theâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
â isâonâaverageâassociatedâwithâaâ0.43 s.d.â (0.21 s.d.)â increaseâ inâ theâ
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
for countriesâwithâaânegativeâ (positive)ânetâexternalâposition.â Inâotherâwords,â theâdomesticâfinancialâconditionsâofâcountriesâ thatâ haveâ aâ negativeâ netâ externalâ positionâ comoveâ approximatelyâ twiceâ asâ strongâ withâ theâ globalâfinancialâcycleâthanâinâtheâcountriesâwithâaânon-negativeânetâposition.âSpecificationâ(4)âsuggestsâthatâthisâisâmoreâgenerallyâ trueâ forâ countriesâwithâaâ relativelyâ smallânetâexternalâposition 2.âCountriesâwithâaâ relativelyâ largeânetâpositionâseemâtoâbeâinsulatedâfromâtheâglobalâfinancialâcycle.â
1â Aâbatteryâofârobustnessâtestsâallâconfirmâourâbaselineâresultsâ:âe.g.â(i)âtrimming,â(ii)âtheâinclusionâofâfinancialâcentres,â(iii)âtheâuseâofâanâalternativeâmeasureâforâ
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!").
âtakenâfromâMiranda-AgrippinoâandâReyâ(2019),â(iv)âinclusionâofâlagsâofâtheâdependentâvariableâandâ(v)âDriscoll-Kraayâstandardâerrorsâ(toâaccountâforâpossibleâcross-sectionalâandâtemporalâdependenceâinâtheâerrorâtermâ
16
4. Empirical results on sensitivity to the global financial cycle
This section provides empirical evidence on (i) whether domestic financial conditions in the euro area countries move in line with the global financial cycle and (ii) to what extent this co-movement is magnified or attenuated by features of their external funding. Therefore, we let various variables âinteractâ with the GFC, such as the gross and net external position, as well as the gross and net capital flows. To that end, we estimate the following panel regression specification:
đšđšđšđšđšđš!" = đźđź! + đ˝đ˝đşđşđşđşđşđş! + đżđżđşđşđşđşđşđş!Ăđđ!" + đžđž!đĽđĽ!",!
!
!!!
+ đđ!"
in which đšđšđšđšđšđš!" denotes the domestic financial conditions index of country đđ and đşđşđşđşđşđş! is the global financial cycle (taken from Habib and Venditti, 2019). To gain insight into what drives countriesâ sensitivity to the GFC, đđ!" captures the various features of their external funding, which we let interact, one-by-one, with đşđşđşđşđşđş!. E.g. if đđ!" = đđđđđđđđ!", defined as assets minus liabilities scaled by GDP, đżđż indicates the degree to which countriesâ sensitivity to the GFC depends on their net international investment position. We make a similar assessment in the other regressions that test for the relevance of the gross external position (đđ!" = đşđşđşđşđşđşđşđş!", sum of external assets and liabilities scaled by GDP), gross flows (đđ!" =đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđśđśđśđś!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). ).
2â Countriesâwithâaânetâexternalâpositionâbelowâtheâfirstâquartileâareâconsideredâtoâhaveâaârelativelyâsmallâposition.âWeâalsoâtestedâtheâco-movementâwithâtheâGFCâofâcountriesâwithâNIIP<-35â%âGDPâasâthisâisâtheâthresholdâusedâinâtheâMIPâtoâidentifyâmacroeconomicâimbalances.âItâturnsâoutâthatâthoseâcountriesâareâtheâmostâsensitiveâtoâtheâGFC,âwhichâprovidesâanâadditionalâjustificationâforâcloseâmonitoringâofâtheseâcountries.
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
Chart 5
Other investment less stable than other funding sources
Direct investment Portfolio investment Other investment0
1
2
3
4
5
6
7
8
9
10
Pre-crisis Post-crisis
Average of capital in- and outflow 1(in % GDP)
Sourcesâ:âECB,NBB.1 Averagesâoverâpre-crisisâ(2002Q4-2007Q4)âandâpost-crisisâ(2008Q1-2018Q4)âperiodâforâeuroâareaâcapitalâin-âandâoutflows.
![Page 15: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/15.jpg)
15NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
The importance of the gross position
15
(đđ!" = đşđşđşđşđşđşđşđşđşđş đđđđđđđđđđ!" , average of in- and outflows scaled by GDP) and net flows (đđ!" = đđđđđđ đđđđđđđđđđ!", difference between out- and inflows scaled by GDP). đźđź! captures country fixed effects and đđ!" is a vector of lagged macroeconomic control variables taken from the literature (domestic and global inflation, domestic and global real growth, real and financial openness of the country; see Rey, 2015; Obstfeld et al., 2017; Davis et al., 2019; Habib & Venditti, 2019 who use a similar framework to analyse the impact of the GFC).
All models are estimated using ordinary least squares on a sample of euro area countries. Standard errors are clustered at the country level. All variables are at the quarterly frequency running from 1990Q1-2017Q4. Stationarity is verified along the lines of Im, Pesaran & Shin (2003). We drop the countries identified as âfinancial centresâ from the analysis as their gross capital flows materially exceed GDP and are typically very volatile.1 All models include quarterly dummies and a linear time trend. In unreported results, we document the results presented below to be robust to various modifications to aforementioned set-up.2
Table 1 (in annex) summarises the main results. Column (1) shows that domestic financial conditions in the euro area are positively related to the global financial cycle. Quantitatively, a 1.0 standard deviation (s.d.) increase in đşđşđşđşđşđş! is associated with a 0.27 s.d. increase in the đšđšđšđšđšđš!". It is worth emphasising at this point that the regression coefficient indicates correlation â not causality (see also Rey, 2015 for a discussion). Importantly, column (2) reveals that this co-movement is stronger for countries that have a negative net external position (negative coefficient). In order to better appreciate the quantitative significance of this result, specification (3) replaces the interaction term with đşđşđşđşđşđş!Ăđđđđđđđđ!",!!, where
đđđđđđđđ!",!! =1 đđđđ đđđđđđđđ!" < 0 0 đđđđ đđđđđđđđ!" ⼠0
The interaction term in column (3) indicates that this increased co-movement is sizeable: in absolute terms, a 1.0 s.d. increase in the đşđşđşđşđşđş! is on average associated with a 0.43 s.d. (0.21 s.d.) increase in the đšđšđšđšđšđš!" for countries with a negative (positive) net external position. In other words, the domestic financial conditions of countries that have a negative net external position comove approximately twice as strong with the global financial cycle than in the countries with a non-negative net position. Specification (4) suggests that this is more generally true for countries with a relatively small net external position3. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI4. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most
1 We consider Luxemburg, Malta, Cyprus, Ireland and the Netherlands as financial centres since the sum of their external asset and liabilities exceeds 1000% GDP. Note that
Belgium is a borderline case with the gross position amounting to 826% GDP at the end of 2018 (due to a large share of intragroup loans). Also, Slovenia is excluded from our analysis due to a lack of sufficient ly long series.
