Erdinc Akyildirima,b,g, Shaen Corbetc,h, Duc Khuong Nguyend,e,∗, Ahmet Sensoyf a Department of Banking and Finance, University of Zurich, Zurich, Switzerland b Department of Mathematics,
Trang 1Erdinc Akyildirima,b,g, Shaen Corbetc,h, Duc Khuong Nguyend,e,∗, Ahmet Sensoyf
a Department of Banking and Finance, University of Zurich, Zurich, Switzerland
b Department of Mathematics, ETH, Zurich, Switzerland
c DCU Business School, Dublin City University, Dublin 9, Ireland
d IPAG Business School, Paris, France
e International School, Vietnam National University, Hanoi, Viet Nam
f Faculty of Business Administration, Bilkent University, Ankara, 06800, Turkey
g Center for Financial Application and Research, Bo˘gazic¸ i University, Istanbul, Turkey
h School of Accounting, Finance and Economics, University of Waikato, New Zealand
Article history:
Received 11 July 2019
Received in revised form 4 December 2019
Accepted 31 March 2020
Available online 23 May 2020
JEL classification:
C58
F36
G01
G18
Keywords:
DCC-MIDAS
European Union
Sovereign bonds
Regulation
Financial crisis
Weestimatethetime-varyinglong-runcorrelationsofEuropeansovereignbondmarketstoidentify specificeffectsthatareattributedtochangingEuropeanregulatoryandpoliticaldynamicsoverthelast twentyyears.OurempiricalresultsfromusingtheDCC-MIDASmethodologyindicatethatregulatory changesinEuropehavecreatedsignificantandnegativeimpactonthelong-runcorrelationswithinthe monthwheretheregulationisdecidedtobetakenintoaction.Thisimpactstillremainsinthefollowing monthsandrobustwithrespecttothetrendcomponentofthelong-runcorrelations.Adirectimplication
isthatthemoreregulationstheEUattemptstoputinplace,thelowerthelong-runconvergenceprocessof sovereignbondmarketsis.Wethenanalysethestructuralshiftsinthelong-runcorrelationdynamicswith penalizedcontrastsmethodologyandtrytofindoutthereasonsoftheseseverechanges.Accordingly, someofthestructuralshiftsoverlapwiththedatesofalimitednumberofregulatorychanges,inaddition
tothemajorglobaleconomicandpoliticalevents
©2020ElsevierInc.Allrightsreserved
TheEuropeanUnionhaswitnessedsubstantialstructural,
reg-ulatoryandpoliticalchanges inthepasttwentyyearssincethe
introductionoftheeuro.Muchresearchhasfocusedonthe
devel-opmentofabroadconvergenceintheyieldsofEuropeanbonds
afterthedevelopmentofastrongmonetaryunion(Codognoetal.,
2003; Kim etal., 2006; Christiansen,2007).This is particularly
important due to the broad diversification effects that existed
throughthecreationofsuchacohortofsovereignstates,each
offer-ingquiteuniquestrengthsandskillstotheunion,withthesmallest
countriesseekingaddedeconomicsecuritythroughdiversification,
sharedskills,experiences, financingsources,and thereinforced
∗ Corresponding author at: IPAG Business School, 184 Boulevard Saint-Germain,
75006 Paris, France.
E-mail address: duc.nguyen@ipag.fr (D.K Nguyen).
bargainingstrengththatwasprovidedthroughsuchalarge num-berofcountrieswhennegotiatinginternationaltradeagreements However,thesecountriesalsoincorporatedbroadstructuralissues towardstheEuropeanMonetaryUnion(EMU),manifestinginwhat canonlybedescribedasoneoftheworstsovereigndebtcrises tak-ingplaceincountriessuchasGreece,Ireland,Portugal,Italy,Cyprus andSpain.Amongthesecountries,bothGreeceandIreland neces-sitatedthird-partyfinancialsupportandinterventionduetothe deep-rootednatureoftheirsovereignbankingcrises.1The devel-opmentoftheEMUhasalsowithdrawnbothmonetaryandmany
1 Specifically focusing on the Irish economic collapse, Corbet (2016) discusses the broad regulatory deficiency that existed in Ireland during the generation of the
‘Celtic Tiger,’ a period synonymous with the rapid expansion of the Irish economy, where the actions of its regulators and policy makers undoubtedly generated not only a catalyst to financial ruin, but also an incubator to strengthen its severity Banks were found to be firmly leveraged towards the Irish property market and the role
of leverage in financial markets created mispricing, to which the basic principles of https://doi.org/10.1016/j.irle.2020.105907
0144-8188/© 2020 Elsevier Inc All rights reserved.
