Journal of International Financial Markets, Institutions & Money

J. Int. Financ. Markets Inst.
Money 52 (2018) 17–36
Contents lists available at ScienceDirect
Journal of International Financial

Markets, Institutions & Money
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / i n t fi n
Volatility co-movements and spillover effects within the

Eurozone economies: A multivariate GARCH approach using the
financial stress index
Ronald MacDonald a, Vasilios Sogiakas a, Andreas Tsopanakis b,⇑
a
Economics Section, Adam Smith Business School, University of Glasgow, Glasgow, UK
b
Economics Section, Cardiff Business School, Cardiff University, Cardiff, UK
a r t i c l e i n f o a b s t r a c t
Article history: The Eurozone crisis is one the most important economic event in recent years. At its peak,
Received 1 July 2016 the effects of the crisis have put at serious risk the outcome of the euro project, exposing
Accepted 12 September 2017 the inherent weaknesses and vulnerabilities of the monetary union. As the degree of eco-
Available online 28 October 2017
nomic and financial integration of these countries is significant, we aim to investigate in
details the potential cross-covariance and spillover effects between the Eurozone econo-
JEL classifications: mies and financial markets. In order to do this, we employ financial stress indexes, as sys-
C43
temic risk metrics in a multivariate GARCH model. This method is able to capture markets’
C58
G01
dependencies and volatility spillovers and is employed on a single market level as well as
G15 on the full spectrum of Eurozone markets. The empirical results have shown the important
and intensive stress transmission on banking and money markets. Moreover, the role of
Keywords: peripheral countries as stress transmitter is verified, but only for particular periods. The
Financial crisis significant spillover effects from core countries are also evident, indicating their important
Volatility spillover effects role in the Euro Area and its overall financial stability. The ‘‘decoupling” hypothesis is
Systemic risk
empirically verified, underling the gradually decreasing intensity of spillovers between
GARGH-BEKK model
Euro Area countries. Overall, this paper exhibits the complex structure of spillover effects
for Eurozone, along with a clustering effect in the most recent times.
Ó 2017 Elsevier B.V. All rights reserved.
1. Introduction
There is no doubt that the recent Eurozone financial and sovereign crisis is one of the most important economic events of
the last decade. It created an unprecedented reaction, in terms of unconventional monetary and fiscal policies from the glo-
bal and local policy makers. Starting from the US, after the collapse of Lehman Brothers, governments provided state funds
for the rescue of insolvent financial institutions and for the stabilization of the financial system in general. The size of this
intervention is vast, with ECB’s total assets reaching almost 21% of Euro Area GDP in 2015 (as opposed to 9% in 2007), while
similar situation prevails in the rest of the world (for instance, FED holds assets equal to 25% of US GDP, with outright
purchases being almost 99% of these assets). At the peak of the crisis, US authorities spent almost 20 trillion US dollars
for rescuing banks, while in Europe, governments spent around 312 billion euros for bailing out financial institutions and
2.92 trillion euros for implicit guarantees (Fratzscher, et al., 2016; Hryckiewicz, 2014; Kizys et al., 2016). The main reason
⇑ Corresponding author.
E-mail addresses: Ronald.MacDonald@glasgow.ac.uk (R. MacDonald), Vasilios.Sogiakas@glasgow.ac.uk (V. Sogiakas), TsopanakisAG@cardiff.ac.uk
(A. Tsopanakis).
https://doi.org/10.1016/j.intfin.2017.09.003
1042-4431/Ó 2017 Elsevier B.V. All rights reserved.
18 R. MacDonald et al. / J. Int. Financ. Markets Inst. Money 52 (2018) 17–36
to proceed to such a large scale bail out programs, for financial institutions and a number of heavily indebted countries in the
Eurozone case, was the heightening uncertainty for the already identified financial, fiscal and real economic meltdown. Fur-
thermore, the lack of a consistent and supra-national macro-prudential and crisis mitigating framework, led to even more
uncertainty.
Bearing all the above in mind and, given the lack of conclusive and clear cut evidence for the potential risk transmission
channels within the Eurozone economies, we aim to shed further light in the issue of volatility co-movement and spillover
effects among the EMU countries. In contrast to the existing literature, we move beyond the usual focus on sovereign and,
sometimes, banking risk channels. Instead, we try to investigate a full set of potential volatility transmission channels, by
implementing a number of financial stress indices for a wide group of financial markets.
This paper contributes to the literature in several dimensions. As implied in the previous paragraph, we provide a detailed
account of potential spillover effects for a wide number of Eurozone countries and financial markets. We do not limit this
research to sovereign or banking risks only. Instead, a broader and inclusive approach is adopted, by studying the effects
of the banking sector, along with the money, equity and bond markets. Financial stress indexes, which are aggregate metrics
of systemic risk and potential instability in the markets, are used as proxies of these market conditions. Additionally, the
econometric modelling is based on multivariate GARCH-BEKK models, which is another innovative characteristic to this
research area.
The empirical work is conducted into two directions: ‘‘within” each one of the markets, we produce financial stress
indexes for the banking sector, money, equity and bond markets and ‘‘between” all of the above markets and countries. This
cross-market approach is another novel feature of this study, providing further decomposed evidence for the Eurozone crisis
spillover effects. Additionally, the empirical modelling is also materialized on a regional analysis, where regional (core –
periphery) stress indices are used, together with sub-sample analysis. The latter is useful, in order to identify changing
patterns to the stress transmission channels due to the crisis outbreak. The findings from the baseline model are further
reinforced by a range of robustness checks and further evidence analysis. Alternative volatility specifications are employed,
along with additional multivariate GARCH modelling approaches, together with alternative sets of financial stress indices.
Our findings show that multiple channels of interconnectedness exist in Eurozone, with an eminent role for banking and
money markets. The direction of these spillover effects is towards both types of countries, core and peripheral, depending to
the time period in some cases. Even though the most heavily affected countries from the Eurozone crisis (Greece, Spain, Por-
tugal and Ireland) are occasionally among the major contributors of volatility transmission, they are also receivers of such
effects from core economies. The structure of transmission channels indicates the existence of clusters of countries, in the
sense that countries are more vulnerable and exposed to spillovers from their own group (core or periphery). In the course
of the Eurozone crisis, we can also identify the flight to quality and flight to liquidity phenomena taking place, as the clus-
tering effect is more prominent after the outbreak of the Euro Area crisis. The asymmetric nature of the results is also verified
by a battery of alternative modelling specifications and robustness checks.
The structure of this paper is as follows. First, a section where we discuss some relevant papers examining the Eurozone
volatility spillovers or contagion issues is provided. Thereafter, a description of our financial stress indices and their compo-
nents, along with the aggregation method, is presented. Moreover, the GARCH-BEKK modelling approach is analyzed. Sec-
tion 4 is where the estimations outcome is discussed, for the market, country and regional level. Part 5 provides further
evidence and robustness tests for our main findings. The last section concludes and provides some discussion on the poten-
tial policy recommendations stemming from this work.
2. Eurozone crisis and modelling of spillover effects
The study of contagion and spillover effects among markets and countries is a topical research area in recent years, given
the current multi-faceted crisis in Eurozone. Recently, the interest is focused on the potential deleterious effects of the global
financial crisis. For the case of Europe, most researchers have focused on the issues of sovereign risk transmission. For
instance, Brutti and Saure (2015) employ SVAR analysis of sovereign CDS for eleven Eurozone countries. They find that expo-
sure to Greek sovereign debt and Greek banks assets are sources of intensive transmission of risk. On the other hand,
Kohonen (2014) uses ten-year government bond yield differentials for the peripheral Euro Area countries, again into a SVAR
framework. Here, the author suggests that there was a default risk transmission from the Greek bonds, but only at the begin-
ning of the crisis. He also suggests that this was not the only risk channel within the countries under scrutiny. More recently,
VAR modelling in the spirit of the work developed by Diebold and Yilmaz (2012, 2014) has been used to examine the finan-
cial spillover effects between different groups of markets or economies. For instance, Apostolakis and Papadopoulos (2014,
2015) study the effects between the G7 banking, securities and foreign exchange markets, identifying some interrelations
within them. Antonakakis and Vergos (2013) examine sovereign risk transmission for some Eurozone countries, showing
that most of them are mostly responsive to their own government bond yield variations. Using a range of econometric tech-
niques, Caporin et al. (2013) indicate that contagion effects were not that intensive, even though peripheral countries went
through serious difficulties because of their heightening fiscal burden. On the other hand, Metiu (2012) identifies strong con-
tagion effects for the period 2008–2012, using the canonical contagion model. Overall, literature shows a lack of consensus
on the actual distress transmission effects among the Eurozone countries (Alter and Beyer, 2014; Alter and Schuler, 2012;
Arezki et al., 2011; Barth et al., 2012; Bruyckere et al., 2013; Caceres et al., 2010; Louzis, 2015).
R. MacDonald et al. / J. Int. Financ. Markets Inst. Money 52 (2018) 17–36 19
Another popular type of models to identify spillover effects is the multivariate volatility models. Audige (2013) employs a
smooth transition conditional correlation (STCC-GARCH) model, with long term governmental bond yields, in order to check
for spillover effects from the Greek crisis. The author identifies contagion effects from Greece to Ireland and Portugal in 2010,
while such effects weaken after that period. In a similar vein, Grammatikos and Vermeulen (2012) examine the transmission
of financial and sovereign debt shocks through the Eurozone stock markets, for the period 2007–2010. In order to do this,
GARCH modelling of stock returns are employed, with the US markets effects taken into consideration. They split EMU into
three groups of countries, namely the North, South and Small economies. Their findings show strong crisis transmission from
US non-financials to European non-financials, with the financial entities from both sides of the world showing not significant
interconnections. Moreover, Greek CDS spreads seem to play a much more important role in the period after the Lehman
collapse, but not for the non-financial firms. Another interesting paper is by Dajcman (2012), who uses a flight-to-quality
indicator to examine the co-movements of stock returns with bond yields for Germany and the most debt ridden Euro Area
economies. The results, using a DCC modelling approach, are concurrent with Kohonen (2014) and Caporin et al. (2013). Also,
the flight-to-quality indicator has higher value prior to 2010, indicating increasing uncertainty for investors, who turned
towards the safe haven of German Bunds.
3. Data and methodology
3.1. Financial stress index construction
The analysis of the financial stress is accomplished by creating aggregated financial stress indexes (FSIs). These indexes
provide information on the financial markets conditions, based on a range of stand-alone indicators representing important
market features. Our focus is on Eurozone crisis and, thus, our sample consists of eleven Eurozone economies: Austria, Bel-
gium, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal and Spain. Our interest is in the calculation of
four market-level indicators, along with a systemic risk index for each individual country. The four markets for which finan-
cial instability indexes are provided are the banking sector, money, equity and the bond market. The reason is that these are
the markets that exhibited the biggest uncertainty and distress level during the Eurozone crisis. Banks were highly distressed
and a number of defaults or near defaults take place in the European banking system the last few years. Money market, as it
is later shown, predominantly represents the interbank funding and liquidity conditions. It is evident from the recent finan-
cial events that these were major issues for European economies. Finally, stock markets are also important, while bond mar-
kets have to do with sovereign risk issues, together with sustainable long term funding and investments to the private sector.
The data used to develop our FSIs are retrieved on a weekly frequency. The reasons are twofold: first, we aim to explore
the transmission channels existing between different markets and different countries. In this way, the implemented analysis
accounts for possible transmission channels on both a country level, through the aggregated country FSIs, and on a specific
market of a country, through the aggregated market FSIs. Finally, we aim to investigate the cross-market spillover effects.
Secondly, we employ multivariate GARCH modelling and, henceforth, there is a necessity to employ high frequency data.
For this reason, the variables chosen are restrained to those that can be collected in such time frequency. Table 1 summarizes
the variables included in the financial stress indices of the economies in our sample. The time period covered is from January
2001 until the 20th of September 2013. In total, there are 664 observations covering the pre- and post-crisis period. We do
not use daily data in order to avoid potential mismatches in public holidays and trading days (Yiu et al., 2010). In this way, a
uniform dataset is created, without any discrepancies in the countries’ series used. The choice of variables relies on the
relevant financial stress literature, covering a broad range of individual indicators that provide useful information for the
Table 1
Data for constructing market-level financial stress indices.
Variables Used in Financial Stress Indices

Banking Sector Money Market
Dividend Yield TED Spread
Market Value Inverted Term Spread
Turnover by Volume Treasury Bill Realized Volatility
P/E Main Refinancing Rate – 2 yr Government Bond Yield
Bank Equities Realized Volatility
Banking Sector Beta Main Refinancing Rate – 5 yr Government Bond Yield
Bank Equities Returns
Equity Market Bond Market

