Banking Reform in the UK
stg1993
Measuringthesystemicimportanceofbanks.pdf
BANK OF GREECE EUROSYSTEM
Working Paper
Economic Research Department Spec ia l S tud ies D iv i s ion 21, E. Venizelos Avenue G R - 1 0 2 5 0 , A t h e n s
Tel.:+ 3 0 2 1 0 3 2 0 3 6 1 0 Fax:+ 3 0 2 1 0 3 2 0 2 4 3 2 w w w . b a n k o f g r e e c e . g r
BANK OF GREECE EUROSYSTEM
WORKINGPAPERWORKINGPAPERWORKINGPAPERWORKINGPAPERISSN: 1109-6691
Georgios Moratis Plutarchos Sakellaris
Measuring the systemic importance of banks
2 DECEMBER 2017WORKINGPAPERWORKINGPAPERWORKINGPAPERWORKINGPAPERWORKINGPAPER
40
BANK OF GREECE Economic Analysis and Research Department – Special Studies Division 21, Ε. Venizelos Avenue GR-102 50 Athens Τel: +30210-320 3610 Fax: +30210-320 2432 www.bankofgreece.gr Printed in Athens, Greece at the Bank of Greece Printing Works. All rights reserved. Reproduction for educational and non-commercial purposes is permitted provided that the source is acknowledged. ISSN 1109-6691
MEASURING THE SYSTEMIC IMPORTANCE OF BANKS
Georgios Moratis Athens University of Economics and Business
Plutarchos Sakellaris
Athens University of Economics and Business
Abstract
We measure the systemic importance of all banks that issue publicly traded CDS
contracts among the world’s biggest 150. Systemic importance is captured by the
intensity of spillovers of daily CDS movements. Our new empirical tool uses Bayesian
VAR to address the dimensionality problem and identifies banks that may trigger
instability in the global financial system. For the period January 2008 to June 2017,
we find the following: A bank’s systemic importance is not adequately captured by
its size. European banks have been the main source of global systemic risk with
strong interconnections to US banks. For the global system, we identify periods of
increased interconnections among banks, during which systemic and idiosyncratic
shocks are propagated more intensely via the network. Using principal components
analysis, we identify a single dominant factor associated with fluctuations in CDS
spreads. Individual banks’ exposure to this factor is related to their government’s
ability to support them and to their retail orientation but not to their size.
JEL classification: E30, E50, E58, E60, G15 Keywords: Macroprudential Policy, Systemic Risk, Financial Markets Acknowledgements: We would like to thank Heather Gibson, Tryphon Kollintzas, Dimitris Korobilis, Dimitris Malliaropoulos, Petros Mygiakis, Frank Smets and seminar participants at Bank of Greece, University of Crete, University of Birmingham and Athens University of Economics and Business for useful comments. This research was conducted when Georgios Moratis was visiting Bank of Greece on the Bank’s programme of cooperation with universities. Correspondence: Georgios Moratis Athens University of Economics and Business 76, Patission Str., 10434, Athens email: [email protected]
3
1. Introduction
Macroprudential policy entails bank supervision from a system-wide perspective,
rather than that of the individual institution. The objective is to limit the risk of
system-wide financial crisis as well as to contain the costs to the real economy, if a
crisis erupts. In order to ensure that each institution pays for the externality it
imposes on the global system, the measures applied to each bank should be
calibrated to the systemic importance of each bank. In this paper, we provide a
measure of systemic importance of all banks that issue publicly traded CDS contracts
among the world’s biggest 150 banks, for the period January 2008 to June 2017. We
capture systemic importance by the intensity of spillovers of daily CDS movements.
This measure captures institutional externalities such as “too big to fail”, or “too
correlated to fail.”
We obtain some strong and, in some respects, surprising results. A bank’s
systemic importance is not adequately captured by its size. In addition, a
considerable number of banks officially designated as GSIBs are not ranked high in
terms of our novel measure of systemic importance.1 Throughout the examined
period, European banks have been the main source of global systemic risk with
strong interconnections to US banks. Looking at the time dimension for global
systemic risk, we identify periods of increased interconnections among banks, during
which systemic and idiosyncratic shocks are propagated more intensely via the
network. In a complementary but separate approach, we use principal components
analysis in order to identify a single dominant factor associated with fluctuations in
banks’ CDS spreads. Individual banks’ exposure to this factor is related to their
government’s ability to support them and to their retail orientation but not to their
size. In particular, regarding the bank-sovereign nexus, our evidence is consistent
with fiscally strong sovereigns insulating their large banks from this dominant factor
throughout this period.
Our novel measure of bank systemic importance identifies separately the degree
of externalities originating in a bank from its vulnerability to the system.
1 See FSB (2013) for a description of the methodology for assessing the systemic importance of global systemically important banks (GSIBs) and the higher loss absorbency requirements imposed on them.
4
Externalities are captured by the degree to which a shock experienced by a bank is
propagated to each individual bank in the global bank system. Vulnerability is
captured by the shocks it receives from each bank in the global system. In particular,
we find that more systemically important banks display relatively higher externalities
than vulnerability to the global system. This decomposition better allows the
macroprudential supervisor to differentiate the “cure” according to the individual
bank’s systemic “disease”. The “cure” usually consists of a combination of capital
requirements, quantitative restrictions, and supervisory review actions. Arguably,
our decomposition facilitates an improved approach to safeguarding financial
stability.
Our methodology is based on two pillars. First, we use market information
incorporated in CDS spreads as a reduced-form measure of the linkages among
banks.2 CDS spreads are a better measure of credit risk than bond spreads, equity
returns or other market variables. Second, we use Bayesian VAR to confront the
high dimensionality of bank networks. Past work on this topic had to limit attention
to a subset of global banks because of the dimensionality problem.3 The closest to
our approach is Alter and Beyer (2013), which builds upon the framework of Diebold
and Yilmaz (2009, 2011). We deviate from common practice in the literature by
removing any market-wide shocks through the inclusion of a set of common external
systemic variables. Thus, we allow each bank to become a source of systemic risk
after idiosyncratic shocks through spillovers.
The remainder of this document is structured as follows. Section 2 presents the
existing literature and section 3 describes the process of measuring systemic risk, the
existing frameworks and the motivation. Section 4 presents the methodology and
the data, while section 5 presents the results and section 6 concludes.
2 These linkages may arise from correlated exposures, counterparty relationships or other structural channels. 3 There are two exceptions that address the dimensionality problem using LASSO methods applied to stock return data: Demirer et al. (2017) for the global bank system and Basu et al. (2016) for the U.S. financial system.
5
2. Relevant literature
Our paper is closely related to four literature strands. First, it is related to
studies concerning macroprudential policy. The aim of macroprudential policy is to
increase the resilience of individual financial institutions and of the financial system
as a whole, by limiting the build-up of vulnerabilities in order to mitigate systemic
risk (ECB, 2016). It is also used to smooth-out the financial cycle, driven by
fluctuations in credit, leverage and asset prices, which may otherwise result in a
pattern of boom and bust (Dell’Ariccia et al., 2013; Elliot et al., 2013; Cerutti et al.,
2015). Appropriate policies should be designed toward limiting the ex ante
externalities that lead to an excessive build-up of systemic risk, and the ex post
externalities that can generate inefficient failures of otherwise sound institutions in a
crisis. All in all, macroprudential policy is the usage of primarily prudential tools to
limit systemic risk (Crockett, 2000; FSB/IMF/BIS, 2011; IMF 2013).
The literature on macroprudential policy is growing at a fast pace but is still at
an early stage and historical experience is thin. The most relevant part of the
literature aims at assessing the systemic importance of GSIBs. The most important
framework is the one developed by Basel Committee on Banking Supervision (BCBS).
The framework compares each bank’s activity over twelve indicators and finally
assigns a score to each bank. The indicators include the size, the interconnectedness,
the substitutability, the complexity and the cross-jurisdictional activity of each bank.
The BCBS methodology has also been used by Financial Stability Board for the
identification of GSIBs. This methodology has been transposed in the EU regulatory
framework (see Article 131 of the Capital Requirements Directive IV (CRDIV)), which
defines global systemically important institutions or G-SIIs. The BCBS/FSB framework
for determining systemic risk has some deficiencies. It assigns primal importance to
size, as all bank characteristics considered are directly related to size. However, the
premise that the biggest banks are the most dangerous ones for financial stability is
not necessarily backed by empirical evidence. In addition, the weights assigned to
the characteristics are arbitrary. Finally, it does not provide any information on the
degree of externalities between a particular systemically important bank and any
other one in the system. Our contribution is to use direct observations on credit risk
6
to measure externalities between any two banks in the global system. In this way,
we quantify the degree of danger that any bank may pose to the financial system or
parts of it. Our methodology flexibly updates the classification dynamically as new
information is obtained.
The second relevant field of literature entails alternative systemic risk rankings
for financial institutions. There are many methodologies for calculating the exposure
of financial institutions to changes in current economic conditions, how
concentrated the risks are among the financial institutions and how closely linked
they are with each other. One grouping of methodologies employs price-based
systemic risk rankings such as banks’ VaR (Adams, Fuss, and Gropp, 2014; White,
Kim, and Manganelli, 2015), ∆CoVaR (Adrian and Brunnermeier, 2014; Castro and
Ferrari, 2014) and MES (Acharya, Pedersen, Philippon, and Richardson, 2010). These
measure the VaR or MES of financial institutions conditional on the entire set of
institutions performing poorly.