Trang 2mit-igatedandalleviated.Thiswasevidentintheeconomiccollapse
ofthecountriesdenotedas‘PIIGS’asmonetarypolicyhadtobe
tailoredtotheneedsoftheEMUratherthantheneedsofspecific
nations
TheregulatoryresponsesmadebytheEuropeanUnionhave
beenquitestrongin thepost-crisis era.Followingtheoutbreak
ofthefinancialcrisis,Europeanregulatoryreformshavefocused
onfourkeyareas:(1)thestrengtheningoffinancialsupervision;
(2)thecreationoftoolstosupportbankrecoveryandresolution;
(3)thecreationofamoreeffectivedepositprotectionsystem;and
(4)thecreationofanimprovedregulatoryframeworkforbanks,
insurancecompanies, securitiesmarkets and othersectors We
mustfocusonanalysingastohowthesereformshavemadethe
financialsystemmorestableandresilientandastowhetherthey
haveinfluencedtheperceptionsofbondtradersasmeasuredbythe
yieldsofsovereigndebt.Suchregulatoryrestructuringnecessitates
re-evaluationofthemanywaysinwhichEuropeancorporations
interact,particularlycross-borderentitiesthatarepartofthesame
institution.Suchreformsintroducedafterthecrisisalsoneedto
bemonitoredtocheckwhethertheyaredeliveringintended
out-comesandtoassesswhetherthenewruleshaveanyunintended
consequences.AppendixAprovidesadetailedoverviewofthekey
completedreformsthathavebeenintroducedalongwiththe
ratio-nalesupportingtheirintroduction
Someoftheearliestcommissionedandnowcompleted
finan-cialreformsincludethoserelatedtotherisk-basedprudentialand
solvencyrulesforinsurers(SolvencyII),AIFMD,CRDIII,the
estab-lishmentoftheEuropeanSupervisoryAuthority,depositguarantee
schemes,derivativereformthroughEMIR,thecreationofthe
Sin-gleEuroPaymentsArea(SEPA),MIFID,andawiderangeofmarket
abuse and transparency reforms among others.2 The European
Commissiondevelopedsuchreformsthroughtheestablishmentof
anumberofpolicyadvisingexpertgroups,3representing
consulta-tivebodiessetupbytheCommissiontoprovideadvicesinrelation
tothepreparationoflegislativeactsandpolicyinitiativesusually
composedofexpertsappointedbyEUgovernments
Asfarasthesovereigndebtcrisisisconcerned,acommon
wis-domisthattheregulatorychangesaffectthedynamicsofsovereign
risksandthewaysthesovereignbondmarketsco-moveovertime
Inthispaper,wegiveacloselooktothisissuebyconsideringabroad
rangeoftheEuropeanregulatoryreformsaspotentialsourcesof
changingtime-varyingbond market behaviour.We alsodevote
ourattentiontosomeofthemanysignificantpoliticaleventsthat
haveoccurredduringthepasttwentyyearsinEuropeaspolitical
developmentsintheEuropeanUnion,whichhavebeenparticularly
extraordinaryinmorerecenttimes,playapivotalroleon
regula-tions.CorbetandLarkin(2018)brieflyreviewthesepoliticalshifts
the efficient market hypothesis (EMH) failed This miscalculation of risk was severe
and destructive for the real economy.
2 We must note that there are a wide-range of actions that have been established
but have not yet been completed These include a number of structural reforms on
banks, the creation of the European deposit insurance scheme (EDIS), rules on capital
requirements, the development of a EU framework on covered bonds, addressing
risks related to NPLs, insurance companies and sovereign bond-backed securities,
and the strengthening of bank recovery and resolution (BRRD) among others A
summary of these development can be found in Appendix B.
3 The key financial regulation groups established in accordance with
Declara-tion 39 on Article 290 of the Lisbon Treaty include the expert groups on Banking,
Payments and Insurance; Sustainable Finance; Corporate Bond Market Liquidity;
Cross-border redress in financial services; Derivatives and Market Infrastructures
Member States; European Crowdfunding Stakeholders Forum; European Post Trade
Forum; the European Securities Committee; the Expert Group on barriers to free
movement of capital; intra-EU cross border investment environment; the
evalua-tion of the IAS Regulation; Retail Financial Services; Mortgage Credit; the Group of
representatives of financial services employees (UNI Europa); the Payment Systems;
andshowthatthelatterhavedevelopedwithinthewidespread financialcrisesandbeenexacerbatedinsomestatesbythe dra-maticinfluxofillegalimmigration.Asaresult,Europefindsitself
atacrossroadsinspiredbypoliticalspectrumshiftstotheleftand theright,withfear,uncertaintyfuellingnationalistrevoltacrossa hostofEuropeannations.Suchpoliticalshiftshavealsomanifested
inthenationalist-baseddecision-makingresultinginthegrowth
ofright-baseddecisionssuchastheBrexitortheItalianbudgetary issueswitnessedinrecentyears.Muchevidencehasbeenprovided thatcontagioneffectsexistsinsuchpoliticaldecision-making(Mei andGuo,2004;Rajsingh,2016).Itisthusopportunetoidentifyas
towhethersuchpoliticaldevelopmentsareasourceofinstability fortime-varyingsovereigndebtinstability,withfurtheremphasis
onthepresenceofcontagioneffects
Oneofthekeydatathroughwhichwecanidentifyboththe severityandcontagioneffectsofcrisesisthroughsovereignbond yields.Ourresearchshiftsattentiontothelong-runrelationship