Stock Returns Sovereign Spread
Dividend Yield Government Bond Realized Volatility
P/E Corporate Spread
Stocks Realized Volatility Government Bond Duration
Market Value
markets. On the other hand, the fact that this study aims to examine a range of economies limits somehow the total number
of variables that can be uniformly picked for the sample.
Getting into more details, the banking sector index comprises of seven variables, while five variables are used in the case
of money and equity markets and four for the case of bond market. Focusing to the banking market, variables representing
banks’ sensitivity to market conditions, along with their profitability and risk level are included. Dividend yield is negatively
related to fundamentals of banking institutions and, thus, excessive dividend yields can be a signal of increasing default risk
for them. On the same time, market value is also important, since its size directly affects the stability of the market. Increas-
ing uncertainty can lead to a significant adverse effect to market value, which is also tied to these institutions’ book value.
Thus, their financial health is at stake. A strong indicator of instability for banks is the turnover by volume. This increases,
according to market sentiment and the perceived level of risk and uncertainty by the investors. Profitability is also an impor-
tant metric here, represented by the P/E ratio. Here, since banks’ operational efficiency and profitability is indicated by higher
values for the aforementioned ratio, we impose a negative sign to this variable (so, higher P/E ratio coincides with increasing
financial stress). Finally, the last three variables here (realized volatility of banks equity index, beta coefficient of the same
index and the (negative) stock returns) depict the risk perception and the volatility level of this market. In accordance to the
stock market bubble literature, increasing stock prices reflect imbalances building up, while a swift decrease in prices is evi-
dence of a market crash (Grimaldi, 2010).
The aggregate index for the money market sector includes some of the most important liquidity, credit and counterparty
risk indicators. TED spread (the difference between the 3-month Euribor and the respective Treasury bill of the same matu-
rity) is one of these measures, reflecting the flight to quality and flight to liquidity phenomena. It is expected to observe
increasing values for this spread, in periods of worsening financial conditions. In such times, interbank funding markets seize
to operate smoothly, while the risk perception reaches unprecedented levels. In the same line of thought, inverted term
spread is incorporated, as indicator of interest rate setting expectations, along with the representation of default risk and
increasing information asymmetry in money markets. Moreover, the spreads of the main refinancing rate from the short
term governmental bills yield is another indicator of deteriorating liquidity conditions. Negative values in these spreads
coincide with higher financial stress and, hence, the need to incorporate them in our aggregate index with a negative sign.
Finally, the treasury bill realized volatility depicts the market volatility risk.
The conditions in the equity markets are captured by five variables. The (negatively signed) stock returns is an indication
of investors’ sentiment and lack of trust to listed firms’ fundamentals. In periods of increasing financial stress, higher volatil-
ity is expected in the stock markets. Then, market value is included and the dividend yield as well. The rationale is similar to
the case of the banking sector, emphasizing the level of default risk, as well as the lack of credibility and funding sources in
the market. The P/E ratio is included again in a similar fashion, while the realized volatility of the general equity market
index is indicative of the historical risk perception on each equity market.
Turning now to the bond market case, we employ the sovereign bond spread, vis-à-vis German bond yield, which is con-
sidered as a safe haven for bond market investors. This is a strong and popular indicator of the perceived sovereign risk of the
countries under investigation. Then, the realized volatility of the long term governmental bond yields is used, as another
variable illustrating the markets’ risk aversion. Moreover, the corporate bond spread (defined as the yield difference between
corporate bonds and government bonds of the same maturity) is a factor showing the default risk and the financial obstacles
that firms face. Government bond duration is also included. It is expected that decreasing credit ratings and increasing con-
cerns for the countries solvency, will lead to lower duration for their bonds (Lee et al., 2011). Hence, decreasing duration
represents increasing financial stress and uncertainty.
The FSIs are computed, following the variance-equal aggregation method. Based on this approach, an equal weight is
attributed to all variables in each of the markets. In this way, the market – level indices are computed (Eq. (1)), while the
same approach is followed for the country – wide (Eq. (2)) and regional ones (Eq. (3)). Before the aggregation, each one
of the single indicator is standardized. That is, its mean value is subtracted by each observation and, then, divided by its stan-
dard deviation, avoiding mis-measurement issues. All series are expressed as deviations from their long run mean value.
Based on the above, the original discrepancies in variables units disappear. The variance – equal approach is frequently used
in the relevant literature (Cardarelli et al., 2011). The reason for this is the simplicity of the relevant calculations and the
effective representation of the prevailing financial conditions in the markets. The indexes are presented in Fig. 1, where their
fluctuations through time follow the major financial events.
X
l
1 i
FSImarket
t
i
¼ xjtstandardized ð1Þ
j¼1
l
n o
FSIcountry
t
i
¼ 0:25 FSIbank
t þ FSImoney
t þ FSIequity
t þ FSIbond
t ð2Þ
X
n
1 i
FSIregion
t
i
¼ FSIcountry
t
j
ð3Þ
j¼1
n
It should be noted here that an additional reason for constructing this set of stress indices is the lack of an appropriately
developed dataset for the countries and financial markets that we examine. Individual researchers or institutions, such as the
Panel A : Banking FSIs Panel B : Money Market FSIs

4 3
2.5
Austria Austria
3
2
Belgium Belgium
1.5
2 Finland Finland
1
France France
1 0.5
Germany Germany
0
Greece Greece
0
-0.5
Ireland Ireland
-1
-1 Italy Italy
-1.5
Netherlands Netherlands
-2 -2 Portugal
Portugal
2001-01-05
2006-01-05
2001-07-05
2002-01-05
2002-07-05
2003-01-05
2003-07-05
2004-01-05
2004-07-05
2005-01-05
2005-07-05
2006-07-05
2007-01-05
2007-07-05
2008-01-05
2008-07-05
2010-01-05
2011-01-05
2013-01-05
2009-01-05
2009-07-05
2010-07-05
2011-07-05
2012-01-05
2012-07-05
2013-07-05
2001-08-05
2002-03-05
2001-01-05
2002-10-05
2003-12-05
2003-05-05
2004-07-05
2005-02-05
2005-09-05
2006-04-05
2006-11-05
2007-06-05
2008-01-05
2010-05-05
2010-12-05
2011-07-05
2008-08-05
2009-03-05
2009-10-05
2012-02-05
2012-09-05
2013-04-05
Spain Spain
Panel C : Equity FSIs Panel D : Bond FSIs

2.5 3
2.5
2 Austria Austria
2
1.5 Belgium Belgium
1.5
Finland Finland
1 1
France France
0.5 0.5
Germany Germany
0
0 Greece Greece
-0.5
-0.5 Ireland Ireland
-1
Italy Italy
-1
-1.5
Netherlands Netherlands
-1.5 -2
Portugal Portugal
2006-11-05
2010-12-05
2011-07-05
2012-09-05
2013-04-05
2001-01-05
2004-01-05
2008-07-05
2013-01-05
2001-01-05
2001-08-05
2002-03-05
2002-10-05
2003-05-05
2003-12-05
2004-07-05
2005-02-05
2005-09-05
2007-06-05
2006-04-05
2008-01-05
2008-08-05
2009-03-05
2009-10-05
2010-05-05
2012-02-05
2001-07-05
2002-01-05
2002-07-05
2003-01-05
2003-07-05
2004-07-05
2006-01-05
2005-01-05
2005-07-05
2006-07-05
2007-01-05
2007-07-05
2008-01-05
2009-01-05
2009-07-05
2010-01-05
2010-07-05
2011-01-05
2011-07-05
2012-01-05
2012-07-05
2013-07-05
Spain Spain
Panel E : Country FSIs

1.5
Austria
1
Belgium
Finland
0.5
France
Germany
0
Greece
Ireland
-0.5 Italy
Netherlands
-1 Portugal
2003-01-05
2004-07-05
2006-01-05
2007-01-05
2007-10-05
2010-01-05
2010-10-05
2011-10-05
2013-01-05
2001-01-05
2001-04-05
2001-07-05
2001-10-05
2002-01-05
2002-04-05
2002-07-05
2002-10-05
2003-04-05
2003-07-05
2003-10-05
2004-01-05
2004-04-05
2004-10-05
2005-01-05
2005-04-05
2005-07-05
2005-10-05
2006-04-05
2006-07-05
2006-10-05
2007-04-05
2007-07-05
2008-01-05
2008-04-05
2008-07-05
2008-10-05
2009-01-05
2009-04-05
2009-07-05
2009-10-05
2010-04-05
2010-07-05
2011-01-05
2011-04-05
2011-07-05
2012-01-05
2012-04-05
2012-07-05
2012-10-05
2013-04-05
2013-07-05
Spain
Fig. 1. Financial stress indices for Eurozone markets and countries.
IMF or ECB, have sporadically being involved in the study and development of similar measures. Still, the composition, types
of market characteristics and the detailed coverage of the Euro Area markets diversifies this work from the rest.1
3.2. Volatility transmission models
Ambition of this paper is the empirical investigation of potential interdependencies and spillover effects of financial dis-
tress in the Euro Area, on a market, country and regional level. As a concept, it is closer to the ‘‘meteor showers” hypothesis of
Engle et al. (1990), than the idea of contagion as developed by other economists (for instance, Forbes and Rigobon, 2002). The
multivariate GARCH family of models has been extensively applied in the past for the examination of spillover effects
between financial data providing a reliable mechanism for examining the significance, the magnitude and the direction of
potential interrelationships of the second moment of time series data.
The mean equation of the FSIs (kxT: k series, T weeks) under consideration is modeled through an unstructured VAR(p⁄)
equilibrium specification:
p
X
DFSIt ¼ c0 þ Li DFSIti þ et ð4Þ
i¼1
where DFSI is the vector of the first log differences of k response times series variables at time t, c0 is a constant vector of
offsets with k elements, Li are kxk matrices for each lag (i = 1 . . . , p⁄) and et is a vector of serially uncorrelated innovations.
For the purposes of our analysis we apply the full BEKK model of Engle and Kroner (1995), the parameterization of which
ensures a positively definite variance covariance matrix, mitigating the estimation process of the parameter set. This is an
alternative to the multivariate VEC model, proposed by Bollerslev et al. (1988), ensuring the positive definiteness of the con-
ditional variance matrix Ht (Bauwens et al., 2006). The residuals of the mean equation are assumed to follow a T-student
distribution with zero mean and a time-varying variance conditional on the past informational set Xt1:
1
In Section 5, we experiment with the CISS index, which is a sovereign risk index created by ECB, as a robustness check.
et jXt1 T Studentð0; Ht Þ ð5Þ

The k-dimension full BEKK GARCH(p,q) conditional volatility specification has the following form:
X
p X
q
Ht ¼ C 00 C 0 þ A0j etj e0tj Aj þ B0j Htj Bj ð6Þ
j¼1 j¼1
where C0 is the constants matrix, A and B are parameter matrices, et1 is the innovation matrix (lagged disturbance vector)
and Ht1 is the lagged variance covariance matrix. The constants matrix is restricted to be a lower triangular matrix, while
the A and B parameter matrices are not restricted. As emphasized by Bollerslev (2010), this quadratic parameterization guar-
antees that the covariance matrix is positive definite, while the number of parameters to be estimated is more compact, com-
pared with its initial version proposed by Bollerslev et al. (1988).2
The relationship between the k FSIs’ volatilities is captured by the elements of A and B matrices. The elements of A
matrix’s coefficients depict the effects of lagged innovations to the conditional variance covariance matrix. As it is commonly
said in the relevant literature, matrix A provides information on ‘‘news effect”, while matrix B captures the ‘‘volatility
spillover” effect (Kim et al., 2015). Both effects can provide important insights for the potential volatility transmission chan-
nels of financial distress within the Euro Area.
For encountering more efficiently the leptokurtic distributional form of the FSI series we use the T-Student distributional
(m degrees of freedom) form in the MLE estimation process:
( )
v þ1 v 1 XT X
T
e2t =ht
likelihood ¼ T log C log C logðht p ðv 2ÞÞ ðv þ 1Þ log 1 þ
2 2 2 t¼1 t¼1
v 2
ð7Þ
⁄
As the required number of parameters for the VAR(p ) – Full BEKK Multivariate GARCH(p,q) model, of a k-dimensional
dataset (kxT), is equal to k + k2p⁄ + (k/2)(k + 1) + pk2 + qk2 plus the degrees of freedom of the T-student distribution, we
choose the parsimonious representation of one lag for both the VAR and GARCH specifications resulting to (3k + 7 k2)/2
parameters for estimation.
3.2.1. Market level

Our first research aim is the investigation of the financial distress spillover effects on a market level across countries, i.e.
the banking sector, money market, equity market and the bond market. Considering the high dimensionality of the param-
eter set and the ensuing computational procedures, we examine the transmission channels using all possible pairwise com-
binations with the estimation of bivariate full BEEK models instead of running a 11-dimensional multivariate GARCH model.
Consequently, with 11 countries in our dataset, 55 different pairwise samples are under investigation for each of the four
markets, having excluded the case of examining own effects. Thus, the volatility specification of the 55 VAR(1) – Full BEKK –
GARCH(1,1) models of each market would be represented by the following equations:
0 !
a11 a12 e21;t1 e1;t1 e2;t1 a11 a12 b11 b12 0
b11 b12

