Description
The paper is to understand how the financial system is influenced by macroeconomic
shocks and how the financial stance, in turn, feeds back into the macroeconomic environment is key
for policy makers. The most recent financial crisis has demonstrated the need for a deeper
understanding of these interdependencies. The purpose of this paper is to analyze what
macroeconomic shocks affect the soundness of the German banking system
Journal of Financial Economic Policy
What macroeconomic shocks affect the German banking system?: Analysis in an
integrated micro-macro model
Sven Blank J onas Dovern
Article information:
To cite this document:
Sven Blank J onas Dovern, (2010),"What macroeconomic shocks affect the German banking system?",
J ournal of Financial Economic Policy, Vol. 2 Iss 2 pp. 126 - 148
Permanent link to this document:http://dx.doi.org/10.1108/17576381011070193
Downloaded on: 24 January 2016, At: 21:38 (PT)
References: this document contains references to 30 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 777 times since 2010*
Users who downloaded this article also downloaded:
Dilip K. Das, (2010),"Financial globalization: a macroeconomic angle", J ournal of Financial Economic
Policy, Vol. 2 Iss 4 pp. 307-325http://dx.doi.org/10.1108/17576381011100847
Donald J . Mullineaux, Mark K. Pyles, (2010),"Bank marketing investments and bank performance", J ournal
of Financial Economic Policy, Vol. 2 Iss 4 pp. 326-345http://dx.doi.org/10.1108/17576381011100856
Subrata Sarkar, Rudra Sensarma, (2010),"Partial privatization and bank performance:
evidence from India", J ournal of Financial Economic Policy, Vol. 2 Iss 4 pp. 276-306 http://
dx.doi.org/10.1108/17576381011100838
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
What macroeconomic shocks
affect the German banking
system?
Analysis in an integrated micro-macro model
Sven Blank
Economics Department, University of Tuebingen, Tuebingen, Germany, and
Jonas Dovern
Kiel Economics Research & Forecasting GmbH & Co. KG, Kiel, Germany
Abstract
Purpose – The paper is to understand how the ?nancial system is in?uenced by macroeconomic
shocks and how the ?nancial stance, in turn, feeds back into the macroeconomic environment is key
for policy makers. The most recent ?nancial crisis has demonstrated the need for a deeper
understanding of these interdependencies. The purpose of this paper is to analyze what
macroeconomic shocks affect the soundness of the German banking system.
Design/methodology/approach – The paper draws on a micro-macro stress-testing framework for
the German banking system in which macroeconomic and bank-speci?c data are used to identify the
effects of various shocks in a structural vector autoregressive model, which includes main
macroeconomic variables and an indicator of stress in the banking system. To this end, the
sign-restriction approach is applied.
Findings – First, it is found that there is a close link between macroeconomic developments and the
stance of the banking sector. Second, monetary policy shocks are the most in?uential shocks for
distress in the banking sector. Third, ?scal policy shocks and real estate price shocks have a
signi?cant impact on the distress indicator, while evidence is mixed for the exchange rate. Fourth, for
the identi?cation of most shocks it is essential to work in the integrated model that combines the
micro- and the macro-sphere.
Originality/value – The paper analyzes various shock scenarios in an integrated micro-macro
framework that takes the mutual relationship between the ?nancial stance and the macroeconomic
environment into account.
Keywords Banking, Monetary policy, Germany, Macroeconomics, Risk analysis
Paper type Research paper
1. Introduction
We analyze what macroeconomic shocks affect the soundness of the German banking
system and how this, in turn, feeds back into the macroeconomic environment.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1757-6385.htm
JEL classi?cation – C32, E44, G21
Financial support by “Stiftung Geld und Wa¨hrung” is highly appreciated by both authors.
The authors would like to thank the Banking Supervision Department of the Deutsche
Bundesbank for its hospitality and for access to its bank-level data; also Claudia Buch, Ferre De
Graeve, Thomas Kick, Michael Koetter, and Thomas Laubach for very helpful comments. Any
errors are of course, the responsibility of the authors. The views presented in this paper re?ect
the authors’ opinion, and do not necessarily coincide with those of the Deutsche Bundesbank.