ofsovereignbondmarkets,insteadoftheirdivergencesincethe rapiddevelopmentofboththeEuropeanregulatoryandpolitical environments.Inparticular,ourmethodologicalchoicefocuseson theinherenttimevariationsinsuchstructuraldestabilizations,i.e.,
astowhetherEuropeansovereignbondmarketsexperience long-termstructuraldestabilizationintheaftermathofchangesinthe regulatoryandpoliticalenvironments.Thisbuildsontheworkof Colacitoetal.(2011)whointroducedtheDCC-MIDASmethodology
toanalysethelong-runcorrelationcomponentsbetweenfinancial timeseries.Accordingtoourperspective,whilepoliticalinstability hasquitestrongtheoreticalgroundingforproducinginfluenceon sovereignbondmarkets,itisveryimportanttofurtherunderstand
astowhetherfinancial marketsthemselveswerein agreement withtheEuropeanCentralBank’sviewsthatitsregulatoryactions wereinfactfosteringtheresilientandsustainabledevelopmentof Europe’sfinanciallandscape.Withinthiscontext,suchregulatory interventionmightbeobservedasbeneficialtofinancialstability, butcouldalsobedetrimentaltosectoralandregionalprofitability, growthanddevelopment
Accordingtoouranalysis,wecontributetotheliteratureby showingthattheregulatorychangesinEuropehavesignificantand negativeimpactonthelong-runcorrelationsofthesovereignbond marketsofthemajoreurozonecountries.Thesecorrelationsalso changedrasticallywithinthemonthwheretheregulationis imple-mentedandthischangeispreservedwithinthefollowingmonths, showingthateffectsofregulatorychangesintheEUarenot transi-toryanddosustainonthecorrelationdynamicsofthesesovereign bondmarkets.Wecheckwhetherthisfindingisdistortedbythe trendcomponentsofthelong-runcorrelationsornot,andreveal thattheresultsarerobust.Adirectimplicationandoneofthemain contributionsofthepaperisthefindingthatthemoreregulations theEUattemptstoputinplace,thelowerthelong-runconvergence processofsovereignbondmarketsis.Nextand asarobustness check,wefocusondetectingpotentialstructuralshiftsinthe long-runcorrelationsandexaminingwhethertheyareassociatedwith regulatorychanges.Byapplyingthepenalizedcontrasts methodol-ogyofLavielle(2005)todetectthechangepoints,weshowthatthe structuralshiftsinthelong-runcorrelationsoccuraroundthetimes whenmajorregulatorychangesorimportantpoliticaleventstake place(suchasthecriticalstagesoftheBrexitprocess),supporting ourviewthatbothpoliticaluncertaintiesandregulatoryactionsare driversofthelong-runrelationshipbetweenthesovereignbonds
ofmajoreurozonemembercountries
Therestofthepaperisorganizedasfollows.Section2presents
aconcisereviewoftheliteraturebasedonsovereignbond dynam-ics,structuralchangesinEuropeanregulatorydynamicsandthe influenceofbroadpoliticaleventsonfinancialmarkets.We pro-videabriefreviewoftheDCC-MIDASmethodologyinSection3 Section4reportsthedatathatweuseandthecorresponding
Trang 3empir-Fig 1. Fully and partially implemented regulatory changes in Europe, 2006–2018 Note: The above data represents all proposals of financial reform that are finalized-implemented or being planned by the European Commission Data available at https://ec.europa.eu.
icalresults.Section5providessomediscussionsandconcluding
remarks
TheEuropeanresponsetotheinternationalfinancialcriseshas
generated a broad range of both anticipated and unanticipated
consequencesformultiplesovereignstatesacrossarangeofboth
economicandpoliticalenvironments.Withinthissection,we
pro-vide a thorough overview of thekey dynamicsthat have been
observedwithinEuropeansovereigndebtmarkets,whichfurther
suggeststhekeyidentifieddriversofinstabilitysourcedwithin
eco-nomicandpoliticaldriversofbondmarketvolatilityandcontagion
Kimetal.(2006)wereamongthefirstresearchersthat
empir-ically investigate the influence of the EMU ontime variations
ininter-stock-bondmarketintegration/segmentationdynamicsto
findthatrealeconomicintegrationandthereductionof
currency-marketrisksupportedfinancialintegration,butinfactgenerated
aflight-to-safetyeffectduetobroadfearsaboutthefutureofthe
EMU.Christiansen(2007)echoedsuchevidenceofEMU
integra-tionintheperiodaftertheintroductionoftheeurowiththekey
driveridentifiedtobethatofinterestrates.Corbet(2014)found
thatEuropeansovereigndowngradesarefoundtobeassociated
withanincreaseinequityreturnsandcausesignificantincreasesin
thecostofinsuringdebtthroughCDSandtheyieldofgovernment
debt.Inarecentstudy,Sensoyetal.(2019)uncoveredahighdegree
ofsovereigndebtmarketintegrationbetweentheEMUmembers
overtheperiodprecedingtherecentfinancialcrises,while
seg-mentationisfoundafterwards.However,theFed’staperingpolicy
announcementin2013generatedanimpacttowardsanintegration
ofthesemarketsagain
Bessembinderetal.(2006)foundthatchangesinmarketdesigns
throughregulationscanhavefirst-order effectsontrade
execu-tioncostsonbondsevenforsophisticatedinstitutionalinvestors
HeathcoteandPerri(2002)foundthatthefinancialautarkymodel
can generate volatility in the terms of trade when
construct-ing a two-country, two-good model, to account for observed
cross-countryoutput,consumption,investmentandemployment
correlations.Sucharesultidentifiedthatinternationalcapitalflows areexceptionallyimportantfortheinternationalbusinesscycle ThecreationoftheEMUwouldhavegreatlyincreasedthiseffect Theseverityofthe2008–2009globalfinancialcrisisand the Europeansovereigndebtcrisisthatfollowedwaswidelyobserved