2 A battery of robustness tests all confirm our baseline results: e.g. (i) trimming, (ii) the inclusion of financial centres, (iii) the use of an alternative measure for đşđşđşđşđşđş! taken from
Miranda-Agrippino and Rey (2019), (iv) inclusion of lags of the dependent variable and (v) Driscoll-Kraay standard errors (to account for possible cross-sectional and temporal dependence in the error term đđ!"). 3
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
4 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
,âforâtheâeuroâareaâcountriesââsensitivityâtoâtheâGFCâisâanalysedâinâcolumnâ(5).âContraryâtoâourâexpectationsâandâtheâliteratureâ(Rey, 2015),âweâfindâtheâgrossâpositionâinsignificantâforâ theâ co-movementâ ofâ domesticâ financialâ conditionsâ withâ theâ GFC.â Aâ possibleâ explanationâ mightâ beâ thatâfinancialâintegrationâinâtheâeuroâareaâhasâreachedâsuchâaâhighâlevelâthatâanâadditionalâincreaseâorâdecreaseâmakesânoâdifferenceâ forâ theâ transmissionâofâglobalâ factors.â Thisâmightâalsoâexplainâ theâdifferencesâ inâ relationâ toâ theâfindingsâ inâ theâ literatureâwhichâmainlyâholdâ forâemergingâeconomies,âwhichâareâ farâ lessâfinanciallyâ integratedâthanâtheâeuroâareaâcountries.â
InâTableâ2,âweâdisentangleâtheânetâexternalâpositionâofâtheâcountryâintoâthreeâsub-categoriesâ:âotherâinvestmentâ(OI),â directâ investmentâ (DI)â andâportfolioâ investmentâ (PI).âColumnâ (2)â revealsâ thatâ theâ increasedâ co-movementâarisesâmainlyâasâaâresultâofânetâpositionsâinâOIâandâââtoâaâsmallerâextentâââDI 1.âTheâcrucialâroleâofâotherâinvestmentâinâcountriesââsensitivityâtoâtheâGFCâisânotâsurprisingâasâtheâliteratureâalsoâfoundâthatâtheâGFCâhadâtheâstrongestâimpactâonâ theâotherâ investmentâ capitalâflowsâ (HabibâandâVenditti, 2019).âMoreover,âBroner et al.â (2013)âandâBussière et al.â(2016)âshowedâthat,âaroundâcrises,âotherâinvestmentâexperiencesâtheâsharpestâdrop.âInâparticular,âtheâbanksââdebtâfundingâflowsâprovedâtheâmostâsensitiveâtoâtheââsuddenâstopââduringâthe 2008 financialâcrisisâ(Milesi-FerretiâandâTille, 2011).âConsistentâwithâtheâfindingsâofâBussière et al.â(2016),âotherâinvestmentâdisplayedâtheâhighestâvolatilityâofâallâcapitalâflowsâinâtheâeuroâareaâduringâtheâfinancialâcrisis,âwhileâdirectâinvestmentâwasâfarâmoreâstableâ(seeâChart 5).âConsequently,âfinancialâconditionsâinâcountriesâthatâfinanceâthemselvesâmoreâthroughâotherâinvestmentâareâmoreâlikelyâtoâreflectâanâinherentâboomâ/âbustâprofile.
Whileâanâadvantageâofâourâanalysisâ isâ thatâweâaggregateâallâfinancialâconditionsâ intoâoneâfigure,â itâmightâalsoâbeârelevantâtoâlookâatâtheâreactionâofâtheâvariousâfinancialâsub-indexes.âRememberâthatâtheâ
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
is a composite indicatorâwhichâaggregatesâfiveâriskâdimensionsâ(creditâdevelopments,ârealâestate,âprivateâsectorâdebt,âbankingâsectorâ andâ financialâ marketâ conditions).â Weâ thereforeâ decomposeâ ourâ resultsâ fromâ Tableâ 1 columnâ (3)â andâreplaceâ
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
âwithâeachâofâ theseâfiveâ riskâdimensions.â Figureâ6 plotsâ theâ coefficientâofâ
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
and
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
+
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
for each subcategoryâofâ
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
.âCoefficientâ
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
(
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
+
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1 Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with
NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
)âthenâquantifiesâtheâco-movementâofâtheâvariousâfinancialâconditionsâwithâthe globalâfinancialâcycleâforâcountriesâwithâaâpositiveâ(negative)ânetâexternalâposition.
Interestingly,âtheâdecompositionâshowsâaâdiverseâpicture.âCreditâdevelopmentsâandâprivateâsectorâleverageâtendâ toâ respondâ asâ expectedâ:â lendingâ andâ leverageâ increaseâ inâ countriesâ withâ aâ negativeâ NIIPâ andâ theâimpactâofâanâupturnâ inâtheâGFCâ isâmagnified.âNoteâthatâtheââshieldingââofâcountriesâwithâaâpositiveâNIIPâisâstrongestâ forâcreditâdevelopments,âalthoughâtheâcoefficientâ isânotâsignificant.âAlso,â realâestateâmarketsâinâ countriesâ withâ aâ negativeâ NIIPâ areâ moreâ vulnerableâ toâ GFCâ movements.â However,â theâ responseâ isâsmallerâthanâinâtheâcaseâofâcreditâandâleverageâdevelopments.âMoreover,âasâshownâbyâtheâcountercyclicalâreactionâinâcountriesâwithâaâpositiveâNIIP,âhouseâpricesâ inâtheâeuroâareaâtendâtoâbehaveâdifferently,âwhichâconfirmsâtheâevidenceâthatâtheseâmarketsâinâtheâeuroâareaâareââseparatedâalongânationalâlinesâ.âTheâmostâcounter-intuitiveâresultsâareâfoundâforâtheâbankingâsectorâandâtheâfinancialâmarkets.âForâtheâbankingâsector,âstatisticalâsignificanceâmightâplayâaâroleâasâthisâsub-indicatorâholdsâlessâobservationsâthanâtheâotherâindices.âInâtheâcaseâofâfinancialâmarkets,âsafe-havenâflowsâaddressedâtoâcountriesâwithâaâpositiveâNIIP,âinâparticularâGermany,âmightâplayâaârole.