Ht ¼ C 00 C 0 þ þ H t1 ð8Þ
a21 a22 e2;t1 e1;t1 e22;t1 a21 a22 b21 b22 b21 b22
h11;t ¼ c11 þ a211 e21;t1 þ a221 e22;t1 þ 2a11 a21 e1;t1 e2;t1 þ b211 h11;t1 þ b221 h22;t1 þ 2b11 b22 h12;t1
ð9Þ
h22;t ¼ c22 þ a212 e21;t1 þ a222 e22;t1 þ 2a12 a22 e1;t1 e2;t1 þ b212 h11;t1 þ b222 h22;t1 þ 2b11 b22 h21;t1
In this framework the transmission channels between country’s market FSIs are examined through the 55 estimations of
coefficients a21, b21 and a12, b12. The former two coefficients represent the volatility spillover from the second country’s mar-
ket FSI to that of the first one while the latter two represent the opposite direction effect. It should be noted that we inves-
tigate the transmission channels based on the squared innovations (et12) and variances (ht1) of Eq. (9) and this implies that
the coefficients’ sign would not affect the direction, the significance and the economic justification. However, the transmis-
sion channels are further examined, for the regional case, by the adoption of the ‘‘News Impact Surface” approach as a
robustness check.
3.2.2. National level

At a national level (Eq. (2)) we adopt a similar approach estimating 55 VAR(1) – Full BEKK models. Coefficients a21, b21
and a12, b12 reflect the potential distress transmission channels between EU countries for all possible pairwise combinations
between the 11 examined countries. Thus the 55 estimated a21 and b21 coefficients examine the effect of the second series on
the first one for all possible 55 pairwise coefficients while the 55 estimated a12 and b12 coefficients examine the opposite
direction effect.
2
In this way, model’s convergence is more easily achieved. See, among others, Bauwens et al. (2006).
3.2.3. Regional analysis