as a criticalpoint in thesharp changes inregulatorydynamics thatfollowedinEurope.ThisisobviousinFig.1whichpresents evidence of the timeline of introduction of regulatory changes
in Europe,whilefurtherconsideringthe announcementof reg-ulatorychangesthathavenotbeenimplementedyet.Mohland Sondermann (2013) foundthat statementsabout restructuring, bailout andtheinvolvementof theEuropeanFinancial Stability Facility (EFSF) have impacted bond spreadsof countries in the peripheryoverGermany,indicatingthatthemoredifferenteuro areagovernmentsissuedstatementsatthesametime,themore bondspreadshaveincreased.Furthermore,theauthorsfindthat statementsfrompoliticiansfromAAA-ratedcountriesseemedto haveaparticularlystrongimpactonspreads.LierseandSeelkopf (2016) foundthat in thecontext of financial market pressures
intheformofrising bondyields,Europeangovernments raised theirtaxes,especiallyinthemoreregressivefieldofindirecttaxes, suggesting thatcapitalist democracieshave littlepolitical room
to maneuver and to conduct redistributive politics at times of highfiscal stress.Katsikas (2011)foundthat theEU’sdecisions
toadoptthestandardsproducedbytheInternationalAccounting StandardsBoard(IASB),andtoestablishanew,differentiated Euro-peanaccountingregulatorymechanism,weredrivenbyitsdesire
tobolsterEuropeaninfluence
DeGrauweetal.(2017)foundevidencethatasignificantpartof thesurgeinthesovereignbondspreadsoftheperipheralEurozone countrieswasdeterminedfromabroaddisconnectionfrom under-lyingfundamentalsandparticularlyfromacountry’sdebtposition Thiswasfoundtobemorelikely tobeassociatedwithmarket sentimentsandliquidityconcerns.Butlong-termpoliticalchanges havealsomanifestedintheincredibleeconomiceventsthathad taken placein countriessuchasCyprusand Italy (Michaelides, 2014;Deeg,2005).Benediktsdottiretal (2011)foundthat Ice-landicauthoritiesasamatterofpolicyencouragedthecreation
Trang 4deregulationofthebankingsystem,rules andregulationsbeing
relaxedandtheneglectof financialsupervision Thisinevitably
reducedsovereignfinancialdiversification
Mugge(2011)foundthatthreekeyresultswereevidentafterthe
EUhadtakenaroleinglobalfinancialgovernance.First,theEUhas
stabilized,ratherthanchallenged.Second,theEUcontinuestobe
oneoftwocentralnodesinGFG,whichessentiallystillisa
transat-lanticaffair,confoundingexpectationsthatEuropewouldfinditself
inamuchmoredispersedweboflinkswithotherregulatory
pow-ersaroundtheworld.Third,givenitsspecialinstitutionalcharacter,
therearesignsthataprominentEUmaytransformgovernance,but
itstillremainsunclearhowpronouncedthesedynamicswillbe
CorbetandLarkin(2017)foundthatEuropeancountrieswith
morelocalbankingnetworksintheformofcreditunions,
pub-lic banksor savings banks,generate greater levels of volatility
whencomparedtothatoftheircommercialcounterparts,
partic-ularlyincountrieswithmoremonopolisticsectors Further,the
announcementsoftheEuropeanBankingAuthoritygenerate
signif-icantvolatilityeffectsfortheEuropeanbankingsectoratlarge,with
particularemphasisonstresstestingresults,butalso
announce-mentsbasedonrecapitalization,regulationandtransparency.The
resultsindicatethatuniformityofregulationmayinfactbe
hin-dering and restricting the growthof somedomestic and more
peripheralandlocallydesignedbankingsectorsintheformofrules
designedforcommercialbankingoperations
Regardingourmethodology,severalstudieshaveusedthe
DCC-MIDAS technique to investigate theinteractions betweenEMU
markets.TheDCC-MIDASmainly differsfrom standard
GARCH-family models as it allows a baseline variance to vary slowly
throughoutthetimeperiodanalysed.VirkandJaved(2017)focused
specificallyonEuropeanstockmarketsbetween1990and2013
usingDCC-MIDAStoidentifyevidenceofsubstantialdivergence
fromGreekriskduringtheEuropeanfinancialcrisisperiod.In
par-ticular,cross-countryjoint relationshipsofconditionalvariance
andreturncorrelationsarefoundtobetypicallypositive.Boffelli
etal (2016) focused onboth thehigh and low frequency
cor-relations inEuropean governmentbonds viaDCC-MIDAS while
consideringtheireconomicdrivers.Theyfindstronglinksbetween
spreads’volatilityand worseningmacroeconomicfundamentals
Accordingly,relativespreadsmovetogetherinpresenceof
sim-ilarmacroeconomicfundamentals;yettheincreasingcorrelation
inspreadsduringtheburstofthesovereigndebtcrisiscannotbe
entirelyascribedtomacroeconomicfactorsbutrathertochanges
in market liquidity Nitoi and Pochea (2019) analysed the
co-movementsand contagionin 24EuropeanUnionstockmarkets
from2004to2016usingtheDCC-MIDASmethodologyandemploy
agravity-typeregressiontoinvestigatethedeterminantsof
long-term correlations They obtained mixed findings for long-term
correlations’driversincontagiontimes,revealingapurecontagion
thatisnotexplainedbyfundamentalsandawake-upcallinterms
ofcross-borderbankflows
Asstatedintheintroductionsection,themajorgoalofthispaper
istoexaminetheimpactofregulatorychangesonthestructural
interdependenciesofEMUsovereignbondmarketsaswellasto
discussitsimplicationsforthefutureofregulations.Weempirically