Inâsum,âourâresultsâpresentedâinâthisâsectionâshowâthatâdomesticâfinancialâconditionsâinâtheâeuroâareaâtendâtoâco-moveâstronglyâwithâ theâglobalâfinancialâ cycle,â inâparticularâ inâ thoseâcountriesâ thatâhaveâaânegativeânetâ internationalâ investmentâ positionâ andâ financeâ themselvesâ throughâ otherâ investment.â Theâ impactâ ofâcapitalâ flowsâonâ sensitivityâ toâ theâGFCâ isâ analysedâ inâ theânextâ sectionâ againstâ theâbackgroundâofâ recentâdevelopments.
1â Weâperformedâaâsimilarâestimateâwithâtheâbreakdownâofâtheâgrossâposition.âAllâdetailedâgrossâpositionsâareâinsignificantâandâthusâconfirmâtheâresultâforâtheâtotalâgrossâposition.
![Page 16: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/16.jpg)
16NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
5. Recent developments and role of capital flows
Theâ effectsâ ofâ theâGFCâ andâ sensitivityâ toâ itsâ boomâ/âbustâ profileâwereâmostâ evidentâ duringâ theâ financialâ crisisâof 2008.âToâestimateâ theâ impactâofâaâ futureâglobalâ shock,â itâ isâworthwhileâ toâanalyseâwhetherâpoliciesâ inâ theâeuroâareaâsinceâthenâhaveâ(intentionallyâorâunintentionally)âcontributedâtoâaâreductionâinâsensitivityâtoâtheâGFC.â
Regardingâ theâNIIP,âChart 7 (leftâpanel)â showsâaâmixedâpicture,âwithâaboutâhalfâofâ theâcountriesâ recordingâanâimprovementâsinceâtheâcrisis.âWhileâinâtheâmajorityâofâcountriesâaâflowâadjustmentâtookâplaceâ(i.e.âanâimprovementâinâtheâcurrentâaccount),âstockâimbalancesâasâmeasuredâbyâtheâNIIPâhaveâbeenâpersistentâinâtheâeuroâarea.âSomeâofâtheâlargestânetâdebtorsâevenâsawâaâfurtherâdeteriorationâinâtheirânegativeâNIIP,âgivenâslowâeconomicâgrowthâandâ theâcostâofâ theâdebtâburden 1.â Inâ theâ lightâofâ this,â asâmentionedâbefore,â theâNIIPâ isâmonitoredâunderâ theâMacroeconomicâ ImbalanceâProcedureâ (MIP),âandâNIIP<-35%âGDPâcanâbeâconsideredâasâexcessive.âDespite thisâmonitoringâframework,âlargeânegativeâNIIPâvaluesâcontinueâtoâexist.
Aâgeographicalâbreakdown,âshowsâthatâaâlargeâpartâofâtheânetâclaimsâandâliabilitiesâisâvis-Ă -visâotherâeuroâareaâcountries 2,âwhichâalsoâexplainsâwhyâatâtheâlevelâofâtheâeuroâarea,âtheâconsolidatedâNIIPâisâonlyâslightlyânegative.âThisâ isâ important,âasâ itâ showsâthatâ theâsignificanceâofâ theâNIIPâ forâsensitivityâ toâ theâGFCâ isâ ratherâdueâtoâdebtâsustainabilityâissuesâ(theâoverallâNIIPâfigure),âthanâtoâspill-overâeffectsâcomingâfromâtheâdirectâholdingsâofâexternalâassetsâandâliabilitiesâ(theâextra-euroâareaâpartâofâtheâNIIP).âSo,âevenâifâaâcountryâdoesânotâholdâextra-euroâareaâassets,âmostâlikelyâitâwouldâstillâfindâitselfâvulnerableâtoâtheâglobalâfinancialâcycleâdueâtoâdebtâsustainabilityâissues.âTheâ importanceâofâ intra-euroâareaâbalancesâandâcapitalâflowsâ forâfinancialâ conditionsâhasâalsoâbeenâ raisedâbyâMerlerâ(2015).
1â Thisâcostâisâreflectedâinâdevelopmentâofâtheâinvestmentâincomeâbalanceâ(partâofâtheâbalanceâofâpayments).âItâshouldâbeânotedâthatâinâtheâaftermathâofâtheâfinancialâandâsovereignâdebtâcrisis,âpoliciesâinâtheâeuroâareaâlimitedâtheseâcostsâforâtheânetâdebtorsâviaâmonetaryâpolicyâandâtheââofficialââESMâfunding.
2â BasedâonâtheâFinflowsâdatabaseâofâtheâEuropeanâCommissionâ(JRC-ECFIN),âpre-releaseâversionâofâJuly 2019.âTheâFinflowsâdababaseâcontainsâyearlyâbilateralâfinancialâinvestmentâpositionsâbetweenâOECD,âEU,âandâoffshoreâcountriesâ(stocksâandâflows)âfrom 2001 to 2017.
Chart 6
Some financial conditions are more sensitive than others 1
Credit developments Banking sector Private debt Real estate Financial marketsâ0.008
â0.006
â0.004
â0.002
0
0.002
0.004
0.006
0.008
0.01
0.012
Positive NIIP Negative NIIP
Sourceâ:âNBB.1 Y-axisâshowsâtheâcoefficientâonâtheâglobalâfinancialâcycle,âcalculatedâasâ
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
(
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
+
17
small net external position1. Countries with a relatively large net position seem to be insulated from the global financial cycle.
The importance of the gross position (đđ!" = đşđşđşđşđşđşđşđş!"), for the euro area countriesâ sensitivity to the GFC is analysed in column (5). Contrary to our expectations and the literature (Rey, 2015), we find the gross position insignificant for the co-movement of domestic financial conditions with the GFC. A possible explanation might be that financial integration in the euro area has reached such a high level that an additional increase or decrease makes no difference for the transmission of global factors. This might also explain the differences in relation to the findings in the literature which mainly hold for emerging economies, which are far less financially integrated than the euro area countries.
In Table 2, we disentangle the net external position of the country into three sub-categories: other investment (OI), direct investment (DI) and portfolio investment (PI). Column (2) reveals that the increased co-movement arises mainly as a result of net positions in OI and â to a smaller extent â DI2. The crucial role of other investment in countriesâ sensitivity to the GFC is not surprising as the literature also found that the GFC had the strongest impact on the other investment capital flows (Habib and Venditti, 2019). Moreover, Broner et al. (2013) and Bussière et al. (2016) showed that, around crises, other investment experiences the sharpest drop. In particular, the banksâ debt funding flows proved the most sensitive to the âsudden stopâ during the 2008 financial crisis (Milesi-Ferreti and Tille, 2011). Consistent with the findings of Bussière et al. (2016), other investment displayed the highest volatility of all capital flows in the euro area during the financial crisis, while direct investment was far more stable (see Chart 5). Consequently, financial conditions in countries that finance themselves more through other investment are more likely to reflect an inherent boom/bust profile.