The regional analysis refers to the distress index of Eq. (3); that is the core and periphery countries’ FSI per se and by mar-
ket. Thus, our sample consists of five series for the core region and five for the periphery one, i.e. total, bank, money, equity
and bond. The 10 regional series result to 45 potential pairwise combinations having excluded the effect of a series to itself
(main diagonal). Thus, in this part of the analysis we examine the potential transmission channels between and/or within
regions, i.e. core banking with periphery banking, core banking with core bond, core banking with periphery equity etc.
Moreover, the regional transmission channels are further examined by the adoption of the ‘‘News Impact Surface” approach
as a robustness check. In this aspect, we aim to examine whether a past shock in a region’s financial distress index (FSI)
affects the forecasted volatility of another region’s FSI. This is implemented within a multivariate Full-BEKK-GARCH frame-
work allowing for asymmetric dynamic covariance responses due to past and current shocks in one of the determinant
series.
4. Discussion of results
4.1. FSI descriptive statistics
Table 2 provides several descriptive statistics for the FSIs across markets (Panel A-D) and countries (Panel E). These statis-
tics refer to the first four moments of the series, their normality, autocorrelation, heteroscedasticity and stationarity. Accord-
ing to the standard deviation of the series, the Banking and Equity FSIs embed the higher risk. Most of the series exhibit a
positive skewness, while their distributions are leptokurtic. Further evidence of non-normal distributional forms for the FSI
series is provided by the high J-B statistic. Moreover, the Ljung-Box statistics (Q, Q2) are in favour of serially correlated series,
exhibiting higher order correlations and non-linear dependencies. The only exception is the Dutch bond market FSI, while
the Ljung-Box Q2 test fails to provide relevant evidence for the cases of Greece, Portugal and Spain (in the case of the bond
markets). The same holds for the money market FSIs of Greece and Finland. Nevertheless, the aforementioned are limited
exceptions to the general conclusion of the autocorrelation existence. The Engle’s ARCH test coincides with the previous find-
ings, underlying the necessity of employing a time varying volatility model for the implementation of our empirical work
that aims to study the spillover effects of financial stress among the Euro Area. Finally, the ADF test for the first log differ-
ences of FSIs could not accept the existence of a unit root.
4.2. Transmission channels across markets
We commence the analysis of the econometric results from the banking sector. Tables 3a and 3b present the outcome of
the estimation in this case. Banking market, along with the money market are the most interconnected sectors, in the sense
that a range of significant interactions can be identified, both for news surprises as well as for spillover effects coefficients.
The prominent role of the peripheral economies is evident. Ireland, together with the Italian and Portuguese banking sectors
are the major volatility risk recipients and transmitters. Similar vulnerability is indicated for the case of Austria, even though,
on a more limited size compared to the aforementioned cases. French banking system contributes to the heightening stress
transmission to the rest. It is interesting to underline that peripheral banking systems are more exposed to effects from
banks of the same group of countries, as it is the case for the major Euro Area economies. This is an indication of a fragmen-
tation, in terms of potential vulnerability transmission effects. It can also be an indication of divergent policies and response
reactions to the crisis outbreak from banks and governments in the common currency area. The Greek case does not seem to
constitute a serious threat in this particular market.
Money market volatility spillovers are depicted in Tables 4a and 4b. The main recipients of the relevant effects are,
mainly, Greece, Ireland and Finland (for the case of ‘‘news effects” coefficients), with France, Ireland and Portugal (apart from
Italy) to take the lead in the volatility transmission risks. Nevertheless, the strong statistical results are accompanied by
small parameter values. Greek case is rather distinctive for the ‘‘news surprises” effect, something reasonable given that
Eurozone crisis has escalated due to the economic event taken place in this particular case. The decreasing significance of
Greece as volatility transmitter, as indicated in the second panel of Table 4, shows that more vulnerability and uncertainty
stemmed from other markets. Given the nature of money market stress indicators, representing funding, liquidity and inter-
bank markets considerations, also contributes to understand the substantial number of statistically significant aij coefficients
in this market case. Countries clustering can be observed, as weaker economies are more exposed to spillover effects from
their peers, as it is the case for core countries’ money markets.
A very different situation appears in the case of equity markets’ spillovers (see Tables 5a and 5b). In essence, the identified
links are limited. In both the cases of cross-innovations and variance volatility transmission parameters, few statistically sig-
nificant results exist. In the case of information shocks, there is no a market with prominent role. There are effects from the
Greek stock market, as well as Spain and Belgium. In terms of the persistence parameters (Table 5b), some interactions can
be identified, with Belgium being the most eminent towards its core peer markets. In contrast to the previous cases, a clear
pattern in core-periphery disaggregated effects is not that evident here. Overall, this lack of significant effects in the equity
markets could, probably be a sign of the, rather limited, financial risk propagation taken place through that market in the
Eurozone case.
Table 2
Descriptive statistics for market and country FSIs.
Descriptive Statistics for Financial Stress Indices across countries and markets
Countries Austria Belgium Finland France Germany Greece Ireland Italy Netherlands Portugal Spain
Panel A: Banking Sector
Mean 0.0000 0.0001 0.0016 0.0007 0.0009 0.0008 0.0008 0.0002 0.0014 0.0008 0.0021
Std. Deviation 0.2268 0.2204 0.2578 0.2361 0.2328 0.2414 0.2443 0.2263 0.2197 0.2251 0.2344
Skewness 0.257** 0.591** 0.423** 0.283** 0.885** 0.852** 0.077 0.828** 0.948** 0.445** 0.911**
Kurtosis 3.303** 6.676** 10.316** 3.03** 7.001** 10.927** 8.691** 3.93** 13.021** 6.381** 6.277**
J-B Test 309.9** 1270** 2959.6** 262.61** 1440.8** 3378.8** 2087.5** 502.65** 4783.2** 1147** 1180.4**
Q(10) 147.03** 149.63** 178.02** 198.72** 156.15** 169.08** 179.32** 173.05** 118.52** 136.88** 157.4**
Q2(10) 481.28** 305.53** 145.51** 566.97** 387.89** 492.56** 209.03** 155.27** 290.14** 172.64** 160.34**
ARCH(5) 24.67** 20.99** 57.54** 45.99** 33.11** 49.52** 42.78** 11.86** 28.72** 23.77** 21.59**
ADF 21.85** 21.02** 22.24** 22.32** 21.75** 18.67** 22.39** 20.23** 21.25** 22.09** 23.43**
Panel B: Money Market
Mean 0.0008 0.0007 0.0006 0.0006 0.0007 0.0001 0.0001 0.0003 0.0003 0.0010 0.0001
Std. Deviation 0.0867 0.0885 0.0748 0.0980 0.0738 0.0483 0.0578 0.0911 0.1739 0.0437 0.0568
Skewness 1.419** 0.911** 1.731** 1.296** 1.241** 3.215** 3.036** 1.14** 0.861 0.118 0.177
Kurtosis 8.958** 8.602** 11.555** 6.288** 7.973** 77.115** 84.406** 13.077** 30.869** 11.913** 4.431**
J-B Test 2439.8** 2136.3** 4019.3** 1278.1** 1926.7** 165420** 197830** 4867.8** 26405** 3921.8** 545.88**
Q(10) 25.84** 11.30 26.47** 19.01* 21.57* 21.61* 100.95** 15.81 59.74** 37.56** 62.18**
Q2(10) 55.58** 140.53** 7.21 89.33** 77.54** 4.53 75.04** 44.25** 85.82** 297.05** 144.62**
ARCH(5) 2.99* 22.37** 1.31 6.08** 2.88* 0.86 2.72* 4** 22.48** 29.07** 13.42**
ADF 16.42** 14.54** 13.48** 15.74** 15.68** 15.4** 13.88** 14.65** 18.43** 12.99** 15.6**
Panel C: Equity Market
Mean 0.0011 0.0007 0.0023 0.0010 0.0003 0.0013 0.0011 0.0010 0.0011 0.0001 0.0007
Std. Deviation 0.2802 0.2689 0.2610 0.2777 0.2760 0.2636 0.2808 0.2738 0.2759 0.2687 0.2803
Skewness 0.655** 0.469** 0.292** 0.506** 0.51** 0.477** 0.01 0.838 0.367** 0.641** 0.542**
Kurtosis 10.015** 3.601** 2.098** 4.415** 4.73** 4.575** 8.246** 7.547** 4.952** 5.807** 3.859**
J-B Test 2818** 382.75** 131.02** 567.04** 646.89** 603.48** 1878.5** 1651.2** 692.43** 977.27** 444.11**
Q(10) 199.94** 203.66** 209.86** 250.71** 263.4** 167.36** 226.7** 228.91** 217.08** 206.24** 256.23**
Q2(10) 383.69** 206.51** 233.54** 541.81** 539.49** 101.5** 455.26** 256.27** 388.87** 400.24** 459.89**
ARCH(5) 32.92** 21.35** 22.02** 55.04** 67.92** 16.75** 41.68** 42.12** 40.05** 67.98** 64.76**
ADF 25.33** 24.41** 22.34** 23.93** 24.63** 24.65** 27.96** 23.33** 23.95** 24.26** 24.58**
Panel D: Bond Market
Mean 0.0005 0.0010 0.0002 0.0011 0.0007 0.0033 0.0013 0.0009 0.0013 0.0006 0.0010
Std. Deviation 0.0704 0.0642 0.0833 0.0748 0.0982 0.0357 0.0918 0.0561 0.1050 0.0768 0.0555
Skewness 1.331** 0.735** 1.28** 0.87** 0.928** 9.693** 1.873 0.617** 1.947** 0.722** 1.592**
Kurtosis 17.613** 14.634** 20.532** 2.688** 13.601** 173.17** 57.139** 5.564** 16.422** 52.808** 11.774**
J-B Test 8766.1** 5976.2** 11827** 283.41** 5205.2** 838780** 90579** 897.48** 7869.5** 77095** 4109.8**
Q(10) 21.54* 30.26** 22.1* 38.98** 21.77* 172.7** 59.1** 24.65** 6.3 27.67** 26.88**
Q2(10) 2.06 32.38** 109.1** 55.73** 21.64* 0.49 89.04** 208.85** 0.98 0.55 6.73
ARCH(5) 0.35 3.95** 25.1** 6.66** 1.95 0.04 17.98** 15.02** 0.13 0.01 0.76
ADF 13.83** 14.94** 14.65** 13.81** 14.58** 10.63** 12.66** 13.96** 14.36** 15.46** 14.43**
Panel E: National FSIs
Mean 0.0002 0.0002 0.0003 0.0000 0.0007 0.0007 0.0004 0.0003 0.0004 0.0002 0.0010
Std. Deviation 0.0563 0.0520 0.0748 0.0431 0.0609 0.0525 0.0715 0.0489 0.0799 0.0595 0.0503
Skewness 0.039 0.212** 0.073 0.496** 0.074 0.07 0.204* 0.804** 0.152 0.366** 0.787**
Kurtosis 2.71** 2.428** 7.562** 3.99** 3.741** 7.705** 7.201** 4.812** 8.855** 9.067** 6.783**
J-B Test 203.06** 167.9** 1580.7** 467.19** 387.34** 1640.7** 1437.4** 711.26** 2169.1** 2286.3** 1339.6**
Q(10) 67.78** 29.06** 102** 20.83* 61.15** 89.91** 116.81** 24.12** 56.16** 113.81** 72.2**
Q2(10) 378.06** 93.87** 121.08** 173.83** 77.18** 396.18** 161.23** 121.66** 77.2** 188.18** 140.33**
ARCH(5) 24** 9.07** 40.45** 18.15** 11.32** 36.94** 28.09** 10.36** 11.3** 30.68** 23**
ADF 18.51** 15.23** 18.51** 14.53** 18.54** 17.65** 19.39** 15.63** 18.89** 17.05** 17.73**
Notes: J-B test is the Jarque-Bera test for normality. Q(10) and Q2(10) is the Ljung-Box statistic for serial correlation in raw series and squared series,
respectively. ARCH(5) is the Engle’s ARCH effects test. ADF is the Augmented Dickey-Fuller test for stationarity.
*
Significance at the 5% level.
**
Significance at the 1% level.
The last market examined is the bond market, incorporating effects and risk transmission from, both, sovereign and cor-
porate risks. It is fair to say that there are some effects identified in this market, but not significantly more and different com-
pared to the previous cases. In relation to news surprises, shocks can be identified from, both, major economies (such as
Germany and Belgium or Austria), as well as from smaller economies, like Greece and Spain. Nevertheless, the most signif-
icant results, both for the propagation and receival of spillover effects, come from Germany (Table 6a). For the case of spil-
lover effects (Table 6b), Greece is a major receiver, with several other economies contributing as risk transmitters to the rest
of the markets examined. On average, the news effect is much more prominent in this market, compared to the results
Table 3a
The information flow effect of the FSI transmission channels across countries’ Banks.
Panel A. Banking sector: The news surprises (‘‘a” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
aij AU BE FI FR GE GR IR IT NE PO SP
AU 0.030 0.033 0.016 0.140 0.037* 0.059 0.102 0.068 0.036 0.039***
0.732 0.837 0.881 0.414 0.080 0.631 0.792 0.942 0.716 0.000
BE 0.034 0.103** 0.058** 0.003 0.032 0.054 0.089 0.053 0.068 0.019
0.959 0.033 0.010 0.962 0.516 0.392 0.691 0.483 0.702 0.979
FI 0.033 0.033 0.018 0.017 0.093** 0.007 0.044 0.027 0.123 0.004
0.598 0.580 0.864 0.914 0.046 0.894 0.978 0.935 0.175 0.997
FR 0.009 0.034*** 0.020 0.013 0.065 0.090 0.001 0.001 0.048 0.017
0.970 0.000 0.923 0.922 0.568 0.511 0.887 0.972 0.409 0.985
GE 0.006 0.003 0.009 0.005 0.106 0.051 0.098 0.007 0.076 0.036
0.990 0.949 0.947 0.939 0.865 0.219 0.383 0.987 0.762 0.969
GR 0.006 0.128 0.301** 0.171 0.032 0.013 0.101 0.026 0.028 0.017
0.858 0.127 0.033 0.220 0.948 0.657 0.417 0.924 0.104 0.844
IR 0.064** 0.036 0.009 0.122** 0.080* 0.019 0.032*** 0.165 0.026*** 0.066*
0.040 0.653 0.931 0.043 0.082 0.641 0.000 0.886 0.000 0.081
IT 0.007 0.060 0.163 0.000 0.057 0.176 0.000 0.031 0.021*** 0.030
0.929 0.503 0.934 0.728 0.119 0.378 0.357 0.763 0.000 0.660
NE 0.000 0.041 0.015 0.006 0.008 0.033 0.045 0.032 0.014 0.037
0.917 0.456 0.960 0.904 0.933 0.739 0.959 0.727 0.676 0.379
PO 0.025 0.018 0.042 0.045 0.027 0.198 0.012*** 0.118*** 0.051 0.147
0.874 0.835 0.813 0.604 0.950 0.247 0.002 0.004 0.465 0.270
SP 0.069*** 0.012 0.002 0.017 0.038 0.036 0.051** 0.003 0.008 0.097