A second grouping of such metrics incorporates book values as well and includes
SRISK (Acharya, Engle, and Richardson, 2012; Brownlees and Engle, 2015), leverage
ratio (Fostel and Geanakoplos, 2008; IMF, 2009; Geanakoplos and Pedersen, 2014),
and CAPM beta times market capitalization (Benoit, Colliard, Hurlin, and Perignon,
2015). Finally, the distressed insurance premium (DIP) by Huang et al. (2011)
measures the insurance premium required to cover distressed losses in the banking
system.
These closely related approaches have a key weakness, which is that they do not
provide information on the pairwise directional connectedness. In other words, they
do not describe externalities between any two banks in the global system. In
response to this shortcoming, some papers (see Billio et al, 2012) use Granger
causality as a tool to uncover directionality. However, Granger causality is unable to
consider contemporaneous movements, control for exogenous variables, quantify
intensities of effects, or consider multi-dimensional networks. All these aspects are
enabled by our methodology and measure.
A third group of relevant papers deals with the estimation of high-dimensional
VAR models. Our approach is closely related to the approach developed by Alter and
7
Beyer (2013), which is based on the framework of Diebold and Yilmaz (2009, 2011).
The high-dimensionality problem had forced research on global bank connectedness
to limit analysis to small samples of banks. Needless to say, this is not appropriate
when considering bank importance for the global system. A relevant methodology
has been recently suggested by Demirer, Diebold, Liu and Yilmaz (2017), who use
LASSO methods to shrink, select and estimate the high-dimensional network linking
the publicly-traded subset of global banks. In a similar vein, Basu et al. (2017) use
Lasso penalized Vector Autoregressive model to estimate a model that leverages a
system-wide approach to identify systemically important financial institutions in the
U.S. Our distinct approach is to use Bayesian VAR in order to resolve the
dimensionality problem.
Finally, our paper relates to studies that apply principal components methods to
analyze systemic risk. Billio et al. (2010) suggested that an important symptom of
systemic risk is the presence of sudden regime shifts. Giglio et al. (2015) proposed
dimension-reduction estimators for constructing systemic risk indexes from the cross
section of measures and prove their consistency in a factor model setting. We differ
by examining individual bank loadings on the dominant factor associated with
fluctuations in bank CDS spreads and determining which observable characteristics
are related to these. This approach provides solid empirical basis for using relevant
characteristics as indicators to measure systemic importance indirectly.
3. Definition of systemic importance
Systemic risk may originate in an endogenous build-up of financial imbalances
possibly associated with a booming financial cycle; large aggregate shocks hitting the
economy or the financial system; or contagion effects across markets, intermediaries
or infrastructures. Our study focuses on contagion among banks and measures the
systemic importance of a bank by the amount of spillovers it receives from and sends
to the rest of the banking system. According to Allen et al. (2012) contagion refers to
the risk that the failure of one financial institution leads to the default of others
through a domino effect in the interbank market, the payment system, or through
asset prices. We adopt the “pure-contagion” (Gomez-Puig and Sosvilla Rivero, 2014)
8
definition by controlling only for external common factors through the inclusion of a
set of common external systemic risk factors, and assume that each bank could
become itself a source of systemic risk as a result of an idiosyncratic shock.
The following example illustrates how we measure the systemic importance of
banks (see Figure 1). Assume that there exist three banks. Focusing on bank A as the
source of shocks, figure 1 presents the potential impact of an idiosyncratic shock on
bank A to bank B and to bank C, separately. Bank A sends a ten-unit shock to B and a
seventeen-unit shock to C, a total of 27. Next, we focus on the shocks received by
bank A from the other banks in the system. Bank A receives a twenty-one-unit shock
from bank B and a five-unit shock from bank C, a total of 26. If we sum the shocks
that bank A sends to and receives from the system, we obtain an estimate of the
degree of connectedness for bank A. This is a valid measure of bank A’s systemic
importance. This procedure is repeated in order to calculate the systemic
importance of bank B and bank C.
[Figure 1 here]
[Table 1 here]
Table 1 presents the entire picture for all three banks in the system. Shocks
emanate from row banks to column banks. Each row shows the contagion effects of
an equal-sized impulse to the relevant bank in the first column. In the last column,
we aggregate the total externality effects of each row bank. The columns provide the
picture of vulnerability of each bank to shocks in different banks. The second to last
row is a measure of total vulnerability of a bank to all other banks in the system. It
contains the answer to the question: “If all other banks in the system experienced
simultaneously an idiosyncratic shock of 100 basis points, what would be the impact
on bank X?” In the bottom row, we aggregate the total externality effect and the
total vulnerability effect of each bank. In other words, we lump together shocks sent
and received by an individual bank as a measure of total individual bank
connectedness. In calculating a bank’s systemic importance, we assign equal
weights to shocks it sends as to shocks it receives, as we are agnostic as to whether
one source of systemic instability is more dangerous than the other.
9
There are two aspects of financial contagion due to a bank’s participation in a
banking system that are of relevance to regulators: externalities emanating from a
bank’s failure and individual bank vulnerability to financial contagion. Both
components are important for regulators but their importance may not be equal. If
they are of equal importance, then the regulator would consider the sum of these
two. However, the clear decomposition in Table 1, as well as in our econometric
method, allows the regulator to assign different weights in order to capture the
appropriate measure of systemic importance.
4. Data and methodology
4.1 Data
We study 77 banks from 19 developed and 7 emerging economies. Our selection
procedure is as follows. We started with the list of the world’s top 150 banks, in
terms of total assets in Q4:2016. Using bank names, we matched 77 banks to CDS
prices from Thomson-Reuters Datastream and Bloomberg. CDS spreads cover the
period from January 2008 to June 2017 and are at daily frequency. The sample
contains all banks that are designated as “global systemically important banks”
(“GSIB’s”) by the Basel Committee on Banking Supervision, except for three Chinese
banks (Agricultural Bank of China, Bank of China, and Industrial and Commercial
Bank of China) and one French bank (Group BPCE). Table 2a (in the Appendix)
classifies banks by assets and provides detail on the 77 banks in the sample, such as
home-country and total assets, while table 2b (in the Appendix) classifies banks by
home-country. We note that 40 out of the 77 banks (52%) in the sample are from
Europe while 28 of them (34%) are headquartered in Eurozone members. Tables 3a
and 3b (in the Appendix) provide the regional characteristics of the sample.
[Tables 2a, 2b, 3a, 3b here]
10
4.1.1 Systemic risk factor
We allow for the presence of a global systemic risk factor. This permits us to
interpret robustly the results obtained from our model. Longstaff et al. (2011), for
instance, has argued that credit risk appears related to global rather than country-
specific factors, while Aizenman et al. (2013) has established the importance of
international economic factors in the pricing of credit risk. The variables we chose to
employ in order to capture global financial risk conditions have been widely used in
related studies as control variables (see, among others, De Santis, 2012; Aizenman et
al., 2013; Ang and Longstaff, 2011). The global default risk conditions are
represented by: the CDX, which is the family of CDS indices covering North America,
the VIX volatility index which captures the global capital markets’ “fear” condition,
and the global liquidity conditions, which is represented by the US 3-month treasury
bills. The systemic factor is assumed to affect the endogenous variables
contemporaneously. Table 4 contains the variable definitions and Table 5 provides
descriptive statistics.
[Tables 4, 5 here]
4.1.2 Bank-specific characteristics
A variety of bank- and country-specific variables are used for identifying the
determinants of systemic risk fluctuation over time. The first bank-specific variable is
bank size expressed as each bank’s total assets (in log). According to BIS (2011a) the
larger a bank is, the more likely it is to receive a bailout package. In this sense, we
also take into consideration the “too-big-to-fail” (TBTF) issue (Acharya et al., 2013).
The second bank-specific variable is the loan-to-asset ratio, which provides
information on the bank’s retail orientation. Ayadi et al. (2011) and Köhler (2013)
suggest that retail-orientated banks appeared to be less risky than other banks
during the recent financial crisis. Also, according to Altunbas et al. (2011) the non-
interest income over total revenue is considered to be a measure of each bank’s
diversification, since the less a bank relies on interest income, the less exposed the
11
bank is to a negative shock. Finally, we include each bank’s nonperforming loans
over total loans (see Tables 4 and 5).
4.1.3 Country-specific characteristics
It is important to include country-specific factors, since the impact of
macroprudential policy might differ depending on the underlying economic
conditions of each bank’s home country. For example, the impact of shocks may be
different for economies that were under stress and hence rely more on rescue
packages and foreign financing (IMF 2015a). These economies would not have the
same ability to support effectively their banking systems in times of distress. We
investigate the role of sovereigns through a bank’s home-country GDP growth, the
primary surplus over GDP, and public debt over GDP.
4.1 Connectedness matrix
We estimate a VARX model with two lags (p=2) for the endogenous variables and
contemporaneous exogenous variables (q=0).
The vector of endogenous variables (y) consists of log differences of daily CDS
spreads for the 77 banks. By including the exogenous variables, we account for
common factors that affect at the same time all bank CDS spreads (Bekaert et al.,
2005).
4.2.1 Bayesian VAR
The suggested model has many more parameters than observations, and as a
consequence could perform poorly. Researchers working in the relevant literature
typically use prior shrinkage on the parameters to overcome such over-
parametrization concerns. Most flexible Bayesian priors that result in shrinkage of
high-dimensional parameter spaces rely on computationally intensive Markov Chain
Monte Carlo (MCMC) methods. Their application to recursive forecasting exercises
ttttt uXBYAYAaY 122110
12
can, as a consequence, be prohibitive or even infeasible. The only exception is the
traditional “Minnesota prior”, an empirical-Bayes prior which is due to Littermann
(1979) and co-authors (see, e.g. Doan, Litterman, and Sims, 1984) and still dominates
many applications of VAR models in economics.