Chart 5 - Other investment less stable than other funding sources
Sources: ECB, NBB. (1) Averages over pre-crisis (2002Q4-2007Q4) and post-crisis (2008Q1-2018Q4) period for euro area capital in- and outflows.
While an advantage of our analysis is that we aggregate all financial conditions into one figure, it might also be relevant to look at the reaction of the various financial sub-indexes. Remember that the đšđšđšđšđšđš!" is a composite indicator which aggregates five risk dimensions (credit developments, real estate, private sector debt, banking sector and financial market conditions). We therefore decompose our results from Table 1 column (3) and replace đšđšđšđšđšđš!" with each of these five risk dimensions. Figure 6 plots the coefficient of đ˝đ˝ and đ˝đ˝ + đżđż for each subcategory of đšđšđšđšđšđš!". Coefficient đ˝đ˝ (đ˝đ˝ + đżđż) then
1
Countries with a net external position below the first quartile are considered to have a relatively small position. We also tested the co-movement with the GFC of countries with NIIP<-35% GDP as this is the threshold used in the MIP to identify macroeconomic imbalances. It turns out that those countries are the most sensitive to the GFC, which provides an additional justification for close monitoring of these countries.
2 We performed a similar estimate with the breakdown of the gross position. All detailed gross positions are insignificant and thus confirm the result for the total gross position.
).âVerticalâbarsâindicateâtheâ95%âconfidenceâintervalâsurroundingâtheâestimatedâcoefficients.
![Page 17: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/17.jpg)
17NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Chart 7
Has the net financial position improved since the financial crisis ?
NIIP : stock imbalances remain
NL MT DE LU BE AT FI IT FR SI EE LT LV SK ES PT CY EL IE
â150
â100
â50
0
50
100
2018Q4
Pre-crisis (2007Q4)
DE BE FI FR AT IT EE SK LV LT SI ES
â100
â80
â60
â40
â20
0
20
40
60
80
100
Imbalances also vis-Ă -vis euro area countries 1(2017)
Euro area
Extra euro area
Total economy (i.e. net external position)
Sourcesâ:âEC,âECB,âNBB.
1 GeographicalâbreakdownâbasedâonââFinlowsâdatabaseââofâtheâEuropeanâCommissionâ(JRC-ECFIN).âPre-releaseâversionâofâJuly 2019.
Chart 8
Has the financing mix improved since the financial crisis ?
23
42
35
Direct investment Portfolio investment Other investment
33
39
29
Share of other investment dropped in outstanding liabilities of the euro area
2007Q4
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011
2013
2015
2017
2019
â10
â5
0
5
10
15
20
25
... but recent worrisome decline of direct and portfolio investment flows ?
2018Q4
(in % GDP, average yearly in- and outflow, 1990Q1-2019Q1)
Sourcesâ:âECB,âNBB.
![Page 18: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/18.jpg)
18NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Whileâ theâNIIPâ didâ notâ improve,â adjustmentâ inâ theâ financingâ instrumentsâmightâ haveâ contributedâ toâ aâ lowerâsensitivity.âSinceâtheâfinancialâcrisis,âweâhaveâseenâaâdropâinâtheâdependenceâonâotherâinvestment.âBothâcapitalâin-âandâoutflowsâofâotherâinvestmentâdeclinedârelativeâtoâotherâfinancingâinstruments.âAsâaâresult,âtheâshareâofâotherâinvestmentâinâtheâoutstandingâliabilitiesâofâtheâeuroâareaâwasâdownâfromâ35 %âatâtheâendâof 2007 toâ29 %âin 2018.âMostâofâthisâreductionâcanâbeârelatedâtoâtheâdropâinâtheâcross-borderâfundingâofâbanks,âwithâtheâlatterâre-focusingâonâtheirâdomesticâmarkets.âWhileâthisâisâaâpositiveâtrendâinâviewâofâourâresults,âtheâdevelopmentâofâtheâotherâfundingâsourcesâisânotâirrelevant.âInâthatâcontext,âweânoticeâaârecentâsetbackâinâallâ(gross)âcapitalâflows,âwithânegativeâflowsâforâdirectâandâportfolioâinvestment.â
Inâorderâtoâshedâsomeâ lightâonâtheâ importanceâofâtheârecentâcapitalâflowsâweâperformâadditionalâestimations,âwhereâweâletâtheâdifferentâflowsâinteractâwithâtheâglobalâfinancialâcycle.âWeârunâestimationsâforâbothâgrossâandânetâflows,âandâ forâ theirâbreakdownâbyâ instrumentâ (Tableâ3).âTheâ resultsâ confirmâthatâ sensitivityâ toâ theâGFCâ isâmainlyâdrivenâbyâotherâinvestmentâ(flows).âInâlineâwithâtheâresultâforâtheâpositions,âitâisâtheânetâratherâthanâtheâgrossâflowsâwhichâareâsignificant.âViewedâinâtermsâofâexposureâtoâtheâGFC,âtheâcurrentâsetbackâinâgrossâcapitalâflowsâisâthereforeânotânecessarilyâgoodâorâbadânews,âalthoughâitâdoesâindicateâaâdeclineâinâfinancialâintegration.
TheâfindingâthatânetâflowsâareâmoreâsignificantâcorroboratesâtheâideaâthatâsustainabilityâissuesâareâatâtheârootâofâsensitivityâtoâtheâGFC.âGrossâflows,âtogetherâwithâtheâgrossâposition,âareâlessâimportant.âTheseâfindingsâcontrastâwith those of Farhi et al.â(2012)âandâReyâ(2015)âforâaâsampleâofâemergingâandâadvancedâcountries.âWeâattributeâourâfindingâtoâtheâfactâthatâgrossâflowsâandâpositionsâmightâloseâsomeâofâtheirâsignificanceâinâtheâeuroâareaâgivenâtheâlevelâofâfinancialâintegrationâreachedâandâtheâsmallerâpotentialâforâmismatchesâbetweenâassetsâandâliabilitiesâ(e.g.ânoâexchangeâriskâonâtheâeuroâareaâexposures).
6. Policy implications
Besidesâtheâfactâthatâdomesticâfinancialâconditionsâinâtheâeuroâareaâseemâstronglyâlinkedâtoâtheâglobalâfinancialâcycleâ(GFC),âourâeconometricâresultsâshowâthatâcross-countryâsensitivityâtoâtheâGFCâdependsâcruciallyâonâtheânetâinternationalâinvestmentâposition.âCountriesâwithânetâliabilitiesâreactâtwiceâasâstronglyâasâcountriesâthatâhaveânetâassets.âMoreover,âespeciallyâthoseâwhichâfinanceâthemselvesâbyâotherâinvestmentâ(mainlyâdebtâfundingâofâbanks)âproveâvulnerableâtoâtheâboomâ/âbustâprofileâofâtheâglobalâfinancialâcycle.