JFEP
2,2
126
Journal of Financial Economic Policy
Vol. 2 No. 2, 2010
pp. 126-148
qEmerald Group Publishing Limited
1757-6385
DOI 10.1108/17576381011070193
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The recent turmoil on the international ?nancial markets has shown very clearly that
continuously monitoring the banking system and quantifying the risks to which banks
are exposed is of utmost importance to investors and policy makers. Over the last two
decades, stress testing at the level of ?nancial institutions has become more and more
important in addition to traditional approaches like value-at-risk (Committee on the
Global Financial System, 2001). However, stress testing from a macroeconomic
perspective has attracted attention not until more recent years (Sorge, 2004).
In this paper, we demonstrate how the recently introduced approach by
De Graeve et al. (2008) can be extended to allow for identi?cation of a richer variety of
macroeconomic shocks. Their approach combines the micro-sphere (banks) and
macro-sphere (entire economy) into an integrated model framework for analyzing the
sensitivity of the banking sector to macroeconomic shocks. The two appealing features
of this approach are, on the one hand, that it makes use of both macro- and bank-speci?c
data and, on the other hand, that it allows for contemporaneous feedback effects between
the macro- and micro-level in both directions. Both features are also advantageous when
analyzing shocks different to the ones presented in De Graeve et al. (2008).
Motivated by the ?nancial crises in emerging markets in the late 1990s and the
increasing worldwide integration of ?nancial markets, central banks, and international
institutions took lead in augmenting the micro-perspective at the individual bank by a
macro-perspective that addresses overall ?nancial stability. Even though the weakness
of individual banks possibly will be the trigger to larger crises, it is mostly the
deterioration of macroeconomic environment that makes the single bank fail and may
cause chain reactions in a tightened surrounding (Gavin and Hausmann, 1996). Major
crises in the ?nancial system, therefore, cannot simply be dispatched as a result of
failures in single institutions. It is the interaction between the ?nancial system and the
macroeconomy that drives the dynamics. Increasingly, central banks and international
organizations study this interaction to assess the resilience of the ?nancial system –
especially the banking sector – to extreme but plausible shocks to its operational
environment (European Central Bank, 2006). The most extensive appliance of
macroeconomic stress testing so far has been accomplished by the International
Monetary Fund as part of its Financial System Assessment Programs[1].
Existing evidence on feedback effects between the real economy and the ?nancial
sector suggests that the link between the two parts of the De Graeve et al. (2008) model
is indeed very important (Goodhart et al., 2004, 2006). On the one hand, the well being
of the banking sector – as measured by various balance sheet items – can be affected
by macroeconomic shocks (Dovern et al., 2008). On the other hand, transmission
channels are also working in the opposite direction; the two most important concepts
are the bank lending channel and the ?nancial accelerator effect. The former concerns
the intermediation role of banks in transmitting changes in the monetary policy stance
into the real economy (Bernanke and Gertler, 1989). The second effect refers to the fact
that frictions in the banking sector can amplify business cycle ?uctuations (Kiyotaki
and Moore, 1997; Bernanke et al., 1999; Allen and Saunders, 2004). While we do not
identify the contribution of each channel to the shock transmission in our empirical
analysis below, the integrated micro-macro framework in principle allows both
channels to be prevalent.
The basic model by De Graeve et al. (2008) has been extended recently along several
dimensions. De Graeve and Koetter (2007b) expand the model to allow for endogenous
The German
banking system
127
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
balance sheet adjustments in the banking sector in response to macroeconomic shocks
by making the bank speci?c variables in the baseline model react endogenously to
variations in the other variables. They show that only the share of customer loans
reacts signi?cantly to monetary policy shocks while the other bank speci?c variables
are not affected. De Graeve and Koetter (2007a) propose an identi?cation scheme that
can be used to identify an exogenous ?nancial distress shock in the extended model.