The “Minnesota prior” is based on the natural conjugate prior, an idea that has
recently been exploited by Banbura, Giannone and Reichlin (2010) and Giannone,
Lenza and Primiceri (2012), among others. While this prior allows for an analytical
formula for the posterior, there is a cost in terms of flexibility in that a priori all VAR
equations are treated in the same manner; see Koop and Korobilis (2010) for a
further discussion of this aspect of the natural conjugate prior. For computational
simplicity we restrict the model to use conjugate prior is (whose posterior has the
same distributional family as the prior distribution). This restriction allows for
analytical calculation of the Bayesian VAR, rather than simulation-based estimation
(e.g. the MCMC method). It is also worth noting that the choice of priors does not
imply the need for different Bayesian techniques of estimation. Disagreement over
the priors may be addressed by post-estimation sensitivity analysis evaluating the
robustness of posterior quantities of interest to different prior specifications.
We estimate the coefficients of a VARX(2) for 77 banks using the log differences
of each bank’s CDS. As we explained above, a key concern of users of Bayesian
statistics, and criticism by critics, is the dependence of the posterior distribution on
one’s prior and for this reason we specify hyperparameters for the prior.
The Bayesian VARX(p,s) model can be written as:
𝑦𝑡 = 𝑎0 +∑𝐴𝑗𝑦𝑡−𝑗 +∑𝛩𝑖 ∗𝑥𝑡−𝑖 + 𝜀𝑡
𝑠
𝑖=0
𝑝
𝑗=1
where yt for t = 1,..., T is an M x 1 vector containing observations on M time series
variables, εt is an M x 1 vector of errors, α0 is an M x 1 vector of intercepts and Aj is
an M x M matrix of coefficients. We assume εt to be i.i.d. N (0, Σ). The prior means
for the exogenous coefficients are set to zero.
13
4.2.2 The connectedness matrix framework
The construction of the diagnostic tool is based on a medium-size Bayesian
vector autoregressive model with exogenous variables (Bayesian VARX) that
accounts for common global and regional trends, and is able to include even bank-
specific characteristics. Then, similar to the framework described by Diebold and
Yilmaz (2009, 2012) and the one described by Alter and Beyer (2013), we construct
the spillover matrix in order to capture any potential spillovers among banks. This
methodology relies on Generalized Forecast Error Variance Decomposition (GFEVD)
or on Generalized Impulse Response Functions (GIRF), obtained as shown in Pesaran
and Shin (1998). Therefore, we derive Generalized Impulse Response Functions as
functions of residuals together with the interdependent coefficients. According to
Alter and Beyer (2012), it is of low importance which methodology we select, since
they produce qualitatively similar results.
[Table 6 here]
In table 6, each row variable is an origin of unexpected shock. Column variables
are the respondents that receive the contagion effects. The bottom right cell
represents the total systemic risk index that is defined as the average response per
bank in the connectedness matrix and is calculated as the sum of all non-diagonal
cells divided by the total number of entities. The expression of total systemic risk as
an index makes the overall risk independent of the number of banks in the sample,
making comparison between different samples more precise. The potential
contagion effects from and to each bank are aggregated on each line and column
and represent measures of externalities (To Others) and vulnerability (From Others).
The main diagonal values represent the effect of a variable’s shock on itself, and they
are excluded from any calculations. The possible contagion effects answer the
question “How would bank B evolve in the following period if bank A CDS increased
by one unit shock?”
We use accumulated Impulse Response functions over a 10-step horizon (10-
days). Not all the banks respond to the shocks within the same period but the
majority of the shocks are absorbed within 10 days. Nevertheless, the framework is
flexible and it easily adapts to the needs of the study.
14
5. Empirical results
5.1 Individual bank connectedness
We estimate the connectedness matrix as described in section 4.2.2 for the
whole sample period, 1 January 2008 to 31 June 2017, and estimate individual bank
connectedness. A striking result in Table 7 is that systemic importance of a bank
cannot be adequately captured by its size. According to table 7, the bank that
creates the most systemic risk in the system is Intesa Sanpaolo, a medium-sized
European bank that is ranked 27th in terms of total assets.
[Table 7 here]
Intesa Sanpaolo has total contagion effects of 1.056, which may be further
decomposed into a 0.505 vulnerability score and a 0.551 externalities score. Intesa
Sanpaolo’s externalities score of 0.551, implies that a one-unit shock to Intesa
Sanpaolo’s CDS spreads will have an impact of 55.1% to the system. The 0.505
vulnerability score means that simultaneous one-unit shocks to all other banks in the
system will affect Intesa Sanpaolo by 50.5%. Among the top 20 most connected
banks, we find other smaller banks like BBVA (3rd), Credit Lyonnais (10th), Banca
Monte dei Paschi (16th) and Mediobanca (20th). The largest bank in the sample (Bank
of China) is ranked 50th in terms of systemic importance.
The existing literature suggests that during crises, large banks behave differently
than small or medium-sized banks (e.g. Laeven et al., 2014). This phenomenon could
be partially attributed to some common characteristics shared by large banks that
are associated with higher levels of risk, namely increased portion of market-based
activities, reduced capital adequacy, less stable funding and higher organizational
complexity. This literature does not identify a threshold of bank size above which
systemic importance kicks in. A major contribution of our paper is to categorize large
banks directly for the first time on the basis of their systemic importance.
Table 8 confirms that bank size, while relevant, is by no means adequate in
describing a bank’s systemic importance. When banks are ranked by our measure of
systemic importance, the first quartile comprises 34% of total assets in the system.
15
However, when the sample is ranked by bank assets, the first quartile comprises 62%
of total assets in the system. It is also indicative that only half of GSIBS are classified
in the first quartile when ranked by our measure of systemic importance. Clearly,
there are bank characteristics other than size that influence bank systemic
importance.
The existing frameworks used by regulators and policy makers, such as the
BCBS/FSB one, rely heavily on size, either through size-related indicators or size per
se, to determine systemic importance. Our results suggest that this may lead to
inefficient policies.
[Table 8 here]
Next, we calculate the systemic contribution of each bank in the system as the
ratio between the total individual contagion effects and the total contagion in the
system:
100 TSR
Score onContributi i
i
y
y
Contribution is the individual systemic contribution, Score is the total individual
contagion effects (externalities and vulnerability) and TSR is the total systemic risk in
the system. The bank with the highest ranking, Intesa Sanpaolo, contributes 2.25% of
the total systemic risk, while the bank at the bottom of the list, Turkiye is bankasi,
has almost zero contribution (see Table 9). This transformation of the initial table
allows one to compare rankings among samples that contain a different number of
banks.
[Table 9 here]
We define each bank’s directional connectedness as the ratio between the
individual externalities and the total individual contagion (score):
100
i
i
i
y
y
y Score
iesExternalit lityDirectiona , or
16
When directionality is larger than 50%, the bank’s systemic importance is
externalities-driven. The separation of banks according to the directionality of shocks
allows the supervisor to treat them differently in the context of financial stability.
This is important, as externalities and vulnerabilities may have different
determinants, different impact on financial stability and may require different
macroprudential regulation measures. The results in Table 7 show that the larger
banks in the sample emit to the system larger shocks than those they receive, being
more extrovert than smaller banks..
Table 10 shows that banks with higher systemic importance tend to have
higher externalities ratios. In particular, banks in the first quartile have average
individual directionality of 54%, whereas banks in the last quartile average 30%.
[Table 10 here]
5.2 Regional network connectedness
A stark result in Table 7 is that all banks ranked in the first quartile in terms of
systemic importance are European. Table 11a displays the geographic distribution of
all quartiles. There seems to be strong regional clustering in the different quartiles.
For example, the first quartile of systemic importance comprises almost half the
European banks in the sample (see Table 11b). It is also noteworthy that 75% of the
banks ranked in the first quartile in terms of systemic importance are headquartered
in the Eurozone.
[Table 11a here]
[Table 11ab here]
Our methodology provides estimates of CDS spillovers between any pair of
banks in the system. This allows us to explore further regional effects unveiled by
our estimates. We now investigate the flows of contagion between different
regions. This will provide a deeper understanding of the degree of connectedness
among the different regions (see Table 12).
17
We focus on four regions, Europe, North America, Asia and Africa. Table 12a
presents the externalities of region X to each other region as a percentage of region
X’s total externalities. The largest portion of shocks generated in the European
banking system remains within the region. Of course, European banks are not
immune to the banks outside the European Union, but throughout the examined
period they have been more vulnerable to shocks that emanated from within. This is
consistent with the eurozone crisis being of primal importance. European banks
seem to be also the favorite target of shocks emanating from all other world regions.
It seems that over the period Jan. 2008 – Jun. 2017 European banks absorbed the
majority of shocks that were being transmitted in the global banking system, being
by far the most vulnerable banking block. Table 12b displays evidence that 64% of
the aggregate shocks that are received by U.S. banks are generated in Europe. Thus,
we conclude that the European and U.S. banking sectors have been strongly
interconnected and severely exposed to each other.
[Table 12 here]
The highly interconnected banking system, the feedback loop among sovereigns
and banks, and the high transmission of contagion effects from one country to the
other, put the European banking system in the eye of the storm throughout the
examined period.
5.3 National banking system importance
We turn to examining the systemic importance at the national bank level. We
calculate the average bank systemic risk for each country in the sample (see Figure
2). The eight leading countries are all in Europe, amplifying the evidence that the
majority of turmoil during 2008 – 2017 stemmed from the interior of the European
Union and, in particular, from the eurozone.