Theseâ observationsâ haveâ variousâ importantâ policyâ implicationsâ forâ macroprudential,â monetaryâ andâ structuralâpoliciesâ andâ theâ co-ordinationâ betweenâ theseâ domains.â Inâ thisâ sectionâweâdiscussâ theâ rationaleâ behindâ theseâlessons.
First,â theâ importanceâofâ theâglobalâ financialâ cycleâ forâ euroâ areaâfinancialâ conditionsâ addsâ aânewââtargetââ forâmacroprudentialâ policyâ:âmitigatingâ andâpreventingâ exposureâ toâ theâ boomâ/âbustâ profileâ ofâ theâ globalâ financialâcycle 1.âAnâeffectiveâmacroprudentialâpolicyâ indeedâ requiresâ closeâmonitoringâofâ theâglobalâ factorsâ influencingâdomesticâfinancialâconditions.âMoreover,âitâprovidesâsupportâtoâtheâideaâthatâmacroprudentialâpolicyâinâtheâeuroâareaâ shouldâbeâdifferentiatedâacrossâmemberâ states,â takingâ intoâaccountâ cross-countryâ variationsâ inâ sensitivityâtoâtheâGFC.â
Whileânationalâpoliciesâcanâinâgeneralânotâinfluenceâtheâglobalâcycle,âtheyâcertainlyâcanâtakeâmeasuresâtoâinfluenceâtheirâexposureâtoâthisâcycle.âOurâresultsâclearlyâshowâthatâifâaâcountryâwishesâtoâreduceâitsâexposureâtoâtheâGFC,âitâcouldâeitherâlimitâtheâsizeâofâitsânetâliabilitiesâorâchangeâitsâfinancingâmix.âTheâmostâefficientâwayâtoâimproveâtheâNIIPâmightâbeâbyâreducingâotherâinvestmentâliabilities,âi.e.âtheâcross-borderâdebtâfundingâofâdomesticâbanks.â
1â Macroprudentialâpoliciesâareâstillâatâtheâdevelopmentâstage,âandâinâpracticeâstillâlargelyâinâsearchâofâclearâtargets.âAccordingâtoâSmetsâ(2014),âmacroprudentialâpolicyâshouldâhaveâfourâtargetsâ:âi)âmitigateâandâpreventâexcessiveâcreditâgrowthâandâleverage,âii)âmitigateâandâpreventâexcessiveâmaturityâandâliquidityâmismatch,âiii)âlimitâexcessiveâexposureâconcentrationsâandâiv)âlimitâbail-outâexpectations.âWeâthusâaddâtoâthisâ:âmitigateâandâpreventâexposureâtoâtheâboomâ/âbustâprofileâofâtheâglobalâfinancialâcycle.
![Page 19: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/19.jpg)
19NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Noteâthatâforâbothâremediesâ(i.e.âimprovingâtheâNIIPâandâmakingâitsâcompositionâmoreârobust),âthereâareâalreadyâpoliciesâ inâ placeâwithinâ theâ EU,â althoughâ theyâ doânotâ intentionallyâ âtargetââ theâ exposureâ toâ theâGFC.â TheseâpoliciesâareâpartâofâtheâstructuralâmacroeconomicâframeworkâwithinâtheâEUâ:âtheâEuropeanâSemesterâ(withinâtheâMIP,âNIIPsâ<35 %âGDPâcanâbeâqualifiedâasâexcessive)âandâ initiativesâ suchâasâ theâbankingâunionâorâ theâCapitalâMarketsâUnionâ(CMU),âwhichâaimâtoâbroadenâtheâfinancingâsourcesâinâtheâEUâandâmakeâthemâmoreârobust.â
GivenâtheâchallengesâmacroprudentialâpolicyâmightâexperienceâtoâdirectlyâinfluenceâtheâNIIPâandâitsâcompositionâdueâtoâ theâ limitedâmacroprudentialâ toolkitâandâtheâdifficultyâofâgoingâbeyondâbank-relatedâflows,â theseâotherâ(structural)âpoliciesâhaveâanâimportantâroleâtoâplay.âItâshouldâalsoâbeânotedâthatâwithinâtheâEMU,âtheâmeasuresâshouldâ beâ inâ lineâ withâ theâ freeâmovementâ ofâ capitalâ andâ shouldâ thusâ differâ fromâ capitalâ flowâmanagementâmeasuresâ(CFM) 1.
Secondly,âtheâstrongâcorrelationâbetweenâdomesticâfinancialâconditionsâinâtheâeuroâareaâandâtheâglobalâfinancialâcycleâ tendsâ toâ confirmâ aâ financialâ dilemmaâ forâ theâ euroâ area,â alongâ theâ linesâ ofâ Reyâ (2015)â forâ emergingâeconomies 2.â Suchâ aâ dilemmaâ impliesâ thatâ wheneverâ theâ financialâ accountâ isâ open,â monetaryâ andâ financialâconditionsâ areâ largelyâ inâ theâ handsâ ofâ globalâ factorsâ andâ lessâ inâ thoseâ ofâ anâ independentâ monetaryâ policy.âWe showâthatâthisâdilemmaâinâtheâeuroâareaâisâparticularlyâpresentâwhenâcountriesâhaveâaânegativeânetâexternalâposition.
Consequently,â asâ aâ thirdâ lesson,â apartâ fromâ theâ callâ byâ someâ forâ internationalâmonetaryâ policyâ co-ordinationâ(Rajan, 2014),âthisâcallsâinâtheâeuroâareaâforâco-ordinationâbetweenâmacroeconomicâ(structural),âmacroprudentialâandâmonetaryâ policyâ inâ orderâ toâ reachâ theirâ objectives.â Addressingâ theâ negativeâ externalâ positionâ and,âmoreâbroadly,â ensuringâ debtâ sustainability,â asâ isâ currentlyâ doneâ underâ theâ Europeanâ Macroeconomicâ ImbalanceâProcedureâ(MIP),âwouldâmostâlikelyâhelpâtoâinsulateâtheâcountriesâduringârisk-onâ/ârisk-offâglobalâregimes,âtherebyâalsoâcontributingâtoâfinancialâstabilityâobjectivesâandâindependentâmonetaryâconditionsâinâtheâeuroâarea.