They show that while the effects of monetary policy and aggregate supply shocks
remain the same, those effects attributed to aggregate demand shocks in the original
model are likely to actually re?ect the in?uence of ?nancial distress shocks.
The original model is extended by Blank et al. (2009) along a different dimension. The
authors explore whether shocks originating at large banks affect the probability of
distress (PD) of smaller banks. To this end, they construct a measure of idiosyncratic
shocks at large banks and include this measure into the integrated model proposed by De
Graeve et al. (2008). They conclude that positive shocks at the large banks reduce the PD
for smaller banks.
In this paper, we extend the model along a third dimension. While the contributions
mentioned so far made an attempt to introduce more realistic features into the micro part
of the integrated model, we enlarge the macro part of the model. By doing so, we are able
to identify additional macroeconomic structural shocks which might be of interest in
stress testing the banking system.
Our results support the following conclusions. First, the results show a close link
between macroeconomic developments and the stance of the banking sector. We ?nd
some differences across sub-samples by restricting the data for estimation to only those
observations that corresponds to a certain category of distress events or to certain types
of banks. There is a larger impact of macroeconomic variables for weaker distress
events. In addition, the results are more robust for cooperative banks. Second, monetary
policy shocks are the most in?uential shocks for the development of the distress
indicator. Third, while also ?scal policy shocks and real estate price shocks have a
signi?cant impact onthe distress indicator. However, the evidence is mixedfor exchange
rate shocks. For equity price shocks, we do not ?nd any impact. Fourth, for the
identi?cation of most shocks it is essential to work in the integrated model that combines
micro and macro evidences. In sum, we conclude based on these ?ndings that ?nancial
regulators should look at a broad range of macroeconomic developments rather than
focusing too narrowly on direct stability measures like balance sheet data or current
stress events.
The remainder of the paper is structured as follows. In Section 2, we present the model
framework in which we analyze the stability of the banking sector. We show how the
reduced formintegrated micro-macro vector autoregressive (VAR) model is extended to
allowfor a richer set of structural shocks and howwe identify these shocks. In Section 3,
we brie?y outline the different datasets, which are used in the empirical analysis. In
Section 4, we showsome basic correlation statistics to demonstrate howour indicator of
?nancial distress moves in line with several macroeconomic variables. In Section 5, we
present the empirical results of our analysis. We present results for the microeconomic
model as well as for the combined approach for the full sample and various sub-samples
of our data. Section 6 concludes the paper.
JFEP
2,2
128
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
2. The model
2.1 The basic model
The basic model is presented in De Graeve et al. (2008). Their extension of an approach
proposed by Jacobson et al. (2005) combines macroeconomic and individual data on
single banks in one integrated model framework that allows for feedback effects
between micro and macro data in both directions. The microeconometric part of the
model is given by a binary model that links the PD for a bank to bank speci?c
covariates and macroeconomic variables. The macroeconometric part is modeled as a
VAR model including the most important macroeconomic variables. As shown below,
the two parts are combined to yield an integrated VAR model that inherits all features
described so far.
The key equation at the micro-level explains a bank’s PD as a function of k
1
bank
speci?c covariates, collected in vector X
it
, and k
2
macroeconomic variables, collected in
vector Z
t
:
PD
it
¼
expðb
1
X
it21
þb
2
Z
t21
Þ
1 þ expðb
1
X
it21
þb
2
Z
t21
Þ
ð1Þ
This equation is estimated using yearly data and a pooled logit model[2]. Since we
expect the explanatory variables to have a delayed impact on the probability of a
distress event of a bank, we adopt the speci?cation proposed by De Graeve et al. (2008)
and choose a lag length of one year which is suggested by conventional lag selection
criteria.
The macro part of the model can be written as:
Z
t
¼ P
MM
Z
t21
þP
FM
PD
a
t21
þu
t:
ð2Þ
It models the dynamics of the macroeconomy as an autoregressive process of Z
t
. As an
additional explanatory variable, we include the aggregate probability of bank distress,
PD
a
t
, which is constructed as in De Graeve et al. (2008) and measured by the frequency
of distressed events across all banks in the sample.