proxythestructuralinterdependenciesofthesemarketsbytheir
long-rundynamicyieldcorrelationswhichwillbeobtainedbythe
DCC-MIDASmethodology(Colacitoetal.,2011)
Considerasetofnsovereignbondsandletthevectorofdaily
changesintheiryieldsbedenotedbyrt=[r1,t, ,rn,t]obeying
thefollowingprocess:
rt∼i.i.d.N(,Ht)
Ht=DtRtDt
(1)
where isthevectorof unconditionalmeans,Ht isthe condi-tionalcovariancematrixandDtisadiagonalmatrixwithstandard deviationsonthediagonal,and
Rt=Et −1[tt]
t=D−1t (rt−)
(2)
Themodelaboveisestimatedintwoconsecutivesteps:(i)the conditionalvolatilitiesinDtareestimated,and(ii)theconditional correlationmatrixRtisobtained
3.1 GARCH-MIDASestimation
WestartwiththeworkofEngleetal.(2013)whoproposeto sep-aratevolatilitydynamicsintoshort-andlong-termcomponents Thisstructureusesamean-revertingunitdailyGARCHprocess sim-ilartoEngleandRangel(2008),andaMIDASpolynomialwhich appliestolowerfrequencyvariables
Wedenotetheshort-and long-runvariancecomponentsfor bondibygiandmirespectively.Wekeeplong-runcomponentmi constantacrossthedaysofthelowfrequencyperiod.Ni
vdenotes thenumberofdaysthatweholdmifixed.Thetwoletterstand denotetime-scales.Inparticular,gi,tmovesdailywhereasmi,only onceeveryNi
vdays
Weassumethatforeachbondi,univariatedailyyieldchanges followtheGARCH-MIDASprocesswithtwovariancecomponents:
ri,t=i+
mi,×gi,ti,t where t=(−1)Nvi+1, ,Nvi (3) The short-run variance component of returnsfollows a simple mean-revertingunitGARCH(1,1)process:
gi,t=(1−˛i−ˇi)+˛i(ri,t−1−i)2
mi, +ˇigi,t−1 (4) with˛i>0,ˇi≥0and˛i+ˇi<1forstationarity.Theshort-run componentgi,taccountsfordailyfluctuationsthat areassumed short-lived,i.e.,itrelatestoday-to-dayconcerns
Thelowfrequencycomponentmi,isaweightedsumofKi
vlags
ofrealizedvariances(RV)overalonghorizon:
mi,=mi+i
K i v
l=1
wheremiandiarefreeparameterstobeestimatedwithmi>0 and0≤i<1toguaranteeacovariancestationaryprocess.The
mi, is a trend component and relates to the effects of future expectedglobal/macro-economicvariablesonvolatility
WhilesettingNi
vequaltothenumberoftradingdayswithin
amonth,therealizedvariancesinvolveNi
vdailynon-overlapping squaredreturnsasfollows:
RVi,=
N i v
t =(−1)N i
v +1
Asa weightingfunction, weusea betafunction withdecay parameterωi
v:
ϕl(ωvi)=
(1− l
K i v )ω i
v −1
K i v
j =1(1− j
K i v )ω
i
wheretheweightattachedtopastrealizedvarianceswilldepend
ontwoparametersωiandKi.Forallω >1,theweightingscheme
Trang 5deter-minedbythesizeofωv.Large(small)valuesofωvgeneratearapidly
(slowly)decayingpattern.Byconstruction,ϕl(ωv)arenon-negative
andsumtoone
3.2 DCC-MIDAS
Inthisstep,wecalculatethecorrelationsbasedonthevolatility
adjusted(standardized)residualsi,tobtainedinSection3.1:
qij,t=¯ij,(1−a−b)+ai,t−1j,t−1+bqij,t−1
¯ij,=
Kcij
l =1
ϕl(ωc)cij,−l
cij,=
Ncij k=(−1)N ij
c +1i,kj,k
Ncij
k=(−1)N ij
c +1
2 i,k
Ncij k=(−1)N ij
c +1
2 j,k
(8)
whereaandbarethedrivingparametersofthecorrelation
pro-cesswitha,b>0anda+b<1forstationarity;andtheweighting
schemeϕl(ωc)for correlationsissimilartothatoneusedinEq
(7).AsintheGARCH-MIDASequation,thelong-run(slowly
mov-ing)correlation ¯ij, doesnotvaryatdailyfrequencyt but ata
lowerfrequency,anditisaweightedsumofKcijlagsofrealized
correlations (i.e.,Kcij arespanlengths of historicalcorrelations),
calculatedonNcij dailynon-overlappingreturns(i.e.,Ncij arethe
laglengths).Whereas,thedailyconditionalcorrelationsbetween
sovereignbondsiandjcaneasilybecalculatedbyusingtime
vary-ingcovariancesqij,tasshownbyEngle(2002),i.e.,
ij,t= qij,t
qii,t
Thistwo-componentstructureallowsustoobservetheshort()
andlong(¯)rundynamicsofthecorrelations.Theparametersof
theDCC-MIDASareestimatedbymaximizingthefollowing
quasi-likelihoodfunction
QL =−
T
t =1
(nlog(2 )+2log|Dt|+rtD−2t rt)
−
T
t =1
(log|Rt|+tR−1t t+t)
(10)
Thefirstsumin Eq.(10)contains thedataand thevariance
parameters (coming fromGARCH-MIDAS estimation)while the
secondsumisbasedonvolatilityadjustedresidualsandthe
corre-lationparameters
4.1 Sampledata
We use daily 10-year benchmarksovereign bond yields for
a sample of eleven countries in our analysis.4 Sample
coun-tries are the major and also the earliest eurozone members;
Austria,Belgium,Finland,France,Germany,Greece,Italy,Ireland,
Netherlands,Portugal,andSpain.Thedataisobtainedfrom
Thom-sonReutersDatastreamanditcoversatimeperiodfromJanuary
4, 1999 (the introduction of euro) until May 28, 2019, which
4
specificallyincludesthevariousphasesoffinanciallinkagesinthe Europeansovereignbondmarketsoverthelast20years
Fig.2showsthechangesinsovereignbondyieldsoftheselected countriesoverthesampleperiod.Weobserveaclearconvergence
of yieldswiththeintroductionofeuro wherethis convergence keepsitspatternuntilthebeginningoftheEuropeansovereigndebt crisisin2009.Inparticular,withasharpincreaseinitssovereign bondyield,Greece demonstratesphasesofdivergencefromthe rest.Inthefollowingperiod,asimilardivergenceisalsoobserved betweentheyieldsofthecountriesthatstrugglewithdebt(Ireland, Italy,PortugalandSpain)andthosethatareviewedassafehaven (FranceandGermany),suggestingaperiodofflight-to-qualityby investorsinthesemarkets