Finally,âourâworkâoffersâanâ interestingâbasisâ forâ furtherâanalysisâ inâ theâdomainâofâ internationalâfinance,âandâ inâparticularâtheâtransmissionâofâglobalâshocksâandâtheâpolicyâimplicationsâforâtheâeuroâarea.âItâencouragesâresearchâthatâ looksâ intoâ theâneedâ forâ co-ordinationâbetweenâpolicyâ domainsâ asâwellâ asâ theâneedâ forâ internationalâ co-operation.âAlso,âitâillustratesâtheâpotentialâofâclosingâtheâdataâgaps,âsuchâasâaâdetailedâgeographicalâandâsectoralâbreakdownâofâtheâNIIP.âBasedâonâtheâlatter,âadditionalâinsightsâmightâbeâobtainedâregardingâcountriesââsensitivityâtoâtheâGFCâandâtheâassociatedâtransmissionâmechanisms.
1â CFMsâ(IMF, 2012)âareâdefinedâasâmeasuresâthatâareâdesignedâtoâlimitâcapitalâflowsâviaâadministrativeâandâprice-basedârestrictionsâonâcapital flows.
2â Inâaârecentâupdateâ(Miranda-AgrippinoâandâRey,â2019)âalsoâquestionsâtheâmonetaryâindependenceâofâlargeâandâadvancedâeconomies,âsuchâasâtheâeuroâarea.
![Page 20: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/20.jpg)
20NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Conclusion
Inâ thisâ article,âweâ analysedâwhetherâ domesticâ financialâ conditionsâ inâ theâ euroâ areaâ countriesâ areâ drivenâ byâ aâglobalâfinancialâcycle.âToâmeasureâthisâeffect,âweâconstructedâaâfinancialâconditionsâindexâ(FCI)âforâtheâeuroâareaâcountries,â summarisingâ theirâ domesticâ financialâ conditions,â andâ comparedâ thisâ indexâwithâ aâmeasureâ forâ theâglobalâfinancialâcycleârelyingâonâtheârecentâliteratureâ(HabibâandâVenditti, 2019).
Ourâresultsâcontributeâtoâaâburgeoningâliteratureâonâtheâglobalâfinancialâcycleâ(GFC),âwhichâmainlyâlooksâintoâtheâeffectâofâtheâGFCâonâcapitalâflowsâofâemergingâeconomies.âWeâcomplementâtheseâresultsâwithâfindingsâregardingâtheâimpactâofâtheâGFCâonâdomesticâfinancialâconditionsâinâtheâeuroâareaâcountries.
First,âweâfindâaâclearâfinancialâcycleâforâtheâeuroâarea,âwithâpeaksâthatâcanâbeârelatedâtoâcrisisâevents.âThereâis,âhowever,âsubstantialâheterogeneityâacrossâtheâeuroâareaâcountries.
Secondly,âfinancialâconditionsâ inâtheâeuroâareaâareâstronglyâ linkedâtoâtheâglobalâfinancialâcycle.âHowever,âeuroâareaâcountriesâshowâvaryingâsensitivitiesâtoâtheâglobalâcycle.
Inâthisâarticleâweâlinkâthisâcross-countryâsensitivityâtoâtheâglobalâcycleâtoâvariousâdeterminants,âincludingâtheâsizeâandâ compositionâofâ theâexternalâfinancialâ position.âAâkeyâfindingâ isâ thatâ sensitivityâ seemsâ toâdependâ cruciallyâonâ theânetâ internationalâ investmentâposition.âCountriesâwithânetâ liabilitiesâ seemâ toâ reactâ twiceâ asâ stronglyâ asâcountriesâthatâhaveânetâassets.âAmongâtheâcountriesâwithânetâliabilities,âespeciallyâthoseâwhichâfinanceâthemselvesâbyâotherâ investmentâ (mainlyâdebtâ fundingâofâbanks)âproveâ vulnerableâ toâ theâboomâ/âbustâprofileâofâ theâglobalâfinancialâcycle.
Ourâ resultsâhaveâseveralâpolicyâ implications.âFirst,â itâ isâusefulâ forâmacroprudentialâpolicyâ toâmonitorâ theâglobalâfinancialâcycleâandâ/âorâhelpâtoâaddressâextremeâsensitivityâtoâitsâboomâ/âbustâprofile.âSecondly,âtheâstrongâcorrelationâbetweenâfinancialâconditionsâinâtheâeuroâareaâandâtheâglobalâfinancialâcycleâtendsâtoâconfirmâaâfinancialâdilemmaâforâtheâeuroâarea,âalongâtheâlinesâofâReyâ(2015)âforâemergingâeconomies.âSuchâaâdilemmaâimpliesâthatâwheneverâtheâfinancialâaccountâ isâopen,âmonetaryâconditionsâareâ largelyâ inâtheâhandsâofâglobalâfactorsâandâlessâ inâthoseâofâanâ independentâmonetaryâpolicy.âWeâshowâthatâthisâdilemmaâ inâtheâeuroâareaâ isâparticularlyâpresentâwhenâcountriesâhaveâaânegativeânetâexternalâposition.
Atâtheâsameâtime,âourâresultsâcallâforâco-ordinationâbetweenâmacroeconomicâ(structural),âmacroprudentialâandâmonetaryâ policyâ toâ reachâ theirâ objectives.â Addressingâ theâ negativeâ netâ externalâ positionâ and,â moreâ broadly,âensuringâdebtâsustainability,âasâisâcurrentlyâdoneâunderâtheâEuropeanâMacroeconomicâImbalanceâProcedureâ(MIP),âwouldâmostâlikelyâhelpâtoâinsulateâtheâcountriesâduringârisk-onâ/ârisk-offâglobalâregimes,âtherebyâalsoâcontributingâtoâfinancialâstabilityâobjectivesâandâindependentâmonetaryâconditionsâinâtheâeuroâarea.