0.000 0.983 0.999 0.961 0.814 0.902 0.041 0.967 0.865 0.572
Note: This table reports the ‘‘news surprises” estimated coefficients from Bivariate Full BEKK VAR(1)-MGARCH(1,1). P-values are reported in italics, under
each parameter reported. The elements of this non-symmetric matrix represent the flow effect (‘‘aij” coefficient) with respect to the spillover effect from
country i’s Banking FSI to country j’s Banking FSI. The direction of the effects is from each row towards the columns. Countries are reported, according to
their initials: AU (Austria), BE (Belgium), FI (Finland), FR (France), GE (Germany), GR (Greece), IR (Ireland), IT (Italy), NE (Netherlands), PO (Portugal) and SP
(Spain).
Table 3b
The spillover effect of the FSI transmission channels across countries’ Banks.
Panel B. Banking sector: The spillover effects (‘‘b” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
bij AU BE FI FR GE GR IR IT NE PO SP
AU 0.068 0.010 0.002 0.053 0.002 0.005 0.018 0.023 0.054 0.012***
0.392 0.922 0.921 0.896 0.904 0.812 0.941 0.877 0.683 0.000
BE 0.014 0.097*** 0.113*** 0.016 0.030 0.042*** 0.058 0.018 0.118 0.054
0.974 0.000 0.000 0.871 0.268 0.002 0.358 0.378 0.206 0.940
FI 0.065 0.027** 0.026 0.018 0.013 0.045 0.025 0.016 0.041 0.039
0.404 0.044 0.889 0.918 0.162 0.502 0.991 0.927 0.575 0.977
FR 0.126*** 0.004* 0.010 0.063 0.078*** 0.044** 0.000 0.003 0.007 0.034
0.001 0.070 0.920 0.168 0.008 0.027 0.980 0.927 0.848 0.915
GE 0.005 0.033 0.002 0.000 0.029 0.036 0.011 0.030 0.039 0.006
0.985 0.675 0.981 0.888 0.928 0.514 0.800 0.891 0.801 0.993
GR 0.027 0.053* 0.018 0.049 0.018 0.036 0.042 0.048 0.030** 0.025
0.298 0.096 0.262 0.284 0.959 0.309 0.718 0.808 0.013 0.112
IR 0.012** 0.089** 0.034 0.062*** 0.058 0.016 0.052*** 0.059 0.208*** 0.003
0.011 0.014 0.698 0.001 0.490 0.832 0.000 0.914 0.000 0.746
IT 0.079 0.002 0.096 0.000 0.149*** 0.049 0.013*** 0.001 0.047*** 0.001
0.580 0.878 0.964 0.974 0.001 0.793 0.000 0.994 0.000 0.851
NE 0.012 0.034 0.024 0.003 0.004 0.003 0.046 0.059 0.058*** 0.056
0.830 0.203 0.956 0.951 0.945 0.969 0.944 0.269 0.000 0.263
PO 0.009 0.001 0.012 0.004 0.012 0.152*** 0.062*** 0.113*** 0.067 0.180
0.880 0.980 0.798 0.872 0.937 0.000 0.000 0.000 0.309 0.412
SP 0.054*** 0.000 0.027 0.004 0.026 0.110 0.003 0.008 0.003 0.095
0.000 0.954 0.981 0.977 0.667 0.325 0.453 0.918 0.922 0.605
Note: This table reports the ‘‘spillover effects” estimated coefficients from Bivariate Full BEKK VAR(1)-MGARCH(1,1). P-values are reported in italics, under
each parameter reported. The elements of this non-symmetric matrix represent the spillover effect (‘‘bij” coefficient) with respect to the spillover effect
from country i’s Banking FSI to country j’s Banking FSI. The direction of the effects is from each row towards the columns. Countries are reported, according
to their initials: AU (Austria), BE (Belgium), FI (Finland), FR (France), GE (Germany), GR (Greece), IR (Ireland), IT (Italy), NE (Netherlands), PO (Portugal) and
SP (Spain).
shown for the beta coefficients. A clear pattern cannot be established, in terms of the direction of the effects. The regional
analysis that follows can probably shed some more light in this market case. Some interesting insights are also provided
by the sub-sample analysis presented in Section 5.4.
Table 4a
The information flow effect of the FSI transmission channels across countries’ Money Markets.
Panel A. Money Market: The news surprises (‘‘a” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.030 0.043 0.009 0.021 0.082*** 0.092*** 0.030 0.018*** 0.006 0.014
0.883 0.279 0.956 0.907 0.008 0.005 0.768 0.000 0.823 0.896
BE 0.036 0.073 0.052 0.020 0.007*** 0.028 0.005 0.021 0.028 0.047
0.905 0.394 0.719 0.845 0.001 0.719 0.713 0.723 0.878 0.999
FI 0.013 0.021 0.038 0.026 0.067*** 0.064* 0.002 0.036 0.016 0.057***
0.691 0.676 0.847 0.723 0.000 0.093 0.721 0.911 0.271 0.000
FR 0.022 0.041 0.086 0.032 0.155*** 0.015 0.081 0.036 0.033 0.011
0.972 0.777 0.801 0.772 0.000 0.696 0.837 0.969 0.792 0.845
GE 0.021 0.038 0.029 0.033 0.090*** 0.006 0.062 0.045 0.021 0.088***
0.943 0.814 0.678 0.942 0.000 0.725 0.408 0.858 0.001
GR 0.078 0.034*** 0.072*** 0.103*** 0.035*** 0.090 0.019 0.180 0.112*** 0.031
0.524 0.000 0.000 0.000 0.000 0.769 0.751 0.815 0.000 0.786
IR 0.040 0.005 0.001 0.003 0.051 0.022 0.048 0.008 0.064 0.003
0.659 0.914 0.934 0.757 0.338 0.963 0.376 0.976 0.701 0.880
IT 0.162 0.002 0.089 0.030 0.029 0.075*** 0.049 0.022 0.032 0.005
0.743 0.493 0.188 0.981 0.813 0.000 0.416 0.944 0.541 0.950
NE 0.054*** 0.017 0.043 0.078 0.038 0.050 0.003 0.083** 0.004 0.070
0.000 0.668 0.438 0.303 0.542 0.654 0.882 0.023 0.999 0.918
PO 0.026 0.022 0.036* 0.056 0.004 0.018*** 0.036 0.014 0.002 0.024
0.445 0.784 0.094 0.607 0.000 0.422 0.823 0.996 0.862
SP 0.006 0.030 0.005*** 0.016 0.056 0.005 0.001 0.008 0.013 0.001
0.940 0.999 0.000 0.937 0.166 0.978 0.807 0.926 0.969 0.937
country i’s Money Market FSI to country j’s Money Market FSI. The direction of the effects is from each row towards the columns. Countries are reported,
according to their initials: AU (Austria), BE (Belgium), FI (Finland), FR (France), GE (Germany), GR (Greece), IR (Ireland), IT (Italy), NE (Netherlands), PO
(Portugal) and SP (Spain).
Table 4b
The spillover effect of the FSI transmission channels across countries’ Money Markets.
Panel B. Money Market: The spillover effects (‘‘b” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.068 0.010 0.002 0.053 0.002 0.005 0.018 0.023 0.054 0.012***
0.392 0.922 0.921 0.896 0.904 0.812 0.941 0.877 0.683 0.000
BE 0.014 0.097*** 0.113*** 0.016 0.030 0.042*** 0.058 0.018 0.118 0.054
0.974 0.000 0.000 0.871 0.268 0.002 0.358 0.378 0.206 0.940
FI 0.065 0.027** 0.026 0.018 0.013 0.045 0.025 0.016 0.041 0.039
0.404 0.044 0.889 0.918 0.162 0.502 0.991 0.927 0.575 0.977
FR 0.126*** 0.004* 0.010 0.063 0.078*** 0.044** 0.000 0.003 0.007 0.034
0.001 0.070 0.920 0.168 0.008 0.027 0.980 0.927 0.848 0.915
GE 0.005 0.033 0.002 0.000 0.029 0.036 0.011 0.030 0.039 0.006
0.985 0.675 0.981 0.888 0.928 0.514 0.800 0.891 0.801 0.993
GR 0.027 0.053* 0.018 0.049 0.018 0.036 0.042 0.048 0.030** 0.025
0.298 0.096 0.262 0.284 0.959 0.309 0.718 0.808 0.013 0.112
IR 0.012** 0.089** 0.034 0.062*** 0.058 0.016 0.052*** 0.059 0.208*** 0.003
0.011 0.014 0.698 0.001 0.490 0.832 0.000 0.914 0.000 0.746
IT 0.079 0.002 0.096 0.000 0.149*** 0.049 0.013*** 0.001 0.047*** 0.001
0.580 0.878 0.964 0.974 0.001 0.793 0.000 0.994 0.000 0.851
NE 0.012 0.034 0.024 0.003 0.004 0.003 0.046 0.059 0.058*** 0.056
0.830 0.203 0.956 0.951 0.945 0.969 0.944 0.269 0.000 0.263
PO 0.009 0.001 0.012 0.004 0.012 0.152*** 0.062*** 0.113*** 0.067 0.180
0.880 0.980 0.798 0.872 0.937 0.000 0.000 0.000 0.309 0.412
SP 0.054*** 0.000 0.027 0.004 0.026 0.110 0.003 0.008 0.003 0.095
0.000 0.954 0.981 0.977 0.667 0.325 0.453 0.918 0.922 0.605
from country i’s Money Market FSI to country j’s Money Market FSI. The direction of the effects is from each row towards the columns. Countries are
reported, according to their initials: AU (Austria), BE (Belgium), FI (Finland), FR (France), GE (Germany), GR (Greece), IR (Ireland), IT (Italy), NE (Netherlands),
PO (Portugal) and SP (Spain).
Table 5a
The information flow effect of the FSI transmission channels across countries’ Equity Markets.
Panel A. Equity Markets: The news surprises (‘‘a” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.041 0.129** 0.024 0.055 0.118 0.016 0.003 0.052 0.022 0.030
0.731 0.022 0.623 0.446 0.231 0.891 0.858 0.885 0.768 0.753
BE 0.005 0.103*** 0.110 0.077 0.011 0.061*** 0.103 0.047 0.019 0.005
0.959 0.006 0.194 0.184 0.915 0.004 0.501 0.684 0.910 0.942
FI 0.008 0.009 0.136** 0.009 0.042 0.013 0.002 0.133* 0.083 0.084
0.851 0.517 0.045 0.953 0.453 0.853 0.995 0.061 0.735 0.538
FR 0.001 0.093 0.020 0.039 0.063 0.069 0.116*** 0.051 0.072 0.069
0.877 0.293 0.626 0.988 0.109 0.640 0.004 0.666 0.312 0.619
GE 0.023 0.023*** 0.000 0.011 0.004 0.041 0.093 0.028*** 0.064 0.004
0.621 0.000 0.986 0.994 0.921 0.485 0.543 0.000 0.612 0.964
GR 0.147* 0.009 0.019 0.075 0.024 0.010 0.009 0.003 0.004 0.681***
0.071 0.925 0.835 0.202 0.595 0.963 0.983 0.979 0.972 0.000
IR 0.019 0.081** 0.017 0.053 0.064 0.006 0.361 0.015 0.053 0.050
0.786 0.028 0.863 0.951 0.632 0.952 0.186 0.836 0.888 0.591
IT 0.045 0.146 0.018 0.090*** 0.072 0.006 0.050 0.019 0.068 0.010
0.421 0.409 0.980 0.000 0.740 0.992 0.811 0.997 0.812 0.889
NE 0.078 0.006 0.054 0.012 0.034*** 0.018 0.028 0.019 0.015 0.012
0.686 0.974 0.683 0.949 0.002 0.885 0.835 0.997 0.934 0.958
PO 0.001 0.074 0.010 0.080 0.119 0.165** 0.026 0.037 0.047 0.003
0.831 0.395 0.978 0.799 0.479 0.015 0.890 0.929 0.895 0.800
SP 0.045 0.041 0.130 0.069 0.004 0.102** 0.062 0.002 0.022 0.086***
0.666 0.723 0.180 0.608 0.949 0.015 0.456 0.958 0.908 0.000
country i’s Equity Market FSI to country j’s Equity Market FSI. The direction of the effects is from each row towards the columns. Countries are reported,
Table 5b
The spillover effect of the FSI transmission channels across countries’ Equity Markets.
Panel B. Equity Markets: The spillover effects (‘‘b” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.043 0.031 0.009 0.122*** 0.042 0.002 0.004 0.021 0.018 0.019
0.660 0.183 0.624 0.001 0.240 0.967 0.876 0.948 0.386 0.758
BE 0.030 0.043*** 0.101*** 0.186*** 0.015 0.025 0.027 0.006 0.025 0.007
0.754 0.000 0.001 0.000 0.870 0.103 0.782 0.990 0.499 0.879
FI 0.016 0.056*** 0.029 0.002 0.035 0.010 0.028 0.018 0.070 0.011
0.457 0.000 0.404 0.976 0.204 0.662 0.971 0.530 0.884 0.841
FR 0.008 0.085* 0.001 0.053 0.001 0.108 0.039*** 0.050 0.012 0.027
0.522 0.057 0.592 0.706 0.923 0.237 0.000 0.378 0.919 0.394
GE 0.012 0.004*** 0.004 0.001 0.012 0.074* 0.016 0.033**** 0.021 0.014
0.728 0.000 0.955 0.993 0.923 0.080 0.933 0.000 0.877 0.628
GR 0.039 0.028 0.031 0.004 0.027 0.058 0.006 0.001 0.012 0.996***
0.168 0.644 0.231 0.840 0.921 0.287 0.703 0.925 0.635 0.000
IR 0.073 0.036 0.004 0.046 0.153**** 0.004 0.156*** 0.001 0.071 0.014
0.193 0.137 0.931 0.943 0.000 0.975 0.000 0.942 0.847 0.866
IT 0.023 0.005 0.014 0.071*** 0.013 0.048 0.022 0.000 0.037 0.001
0.264 0.963 0.984 0.000 0.939 0.277 0.268 0.998 0.871 0.970
NE 0.030 0.006 0.041 0.003 0.026*** 0.004 0.035 0.003 0.026 0.015
0.685 0.987 0.109 0.957 0.000 0.951 0.530 1.000 0.847 0.923
PO 0.028 0.004 0.077*** 0.026 0.005 0.004 0.037 0.106 0.026 0.004
0.321 0.956 0.000 0.855 0.984 0.891 0.860 0.538 0.944 0.406
SP 0.044 0.025 0.026 0.025 0.010 0.077*** 0.014 0.030 0.020 0.004
0.153 0.688 0.592 0.731 0.923 0.000 0.777 0.861 0.860 0.725
from country i’s Equity Market FSI to country j’s Equity Market FSI. The direction of the effects is from each row towards the columns. Countries are
reported, according to their initials: AU (Austria), BE (Belgium), FI (Finland), FR (France), GE (Germany), GR (Greece), IR (Ireland), IT (Italy), NE (Netherlands),
PO (Portugal) and SP (Spain).
Table 6a
The information flow effect of the FSI transmission channels across countries’ Bond Markets.
Panel A. Bond Markets: The news surprises (‘‘a” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.063 0.132*** 0.005 0.001*** 0.041 0.048 0.096 0.002 0.119 0.012
0.458 0.000 0.897 0.009 0.594 0.220 0.668 0.704 0.233 0.881
BE 0.013 0.023 0.055 0.099** 0.022 0.000 0.002 0.001 0.122** 0.011
0.941 0.664 0.669 0.029 0.432 0.298 0.684 0.907 0.012 0.800
FI 0.067* 0.071 0.101** 0.035 0.006 0.216 0.060 0.007 0.048 0.021
0.059 0.123 0.025 0.386 0.414 0.442 0.687 0.577 0.499
FR 0.040 0.017 0.039 0.024 0.030 0.065 0.034 0.000 0.043 0.020
0.480 0.949 0.531 0.670 0.994 0.605 0.443 0.991 0.912 0.664
GE 0.010 0.024 0.025 0.001 0.009 0.012 0.021 0.000*** 0.072*** 0.065**
0.723 0.448 0.161 0.865 0.214 0.327 0.988 0.001 0.000 0.010
GR 0.037 0.136*** 0.274 0.446 0.017*** 0.102 0.030 0.125 0.006 0.003
0.826 0.002 0.980 0.000 0.388 0.354 0.567 0.957 0.905
IR 0.023 0.000 0.213 0.244 0.033 0.003 0.012 0.001 0.004 0.005
0.682 0.666 0.790 0.345 0.338 0.923 0.861 0.598 0.974 0.796
IT 0.058 0.112 0.022 0.187** 0.023 0.001 0.035 0.008 0.197 0.079
0.648 0.280 0.925 0.010 0.998 0.617 0.866 0.756 0.264 0.251
NE 0.010 0.026 0.001 0.014 0.000*** 0.009 0.000 0.006 0.001 0.047*
0.876 0.125 0.483 0.936 0.001 0.799 0.697 0.891 0.661 0.089
PO 0.025 0.022 0.004 0.049 0.105 0.045 0.001 0.052 0.016 0.000
0.759 0.801 0.965 0.779 0.281 0.831 0.947 0.144 0.816 0.678
***
SP 0.022 0.005 0.021 0.079 0.094 0.002 0.005 0.064 0.020 0.000
0.920 0.970 0.735 0.237 0.229 0.926 0.820 0.158 0.951 0.565
country i’s Bond Market FSI to country j’s Bond Market FSI. The direction of the effects is from each row towards the columns. Countries are reported,
Table 6b
The spillover effect of the FSI transmission channels across countries’ Bond Markets.
Panel B. Bond Markets: The spillover effects (‘‘b” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.213 0.019 0.018 0.011 0.028*** 0.050 0.075 0.004 0.119 0.258
0.851 0.826 0.921 0.577 0.000 0.796 0.965 0.913 0.960 0.802
BE 0.228 0.033 0.139 0.350** 0.010 0.000 0.003 0.018 0.133 0.031
0.532 0.945 0.605 0.010 0.172 0.253 0.756 0.920 0.772 0.941
FI 0.040 0.060 0.035 0.015 0.088 0.084 0.063 0.003 0.009 0.034
0.193 0.659 0.693 0.235 0.872 0.729 0.324 0.934 0.707
FR 0.003 0.008 0.026 0.010 0.004 0.044 0.023 0.016 0.093 0.137
0.721 0.987 0.875 0.855 0.999 0.685 0.679 0.993 0.704 0.238
GE 0.002 0.040 0.001 0.007 0.029*** 0.000 0.009 0.000 0.015 0.003
0.826 0.674 0.708 0.598 0.000 0.823 0.999 0.165 0.979 0.849