Table1presentsthedescriptivestatisticsofthedailychanges (takenasthedifferenceinyieldsinconsecutivedays),aswellasthe stationaritytestresults.Wecanseethatallyieldshaveanegative dailyaveragechangeshowingthatcostofborrowinghasdecreased forallthesamplecountriesinourstudyperiod.Greecehasthe lowestdailyaverage(−0.0007)asexpectedduetothesustained periodsofhighyields,especiallyduringthe2011–2012sovereign bondcrisisphase.Forthesamereason,italsohasthehighestyield increase(3.947)inasingleday
TheunconditionalvolatilityoftheGreeksovereignbondyields (0.54),measuredbystandard deviations,isalmostfourtimesof Portugal(0.15),thecountryhavingthenexthighest bondyield volatility.Yieldchangedistributionsareskewedtotherightexcept for the Greece, Italy, Ireland and Spain Also, all yield changes exhibitexcesskurtosis(fattails),withGreecehavingan outstand-ingvalueof660.Clearly,skewnessandkurtosiscoefficientsindicate thatreturnseriesarefarfromnormallydistributed.Thisdeparture fromnormalityisconfirmedbytheJarque–Berateststatisticsthat rejectsnormalityatthe1%levelforallseries.5
Table1alsopresentstheunitroottestresultforthe stationar-ityofourdailychangeseries(unitroottestscontainaconstant) AugmentedDickey–Fuller(ADF)testrejectsthenullhypothesisof unitrootforalltheseriesunderconsiderationatthe1%significance level,indicatingthatallthedailyyieldchangeseriesarestationary 4.2 Dynamicsofshort-andlong-runcorrelations
ThetoppanelofTable2displaystheestimationresultsforthe conditionalyieldchangevarianceswherethevaluesinthe paren-thesesbelowarethestandarderrorsoftheestimatedcoefficient Accordingly,mostoftheparametersaresignificantatthe1%level Thesumsof˛andˇvarywithintherangelimitedby0.77and 0.999frombelowandaboverespectively,thereforesatisfyingthe stationarityboundary˛+ˇ<1
Thedecayparameterωvissubstantiallylargerthan1for major-ityofthebonds,indicatingthatweightofthelagsdecreasesrapidly whencalculatingrealizedvariances.Ontheotherhand,this param-eterisalmost1forAustria,FranceandGermany,implyingaflat weightingfunctionforthesecountries.Estimationresultsforthe MIDAS correlationsare provided inthelower panel.Thedecay parameterωc impliesa moderatelevel ofdecreasingweighting function.Theaandbparametersarebothhighlysignificant,and theirsumof0.995whichisverycloseto1,suggestingalong-run correlationwithahighlypersistentstructure
Inourwork,thereare11countriesunderconsideration,which makesthebilateralanalysisimpracticalsincewewouldhaveto analyse55differentcorrelationstructures.Instead,weproceedas follows.Foreachday,wetaketheequallyweightedaverageofthe dailyyieldchangesofthesamplesovereignbonds.Thistimeseries
5 In the tables throughout this paper, *, ** and *** denote significance at the 10%,
Trang 6Fig 2.European Sovereign Bond Yields from 1999 to 201.
Table 1
Summary statistics and results of unit root tests for the first differences in EMU sovereign bond yields.
Notes: Asymptotic critical values for the ADF test are −3.43, −2.86 and −2.57 for 1%, 5% and 10% significance levels respectively We use the standard acronyms in the column tables for the country name abbreviations In the last column, Benchmark refers to the cross-sectional equally-weighted daily yield changes of all sample sovereign bonds.
actuallyshowhow usefulDCC-MIDASmodelscanbein
under-standingthestructuralchangesinthedependenciesbetweenthe
samplesovereignbondyields.Forexample,DCCtakesminimum
andmaximumvaluesof−0.16and0.87respectivelyforGermany, givingusarangegreaterthan1.Ontheotherhand,DCC-MIDASis confinedtotheinterval(0.36,0.76)forthesamecountry.Table3 presentsthedescriptivestatisticsoftheDCCandtheDCC-MIDAS forallsamplecountries.Thestabilityofthelatterisobserved eas-ilywhenwecomparethestandarddeviations.Inmanycases,the unconditionalvolatilityof theDCCisaroundtwiceofthe DCC-MIDAS,andinsomeextremecasessuchasIreland,thisratiocan reachuptoalmost4.Thevolatilemovementof theshort-term correlationcomponentisalsoreflectedinthemeancorrelation val-ues.Withoutanexception,DCCmeanstaysbelowtheDCC-MIDAS
Trang 7Table 2
GARCH-MIDAS and DCC-MIDAS parameter estimates.
Notes: The top panel reports the estimates of the GARCH-MIDAS coefficients for the sovereign bonds The bottom panel reports the estimates of the DCC-MIDAS parameters Standard error of the coefficient estimates are given in the parenthesis The number of MIDAS lags is 20 for the GARCH process and 120 for the DCC process.
Fig 3. Long-run vs short-run dynamic correlations Note: The figure shows the dynamic correlations between individual sovereign bond yields and the aggregate EMU sovereign bond market yield.
correla-tionsbetweenindividualsovereignbondyieldsandtheaggregate sovereignbondmarketyield.Theshadedareasrefertothecalendar monthsoftheregulatorychanges.Thisfiguresuggeststhatthere mightbeasignificantrelationbetweentheregulatoryactionsand
Trang 8Table 3
Summary statistics and results of unit root tests for long-run (DCC-MIDAS) and short-run (DCC) correlations.
Notes: Asymptotic critical values for the ADF test are −3.43, −2.86 and −2.57 for 1%, 5% and 10% significance levels respectively.