![Page 21: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/21.jpg)
21NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Annex : Tables
Table 1
Co-movement of the domestic / global financial cycle and role of NIIP
Dependent variable Baseline NIIP level Negative NIIP High vs. Low NIIP GIIP Level
FCIit (1) (2) (3) (4) (5)
GF Ct 0.058***
(0.01)
0.037***
(0.01)
0.028**
(0.01)
0.041***
(0.01)
0.050**
(0.02)
NIIPit Ă GF Ct â0.0005**
(0.00)
NIIPit,<0 Ă GF Ct 0.030**
(0.01)
NIIPit,low Ă GF Ct 0.041*
(0.02)
NIIPit,high Ă GF Ct â0.032
(0.03)
GIIPit Ă GF Ct 0.000
(0.00)
Domestic inflationitâ1 â0.686
(0.56)
â0.413
(0.57)
â0.487
(0.59)
â0.358
(0.55)
â0.544
(0.60)
Domestic growthitâ1 â0.270
(0.20)
â0.409*
(0.22)
â0.473*
(0.23)
â0.369
(0.21)
â0.531**
(0.22)
World inflationtâ1 0.045
(0.14)
0.041
(0.14)
0.028
(0.14)
0.087
(0.13)
â0.006
(0.15)
World growthtâ1 â0.511
(0.82)
â1.002
(0.94)
â0.894
(0.92)
â1.029
(0.91)
â0.789
(0.89)
N 1.454 1.136 1.136 1.136 1.136
R2 adj. 0.399 0.410 0.396 0.428 0.388
Countries 13 13 13 13 13
Macroeconomic controls x x x x x
Country fixed effects x x x x x
Notes : The asterisks *, **, and *** indicate statistical significance at the 10 %, 5 % and 1 % level, respectively. Standard errors are reported in parentheses. The dependent variable, FCIit, is the domestic financial conditions indicator. GF Ct proxies the global financial cycle and is taken from Habib and Venditti (2019). NIIPit = (External assetsit â External liabilitiesit) / GDPit quantifies the net external position of country i. NIIPit,<0 is a dummy variable taking the value 1 if NIIPit <0. Indicator variable NIIPit,low (NIIPit,high) takes the value 1 if the country has a net position below (above) the first (third) quartile. GIIPit = (External assetsit + External liabilitiesit) / GDPit quantifies the gross external position of country i. The set of national control variables also includes Financial opennessit = (External assetsit â External liabilitiesit) / GDPit and Real opennessit, which is a dummy variable if the sum of a countryâs exports and imports (over GDP) is larger than the cross-sectional mean. All specifications include a linear time trend and quarterly dummies.
![Page 22: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/22.jpg)
22NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Table 2
Sensitivity to GFC and type of external funding
Dependent variable Net position Negative net position
FCIit (1) (2)
GF Ct 0.042***
(0.011)
0.020
(0.023)
Net OIit Ă GF Ct â0.001
(0.000)
Net DIit Ă GF Ct â0.000
(0.000)
Net PIit Ă GF Ct â0.000
(0.000)
Net OIit,<0 Ă GF Ct 0.049**
(0.017)
Net DIit,<0 Ă GF Ct 0.026*
(0.015)
Net PIit,<0 Ă GF Ct 0.004
(0.016)
N 858 858
R2 adj. 0.598 0.556
Countries 13 13
Macroeconomic controls x x
Country fixed effects x x
Notes : The asterisks *, **, and *** indicate statistical significance at the 10 %, 5 % and 1 % level, respectively. Standard errors are reported in parentheses. The dependent variable, FCIit, is the domestic financial cycle indicator. GF Ct proxies the global financial cycle and is taken from Habib and Venditti (2019). Net PIit is the net portfolio investment position of country i, scaled by GDP. A similar definition applies to Net DIit (direct investment) and Net OIit (other investment). Net PIit,<0 is an indicator variable taking the value 1 if Net PIit<0 (similarly for DI and OI). All specifications include a linear time trend and quarterly dummies.
![Page 23: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/23.jpg)
23NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Table 3
Co-movement of the domestic / global financial cycle and role of net capital flows
Dependent variable Direct investment (DI)
Portfolio investment (PI)
Other investment (OI) Total investment
FCIit (1) (2) (3) (4)
GF Ct 0.036***
(0.01)
0.038***
(0.01)
0.036***
(0.01)
0.034***
(0.01)
Net flowsit Ă GF Ct â0.004***
(0.00)
â0.004***
(0.00)
â0.001
(0.00)
0.003
(0.00)
Net DI flowsit Ă GF Ct 0.001
(0.00)
â0.003
(0.00)
Net PI flowsit Ă GF Ct 0.003
(0.00)
â0.004
(0.00)
Net OI flowsit Ă GF Ct â0.004**
(0.001)
â0.008**
(0.00)
N 1.105 1.105 1.105 1.105
R2 adj. 0.430 0.443 0.457 0.460
Countries 13 13 13 13
Macroeconomic controls x x x x
Country fixed effects x x x x
Notes : The asterisks *, **, and *** indicate statistical significance at the 10 %, 5 % and 1 % level, respectively. Standard errors are reported in parentheses. The dependent variable is the domestic financial cycle indicator, FCIit. GFCt proxies the global financial cycle and is taken from Habib and Venditti (2019). Net flowsit is the difference between outâ (+) and inflows (â). Net DI flowsit, Net PI flowsit and Net OI flowsit break down the net capital flows into net flows of direct, portfolio and other investment, respectively.
![Page 24: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/24.jpg)
24NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Table 4
Co-movement of the domestic / global financial cycle and role of gross capital flows
Dependent variable Direct investment (DI)
Portfolio investment (PI)
Other investment (OI) Total investment
FCIit (1) (2) (3) (4)
GF Ct 0.028*
(0.01)
0.028*
(0.01)
0.030*
(0.01)
0.029*
(0.01)
Gross flowsit Ă GF Ct 0.002**
(0.00)
0.002
(0.00)
0.000
(0.00)
0.011
(0.01)
Gross DI flowsit Ă GF Ct â0.002
(0.00)
â0.011
(0.01)
Gross PI flowsit Ă GF Ct â0.001
(0.00)
â0.010
(0.01)
Gross OI flowsit Ă GF Ct 0.002
(0.00)
â0.009
(0.01)
N 1.105 1.105 1.105 1.105
R2 adj. 0.420 0.418 0.420 0.427
Countries 13 13 13 13
Macroeconomic controls x x x x
Country fixed effects x x x x
Notes : The asterisks *, **, and *** indicate statistical significance at the 10 %, 5 % and 1 % level, respectively. Standard errors are reported in parentheses. The dependent variable is the domestic financial cycle indicator, FCIit. GFCt proxies the global financial cycle and is taken from Habib and Venditti (2019). Gross flowsit is the average of the in- and outflows (% GDP) of country i. Gross DI flowsit, Gross PI flowsit and Gross OI flowsit break down the total gross capital flow into direct, portfolio and other investment, respectively.
![Page 25: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/25.jpg)
25NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
Bibliography
BISâ(2017),â87th AnnualâReport.
BorioâC.â(2012),âThe Financial Cycle and the Marco-economy : What have we learnt,âBISâWorkingâPaperâSeries,â395,âDecember.
BronerâF.,âT.âDidier,âA.âErceâandâS.âL.âSchmuklerâ(2013),ââGrossâCapitalâFlowsâ:âDynamicsâandâCrisesâ, Journal of Monetary Economics,â60,â113-133.
BrunoâV.âandâH.âShinâ(2015a),ââCapitalâFlowsâandâtheâRisk-TakingâChannelâofâMonetaryâPolicy,â Journal of Monetary Economics,â71(April),â119-132.
BrunoâV.âandâH.âShinâ(2015b),ââCross-BorderâBankingâandâGlobalâLiquidity,ââReview of Economic Studies,â82(2),â535-564.