The VAR approach is used for three reasons. First, VARs usually perform relatively
well in describing the data generating process of macroeconomic variables. Second, the
general form of the model is not based on any structural assumptions on the way
?nancial distress and the macroeconomy interact. A consensus on the effectiveness of
different transmission channels has not yet emerged (European Central Bank, 2005). It
is therefore appealing to resort to a model that needs as little a priori theorizing as
possible. Finally, the structure of the model allows us to introduce feedback effects in a
very convenient way in the next step.
Note that so far the model does not incorporate the feedback effects between the
micro- and the macro-sphere, since PD
a
t
is assumed to be exogenously determined. To
obtain the integrated model, we need to augment the macroeconometric model by one
equation originating from the micro part that describes the evolution of the PD:
Z
t
PD
a
t
" #
¼
P
MM
P
MF
!
Z
t21
þ
P
FM
P
FF
!
PD
a
t21
þ1
t
ð3Þ
The German
banking system
129
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The elasticity of the PD with respect to the macroeconomic variables is given by the
outcome from the microeconometric part of the model:
P
MF
;
›PðPD
it
¼ 1jZ
t21
; X
it21
Þ
›Z
t
¼
expð
^
b
1
X
it21
þ
^
b
2
Z
t21
Þ
½1 þ expð
^
b
1
X
it21
þ
^
b
2
Z
t21
Þ?
2
^
b
2
ð4Þ
Note that by imposing the estimated effects from the microeconometric model, the
bank speci?c variables retain an important role in the integrated model although they
are assumed to be exogenous at this stage. This is because P
MF
depends on the level of
each of the variables in the micro part of the model[3].
2.2 Extending the basic model
In this subsection, we enhance the basic model by De Graeve et al. (2008) to allow for a
wider set of macroeconomic shocks. We can write an augmented version of the model
presented in equation (3) as:
Y
t
¼ PY
t21
þGW
t
þ1
t
ð5Þ
where Y
t
¼ ½Z
t
PD
a
t
y
1
y
2
. . .y
p
?
0
includes the endogenous variables of the basic model
together with p additional endogenous variables, W
t
¼ ½w
1
w
2
. . .w
q
? denotes a vector
of q exogenous variables, and P and G are properly dimensioned coef?cient matrices.
By extending the model with the appropriate additional exogenous and endogenous
variables, it becomes feasible to identify more speci?c fundamental shocks – such as
?scal shocks (Caldara and Kamps, 2008), or exchange rate and asset price shocks
(Fratzscher et al., 2007) – than in the basic model where the variation could only be
attributed to monetary policy shocks, aggregate supply shocks, or aggregate demand
shocks (De Graeve et al., 2008).
2.3 Identi?cation of structural shocks
Starting from the complete reduced form n-dimensional VAR given in equation (5),
we are interested in the responses of the variables in Y
t
to various structural shocks. To
this end, the vector of prediction errors u
t
has to be translated into a vector of
economically meaningful structural innovations. The essential assumption in this
context is that these structural innovations are orthogonal to each other. Consequently,
identi?cation amounts to providing enough restrictions to uniquely solve for a
decomposition of the covariance matrix of the reduced form VAR: S ¼ A
0
A
0
0
. This
de?nes a one-to-one mapping from the vector of orthogonal structural shocks v
t
to the
reduced-form residuals, u
t
¼ A
0
v
t
. Because of the orthogonality assumption, and the
symmetry of S, only n(n 2 1)/2 restrictions need to be imposed to pin down A
0
.
Like De Graeve et al. (2008), we follow an approach recently proposed by Uhlig
(2005), which achieves identi?cation of the VAR by imposing sign-restrictions on the
impulse responses of a set of variables[4]. Identi?cation of the model is achieved by a
simulation approach[5].
This so-called “pure sign-restriction approach” has one main advantage.