GR 0.042 0.005 0.440 0.025 0.003*** 0.101 0.024* 0.053 0.053 0.000
0.869 0.929 0.998 0.000 0.279 0.063 0.842 0.913 0.971
IR 0.027 0.001* 0.002 0.045 0.020 0.035 0.048 0.000 0.001 0.006
0.848 0.093 0.948 0.851 0.624 0.170 0.360 0.780 0.958 0.897
IT 0.171 0.300 0.030 0.105 0.014 0.016*** 0.034 0.020 0.053 0.145
0.903 0.447 0.911 0.337 1.000 0.001 0.908 0.806 0.885 0.172
NE 0.000 0.021 0.001 0.009 0.000 0.011** 0.001 0.002 0.002 0.030
0.897 0.825 0.818 0.982 0.241 0.030 0.912 0.943 0.863 0.968
PO 0.051 0.495* 0.230* 0.121 0.011 0.044 0.007 0.043 0.004 0.000
0.960 0.080 0.061 0.823 0.995 0.577 0.896 0.829 0.905 0.835
SP 0.179 0.243 0.040 0.256* 0.022 0.002 0.001 0.303*** 0.039 0.000
0.948 0.356 0.800 0.079 0.804 0.930 0.920 0.002 0.894 0.808
from country i’s Bond Market FSI to country j’s Bond Market FSI. The direction of the effects is from each row towards the columns. Countries are reported,
4.3. Transmission channels across countries
Tables 7a and 7b depict the country-wide financial stress spillovers. According to these findings, the most heavily affected
countries are those with the strongest spillovers to the rest of the Euro Area countries. Especially Ireland, Belgium and France
present significant news and volatility spillover effects to the rest of the economies. The previously mentioned countries
Table 7a
The information flow effect of the National FSI transmission channels across countries.
Panel A. National FSI: The news surprises (‘‘a” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.073 0.033 0.057 0.019 0.010 0.041 0.081 0.016 0.021 0.017
0.611 0.885 0.945 0.922 0.995 0.589 0.984 0.934 0.664 0.693
BE 0.043 0.024 0.062*** 0.002 0.035*** 0.059** 0.168* 0.096 0.023 0.086
0.839 0.311 0.005 0.757 0.000 0.043 0.071 0.637 0.956 0.970
FI 0.021 0.155*** 0.043 0.000 0.063 0.006*** 0.073 0.015 0.030 0.001
0.398 0.000 0.954 0.954 0.315 0.000 0.735 0.739 0.956 0.963
FR 0.136 0.003 0.002 0.010 0.025 0.100 0.027 0.015 0.015*** 0.020
0.885 0.859 0.944 0.983 0.557 0.477 0.860 0.863 0.000 0.413
GE 0.017 0.006 0.001 0.001 0.003 0.159*** 0.078 0.046 0.003 0.007
0.950 0.883 0.892 0.894 0.920 0.002 0.945 0.737 0.961 0.859
GR 0.108 0.024*** 0.151 0.029 0.024 0.089 0.031 0.005 0.122 0.025
0.833 0.000 0.256 0.268 0.619 0.989 0.761 0.942 0.712 0.814
IR 0.099 0.164*** 0.046*** 0.111** 0.000 0.157 0.226* 0.032 0.030 0.243***
0.240 0.000 0.000 0.013 0.781 0.960 0.082 0.130 0.623 0.000
IT 0.015 0.017 0.085 0.069 0.001 0.034 0.034 0.009 0.046 0.000
0.991 0.774 0.647 0.597 0.898 0.870 0.473 0.952 0.664 0.964
NE 0.020 0.013 0.007 0.036** 0.011 0.001 0.005 0.033 0.040 0.030***
0.810 0.904 0.893 0.027 0.844 0.935 0.891 0.814 0.325 0.010
PO 0.003 0.070 0.007 0.060*** 0.012 0.048 0.018 0.039 0.020 0.008
0.949 0.883 0.996 0.000 0.904 0.565 0.468 0.560 0.617 0.972
SP 0.000 0.004 0.012 0.011 0.016 0.004 0.126*** 0.004 0.110 0.000
0.889 0.983 0.971 0.821 0.786 0.930 0.000 0.623 0.336 0.974
country i’s National FSI to country j’s National FSI. The direction of the effects is from each row towards the columns. Countries are reported, according to their
initials: AU (Austria), BE (Belgium), FI (Finland), FR (France), GE (Germany), GR (Greece), IR (Ireland), IT (Italy), NE (Netherlands), PO (Portugal) and SP (Spain).
Table 7b
The spillover effect of the National FSI transmission channels across countries.
Panel B. National FSI: The spillover effects (‘‘b” coefficients) of the Bivariate Full BEKK VAR(1)-MGARCH(1,1) models
AU 0.022 0.043 0.002 0.002 0.007 0.027 0.116 0.018 0.023 0.020
0.878 0.846 0.956 0.857 0.990 0.621 0.976 0.957 0.699 0.642
BE 0.022 0.035*** 0.066*** 0.022 0.033*** 0.002 0.011 0.047*** 0.025 0.112
0.901 0.000 0.000 0.569 0.000 0.649 0.787 0.001 0.898 0.707
FI 0.056* 0.108*** 0.066 0.000 0.019 0.163*** 0.036 0.040 0.075 0.029
0.091 0.000 0.937 0.896 0.238 0.000 0.622 0.303 0.960 0.310
FR 0.011 0.004 0.008 0.006 0.033* 0.079*** 0.107* 0.002 0.167*** 0.021*
0.983 0.887 0.911 0.973 0.081 0.003 0.079 0.876 0.000 0.064
GE 0.014 0.001 0.004 0.004 0.030 0.021 0.049 0.041 0.002 0.004
0.824 0.744 0.950 0.893 0.548 0.393 0.934 0.325 0.943 0.874
GR 0.007 0.107*** 0.102 0.003 0.140*** 0.010 0.024 0.001 0.034 0.009
0.888 0.000 0.155 0.871 0.006 0.991 0.774 0.737 0.843 0.900
IR 0.057 0.001 0.008*** 0.025 0.002 0.029 0.103*** 0.002 0.022 0.004
0.410 0.475 0.000 0.303 0.612 0.948 0.000 0.705 0.466 0.759
IT 0.040 0.002 0.078 0.037 0.139 0.054 0.009*** 0.015 0.038 0.009
0.963 0.847 0.306 0.497 0.506 0.199 0.000 0.953 0.709 0.942
NE 0.004 0.041 0.008 0.021** 0.002 0.001 0.144*** 0.038 0.013 0.033***
0.925 0.373 0.641 0.014 0.962 0.787 0.000 0.727 0.825 0.003
PO 0.000 0.021 0.015 0.052*** 0.004 0.032 0.007 0.001 0.001 0.072
0.692 0.829 0.990 0.000 0.895 0.877 0.917 0.869 0.862 0.546
SP 0.005 0.126 0.016 0.006** 0.004 0.034 0.001 0.008 0.113*** 0.003
0.812 0.902 0.958 0.037 0.930 0.907 0.261 0.885 0.000 0.852
each parameter reported. The elements of this non-symmetric matrix represent the spillover effect (‘‘bij” coefficient) with respect to the spillover effect from
country i’s National FSI to country j’s National FSI. The direction of the effects is from each row towards the columns. Countries are reported, according to their
initials: AU (Austria), BE (Belgium), FI (Finland), FR (France), GE (Germany), GR (Greece), IR (Ireland), IT (Italy), NE (Netherlands), PO (Portugal) and SP (Spain).
(except Greece) exhibit spillover effects to the rest of the sample economies. On the same time, Ireland and France are also
the main recipients of financial stress spillovers. It is interesting to notice that Germany is highly immune to financial stress
transmission. Again, a clear pattern cannot be established here, in relation to potential transmission channels. This can be
due to the fact that the effects here are, essentially, influenced by the nature of the used dataset (country indexes are the
average values of the market specific indicators for the set of markets used for each economy). In total, debt ridden countries
do not seem to lead the effects here, while they are also the major recipients of transmission effects from, both, peripheral as
well as the major Euro Area economies.
4.4. Transmission channels across regions
In order to further examine the interconnections and spillover effects between the Euro Area core and peripheral econo-
mies, we proceed to a core-periphery empirical investigation. As before, our estimation strategy is the same. We employ
GARCH-BEKK modelling framework, using financial stress indices in our dataset. This time, the systemic risk indexes are
regional ones, representing the Euro core and periphery. In order to construct them, we use the average of the indices used
for the countries before: the core countries and the GIIPS (Greece, Ireland, Italy, Portugal and Spain) for the peripheral indi-
cators. In this section, we present the results for the whole period, while in the robustness checks we also proceed to a pre-
and post-crisis examination.
According to our results, the prominent role of banking sector and bond markets in the Euro Area is verified. These two
are the main markets where substantial risk transmission channels are identified. Additionally, the intensity of information
flow (‘‘news effect”, Table 8a) is evident to both directions and with marginally greater impact from the periphery to core
countries. In fact, this is more pronounced for the core countries’ banking sector, which used to be one of the major creditors
for peripheral European countries. However, these findings need to be further investigated throughout the crisis period (sub-
sample analysis), as this will indicate the potential impact of the Eurozone crisis outbreak to these spillover channels.
Similarly, the banking and bond markets are the most susceptible sectors to financial spillover effects (Table 8b). The core
banks have major effects on all peripheral markets, while the same holds for the peripheral bond stress, transmitting to all
Table 8a
The information flow effect of the FSI transmission channels across regions.
Panel A. Core – Periphery: The elements of this non-symetrix matrix represent the aij coefficient of the bivariate GARCH model with respect to the
spillover effect from core and periphery’s markets FSIs to another market
aij Core Periphery
Total Banking Money Equity Bond Total Banking Money Equity Bond
Sector Market Markets Markets Sector Market Markets Markets
Core Total 0.011 0.033 0.037 0.008 0.110 0.008 0.004 0.026 0.002
0.746 0.005 0.557 0.536 0.858 0.344 0.957 0.378 0.867
Banking 0.037 0.142*** 0.002*** 0.040 0.018*** 0.165*** 0.158 0.013 0.092***
Sector
0.908 0.000 0.000 0.222 0.000 0.000 0.567 0.177 0.000
Money 0.001 0.036 0.096 0.016* 0.092 0.008 0.007 0.070*** 0.025
Market
0.279 0.341 0.226 0.076 0.730 0.317 0.720 0.000 0.161
Equity 0.006 0.006*** 0.034 0.093** 0.011 0.074** 0.030 0.018 0.058
Markets
0.924 0.000 0.855 0.038 0.664 0.016 0.748 0.829 0.181
Bond 0.027 0.141 0.067 0.430 0.010 0.020 0.092 0.077 0.022***
Markets
0.954 0.125 0.573 0.107 0.933 0.786 0.844 0.383 0.000
Periphery Total 0.104 0.129*** 0.010 0.232 0.017 0.031*** 0.053 0.083*** 0.141***
0.384 0.000 0.976 0.321 0.824 0.008 0.206 0.000 0.000
Banking 0.003 0.007*** 0.018 0.023 0.071 0.086 0.101 0.387*** 0.111
Sector
0.773 0.000 0.813 0.876 0.558 0.727 0.918 0.000 0.738
Money 0.056 0.014 0.080 0.002 0.049 0.043 0.021 0.062 0.009
Market
0.688 0.890 0.605 0.984 0.784 0.890 0.909 0.692 0.865
Equity 0.027 0.220*** 0.093*** 0.165 0.167** 0.052 0.172*** 0.144 0.006
Markets
0.927 0.000 0.000 0.967 0.018 0.966 0.000 0.472 0.871
Bond 0.043 0.197*** 0.022 0.025 0.088*** 0.015*** 0.008 0.017 0.008
Markets
0.973 0.000 0.680 0.954 0.000 0.000 0.804 0.926 0.962
region i’s FSI (by market and aggregated) to region j’s FSI. The direction of the effects is from each row towards the columns.
Table 8b
The spillover effect of the FSI transmission channels across regions.
Panel B. Core-Periphery FSIs: The spillover effect (‘‘b” coefficients) of the Bivariate Full BEKK VAR(1) – MGARCH(1,1) models
bij Core Periphery
Total Banking Money Equity Bond Total Banking Money Equity Bond
Sector Market Markets Markets Sector Market Markets Markets
Core Total 0.005 0.006 0.045 0.001 0.045 0.011 0.038 0.016 0.002
0.797 0.012 0.364 0.977 0.293 0.118 0.697 0.407 0.606
Banking 0.087 0.042 0.031*** 0.023 0.036*** 0.190*** 0.124*** 0.004** 0.012***
Sector
0.462 0.174 0.000 0.122 0.000 0.000 0.000 0.038 0.000
Money 0.025*** 0.010 0.036 0.005*** 0.016 0.004* 0.010 0.005*** 0.033***
Market
0.000 0.308 0.496 0.000 0.955 0.088 0.191 0.000 0.001
Equity 0.000 0.044*** 0.084 0.026 0.024 0.070*** 0.007 0.005 0.020
Markets
0.940 0.000 0.465 0.226 0.282 0.000 0.721 0.987 0.323
Bond 0.049 0.092 0.049** 0.163** 0.035 0.009 0.005 0.034 0.050***
Markets
0.910 0.669 0.037 0.040 0.252 0.780 0.960 0.285 0.000
Periphery Total 0.038 0.024*** 0.003 0.107*** 0.003 0.004 0.014 0.069** 0.502***
0.821 0.000 0.998 0.009 0.939 0.351 0.901 0.027 0.000
Banking 0.020 0.004*** 0.003 0.023 0.025 0.008 0.098 0.995*** 0.138**
Sector
0.782 0.000 0.841 0.817 0.483 0.914 0.793 0.000 0.019
Money 0.146 0.001 0.016 0.003 0.012 0.037 0.014 0.019 0.020
Market
0.774 0.971 0.739 0.931 0.854 0.617 0.957 0.784 0.696
Equity 0.000 0.053*** 0.076*** 0.005 0.005 0.011 0.268*** 0.068 0.006***
Markets
0.645 0.004 0.000 0.990 0.920 0.904 0.000 0.330 0.000
Bond 0.024 0.015*** 0.046** 0.037 0.030*** 0.082*** 0.003 0.001 0.002
Markets
0.957 0.000 0.026 0.780 0.000 0.000 0.820 0.858 0.957
from region i’s FSI (by market and aggregated) to region j’s FSI. The direction of the effects is from each row towards the columns.
core (but not to the equity) financial markets. Moreover, periphery’s bond sector is exposed to spillover effects from core
markets and economies. On top of the above, both regions are interlinked and exposed to their own sources of distress. These
are interesting findings, indicating the complex nature of the interactions between Euro Area markets, indicating the threats
posed by both mature and less developed markets in this monetary union.
5. Further evidence and robustness checks
In order to substantiate our previous findings, we extend our analysis in many different ways. First, we use alternative
FSIs, by adopting a PCA-based aggregation approach and by using the ECB CISS index for sovereign risk within our modelling
framework. Second, we consider the cross-market transmission channels; that is the examination of all possible combina-
tions of countries and markets at the same time. Third, we extend the analysis by recruiting many alternative volatility spec-
ifications. This is executed by estimating a 11-dimensional model instead of the pairwise bivariate models, the analysis of
potential asymmetries at the volatility specifications and, finally, by testing the optimal lag terms (p, q) for the GARCH mod-
els in the aggregated regional analysis. Fourth, our robustness checks accounts for the effect of the financial crisis on the
regional distress transmission channels. Finally, we adopt the News Impact Surface approach for a more insightful examina-
tion of the core-periphery impact channels. Given the large number of figures and tables produced out of these additional
empirical exercises, the detailed results are available as an online appendix.
5.1. Alternative financial stress indices
5.1.1. A PCA-based approach for FSI