[Figure 2 here]
French banks suffered from their exposure to the sovereign debt of peripheral
euro-area countries and the withdrawal of funds by U.S. money market mutual. Italy
and Spain, both of which suffered from banking systems in distress, contributed
18
heavily as well. This verifies the concerns regarding the systemic importance of
relatively smaller banks and their potential effects on the global financial stability.
Banks in non-eurozone countries, i.e. Swiss and UK banks are also of great systemic
importance. The average systemic importance of German banks places them almost
in the middle of the league. Finally, both the Portuguese and especially the Greek
banking system appeared to be isolated from the global banking system, while
shocks emanating from these banks remained largely within the boundaries of their
national banking system.
5.3 Rolling window analysis
5.3.1 Individual Banks
In order to better understand the evolution of systemic risk and how this
fluctuated over the whole period we apply rolling-window Bayesian VAR analysis.
The length of the window is 340 days and the step is 150 days. Figure 3 presents the
evolution of bank systemic importance over the examined period indicating strong
co-movement and interconnections among banks.
There are periods where the cluster of systemic importance lines shrinks in spread.
These are periods when systemic risk becomes more homogeneous across banks.
[Figure 3 here]
We introduce a new metric, which we call systemic risk range, defined as the
difference between the highest and the lowest systemic importance score in the
system (see Figure 4). A lower range signifies more homogeneous systemic
importance scores in the global banking system. It is interesting to note that this
metric attains its lowest value in the period just after the collapse of Lehman
Brothers.
[Figure 4 here]
19
5.3.2 Global banking system
We move on to analyzing aggregate contagion effects. Total Systemic risk (TSR)
is defined as the total sum of the off-diagonal entries in the connectedness matrix,
or as the sum of the “from” column or “to” row measures total connectedness.
N
ij ji
H
ij
H IRTSR 1,
A slightly different but more interpretable and inclusive way of presenting total
systemic risk is the Total Systemic Risk index (TSRI) which is defined as the average
response per bank in the connectedness matrix and is calculated as the sum of all
non-diagonal cells divided by the total number of entities:
N
i j yy ji IR
N TSRI
1 1
1
Since cumulative IRs lie in the interval [0,1], the index will be bound between 0 and
100. A higher contagion index implies a tightening of the nexus among banks (see
Figure 5).
[Figure 5 here]
Total connectedness reached its peak after the collapse of Lehman Brothers.
Another important period was associated with the developments in the European
banking and sovereign debt markets that shocked some EU member countries until
mid-2012. The Greek crisis and then, in early 2011, the inclusion of Italy and Spain to
the countries with stressed banking systems pushed total systemic connectedness
upwards. After early 2012, the actions taken by the ECB contained system-wide
contagion. However, after mid 2015 market concerns about euro area banks’
longer-term profitability prospects as well as higher political uncertainty regarding
UK membership in the EU and the US election, contributed to the sharp increase in
20
the index. Another major concern for global markets was the crisis in Deutsche Bank
as the bank had deep connections to global financial institutions. However, since
early 2017 total systemic connectedness has been decreasing substantially.
5.3.3 Rolling window – regional systemic contribution
In this subsection, we concentrate on the European and U.S. banking systems.
We track the evolution of the systemic importance of the two regions calculated as
the ratio of aggregated systemic importance scores of banks in a region over those in
the complete sample. Figure 6 compares the systemic contribution of these regions.
The share of systemic risk due to European banks increased dramatically after 2014,
while the contribution of U.S. banks started dropping after late 2011 and has
remained lower since. The difference between the systemic contribution of
European and US banks fluctuates between 40% and 55%. The fact that the share of
a region’s systemic contribution increases does not necessarily mean that total
systemic risk increases as well. The sharp increase in European banking sector’s
systemic importance starting in early 2014 is probably related to the persistent
accumulation of non-performing loans and contagion effects from the Deutsche
bank crisis.
[Figure 6 here]
5.4 Principal components analysis
So far in this paper, we used an impulse response methodology to study how one
bank’s risk contaminates another’s. In this section, we depart from these novel
measures of systemic importance for banks. Instead, we study contemporaneous
comovements in the CDS spreads of banks. We identify factors that make banks
move together and observable characteristics that are related to them. We use
principal components analysis (PCA), in which movements banks’ CDS spreads are
decomposed into movements of orthogonal factors of decreasing explanatory power
(see Muirhead, 1982 for an exposition of PCA).
21
PCA produces the decomposition of the variance-covariance matrix of CDS
spreads of the 77 banks contained in the sample into the orthogonal matrix of
loadings (eigenvector of the correlation matrix of CDS spreads) and the diagonal
matrix of eigenvalues. We focus on the first three eigenvalues as they explain most
of the variation in the system. When bank CDS spreads co-move more intensely
together, these three eigenvalues should explain a larger portion of the total
volatility in the system. Therefore, periods, when the first three factors account for
larger share of total volatility suggest increased interconnection among banks.
We apply rolling window PCA, where the length of the window is 200 days and
the step is 100 days, over the period January 2008 to December 2016 (see Figure 7).
The first component accounts for between 30% and 40% of total variance in the
observed variables while component 2 and 3 seem to have limited explanatory
power and limited variation in that. Consequently, we focus our analysis on the first
component. The identification of a single dominant component that determines the
fluctuations of CDS spreads is in line with the evidence in Fontana and Scheicher
(2010). According to Billio et al. (2012), during periods of distress, fewer
components explain a larger part of the volatility in CDS spreads, thus revealing
increased systemic risk.
[Figure 7 here]
We now ask the question: Which observable bank and country characteristics are
related to the exposure of a bank’s CDS movements to the dominant common
factor? We run the following cross-sectional regression each year over the period
2008-2016:
ii
iiiiiii
Debt
SurplusGDPNIINPLsLoansAssetsay
7
654321
yi is bank i’s loading on the first component.
22
[Table 9 here]
The observable bank-specific characteristics we explore are: total value of bank
assets (in logs), its retail orientation (total loans/total assets in levels), NPLs/ total
loans, and non-interest income / total revenue. We also relate the following home-
country characteristics: GDP growth, primary surplus/GDP, public debt/GDP). The
sub-sample used in this regression contains 47 banks that have been matched to
bank-specific characteristics.
Table 13 contains the results. Throughout the examined period, retail
orientation affects strongly and negatively the banks’ loadings in the first
component. This indicates that retail-orientated banks have been less exposed to
secular movements in the dominant common factor over the decade examined in
this paper. Surprisingly, non-performing loans, diversification as well as bank size
seem to have an insignificant effect on bank exposure to systemic movements.
Regarding country-specific characteristics, a strong result is that the home country’s
GDP affects strongly and negatively the exposure of banks to systemic risk. This is
evidence that, in cases of a distressed sovereign, home banks are less exposed to the
common factor. The opposite holds where the sovereign is strongly solvent. In
addition, Debt/GDP is negatively and significantly associated with the exposure to
systemic risk and it seems that banks with weaker sovereigns have less reliance on
the interbank market and are, thus, less affected by global shocks.
[Table 13 here]
6 Conclusions
Macroprudential policy is still in its infancy. Much work is still needed on
developing good and timely analysis, effective policy instrument tools, and effective
implementation. Our paper makes a contribution on the dimension of analysis and
measurement. The key aim of macroprudential policy is to address externalities and
spillovers among financial institutions in an effort to safeguard financial stability.
These interactions are complex. We provide a tool that clarifies and quantifies
these interactions expressing them in a single score. Our measure can guide
appropriate macroprudential policies that aim to internalize these externalities. A
23
key conclusion from our study is that a focus on size does not adequately address
the systemic importance of banks.
24
References
Acharya, V., Engle, R., Richardson, M., 2012. Capital Shortfall: A New Approach to
Ranking and Regulating Systemic Risks. American Economic Review Vol.102, No. 3.
Acharya, V.,, Drechsler, I., Schnabl, P., 2014. A Pyrrhic Victory? Bank Bailouts and
Sovereign Credit Risk Journal of Finance, 2014, vol. 69, issue 6, 2689-2739.
Acharya, V., Pedersen, L., Philippon, T., Richardson, M., 2010. Measuring systemic
risk. Technical Report. Department of Finance, NYU.
Adams, Z., Fuss, R., Gropp, R., 2014. Spillover Effects among Financial Institutions: A
State-Dependent Sensitivity Value-at-Risk (SDSVaR) Approach. Journal of Financial
and Quantitative Analysis (JFQA), Vol. 49, No. 3.
Adrian, T., Brunnermeier, M., 2009. CoVaR. Manuscript, Princeton University.
Aizenman, J., Hutchison, H., and Jinjarak,Y., 2013. What is the Risk of European
Sovereign Debt Defaults? Fiscal Space, CDS Spreads and Market Pricing of Risk.
Journal of International Money and Finance, 34, 37-59.
Allen, F., A. Babus, and E. Carletti (2012). Asset commonality, debt maturity and
systemic risk. Journal of Financial Economics 104(3), 519–534.
Alter, A., Beyer, A., 2013. The dynamics of spillover effects during the European
sovereign debt turmoil. Journal of Banking and Finance, 42, 134-153.
Altunbas, Y., Manganelli, S., Marques-Ibanez, D., 2011. Bank risk during the financial
crisis. Do business models matter? ECB Working Paper, No 1394.
Ang, A., Longstaff, F., 2011. Systemic sovereign credit risk: Lessons from the US and
Europe. NBER discussion paper 16983.
Ayadi, R., Arbak, E., De Groen, W., 2011. Business Models in European Banking: A
pre-and post-crisis screening. CEPS Paperbacks.
Banbura, M., Giannone, D., Reichlin, L., 2011. Nowcasting. Oxford Handbook on
Economic Forecasting, ed. by M. P. Clements, and D. F. Hendry, pp. 63–90. Oxford
University Press.