Fig 4. Long-run dynamic correlations with applied regulatory changes Note: The dynamic correlations above are between individual sovereign bond yields and the aggregate EMU sovereign bond market yield The shaded areas denote the months of the regulatory changes.
¯
InEq.(11),¯tistheDCC-MIDASandtheD1mt isadummy
vari-ablethattakesoneinthecalendarmonththattheregulatoryaction
istaken,andotherwisezero.Wecallthistype1modelandit
basi-callyshowsuswhetherthereisasignificantrelationshipbetween
theregulatoryactionsandthelong-runcorrelationsamongbond
yieldswithinthemonthofactions.Table4displaystheestimation
results.AllcountriesexceptIrelandgeneratesignificantresults.We
furthercheckwhetherthissignificantimpactistransitoryornot,
soweestimatethesamemodeltypebutchangethedummy
vari-ableD1m
t toD2m
t wherethenewdummyvariabletakesoneinnot
onlythecalendarmonthoftheregulatoryactionbutalsointhe
followingcalendarmonth.Table4showsthatalldummyvariables
aresignificant,withhowevertheoppositesigncomparedtothe
estimationresultswhenweuseD1m
t Thispresentsaninteresting caseandmightbeanindicatoroftheinvestors’overreactiontothe regulatorychangeswithinashorttimeframe
Onemightarguethatthesignificantimpactoftheregulatory changesmightariseduetothetrendinthelong-runcorrelations Indeed,itmightactuallybethecasesincesomeofthelong-run correlationshavebeenfoundtobenon-stationaryasdisplayedin Table3.Tocontrolforthetrendeffect,weestimatethemodelin thefollowingequation:
¯
t=c0+c1D1mt +c2t1m+ t (12)
WecallthemodelinEq.(12)type2anditbasicallyshowsus whetherthereisasignificantrelationshipbetweentheregulatory actionsand thelong-runcorrelationsamongbondyieldswithin themonthofactionswhenwecontrolforthetrendinthe corre-lations.AccordingtoTable4,trendtermcoefficientsarefoundto
Trang 9Table 4
Impact of regulatory changes in the long-run dynamic correlations.
Notes: In this table, Model 1 and Model 2 represent the regressions without and with the trend variable, respectively D 1m
t refers to the coefficient of the dummy variable that takes the value one in the calendar month that the regulatory action is taken, and otherwise zero Similarly, D 2m
t refers to the coefficient of the dummy variable that takes the value one in the calendar month that the regulatory action is taken and also the following calendar month, otherwise zero t 1m (t 2m ) refers to the trend coefficient when
we estimate the Model 2.
Table 5
This table shows the dates of the long-run correlation shifts for each sovereign bond detected by Lavielle’s penalized contrasts methodology with allowance for different number of maximum break points.
(a) Panel A: Maximum number of allowed break points is three
(b) Panel B: Maximum number of allowed break points is six
(c) Panel C: Maximum number of allowed break points is nine
Note: This table demonstrates the shift dates in long-run correlations when we allow for maximum number of breaks equal to 3, 6 and 9 In the analysis, we cover all possible break structures when maximum number of breaks runs through 2 to 10.
significant
ofswitchingsignsisstillthere.Allinall,ouranalysisshowsthat
regulatorychangeshavesignificantand negativeimpactonthe
long-runrelationshipofsovereignbondyieldsofthesample
coun-triesand thissignificanceis robustwithrespecttothetrendin
thecorrelations.Hence,themoretheregulationstheEUattempts
toputinplaceoverthelongrun,thelowertheconvergence
pro-cess
4.3 Detectingstructuralshiftsinthelong-runcorrelations
Wenowrunarobustnesscheckoftheresultsintheprevious sectionbydetectingthestructuralshiftsinthelong-run correla-tionsand investigatingwhethertheseshiftsareassociatedwith theregulatorychangestheEuropeanUnionhasundertakenover ourstudyperiod.Todoso,weapplythestateoftheartpenalized contrastsmethodologybyLavielle(2005)tothecorrelationseries
todetectthechangepoints.Thedetailsofthemethodologyare pro-videdinAppendixC.Weprovidethemaximumpotentialnumber
ofchangepointsasaninputandreceivethedatesofchangesasthe output.Table5demonstratesthechange(orbreak)pointdatesfor eachcountry’ssovereignbondyieldcorrelationswhenweallow
Trang 10Table 6
Important events that potentially created shifts in the long-run correlations
contagion and the announcement of a tax
on UK banks and a deposit guarantee schemes for bank failures
https://www.ft.com/content/7e0186ac-71aa-11df-8eec-00144feabdc0 https://www.spiegel.de/international/germany/radical-cutbacks-german-government-agrees-on-historic-austerity-program-a-699229.html 05/07/2011 02/08/2011 30/08/2011 New Greek bailout tranche and a sharp
escalation of the Greek crisis
https://www.telegraph.co.uk/finance/economics/8620735/The-challenges-facing-Christine-Lagarde.html
20/12/2011 Regulation shifts/announcement https://www.telegraph.co.uk/finance/comment/damianreece/8971513/
Eurozone-zombies-follow-Mario-Draghis-cheap-money.html 14/02/2012 First major broad EU credit ratings cut
https://www.forbes.com/sites/afontevecchia/2012/02/13/moodys-cuts-peripherals-scrutinizes-france-and-britains-triple-as/
10/04/2012 Cypriot financial problems escalate sharply https://www.ft.com/content/f209b43c-8316-11e1-929f-00144feab49a 08/05/2012 European austerity measures are reported
to be broadly damaging numerous European real economies
https://www.telegraph.co.uk/news/worldnews/europe/eu/9252941/ Europe-austerity-crisis-Q-and-A.html
03/07/2012 Start of the Brexit process and major
regulatory developments (Insurance, BRRD, PRIPS, IMD, UCITS)
https://www.news24.com/World/News/Cameron-under-pressure-over-EU-referendum-20120702
https://www.independent.co.uk/news/world/europe/what-if-britain-left-the-eu-7904469.html
http://europa.eu/rapid/press-release MEMO-12-516 en.htm?locale=en 25/09/2012 Basel III regulations and rules on high
frequency trading
https://www.bis.org/publ/bcbs229.pdf 20/11/2012 New EU data protection regulation
https://www.lexology.com/library/detail.aspx?g=36c4f233-a484-41b2-9fc1-65e9db2cb0a0 18/12/2012 EU ODR regulation http://europa.eu/rapid/press-release MEMO-12-994 en.htm
15/01/2013 EU credit rating agencies regulation http://europa.eu/rapid/press-release MEMO-13-13 en.htm
10/03/2015 Beginning of EU quantitative easing
https://www.euractiv.com/section/euro-finance/news/ecb-euro-central-banks-begin-qe-stimulus-programme/
07/02/2017 EU securitization problem due to Brexit https://www.ft.com/content/b47104c6-ea32-11e6-893c-082c54a7f539 Note: The first column represents the long-run correlation shift dates that are common for at least five sample countries The second column provides the events that might
be associated with these shifts and the sources of these events are provided in the third column.