In order to test the reliability of our findings, we employ an alternative set of financial stress indices. The difference lies on
the aggregation method used to construct the FSIs. A principal components (PCA) approach is used, as opposed to the
variance-equal method in the baseline framework. In brief, PCA is a statistical method, transforming a set of correlated vari-
ables into a new linear combination of them. It relies on the covariance matrix of these indicators and their eigenvalues for
the decomposition of the principal components. In our case, the first principal component and its ensuing factor loadings are
Fig. 2. CISS sovereign risk sub-indices for core and peripheral countries.
used as weights for the final aggregation of the stand-alone indicators. After constructing the new stress indices, we re-
estimate our models for, both, country and market level analysis. Overall, the results are similar to the one produced from
the baseline model.
5.1.2. European central bank’s CISS indices

ECB has developed a range of systemic risk indicators, similar in nature to our own FSI dataset. The Composite Indicators
of Systemic Risk (CISS) aim to capture systemic risk exposures, as they become evident to a number of markets in Euro Area.
In their indexes, Hollo et al. (2012) incorporate metrics representing the prevailing market conditions in financial interme-
diaries, money, securities and foreign exchange markets. Since this particular dataset does not cover the whole spectrum of
markets and economies we focus, we decide to examine the potential sovereign risk transmission channels, as documented
by the employment of the CISS sovereign risk sub-indices for some core and peripheral European countries. The indices,
country-specific as well as the regional ones constructed by us, are shown in Fig. 2. The indices clearly exhibit a pattern
in accordance to the development of the economic events creating havoc in the sovereign debt conditions of, both, core
and peripheral European economies.
We conduct an empirical investigation, using all possible pairwise combinations for examining the sovereign risk trans-
mission channels in Eurozone’s core and periphery. A bivariate VAR(1) – Full BEKK GARCH(1,1) model is estimated and the
relevant spillover effects are shown in Table 9 of the online appendix. Panel A refers to the news surprises effect while Panel
B to the spillover effects per se. The basic outcome of this exercise is the lack of concrete evidence in favour of spillover
effects between the countries and peripheries of the sample. Only weak evidence towards regional transmission channels
are identified, even though not in the degree indicated to the baseline modelling framework. This weak regional clustering
effect is evident when the spillover effects are accounted for.
5.2. Cross-market analysis
The second robustness check refers to the consideration of all possible transmission channels of financial distress
‘‘between” and ‘‘within” the four markets (banks, money, equity and bonds) of the 11 Eurozone countries. Thus, our analysis
refers to the country- specific and market FSIs, consisting of the full 55 series of our dataset. All possible pairwise combina-
tions count to 1430, excluding each market’s own effect.
The estimated coefficients of all the bivariate Full BEKK-GARCH(1,1) models are used for calculating their joint distribu-
tions with respect to their significance. The joint distribution is classified as, either, ‘‘causer” or ‘‘receiver” of financial stress,
depending on the direction of the implied transmission channel between each combination of series. For each of the 55 ser-
ies, we estimated the percentage of the significant ‘‘causer” effect of volatility spillovers to the rest of the series. Similarly, we
also obtain the percentage of the significant ‘‘receiver” effect of volatility spillovers from the rest of the sample. Bar charts are
employed for the provision of more lucid and insightful results.3
Fig. 3 of the appendix represents the causal relationship between all FSIs, either the news impact effect (Panel A) or the
spillover effect (Panel B). On the country level, Ireland and Spain seem to have prominent role on the ‘‘news” effects. On the
3
The estimated parameters tables are available upon request.
other hand, several peripheral economies (Greece, Spain, Ireland), along with France, seem to contribute substantially to
financial distress volatility. Moreover, the periphery is more susceptible to such effects. Once again, the importance of bank-
ing and bond markets is underlined, with the role of most peripheral economies’ banking systems to be imminent. Again, the
major economies are also among the major bond risk transmitters. In the case of equity market, the results are again poor,
while the money market case is similar with the previous section analysis.
Overall, the market-level analysis provides some useful insights to the prevailing conditions to Euro Area economies and
markets. We find evidence of strong spillover effects among most of the economies under scrutiny. Moreover, the most vola-
tile and vulnerable to risk transmission are the bond and bank markets. A notable exception is the equity markets analysis,
where no convincing evidence for volatility spillover is detected. Regarding the main risk spillover propagators, results vary
but, again, there is no clear cut evidence whatsoever against a specific country or group of countries as the major financial
stress contributors.
5.3. Alternative volatility specifications
As the main body of the analysis is based on bivariate volatility models, our estimations might suffer from an omitted
variable bias, since there are third countries or markets that transmit or receive volatility from each examined pair in the
estimated bivariate models. For this reason, we employ a multivariate approach for the full set of country level FSIs under
consideration. In order to accomplish this task, a 11-dimensional VAR(1) – Full BEKK GARCH(1,1) model is estimated, in
accordance to the baseline model’s characteristics. The resulting estimation consists of 308 estimated coefficients for, both,
‘‘news” and persistence effects.
The results (Table 10 in the appendix) do not differ from the previous findings. There is hardly any statistically significant
alpha, while only a few significant spillover effects are identified in the second panel of this table. In fact, most of these
effects are own effects, as it is the case for Austria and Netherlands. Moreover, there are some bidirectional effects between
countries from the same group, such as Belgium and Finland. France and Portugal also seem to transmit financial distress
between each other. Overall, these findings should be assessed with caution, as they are computationally sensitive.4
Further robustness checks are provided for potential asymmetries within the volatility specifications (leverage effect), in
the case of regional stress transmission analysis. In order to do so, a range of alternative GARCH models is employed, namely
the EGARCH, the GJR and the APARCH models. All these models are extensions to the GARCH model which is our baseline
model in this paper. Again, on the basis of the pairwise bivariate framework, we consider the transmission channels between
the regional FSIs (i.e. core and periphery FSIs). The asymmetric volatility models are applied using a non-linear two-stage
estimation process that involves the orthogonalisation of our pairwise data within a PCA analysis. In the EGARCH model
specification (Nelson, 1991), our interest is focused on the h1 and h2 coefficients with h2 representing the leverage effect
(when h2 < 0 the leverage effect is taken place).
!
e e et1
t1 t1
lnðht Þ ¼ a0 þ h1 pffiffiffiffiffiffiffiffiffi E pffiffiffiffiffiffiffiffiffi þ h2 pffiffiffiffiffiffiffiffiffi þ b lnðht1 Þ ð10Þ
ht1 ht1 ht1
Another popular way to model the asymmetry of positive and negative innovations is the use of indicator functions
according to the GJR-GARCH(p,q) model (Glosten et al., 1993):
X
q X
q X
p
ht ¼ a0 þ ðai e2ti Þ þ ðci dðeti < 0Þ e2ti Þ þ ðbj htj Þ ð11Þ
i¼1 i¼1 j¼1
Here, ci, i = 1 . . . q, are parameters for estimation while d(.) denotes the indicator function (i.e. d(eti < 0) = 1 if eti < 0, and
d(eti < 0) = 0 otherwise). The GJR model allows for good news (eti > 0), and bad news (eti < 0), to have different effects on
the conditional variance. Therefore, in the case of the GJR(0,1) model, good news has an impact of a1, while bad news has an
impact of a1 + c1, meaning that for negative c1 the ‘‘leverage effect” exists.
According to Ding et al. (1993), we also employ the Asymmetric Power ARCH, or APARCH(p,q) model, which includes
seven ARCH models as special cases (ARCH, GARCH, A-GARCH, GJR, T-ARCH, N-ARCH and log-ARCH), with the following con-
ditional variance:
d X
q X
p
d
h2t ¼ a0 þ ai ðjet1 j c1;i eti Þd þ bj h2tj ð12Þ
i¼1 j¼1
This model imposes a Box-Cox power transformation of the conditional standard deviation process and the asymmetric
absolute innovations while the leverage effect is captured by the parameter c1.
The results of this analysis are presented on Table 11 of the Appendix. It is obvious that the leverage effect exists in most
cases while the model fit of asymmetric models is marginally better than that of our benchmark GARCH model. However, it
reflects the necessary transmission channels adequately (model diagnostics) and, thus, it is preferred for parsimonious
reasons.
4
In fact, our effort to estimate the same model for the four types of markets examined here brought no success. The only model that worked was this one.
Finally, we conduct an analysis for the optimal lag structure of the GARCH(p,q) bivariate specification. It is a necessary
step, in order to verify the validity and reliability of our chosen specification in the baseline model. Once more, the analysis
is performed on the core-periphery case. The evidence (Table 12) is in favour of the GARCH(1,1) model, compared to a higher
lagged specification.
5.4. Financial crisis and its long and short term dynamics
A sub-sample empirical investigation is conducted, in order to identify whether the Eurozone crisis outbreak led to a
shifting behaviour in the structure of spillover effects between the Euro Area markets. The estimation process is the same
as in the baseline approach, with May 2010 being the cut-off point for the pre- and post-crisis period analysis. The break
point coincides with the time Greece sought for financial assistance from its Eurozone partners.
As it can be seen from Table 13, there is a clear distinction on the interconnections between core and peripheral econo-
mies, before and after the crisis outbreak. In the first period, there are significant ‘‘news surprises” effects between these two
groups of countries, with strong bidirectional links. As it is also evident from Table 13.D, the persistence of these effects is
important, especially for the core-to-periphery direction. Interestingly, the situation is rather different in the second half of
the examined period (post-crisis time). The number of statistically significant coefficients, for both aij’s and bij’s, is signifi-
cantly lower, while a clustering effect is pronounced. Core economies and markets are susceptible to stress transmission
effects from core countries and the same holds for the periphery case. It is a clear indication of market decoupling taking
place, while market participants flee from the more vulnerable economies and adjust their portfolio positions towards safer
investments (flight to safety and flight to quality phenomena).
Beyond the sub-sample analysis, our aim is to provide some insights to the time varying conditional correlation behaviour
of the two regions’ markets discussed above. In order to do so, we employ the well-known DCC model, as developed by Engle
and Sheppard (2001) and Engle (2002). This MGARCH approach is based on a two-step procedure, where standardized resid-
uals, produced from univariate GARCH models in the first step, are subsequently incorporated to the estimation of the con-
ditional correlation estimator in the second stage. The graphical exposition of the dynamic conditional correlation for the
regions and markets of interest is provided in Fig. 4 of the appendix. Overall, the total conditional correlation is mild and
relatively stable in size, throughout the period under investigation. Also, with the introduction of euro as the common cur-
rency for these groups of countries, we can see a significant increase in the conditional correlation for these markets, espe-
cially for the money market case. This is logical, given the nature of the money market indicators, which incorporate
liquidity, funding and interbank market indicators. All these metrics are expected to be strongly interrelated for countries
sharing the same monetary policy. Moreover, the banking sector exhibits an increasing trend in its conditional correlation,
an outcome related again with the common currency effect. It is noteworthy the opposite reaction, in terms of the degree of
correlation that can be identified for the money market and the bond market of the two regions right before and during the
peak of Eurozone crisis. Money market conditional correlation plummets, remaining at fairly low level and with strong
swings between positive and negative values, until the end of the sample. On the other hand, bond markets turn from neg-
ative into positive correlation, remaining like this from this point onwards. Once more, both markets clearly indicate their
importance and their degree of interconnectedness for the two regions under investigation.
5.5. Volatility surfaces
The usage of impact news surfaces is the last empirical exercise. The purpose of this test is to examine whether and by
how much the conditional variance of the ensuing GARCH model is affected by its own lagged innovations, as well as the
other market innovations (Kroner and Ng, 1998; Savva, 2009). Additionally, potential asymmetric effects can also be cap-
tured, in relation to the potential shocks on the stress transmission volatility of the markets (Martens and Poon, 2001). Once
more, the news impact surface for the regional FSIs are constructed and presented in Figs. 5A and 5B of the appendix for the
core and peripheral regions respectively. Undoubtedly, there are bidirectional transmission channels that explain the way
periphery and core countries’ distress is distributed to each other. A more careful consideration of these figures (and under
alternative parametrizations) reveals that there is a tendency for these channels to have a prominent effect from the core
countries to the peripheral ones. This is because peripheral positive distress shocks are transmitted to the core countries
when their magnitude is big enough. It is noteworthy that some asymmetric behaviour can be identified, especially for
the case of the core countries conditional variance, as well as for the periphery case (but much milder in this case). More
specifically, the conditional variance of core countries’ FSI is increasing smoothly (Fig. 5.A) while current and past core-
periphery distress shocks are taking place. On the other hand, the conditional variance of periphery distress (Fig. 5.B) is
affected rapidly during turbulent financial conditions within core economies.
6. Concluding remarks
Our aspiration for this paper is the Eurozone crisis that, when was fully fledged and on its peak, rendered the European
economy under severe strains. A prolonged recessionary period is its real economy reflection. Both governments and market
participants were alerted for the eventuality of crisis transmission from the most vulnerable EMU economies to the rest.
Thus, there is a growing research interest, focusing on the examination of contagion and channels of interconnectedness
between the major protagonists of the Euro crisis.
This paper aims to extend the relevant literature in several ways. First of all, our interest is to study the crisis to its fully
diverse nature. That is, we do not limit our study only to the sovereign risk or the banking instability issues, as most of the
research has done until now. Instead, we try to encapsulate the necessary information into a number of metrics that are able
to provide clear cut insights to the crisis and its constituents. In order to do it, we employ a set of financial stress indices.
These are aggregate indicators, representing the level of systemic risk in each one of the markets we analyze. These are,
the banking sector, money market, equity and bond markets, while we also provide an index for each national economy.
The dataset contents and development is rather unique, towards other similar aggregate measures existing in the literature
or developed by policy making institutions. The individual indicators used cover several aspects and sources of financial risk,
while the degree of disaggregation in countries and markets renders our dataset distinctive. The next important extension is
the adoption of a multivariate GARCH framework for the empirical investigation of potential spillover effects among the
aforementioned markets. To the best of our knowledge, it is the first time that such a modelling approach is used in conjunc-
tion to such successful systemic risk indicators. It is an innovative combination, given the very nature of the financial stress
indexes and the ability of the MGARCH type of models to estimate time-varying co-variances. Another important feature of
our research is the simultaneous assessment of potential volatility transmission channels between and within the previously
mentioned markets and countries. The Eurozone case is analyzed and discussed in a detailed level and markets’ decompo-
sition. Moreover, a core – periphery modelling is provided, together with sub-sample analysis to take into account potential
changes to spillover effects due to the Eurozone crisis outbreak. A whole battery of robustness checks and further economet-
ric specifications are employed, in order to provide further credibility and substantiate the baseline results produced by our
BEKK modelling framework. Finally, our dataset covers the Eurozone crisis until very recently.
The results shed new light into the Euro Area’s volatility transmission literature. There is strong evidence that there exist
multiple links between the EMU financial markets. Depending on the sector discussed, the main receivers and transmitters
of the spillover effects vary. For instance, it is true that the GIIPS countries significantly contribute to the cross-volatility,
especially in the case of the country level analysis and the banking and money markets. On the same time, the core is also
an important channel of variance volatility transmission, both within the North European countries, but also towards the
peripheral ones. Such a, somewhat surprising, result for part of the profession is in accordance to latest findings
(Antonakakis and Vergos, 2013; Kohonen, 2014). Moreover, we find strong bidirectional effects between countries of the
same group. Equity market, on its single market analysis, does not provide convincing evidence as a sector where volatility
spillovers take place. Also, the case of money market is interesting. Given the representation of the interbank funding con-
ditions, along with the relative volatility measures and the yield curve, this sector manifests itself as one which central bank-
ers should pay special attention to. An important finding, directly relevant to policy making decisions, is the one coming from
the sub-sample analysis. The decoupling hypothesis for the Euro Area markets clearly holds and it is manifested with the
crisis outbreak. The, initially, highly interconnected European markets exhibit a clear disaggregation when the adverse eco-
nomic events took place in the peripheral economies, leading to markets interacting and being susceptible to effects from
their particular group of countries (either, the core or the peripheral ones). Investors seem to fled from the debt ridden
economies, looking for alternative and safe investment positions to safer markets (flight to quality and flight to liquidity phe-
nomena). These findings are confirmed by the robustness checks, using alternative volatility specifications, together with
techniques verifying the existence of asymmetries in the market under investigation.
We believe that these facts underline the direction towards which macroprudential policies should aim to. Such policies
should be formulated in a way to accommodate the multifaceted nature of modern financial systems, taking into account
and monitor potential risks that can source to different markets. Moreover, the clustering effect identified in our econometric
investigation should be taken into serious consideration from Eurozone policy makers. Divergent policies should be followed
and applied to different regions or countries, according to their special characteristics and vulnerabilities and applied on an
ad hoc basis, depending on the prevailing market and macroeconomic conditions. The mix of standardized policies, as those
currently applied for monitoring and regulating Euro Area banks, together with tailored made, country level prudential poli-
cies, can be the appropriate way forward for ameliorating policy responses to future events of economic instability.
Acknowledgements
We are grateful to our discussants and the participants to the 4th International Conference of the Financial Engineering
and Banking Society (June 2014), 4th International Symposium in Computational Economics and Finance (April 2016) and
the 19th International Conference on Macroeconomic Analysis and International Finance (May 2016) for their fruitful discus-
sion and suggestions. Andreas Tsopanakis also thanks seminar participants at Sabanci University seminar series. We also
thank the anonymous reviewers for their helpful and insightful comments.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.intfin.
2017.09.003.
References
Alter, A., Beyer, A., 2014. The dynamics of spillover effects during the European sovereign debt turmoil. J. Bank. Financ. 42, 134–153.
Alter, A., Schuler, Y.S., 2012. Credit spread interdependencies of European states and banks during the financial crisis. J. Bank. Financ. 36, 3444–3468.
Antonakakis, N., Vergos, K., 2013. Sovereign bond yield spillovers in the Euro zone during the financial and debt crisis. J. Int. Financ. Markets Inst. Money 26,
258–272.
Apostolakis, G., Papadopoulos, A.P., 2014. Financial stress spillovers in advanced economies. J. Int. Financ. Markets Inst. Money 32, 128–149.
Apostolakis, G., Papadopoulos, A.P., 2015. Financial stress spillovers across the banking, securities and foreign exchange markets. J. Financ. Stab. 19, 1–21.
Arezki, R., Candelon, B., Sy, A.N.R., 2011. Sovereign Rating News and Financial Markets Spillovers: Evidence from the European Debt Crisis. IMF Working
Papers Series, 68
Audige. H., 2013. A New Approach of Contagion based on Smooth Transition Conditional Correlation GARCH Models: An Empirical Application to the Greek
Crisis. EconomiX CNRS Working Paper Series, 2
Barth, J.R., Prabha, A., Yun, G., 2012. The Eurozone financial crisis: role of interdependencies between bank and sovereign risk. J. Financ. Econ. Policy 4, 76–
97.
Bauwens, L., Laurent, S., Rombouts, J.V., 2006. Multivariate GARCH models: a survey. J. Appl. Economet. 21, 79–109.
Bollerslev, T., 2010.Glossary to ARCH (GARCH). In: Bollerslev, T., Russel J., Watson, M. (Eds.), 2010.Volatility and Time Series Econometrics: Essays in Honor
of Robert Engle, Oxford University Press, Oxford (Chapter 8).
Bollerslev, T., Engle, R.F., Wooldridge, J.M., 1988. A capital asset pricing model with time varying covariances. J. Polit. Econ. 96, 116–131.
Brutti, F., Saure, P., 2015. Transmission of sovereign risk in the Euro crisis. J. Int. Econ. 97 (2), 231–248.
Bruyckere, V.D., Gerhardt, M., Schepens, G., Vennet, R.V., 2013. Bank/sovereign risk spillovers in the European debt crisis. J. Bank. Financ. 37, 4793–4809.
Caceres, C., Guzzo, V., Segoviano, M., 2010. Sovereign Spreads: Global Risk Aversion, Contagion or Fundamentals? IMF Working Paper Series, 120
Cardarelli, R., Elekdag, S., Lall, S., 2011. Financial stress and economic contractions. J. Financ. Stab. 7, 78–97.
Caporin, M., Pelizzon, L., Ravazzolo, F., Rigobon, R., 2013.Measuring Sovereign Contagion in Europe. NBER Working Paper Series, 18741
Dajcman, S., 2012. Comovement between stock and bond markets and the flight-to-quality during financial market turmoil – a case of the Eurozone
countries most affected by the sovereign debt crisis of 2010–2011. Appl. Econ. Lett. 19, 1655–1662.
Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive directional measurement of volatility spillovers. Int. J. Forecast. 23, 57–66.
Diebold, F.X., Yilmaz, K., 2014. On the network topology of variance decompositions: measuring the connectedness of financial firms. J. Economet. 182, 119–
134.
Ding, Z., Granger, C.W.J., Engel, R.F., 1993. A long memory property of stock market returns and a new model. J. Empiric. Financ. 1, 83–106.
Engle, F.R., 2002. Dynamic conditional correlation: a simple class of multivariate generalized autoregressive conditional heteroskedasticity models. J. Bus.
Econ. Stat. 20, 339–350.
Engle, R.F., Ito, T., Lin, W.-L., 1990. Meteor showers or heat waves? Heteroskedastic intra-daily volatility in the foreign exchange market. Econometrica 58,
525–542.
Engle, F.R., Kroner, F.K., 1995. Multivariate simultaneous generalized ARCH. Economet. Theory 11, 122–150.
Engle, F.R., Sheppard, K., 2001, Theoretical and Empirical properties of Dynamic Conditional Correlation Multivariate GARCH. NBER Working Papers, 8554
Forbes, K., Rigobon, R., 2002. No contagion, only interdependence: measuring stock market comovements. J. Financ. LVII, 2223–2261.
Fratzscher, N., Lo Duca, M., Straub, R., 2016. ECB unconventional monetary policy: market impact and international spillovers. IMF Econ. Rev. 6, 36–74.
Glosten, L.R., Jagannathan, R., Runkle, D.E., 1993. On the relation between the expected value and the volatility of the nominal excess return on stocks. J.
Financ. 48 (5), 1779–1801.
Grammatikos, T., Vermeulen, R., 2012. Transmission of the financial and sovereign debt crises to the EMU: stock prices, CDS spreads and exchange rates. J.
Int. Money Financ. 31, 517–533.
Grimaldi, M.B., 2010. Detecting and Interpreting Financial Stress in Euro Area. ECB Working Paper Series, No. 1214
Hollo, D., Kremer, M., Lo Duca, M., 2012. CISS – A composite indicator of systemic stress in the financial system. ECB Working Paper Series, No. 1426.
Hryckiewicz, A., 2014. What do we know about the impact of government interventions in the banking sector? An assessment of various bailout programs
on bank behavior. J. Banking Financ. 46, 246–265.
Kim, B.-H., Kim, H., Lee, B.-S., 2015. Spillover effects of the US financial crisis on financial markets in emerging Asian countries. Int. Rev. Econ. Financ. 39,
192–210.
Kizys, R., Paltalidis, N., Vergos, K., 2016. The quest for banking stability in the euro area: the role of government interventions. J. Int. Financ. Markets Inst.
Money 40, 111–133.
Kohonen, A., 2014. Transmission of government default risk in the Eurozone. J. Int. Money Financ. 47, 71–85.
Kroner, K., Ng, V., 1998. Modelling asymmetric comovement of asset returns. Rev. Financ. Stat. 11, 817–844.
Lee, H.W., Xie, Y.A., Yau, J., 2011. The impact of sovereign risk on bond duration: evidence from Asian sovereign bond markets. Int. Rev. Econ. Financ. 20,
441–451.
Louzis, D.P., 2015. Measuring spillover effects in Euro area financial markets: a disaggregate approach. Empiric. Econ. 49 (4), 1367–1400.
Martens, M., Poon, S., 2001. Returns synchronization and daily correlation dynamics between international stock markets. J. Bank. Financ. 25, 1805–1827.
Metiu, N., 2012. Sovereign risk contagion in the Eurozone. Econ. Lett. 117, 35–38.
Nelson, D.B., 1991. Conditional heteroskedasticity in asset returns: a new approach. Econometrica 59, 347–370.
Savva, C.S., 2009. International stock markets interactions and conditional correlations. J. Int. Financ. Markets Inst. Money 19, 645–661.
Yiu, M.S., Ho, W.-Y., Choi, D.F., 2010. Dynamic correlation analysis of financial contagion in asian markets in global financial turmoil. Appl. Financ. Econ. 20,
345–354.