25
Bank for Internaional Settlements, 2010. Guidance for national authorities operating
the countercyclical capital buffer.
Bank for Internaional Settlements, 2011. Global systemically important banks:
Assessment methodology and the additional loss absorbency requirement.
Basel Committee on Banking Supervision, 2010. Basel III: A global regulatory
framework for more resilient banks and banking systems. Bank for International
Settlements.
Basu, S., Das, S., Michailidis, G., Purnanandam, A., 2017. A system-wide approach to
measure connectivity in the financial sector. Unpublished working paper.
Benoit, S., Colliard, J.-E., Hurlin, C., Perignon, C., 2015. Where the risks lie: A survey
on systemic risk. Unpublished working paper.
Billio, M., Getmansky, M., Lo, A. W., Pelizzon, L., 2012. Econometric measures of
systemic risk in the finance and insurance sectors. Journal of Financial Economics
104, 535–559.
Brownlees, C.T., Engle, R.F., 2010. Volatility, correlation and tails for systemic risk
measurement. Manuscript, Stern School of Business, New York University.
Canova, F. and Pappa, P., 2007. Price Differentials in Monetary Unions: The Role of
Fiscal Shocks. Economic Journal, 117, 2007, 713-737.
Castro, C., Ferrari, S., 2014. Measuring and testing for the systemically important
financial institutions. Journal of Empirical Finance, 2014, vol. 25, issue C, 1-14.
Cerutti, E., Claessens, S., Laeven, L., 2015. The Use and Effectiveness of
Macroprudential Policies: New Evidence. IMF Working Paper No. 15/61.
Crockett, A., 2000. Marrying the Micro- and Macro-prudential Dimensions of
Financial Stability. Eleventh International Conference of Banking Supervisors, held in
Basel, 20-21.
De Santis, R. (2012). The euro area sovereign debt crisis: Safe haven, credit rating
agencies and the spread of the fever from Greece, Ireland and Portugal. ECB Working
Paper 1419.
26
Dell’Ariccia, G., Igan, D., Laeven, L., Tong, H., Bakker, B., Vandenbussche, J., 2012.
Policies for Macrofinancial Stability: How to Deal with Credit Booms. IMF Staff
Discussion Note 12/06.
Demirer, M., Diebold, F., Liu, L., Yilmaz, K., 2017. Estimating global bank network
connectedness. Journal of Applied Econometrics.
Diebold, F., X., Yilmaz, K., 2009. Measuring Financial Asset Return and Volatility
Spillovers, with Application to Global Equity Markets. The Economic Journal
119(534), 158-171.
Diebold, F. X., Yilmaz, K., 2012. Better to give than to receive: Predictive directional
measurement of volatility spillovers. International Journal of Forecasting, Issue 28,
pp. 57-66.
Doan, T., Litterman, R., Sims, C.A., 1984. Forecasting and Conditional Projections
Using Realistic Prior Distributions. Econometric Reviews, 3(4), 1– 100.
Elliott, D. J., Feldberg, G., Lehnert, A., 2013. The History of Cyclical Macroprudential
Policy in the United States. Office of Financial Research Working Paper No. 8.
European Central Bank, 2016. Macroprudential Bulletin Issue 1/2016.
Financial Stability Board, 2011. Measures to address systemically important financial
institutions.
Fontana, A., Scheicher, M., 2010. An analyisis of euro area sovereign CDS and their
relation with government bonds. ECB Working Paper Series, Issue 1271.
Fostel, A., Geanakoplos, J. 2008. Leverage Cycles and the Anxious Economy.
American Economic Review, 98 (4), 1211–1244.
Geanakoplos, J., Pedersen, L. H., 2014. Monitoring leverage. In: Brunnermeier, M. K.,
Krishnamurthy, A. (Eds.), Risk Topography: Systemic Risk and Macro Modeling. NBER,
pp. 175–182.
Giannone, D., Lenza, M., Primiceri, G.E., 2012. Prior Selection for Vector
Autoregressions. CEPR Discussion Papers 8755, C.E.P.R. Discussion Papers.
27
Giglio, S., Kelly, B. T., Pruitt, S., 2015. Systemic risk and the macroeconomy: An
empirical evaluation. Journal of Financial Economics, 2016, vol. 119, issue 3, 457-
471.
Gómez-Puig, M., Sosvilla-Rivero, S. 2013. EMU sovereign debt markets crisis:
Fundamental-based or pure contagion? Institut de Recerca en Economia Aplicada
(IREA) Working Papers. Universitat de Barcelona. 2014/02.
Huang, X., Zhou, H., Zhu, H., 2012. Systemic Risk Contributions. Journal of Financial
Services Research, 2012, vol. 42, issue 1, 55-83.
International Monetary Fund, 2013a. The Interaction of Monetary and
Macroprudential Policies. Washington: International Monetary Fund.
Köhler, M. 2013. Does non-interest income make banks more risky? Retail- versus
investment oriented banks. Deutsche Bundesbank, Discussion Paper, No.17/2013.
Koop, G., Korobilis, D., 2010. Bayesian Multivariate Time Series Methods for
Empirical Macroeconomics. Foundations and Trends in Econometrics: Vol. 3: No. 4,
pp 267-358
Litterman, Robert B., 1979. Techniques of forecasting using vector autoregressions.
Research Department Working Paper 115. Federal Reserve Bank of Minneapolis,
Minnesota.
Longstaff, F., Pan, J., Pedersen, L., Singleton, K., 201. How Sovereign Is Sovereign
Credit Risk? American Economic Journal: Macroeconomics, 2011, vol. 3, issue 2, 75-
103.
Muirhead, R.J., 1982. Aspects of Multivariate Statistical Theory. Wiley, New York.
Pesaran, H., Shin, Y., 1998. Generalized impulse response analysis in linear
mulitvariate models. Economic Letters, 58(1), 17-29.
White, H., Kim, T.H., Manganelli, S., 2008. Modeling autoregressive conditional
skewness and kurtosis with multi-quantile CAViaR. In: Russell, J., Watson, M. (Eds.),
Volatility and Time Series Econometrics: A Festschrift in Honor of Robert F. Engle.
28
APPENDIX
Table 1: Directional connectedness matrix
Shock/Response Bank A Bank B Bank C To Others
(Externalities)
Bank A - 10 17 27
Bank B 21 - 28 49
Bank C 5 19 - 24
From Others (Vulnerability) 26 29 45 100
Score 53 78 69
Note: Variables in the first column are the impulse origin, while variables on the top row are the respondents to the shock.
29
Table 2a: Banks ordered by Total Assets (in US $ billion)
Rank Bank Name Country Total assets,
US$B
1 Bank of China China 2,613 2 Mitsubishi UFJ Financial Group Japan 2,597 3 JPMorgan Chase & Co USA 2,490 4 HSBC Holdings UK 2,374 5 BNP Paribas France 2,196
6 Bank of America USA 2,187 7 Wells Fargo USA 1,930 8 China Development Bank China 1,904 9 Credit Agricole Group France 1,821 10 Citigroup USA 1,792
11 Mizuho Financial Group Japan 1,757 12 Deutsche Bank Germany 1,682 13 Sumitomo Mitsui Financial Group Japan 1,654 14 Barclays PLC UK 1,490 15 Societe Generale France 1,461
16 Banco Santander Spain 1,416 17 Lloyds Banking Group UK 1,004 18 Norinchukin Bank Japan 984 19 Royal Bank of Scotland Group UK 981 20 UBS Group AG Switzerland 919
21 Unicredit S.p.A. Italy 908 22 ING Groep NV Netherlands 893 23 Goldman Sachs Group USA 860 24 Morgan Stanley USA 814 25 Credit Suisse Group Switzerland 806
26 BBVA Spain 773 27 Intesa Sanpaolo Italy 766 28 Commonwealth Bank of Australia Australia 703 29 Rabobank Group Netherlands 700 30 Australia & New Zealand Banking Group Australia 661
31 Nordea Sweden 651 32 Standard Chartered Plc UK 646 33 Westpac Banking Corp Australia 607 34 National Australia Bank Australia 562 35 Commerzbank Germany 549
36 Danske Denmark 495 37 State bank of India India 492 38 U.S. Bancorp USA 445 39 The Export-Import Bank of China China 427 40 Sberbank of Russia Russia 420
41 Resona Japan 412 42 Sumitomo Mitsui T.H. Japan 406 43 Nomura Holdings Japan 370 44 PNC Financial Services USA 366 45 Capital One Financial Corporation USA 357
46 DBS Group Holdings Singapore 332 47 Shinhan Financial Group South Korea 328 48 KBC Group NV Belgium 291 49 Svenska Handelsbanken Sweden 289 50 Skandinaviska Enskilda Banken Sweden 289
51 Hana Financial Group South Korea 288 52 Nationwide Building Society UK 276 53 Korea Development Bank South Korea 268
30
Table 2a – Continued from previous page
54 Woori Bank South Korea 257 55 Landesbank Baden-Wurttemberg Germany 257
56 Cathay Financial Holding Taiwan 252 57 Swedbank Sweden 237 58 United Overseas Bank (UOB) Singapore 235 59 Dexia Belgium 225 60 Banco Sabadell Spain 224
61 Bayerische Landesbank Germany 224 62 Erste Group Bank AG Austria 220 63 Banco Popular Espanol Spain 204 64 Industrial Bank of Korea South Korea 196 65 Bank of Ireland Ireland 182