forthenumberofmaximumbreakpointsas3(PanelA),6(Panel
B)and9(PanelC)
Inourextendedanalysis,weallowforthemaximumnumberof
breakpointstorunfrom2to10,andthenselectthedatesthatare
commonforatleast5samplecountries.Table6providesthebreak
dateswiththepotentialreasonscausingtheshifts.Itisclearthat
thebreakdatesmostlyfallintotheyears 2011and2012,when
theEuropeansovereigndebtcrisisreacheditspeakwithvarious
regulationsbeingputintoplacetocontrolthesituation.Whenwe
takeadetailedlookatthepotentialsources,weseeavarietyof
regulatoryactionstakenbytheEuropeanCommissionandtheBasel
Committee.Inadditiontothose,majoreconomicaleventssuchas
thefearofcontagioninEurope,Greekbailoutprogramme,credit
ratingcutsintheEUandthequantitativeeasingintheeurozone
standoutaspotentialsources.Finally,weseethatpoliticalevents,
inparticularvariousstagesoftheBrexitprocess,alsoseemtohave
asignificantimpactonshapingthelong-runcorrelationsbetween
thecoreeurozonecountries’sovereignbondmarkets
Tosumup,long-runcorrelationsbetweentheEMUsovereign
bond markets are characterized by occasional structural shifts
whichmostlytookplaceduringthetimesofregulatorychanges
inordertodealwiththesovereigndebtcrisisorimportant
eco-nomicandpoliticalevents(e.g.,creditratingdowngradesinEurope
andtheBrexit).Thisfindingstrengthenstheargumentthatboth
political and economicuncertainties as wellas thekey
regula-toryactionsaredriversofthelong-runrelationshipbetweenthe
sovereignbondsofmajoreurozonemembercountries
ThisresearchidentifiesandexaminestheimpactofEuropean
regulatorychangesonthestructuralinterdependencies ofEMU
sovereignbondmarketsaswellastodiscussitsimplicationsfor
thefutureofregulations.Tocompletethistask,weutilizethe
DCC-MIDASmethodologywhichallowsforbaselinecorrelationlevels
tovaryslowlythroughouttheperiodunderinvestigation.Oneof
thekeyissuesidentifiedduringtheprocessofEuropeanintegration wasbasedonthefactthatbroadregulation,withparticular empha-sisonitsuniformity,mightactuallybehinderingbroadgrowthof somedomesticandmoreperipheralandlocally-designedmarkets OurselectedcountriesincludenotonlycoreEuropeanstatessuch
asFrance,Germany,Austria,BelgiumandtheNetherlands,butalso themoreproblematicandperipheralstatesreferredtoasthePIIGS; i.e.,Portugal,Italy,Ireland,GreeceandSpain
The empirical results obtained from using the DCC-MIDAS frameworkshowthatregulatorychangeshavesignificantimpacts
onthelong-runrelationshipamongsovereignbondyieldsofthe samplecountriesandthissignificanceisrobustwithrespecttothe trendinthecorrelations.Themethodologicalselectionisvalidated whendocumentingthedifferingbehaviourofboththelong-and short-termcorrelationcomponentsbetweenindividualsovereign bondyields.Astotheexceptionalnatureoftheinfluenceofthe financial crises and sovereigndebt crises that affected Ireland, Greece,Spain,PortugalandItaly,wefindsubstantialevidenceof significanteffectsofregulatoryannouncementsduringtheperiod analysed
Our analysis also examines the structural shifts in the long-runcorrelationswithrespecttoregulatorychange announce-ments(boththeregulatoryannouncementsthathavebeenboth announced andfully implementedand thosethat arecurrently beingimplementedandhavenotyetreachedconclusion).Itwas broadly assumed that market responses to regulatory change announcementswouldbesubstantialatthepointthatsuch infor-mationof largestructural changesbeingannounced Thisturns outtobethecase,withsharpresponsesintheDCC-MIDAS frame-workobservedduringkeyeventssuchastheprovisionoffinancial support for Greece The detailed analysis of the results shows thatEuropeanbondmarketsweresharplyinfluencedthroughthe implementationof keyregulationsundertakenbytheEuropean CommissionandtheBaselCommittee,suchasBRRD,PRIPS,IMD, UCITS,BaselIII,dataprotectionregulation,EUODRregulation,and theregulationofEUcreditratingsagencies.Thesesubstantialshifts