Journal of International Financial Markets, Institutions & Money

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of International Financial Markets, Institutions & Money

Uploaded by

Copyright:

Available Formats

J. Int. Financ. Markets Inst.

Money 52 (2018) 17–36

Contents lists available at ScienceDirect

Journal of International Financial

Volatility co-movements and spillover effects within the

2. Eurozone crisis and modelling of spillover effects

3. Data and methodology

3.1. Financial stress index construction

Variables Used in Financial Stress Indices

Equity Market Bond Market

Panel A : Banking FSIs Panel B : Money Market FSIs

Panel C : Equity FSIs Panel D : Bond FSIs

Panel E : Country FSIs

Fig. 1. Financial stress indices for Eurozone markets and countries.

3.2. Volatility transmission models

et jXt1  T Studentð0; Ht Þ ð5Þ

3.2.1. Market level

3.2.2. National level

3.2.3. Regional analysis

4.1. FSI descriptive statistics

4.2. Transmission channels across markets

4.3. Transmission channels across countries

4.4. Transmission channels across regions

5. Further evidence and robustness checks

5.1. Alternative financial stress indices

5.1.1. A PCA-based approach for FSI

5.1.2. European central bank’s CISS indices

5.2. Cross-market analysis

5.3. Alternative volatility specifications

5.5. Volatility surfaces

Appendix A. Supplementary material

You might also like

et jXt1 T Studentð0; Ht Þ ð5Þ