BussièreâM.,âSchmidtâJ.âandâN.âVallaâ(2016),âInternational financial flows in the new normal : Key patterns (and why we should care),âEuropeanâInvestmentâBank,âWorkingâPaper 2016â/â02.
CeruttiâE.,âS.âClaessensâandâA.K.âRoseâ(2017),âHow Important is the Global Financial Cycle ? Evidence from Capital Flows,âNationalâBureauâofâEconomicâResearch,âWorkingâPaperâ23699.
DavisâJ.S.,âG.âValenteâandâE.âvanâWincoopâ(2019),âGlobal drivers of gross and net capital flows,âFederalâReserveâBankâofâDallas,âGlobalizationâInstitute,âWorkingâPaperâ357.
DrehmannâM.,âC.âBorioâandâK.âTsatsaronisâ(2012),âCharacterising the financial cycle : donât lose sight of the medium term,âBISâWorkingâPaperâSeries,â380,âJune.
ECBâ(2019),âCyclical systemic risk measurement,âMPAGâWorkâStreamâonâCyclicalâRiskâAnalysis,âOccasionalâWorkingâPaper,âforthcoming.
ESRBâ(2014),âRecommendation of the European systemic risk board of 18 June 2014 on guidance for setting countercyclical buffer rates,â(ESRBâ/â2014â/â1).
ForbesâK.J.âandâF.E.âWarnockâ(2012),ââCapitalâFlowâWavesâ:âSurges,âStops,âFlight,âandâRetrenchment", Journal of International Economicsâ88,â235-251.
GhoshâA.,âM.âQureshi,âJ.âKimâandâJ.âZalduendoâ(2014),ââSurgesâ,âJournal of International Economics,â92(2),â266-285.
GourinchasâP.-O.âandâH.âReyâ(2014),ââExternalâAdjustment,âGlobalâImbalances,âValuationâEffectsâ,â4 of Handbook of International Economics,â585-645,âElsevier.
HabibâM.âandâF.âVendittiâ(2018),âThe global financial cycle : implications for the global economy and the euro area,âECB,âEconomicâBulletin,âIssueâ6â/â2018.
HabibâM.âandâF.âVendittiâ(2019),âThe global capital flows cycle : structural drivers and transmission channels,âECB,âWorkingâPaperâ2280.
HollĂłâD.,âM.âKremerâandâM.âLoâDucaâ(2012),âCISS â A Composite Indicator of Systemic Stress in the Financial System,âECB,âWorkingâPaperâ1426.
![Page 26: Are we riding the waves of a global financial cycle in the](https://reader031.vdocuments.net/reader031/viewer/2022011803/61d37879e9161555ba445e59/html5/thumbnails/26.jpg)
26NBB Economic Review ÂĄ September 2019 ÂĄâ Areâweâridingâtheâwavesâofâa global financial cycleâinâtheâeuroâareaâ?
IMFâ(2012),âThe Liberalization and management of capital flows : an institutional view,âInternationalâMonetaryâFund,âWashingtonâDC.
KleinâM.âandâJ.âShambaughâ(2015),ââRoundingâtheâCornersâofâtheâPolicyâTrilemmaâ:âSourcesâofâMonetaryâPolicyâAutonomy,ââAmerican Economic Journal : Macroeconomics,â7(4),â33-66.
KrippnerâL.,â(2013),ââMeasuringâtheâStanceâofâMonetaryâPolicyâinâZeroâLowerâBoundâEnvironments,ââEconomics Letters,â118(1),â135-138.
LaneâP.âandâG.âM.âMilesiâFerrettiâ(2001),ââTheâexternalâwealthâofânationsâ:âmeasuresâofâforeignâassetsâandâliabilitiesâforâindustrialâandâdevelopingâcountriesâ,âJournal of International Economicsâ55(2),â263-294.
LaneâP.âandâG.âM.âMilesiâFerrettiâ(2017),âInternational Financial Integration in the aftermath of the global financial crisis,âIMF,âWorkingâpaper,âWPâ/â17â/â115,âMay.
LaneâP.âandâP.âMcQuadeâ(2013),âDomestic credit growth and International capital flows,âECB,âWorkingâ papersâ1566,âJuly.
LoâDucaâM.,âA.âKobanâ,âM.âBasten,âE.âBengtsson,âB.âKlaus,âP.âKusmierczyk,âJ.âLang,âC.âDetkenâandâT.âPeltonenâ(2017),âA new database for financial crises in European countries,âECB,âOccasionalâPaperâSeriesâ194.
MerlerâS.â(2015),âSquaring the cycle : financial cycles, capital flows and macroprudential policy in the euro area,âBruegel,âWorkingâPaper 2015â/â14.
Milesi-FerrettiâG.-M.âandâC.âTilleâ(2011),ââTheâGreatâRetrenchmentâ:âInternationalâCapitalâFlowsâduringâtheâGlobalâFinancialâCrisisâ,âEconomic Policyâ26(66),â285-342.
Miranda-AgrippinoâS.âandâH.âReyâ(2019),âUS Monetary Policy and the Global Financial Cycle,âNationalâBureauâofâEconomicâResearch,âWorkingâPaperâ21722.
ObstfeldâM.,âJ.âD.âOstryâandâM.âS.âQureshiâ(2017),âA tie that binds : revisiting the trilemma in Emerging Market Economies,âIMF,âWorkingâpaperâ17â/â130,âJune.
ObstfeldâM.,âJ.âShambaugh,âandâA.âTaylorâ(2005),ââTheâTrilemmaâinâHistoryâ:âTradeoffsâamongâExchangeâRates,âMonetaryâPolicies,âandâCapitalâMobilityâ,âReview of Economics and Statistics,â87(3),â423-438.
RajanâR., 2014,âCompetitive Monetary Easing : Is It Yesterday Once More ?âRemarksâmadeâatâtheâBrookingâInstitution,âWashingtonâDC,âAprilâ10, 2014.
ReyâH.â(2015),âDilemma not Trilemma : The Global Financial Cycle and Monetary Policy Independence,â NBERâWorkingâPaperâSeries,â21162,âMay,âRevisedâFebruary 2018.
SchĂźlerâY.,âP.âHiebertâandâT.âPeltonenâ(2015),âCharacterising the financial cycle : a multivariate and time-varying approach,âECB,âWorkingâPaperâ1846.
ShambaughâJ.â(2004),ââTheâEffectâofâFixedâExchangeâRatesâonâMonetaryâPolicy,ââThe Quarterly Journal of Economics,â119(1),â301-352.
SmetsâF.â(2014),âFinancial Stability and Monetary Policy : How Closely Interlinked ?,âInternationalâJournal ofâCentralâBanking.