66 Malayan Malaysia 161 67 Standard Bank Group South Africa 161 68 Banca Monte dei Paschi di Siena Italy 161 69 American Express USA 158 70 National Bank of Greece Greece 153 71 Macquarie USA 143
72 Credit Lyonnais France 120 73 Comercial Portuguese Portuguese 113 74 Banco Espirito Santo Portugal 112 75 Turkiye is bankasi Turkey 112 76 Mediobanca Italy 95
77 Landesbank Hessen Germany 92
31
Table 2b: Banks ordered by Country
Rank Bank Country Total assets,
US$B
1 Commonwealth Bank of Australia Australia 703 2 Australia & New Zealand Banking Group Australia 661 3 Westpac Banking Corp Australia 607 4 National Australia Bank Australia 562
5 Erste Group Bank AG Austria 220
6 KBC Group NV Belgium 291 7 Dexia Belgium 225
8 Bank of China China 2,613 9 China Development Bank China 1,904
10 The Export-Import Bank of China China 427
11 Danske Denmark 495
12 BNP Paribas France 2,196 13 Credit Agricole Group France 1,821 14 Societe Generale France 1,461 15 Credit Lyonnais France 120
16 Deutsche Bank Germany 1,682 17 Commerzbank Germany 549 18 Landesbank Baden-Wurttemberg Germany 257 19 Bayerische Landesbank Germany 224 20 Landesbank Hessen Germany 92
21 National Bank of Greece Greece 153
22 State bank of India India 492
23 Bank of Ireland Ireland 182
24 Unicredit S.p.A. Italy 908 25 Intesa Sanpaolo Italy 766 26 Banca Monte dei Paschi di Siena Italy 161 27 Mediobanca Italy 95
28 Mitsubishi UFJ Financial Group Japan 2,597 29 Mizuho Financial Group Japan 1,757 30 Sumitomo Mitsui Financial Group Japan 1,654 31 Norinchukin Bank Japan 984 32 Resona Japan 412 33 Sumitomo Mitsui T.H. Japan 406 34 Nomura Holdings Japan 370 35 Yamaguchi Financial Group Japan 93
36 ING Groep NV Netherlands 893 37 Rabobank Group Netherlands 700
38 Banco Espirito Santo Portugal 111
39 Sberbank of Russia Russia 420
40 DBS Group Holdings Singapore 332 41 United Overseas Bank (UOB) Singapore 235 42 Standard Bank Group South Africa 161
43 Shinhan Financial Group South Korea 328 44 Hana Financial Group South Korea 288 45 Korea Development Bank South Korea 268 46 Woori Bank South Korea 257 47 Industrial Bank of Korea South Korea 196
48 Banco Santander Spain 1,416 49 BBVA Spain 773 50 Banco Sabadell Spain 224 51 Banco Popular Espanol Spain 204
32
Table 2b – Continued from previous page
53 Nordea Sweden 651 54 Svenska Handelsbanken Sweden 289 55 Skandinaviska Enskilda Banken Sweden 289 56 Swedbank Sweden 237
57 UBS Group AG Switzerland 919 58 Credit Suisse Group Switzerland 806
59 Cathay Financial Holding Taiwan 252
60 Turkiye is bankasi Turkey 112
61 HSBC Holdings UK 2,374 62 Barclays PLC UK 1,490 63 Lloyds Banking Group UK 1,004 64 Royal Bank of Scotland Group UK 981 65 Standard Chartered Plc UK 646 66 Nationwide Building Society UK 276
68 JPMorgan Chase & Co USA 2,490 69 Bank of America USA 2,187 70 Wells Fargo USA 1,930 71 Citigroup USA 1,792 72 Goldman Sachs Group USA 860 73 Morgan Stanley USA 814 74 U.S. Bancorp USA 445 75 PNC Financial Services USA 366 76 Capital One Financial Corporation USA 357 77 American Express USA 158 78 Macquarie USA 143
33
Table 3a: Banks’ home-countries ordered by the sum of total bank assets
Developed Total Assets Developing Total Assets
USA 11547 China 4944 Japan 7867 India 492 UK 6936 Russia 420 France 5599 Taiwan 252 Germany 2994 South Africa 161 Spain 2617 Turkey 112
Australia 2534 Italy 1933
Switzerland 1725
Netherlands 1584 Sweden 1467 South Korea 1340 Singapore 568 Belgium 516 Denmark 495 Ireland 344 Austria 220 Greece 153 Portugal 111 Total Assets of banks that are headquartered in:
Developed 50553.66 Emerging 6382.3 % of total assets 88,8% % of total assets 11,2%
Table 3b: Regional details
Region Number of Banks Total bank
assets % of total
assets
Europe 40 26696,73 47,3
Asia 21 15576,58 27,5
N. America 11 11547,24 20,4
Oceania 4 2534 4,5
Africa 1 161 0,3
34
B. Data Definitions and Descriptive Statistics
Table 4: Data Definitions
Variable Description
Endogenous CDS CDS 5-year spread
Exogenous Systemic risk
CDX The family of CDS indeces covering North America
VIX The volatiliy index of S&P 500
US 3-month T Bill The short-term obligation backed by the Treasury Dept. of the U.S. goverment
Bank-specific Size Total assets
Retail orientation Total Loans / Total assets
Diversification Non-interest income / Total revenues
NPLs Non-performing loans / Total Loans
Country-specific
GDP Each bank’s home-country GDP growth
Budget Balance Current Account/GDP
Public Debt Public Debt/GDP
35
Table 5: Descriptive statistics
Panel A: Systemic risk factor
CDX VIX TED
Mean 2.69E-06 -0.000123 -0.00121 Median 0.000 -0.001 0.000 Maximum 0.020 0.176 0.250 Minimum -0.009 -0.152 -0.750
Std. Dev. 0.001 0.031 0.033 Skewness 4.689 0.689 -17.326
Kurtosis 106.047 6.789 377.496 Jarque-Bera 1105896 1679.364 14610377 Probability 0.000 0.000 0.000
Sum 0.007 -0.304 -3.000 Sum Sq. Dev. 0.003 2.427 2.746
Note: CDX and VIX are in log differences. TED spread is in first differences.
Panel B: Bank specific
Assets Loans_to_Assets Non_Interest_Inc. NPLs
Mean 732.382 55.596 24.42 4.601
Median 458009 59.921 23.78 2.411
Maximum 3030645 86.64 86.40 35.217
Minimum 43543,87 9.070 -59.62 0.1082
Std. Dev. 746001.3 17.017 14.81 5.612
Skewness 1.081 -0.641 -0.694 2.376
Kurtosis 30.129 2.584 6.193 9.630
Jarque-Bera 7.023 272.888 1821.17 9995.01
Probability 0.000 0.000 0.000 0.000
Sum 2.64E+09 200370.2 88037.46 16584.30 Sum Sq. Dev. 2.01E+15 1043392. 790803.6 113486.7
Note: Data are in levels
36
Table 6: Contagion/connectedness matrix
Note: Variables in the first column are the impulse origin, while variables on the top row are the respondents to the shock. The impact is bound between 0 and 1. A value of 0.3 means that the response variable would be impacted in the same direction with an intensity of 30% the initial unexpected shock in the impulse variable. The last column presents the aggregated impact sent (To Others) by each row variable and on the bottom row the aggregated spillover received (From Others) by each column variable.
37
Part B – Global Sample
Table 7: Individual systemic importance
Panel A – Ranked by total score Rank by score
Rank by bank assets Bank Name
Home- Country Region
Assets (billion
US $) Score
From others (Aggr.)
To others (Aggr.)
1 27 Intesa Sanpaolo Italy Europe 766 1.056 0.505 0.551
2 16 Banco Santander Spain Europe 1416 1.038 0.464 0.574
3 26 BBVA Spain Europe 773 0.987 0.454 0.533
4 5 BNP Paribas France Europe 2196 0.982 0.445 0.537
5 21 Unicredit S.p.A. Italy Europe 908 0.975 0.489 0.485
6 14 Barclays PLC UK Europe 1490 0.961 0.454 0.507
7 12 Deutsche Bank Germany Europe 1682 0.954 0.428 0.526
8 9 Credit Agricole Group France Europe 1821 0.942 0.424 0.518
9 15 Societe Generale France Europe 1461 0.940 0.427 0.512
10 72 Credit Lyonnais France Europe 120 0.938 0.445 0.493
11 17 Lloyds Banking Group UK Europe 1004 0.925 0.438 0.487
12 25 Credit Suisse Group Switz. Europe 806 0.907 0.376 0.531
13 35 Commerzbank Germany Europe 549 0.904 0.410 0.494
14 20 UBS Group AG Switz. Europe 919 0.890 0.387 0.503
15 32 Standard Chartered Plc UK Europe 646 0.870 0.419 0.451
16 68 Banca Monte dei Paschi Italy Europe 161 0.861 0.429 0.433
17 19
Royal Bank of Scotland Group
UK Europe 981 0.845 0.423 0.423
18 29 Rabobank Group Netherl. Europe 700 0.841 0.376 0.465
19 22 ING Groep NV Netherl. Europe 893 0.823 0.374 0.449
20 76 Mediobanca Italy Europe 95 0.821 0.389 0.433
21 24 Morgan Stanley USA N. Amer. 814 0.752 0.315 0.437
22 69 American Express USA N. Amer. 158 0.736 0.332 0.404
23 6 Bank of America USA N. Amer. 2187 0.716 0.318 0.398
24 23 Goldman Sachs Group USA N. Amer. 860 0.711 0.302 0.410
25 10 Citigroup USA N. Amer. 1792 0.706 0.313 0.393
26 4 HSBC Holdings UK Europe 2374 0.695 0.372 0.323
27 62 Erste Group Bank AG Austria Europe 220 0.688 0.314 0.375
28 49 Svenska Handelsbanken Sweden Europe 289 0.678 0.281 0.397
29 53 Korea Development Bank S. Korea Asia 268 0.677 0.393 0.284
30 60 Banco Sabadell Spain Europe 224 0.670 0.295 0.375
31 36 Danske Denmark Europe 495 0.663 0.296 0.367
32 45
Capital One Financial Corp.
USA N. Amer. 357 0.662 0.260 0.401
33 3 JPMorgan Chase & Co USA N. Amer. 2490 0.658 0.257 0.401
34 7 Wells Fargo USA N. Amer. 1930 0.653 0.259 0.394
35 34 National Australia Bank Australia Oceania 562 0.622 0.376 0.247
Table continued on next page
38
Table 7 - Panel A continued from previous page
36 2 Mitsubishi UFJ Financial Japan Asia 2597 0.621 0.223 0.398
37 47 Shinhan Financial Group S. Korea Asia 328 0.615 0.331 0.283
38 54 Woori Bank S. Korea Asia 257 0.612 0.333 0.278
39 64 Industrial Bank of Korea S. Korea Asia 196 0.603 0.341 0.261
40 28 Commonwealth Bank Australia Oceania 703 0.599 0.388 0.211
41 40 Sberbank of Russia Russia Europe 420 0.598 0.294 0.305
42 30 Australia & N. Zealand Australia Oceania 661 0.597 0.383 0.214
43 50 Skandinaviska Enskilda Sweden Europe 289 0.578 0.262 0.316
44 31 Nordea Sweden Europe 651 0.574 0.289 0.285
45 71 Macquarie USA N. Amer. 143 0.572 0.321 0.251
46 48 KBC Group NV Belgium Europe 291 0.568 0.241 0.327
47 63 Banco Popular Espanol Spain Europe 204 0.558 0.297 0.261
48 61 Bayerische Landesbank Germany Europe 224 0.552 0.249 0.303
49 39 The Export-Import Bank China Asia 427 0.548 0.284 0.264
50 1 Bank of China China Asia 2613 0.548 0.276 0.272
51 66 Malayan Malaysia Asia 171 0.543 0.321 0.223
52 33 Westpac Banking Corp Australia Oceania 606 0.540 0.374 0.166
53 37 State bank of India India Asia 492 0.530 0.238 0.291
54 51 Hana Financial Group S. Korea Asia 288 0.518 0.290 0.227
55 57 Swedbank Sweden Europe 237 0.486 0.253 0.233
56 8 China Development Bank China Asia 1904 0.468 0.260 0.208
57 52 Nationwide Building Society UK Europe 276 0.463 0.217 0.246
58 59 Dexia Belgium Europe 225 0.409 0.197 0.212
59 46 DBS Group Holdings Singapore Asia 332 0.397 0.212 0.186
60 65 Bank of Ireland Ireland Europe 182 0.379 0.197 0.182
61 58 United Overseas Bank Singapore Asia 235 0.373 0.214 0.159
62 11 Mizuho Financial Group Japan Asia 1757 0.367 0.271 0.096
63 67 Standard Bank Group S. Africa Africa 161 0.365 0.123 0.242
64 74 Espirito Santos Portugal Europe 112 0.362 0.186 0.176
65 41 Resona Japan Asia 412 0.357 0.263 0.094
66 13 Sumitomo Mitsui Financial Japan Asia 1654 0.355 0.250 0.104
67 38 U.S. Bancorp USA N. Amer. 445 0.353 0.188 0.165
68 55 Landesbank Baden-Wurtt. Germany Europe 257 0.340 0.170 0.171
69 42 Sumitomo Mitsui T.H. Japan Asia 406 0.323 0.273 0.050
70 56 Cathay Financial Holding Taiwan Asia 252 0.298 0.189 0.109
71 73 Comercial Portuguese Portugal Europe 113 0.279 0.268 0.011
72 70 National Bank of Greece Greece Europe 153 0.220 0.176 0.044
73 18 Norinchukin Bank Japan Asia 984 0.215 0.164 0.050
74 43 Nomura Holdings Japan Asia 370 0.197 0.161 0.035
75 77 Landesbank Hessen Germany Europe 92 0.159 0.089 0.071
76 44 PNC Financial Services USA N. Amer. 366 0.078 0.034 0.043
77 75 Turkiye is bankasi Turkey Europe 112 0.037 0.030 0.007
Note: Results concern the period January 2008 – June 2017 and are of daily frequency
39
Table 8: Distribution of assets when banks ranked by systemic importance and by assets
Banks ranked by:
Quartile Systemic Importance
Assets
1st 34% 62%
2nd 33% 22,4%
3rd 18% 10,5%
4th 14% 5,1%
Note: The systemic importance column shows the distribution by quartile of total bank system assets when banks are ranked in terms of systemic importance. The assets column shows the distribution by quartile of total bank system assets when banks are ranked in terms of assets.
40
Table 9: Systemic contribution per bank (January 2008 – June 2017)
R. Bank Name SC R. Bank Name SC R. Bank Name SC R. Bank Name SC
1 Intesa Sanpaolo
2,25 21 Morgan Stanley 1,60 41 Sberbank of Russia
1,28 61 United Overseas
0,80
2 Banco Santander
2,21 22 American Express
1,57 42 Australia & N. Zealand
1,27 62 Mizuho Financial
0,78
3 BBVA 2,10 23 Bank of America
1,53 43 Skandinaviska Enskilda
1,23 63 Standard Bank Group
0,78
4 BNP Paribas 2,09 24 Goldman Sachs Group
1,52 44 Nordea 1,22 64 Espirito Santos
0,77
5 Unicredit S.p.A.
2,08 25 Citigroup 1,51 45 Macquarie 1,22 65 Resona 0,76
6 Barclays PLC 2,05 26 HSBC Holdings 1,48 46 KBC Group NV 1,21 66 Sumitomo Mitsui
0,76
7 Deutsche Bank
2,03 27 Erste Group 1,47 47 Banco Popular 1,19 67 U.S. Bancorp 0,75
8 Credit Agricole
2,01 28 Svenska 1,45 48 Bayerische Landesbank
1,18 68 Landesbank. 0,72
9 Societe Generale
2,00 29 Korea Development
1,44 49 The Export- Import
1,17 69 Sumitomo Mitsui T.H.
0,69
10 Credit Lyonnais
2,00 30 Banco Sabadell 1,43 50 Bank of China 1,17 70 Cathay Financial
0,64
11 Lloyds Banking
1,97 31 Danske 1,41 51 Malayan 1,16 71 Comercial Portuguese
0,59
12 Credit Suisse Group
1,93 32 Capital One Financial Corp.
1,41 52 Westpac Banking Corp
1,15 72 NBG 0,47
13 Commerzbank 1,93 33 JPMorgan Chase & Co
1,40 53 State bank of India
1,13 73 Norinchukin Bank
0,46
14 UBS Group AG 1,90 34 Wells Fargo 1,39 54 Hana Financial 1,10 74 Nomura Holdings
0,42
15 Standard Chartered
1,86 35 National Australia
1,33 55 Swedbank 1,04 75 Landesbank Hessen
0,34
16 Banca Monte dei Paschi
1,84 36 Mitsubishi UFJ Financial
1,32 56 China Development
1,00 76 PNC Financial Services
0,17
17 RBS 1,80 37 Shinhan Financial
1,31 57 Nationwide Building
0,99 77 Turkiye is bankasi
0,08
18 Rabobank Group
1,79 38 Woori Bank 1,30 58 Dexia 0,87
19 ING Groep NV 1,75 39 Industrial Bank of Korea
1,29 59 DBS Group Holdings
0,85
20 Mediobanca 1,75 40 Commonwealth Bank
1,28 60 Bank of Ireland 0,81
Note: Table presents the systemic contribution of each bank in the system as the ratio between the total individual contagion effects and the total contagion in the system
41
Table 10: Directionality of Bank Systemic Importance
Quartile Directionality
1st 54%
2nd 53%
3rd 47%
4th 30%
Note: Table presents the average directionality for banks in each quartile, when ranked by systemic importance. Directionality is the ratio between the individual externalities and the total individual contagion.
Table 11a: Concentration of banks per region
Quartile Europe N. America Asia Oceania Africa
1st 100% - - - -
2nd 25% 40% 25% 10% -
3rd 50% 10% 30% 10% -
4th 35% 12% 48% - 5%
Note: Table presents the regional concentration of banks for each quartile of systemic importance. Banks are ranked by their systemic importance.
Table 11b: Regional concentration as a percentage of total banks per region
Quartile Europe N. America Asia Oceania Africa
1st 47% - - - - 2nd 12% 73% 25% 25% - 3rd 21% 9% 35% 75% - 4th 16% 18% 40% - 100%
Note: The table presents the concentration of banks as a percentage of the total number of banks for each region for each quartile of systemic importance. For instance, 47% of the European banks can be found in the first quartile of systemic importance.
42
Table 12: Regional Bank Network
Panel A – Externalities of a region as a % of this region’s total externalities
Asia Europe N.America Oceania Sum
Asia 41% 41% 9% 9% 100%
Europe 16% 69% 9.6% 5.4% 100%
N.America 19.5% 58% 16% 6.5% 100%
Oceania 33% 44.5% 9.5% 12% 100%
Panel B - Vulnerabilities of a region as a % of this region’s total vulnerabilities
Asia Europe N.America Oceania
Asia 28% 10% 13% 21%
Europe 54% 75% 64% 59%
N.America 12% 12% 20% 13%
Oceania 6% 3% 3% 7%
Sum 100% 100% 100% 100%
Note: Panel A presents the externalities of region X to each other region as a percentage of region X’s total externalities. Panel B presents the vulnerability of region X to each other region as a percentage of region X’s total vulnerability.
43
Table 13: Determinants of each bank’s loadings in the first component over the nine periods
2008 2009 2010 2011 2012 2013 2014 2015 2016
Size -0.012 -0.002 -0.007 -0.012** 0.001 0.004 0.001 -0.002 -0.003
(0.008) (0.007) (0.006) (0.006) (0.007) (0.006) (0.010) (0.008) (0.009)
Loan-to- assets -0.048 -0.142*** -0.113*** -0.136*** -0.135*** -0.114*** -0.212*** -0.210*** -0.239***
(0.043) (0.051) (0.051) (0.051) (0.054) (0.051) (0.078) (0.069) (0.074)
NPLs 0.844*** 0.414 0.352** 0.035 0.004 -0.097 0.333 0.304*** 0.194
(0.349) (0.253) (0.187) (0.183) (0.054) (0.139) (0.187) (0.147) (0.174)
Non-interest- income 0.000 0.000 0.001** 0.000 0.001 0.000 0.001 0.000 0.000
(0.000) (0.000) (0.006) (0.006) (0.000) (0.000) (0.000) (0.000) (0.000)
GDP -0.004 -0.003 -0.006*** -0.003 -0.009*** -0.009*** -0.011*** -0.004*** -0.007
(0.004) (0.002) (0.001) (0.003) (0.054) (0.003) (0.005) (0.001) (0.006)
Primary Surplus/GDP 0.000 -0.002 0.002 0.000 -0.002 -0.004 -0.009*** -0.014*** -0.012***
(0.002) (0.001) (0.001) (0.170) (0.054) (0.003) (0.004) (0.004) (0.005)
Debt/GDP -0.001*** -0.0005*** -0.0006*** -0.001*** -0.0007*** -0.0005*** -0.0005*** 0.000 -0.000***
(0.000) (0.000) (0.000) (0.000) (0.054) (0.000) (0.000) (0.000) (0.000)
_cons 0.300 0.235 0.297 0.372 0.248 0.229 0.252 0.283 0.323
(0.080) (0.076) (0.068) (0.078) (0.054) (0.072) (0.111) (0.094) (0.107)
Note: ***1%, **5%,*10%. Bank size is each bank’s total assets (log). The loan-to-asset ratio (levels) stands for each bank’s retail orientation. The non-interest income (levels) over total revenue is considered to be a measure of each bank’s diversification. The role of sovereigns is examined through a bank’s home-country GDP growth, the primary surplus over GDP (levels), and public debt over GDP (levels).
44
Figure 1: Example of pairwise directional connectedness
Figure 2: Average systemic risk per bank – Own shocks are excluded
Note: Results cover the period January 2008-June 2017.
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
45
Figure 3: Evolution of Bank Systemic Importance
Note: Each line depicts the total systemic risk (externalities + vulnerability) for each bank in the sample. The length of the rolling window is 340 days and the step is 150 days.
Figure 4: Systemic risk range
Note: Systemic risk range is defined as the difference between the highest and the lowest systemic importance score in the system. The length of the window is 340 days and the step is 150 days.
0.00
0.20
0.40
0.60
0.80
1.00
1.20
Ja n
-0 8
M ay
-0 8
Se p
-0 8
Ja n
-0 9
M ay
-0 9
Se p
-0 9
Ja n
-1 0
M ay
-1 0
Se p
-1 0
Ja n
-1 1
M ay
-1 1
Se p
-1 1
Ja n
-1 2
M ay
-1 2
Se p
-1 2
Ja n
-1 3
M ay
-1 3
Se p
-1 3
Ja n
-1 4
M ay
-1 4
Se p
-1 4
Ja n
-1 5
M ay
-1 5
Se p
-1 5
Ja n
-1 6
M ay
-1 6
Se p
-1 6
Ja n
-1 7
M ay
-1 7
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Ja n
-0 8
M ay
-0 8
Se p
-0 8
Ja n
-0 9
M ay
-0 9
Se p
-0 9
Ja n
-1 0
M ay
-1 0
Se p
-1 0
Ja n
-1 1
M ay
-1 1
Se p
-1 1
Ja n
-1 2
M ay
-1 2
Se p
-1 2
Ja n
-1 3
M ay
-1 3
Se p
-1 3
Ja n
-1 4
M ay
-1 4
Se p
-1 4
Ja n
-1 5
M ay
-1 5
Se p
-1 5
Ja n
-1 6
M ay
-1 6
Se p
-1 6
Ja n
-1 7
M ay
-1 7
46
Figure 5: Total Systemic Risk Index (TSRI)
Note: TSRI is defined as the average response per bank in the connectedness matrix. The length of the window is 340 days and the step is 150 days.
Figure 6: European, US and Euro-area banks’ contribution to total systemic risk
Note: Figure presents the ratio of aggregated systemic importance scores of banks in Europe, Eurozone and the U.S. over those in the total sample. US banks’ contribution is measured on the right axis.
0.00
5.00
10.00
15.00
20.00
25.00
30.00
35.00
40.00
45.00
Ja n
-0 8
M ay
-0 8
Se p
-0 8
Ja n
-0 9
M ay
-0 9
Se p
-0 9
Ja n
-1 0
M ay
-1 0
Se p
-1 0
Ja n
-1 1
M ay
-1 1
Se p
-1 1
Ja n
-1 2
M ay
-1 2
Se p
-1 2
Ja n
-1 3
M ay
-1 3
Se p
-1 3
Ja n
-1 4
M ay
-1 4
Se p
-1 4
Ja n
-1 5
M ay
-1 5
Se p
-1 5
Ja n
-1 6
M ay
-1 6
Se p
-1 6
Ja n
-1 7
M ay
-1 7
0.00
5.00
10.00
15.00
20.00
25.00
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
Ja n
-0 8
M ay
-0 8
Se p
-0 8
Ja n
-0 9
M ay
-0 9
Se p
-0 9
Ja n
-1 0
M ay
-1 0
Se p
-1 0
Ja n
-1 1
M ay
-1 1
Se p
-1 1
Ja n
-1 2
M ay
-1 2
Se p
-1 2
Ja n
-1 3
M ay
-1 3
Se p
-1 3
Ja n
-1 4
M ay
-1 4
Se p
-1 4
Ja n
-1 5
M ay
-1 5
Se p
-1 5
Ja n
-1 6
M ay
-1 6
Se p
-1 6
Ja n
-1 7
M ay
-1 7
EU Euroarea US
47
Figure 7: Rolling Principal Components analysis
Note: Figure presents the first three eigenvalues that explain most of the variation in the system over the period 2008-2016. The length of the window is 200 days and the step is 100 days
0
10
20
30
40
50
60
Ja n
-0 8
Ju n
-0 8
N o
v -0
8
A p
r- 0
9
Se p
-0 9
F eb
-1 0
Ju l-
1 0
D ec
-1 0
M ay
-1 1
O ct
-1 1
M ar
-1 2
A u
g- 1
2
Ja n
-1 3
Ju n
-1 3
N o
v -1
3
A p
r- 1
4
Se p
-1 4
F eb
-1 5
Ju l-
1 5
D ec
-1 5
M ay
-1 6
O ct
-1 6
comp1 comp2 comp3
48
BANK OF GREECE WORKING PAPERS
225. Hondroyiannis G. and D. Papaoikonomou, “The Effect of Card Payments on VAT Revenue in Greece”, May 2017.
226. Mamatzakis E.C. and A.N. Vu, “The interplay between quantitative easing and risk: the case of the Japanese banking”, May 2017.
227. Kosma, T., E. Papapetrou, G. Pavlou, C. Tsochatzi and P. Zioutou, “Labour Market Adjustments and Reforms in Greece During the Crisis: Microeconomic Evidence from the Third Wave of the Wage Dynamics Survey”, June 2017.
228. Gibson D.H, and G. Pavlou, “Exporting and Performance: Evidence from Greek Firms”, June 2017.
229. Papaspyrou S. T. “A New Approach to Governance and Integration in EMU for an Optimal Use of Economic Policy Framework - Priority to Financial Union”, June 2017.
230. Kasimati, E. and N. Veraros, “Is there accuracy of forward freight agreements in forecasting future freight rates? An empirical investigation, June 2017.
231. Rompolis, L., “The effectiveness of unconventional monetary policy on risk aversion and uncertainty”, June 2017.
232. Mamatzakis, C. E., and A. Kalyvas, “Do creditor rights and information sharing affect the performance of foreign banks?”, July 2017.
233. Izquierdo M., J. F. Jimeno, T. Kosma, A. Lamo, S. Millard, T. Rõõm, E. Viviano, “Labour market adjustment in Europe during the crisis: microeconomic evidence from the wage dynamics network survey”, September 2017.
234. Economides, G., D. Papageorgiou, and A. Philippopoulos, “The Greek Great Depression: a General Equilibrium Study of its Drivers”, September 2017.
235. Dellas, H., D. Malliaropulos, D. Papageorgiou, E. Vourvachaki, “Fiscal Policy with an Informal Sector”, October 2017.
236. Dellas, H., G.S., Tavlas, “Milton Friedman and the case for flexible exchange rates and monetary rules”, October 2017.
237. Avramidis, P., I. Asimakopoulos, D., Malliaropulos, and N.G. Travlos. “Group affiliation in periods of credit contraction and bank’s reaction: evidence from the Greek crisis”, December 2017.
238. Karadimitropoulou, A., “Advanced Economies and Emerging Markets: Dissecting the Drivers of Business Cycle Synchronization”, December 2017.
239. Bussiere, M., A. Karadimitropoulou, and M. A. León-Ledesma, “Current Account Dynamics and the Real Exchange Rate: Disentangling the Evidence”, December 2017.