Sign-restrictions are relative mild identifying assumptions compared to conventional
approaches like identi?cation through recursive ordering of the variables or long-run
zero restrictions. The obtained results are, for instance, insensitive to the speci?c
decomposition of S or the ordering of the variables in the VAR[6]. Thus, we do not
JFEP
2,2
130
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
have to take any stance on the question to which particular shocks the banking sector
corresponds contemporaneously. Instead, we let the data speak about the effects of the
shocks on the banking sector.
One major disadvantage of this approach is that the number of restrictions has to be
suf?ciently large to identify the impact of a shock. Hence, in larger systems, the
number of necessary identifying restrictions can grow rapidly. In this study, however,
the simulated models are suf?ciently small to dismantle this concern.
3. Data
3.1 Data sources
3.1.1 Distress indicators. For the estimation of the micro part of the model, we use data
on distress events among German banks between 1994 and 2004 from the annual
distress database of the Deutsche Bundesbank. These data are con?dential and are
available on the premises of the Deutsche Bundesbank only. The construction of the
distress events follows the classi?cation proposed by Kick and Koetter (2007) and used
by De Graeve et al. (2008)[7]. The data includes the following four categories of distress
events. The weakest type of distress (“distress category I”) comprises mandatory
announcements by individual banks to the supervisory authority like a drop of annual
operational pro?ts or liable capital by more than 25 percent. The second category
(“distress category II”) captures of?cial warnings by the German Financial Supervisory
Authority (Bundesanstalt fu¨r Finanzdienstleistungsaufsicht, BaFin). A more severe
sign of banking distress (“distress category III”) are direct interventions into the
ongoing business of a bank by the BaFin, like restrictions to lending or deposit taking.
In addition, this category also comprises capital injections by the insurance scheme of
the respective banking sector. Finally, the worst distress category (“distress category
IV”) comprises all closures of banks and restructuring mergers.
On the aggregate level, the PD is constructed as the ratio of distress events in the
banking sector to the total number of banks in the sample. It measures the
unconditional frequency of distress events across the banking sector[8].
3.1.2 Other bank-speci?c variables. Information on individual bank balance sheets
comes from con?dential data collected by the Deutsche Bundesbank. Since the number
of bank-speci?c covariates is potentially very large, we use a selection of variables that
have been identi?ed by De Graeve et al. (2008) following an approach by Hosmer and
Lemeshow (2000). The selection procedure is oriented at the so-called capitalization,
asset quality, management, earnings, and liquidity taxonomy (King et al., 2006). Details
on the selection process can be found in the paper by De Graeve et al. (2008). The
bank-speci?c variables which are taken into account, are the equity ratio, total
reserves, customer loans, off-balance sheet activities, size, return on equity, and
liquidity. Consequently, the vectors X
it
contains eight variables (including a constant).
3.1.3 Macroeconomic variables. The set of variables representing the macroeconomy
consist of the three standard variables, namely gross domestic product (GDP) growth,
consumer price in?ation, and the short-term interest rate; consequently, the vector Z
t
has dimension three. In the different augmented models that we show in the next
section, we include the German Stock Market Index (DAX-index), government revenues
(GRVs) and expenditures, a price de?ator for construction in the housing sector (which
serves as a proxy for real estate prices), and the nominal effective exchange
The German
banking system
131
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
rate (NXR),respectively. The data – except for the exchange rate and the interest rate –
have been seasonally adjusted prior to differencing.
3.2 Data availability
The sample period is restricted by the availability of bank speci?c data on distress
events and the banks’ characteristics. In its currently available form, these data cover
the period from 1994 to 2004. We use annual data to estimate the model in equation (1),
since the properties of the data on the bank speci?c covariates do not allow to switch to
a higher frequency. Owing to the panel structure of the data with a fairly high
cross-section dimension, estimation is feasible with effectively ten years of data.
The sample for the macroeconomic variables is chosen correspondingly. To enable
us to estimate versions of the extended model given in equation (5), we use quarterly
data and transform the estimates subsequently to conform to the microeconometric
part. Still, the sample size is relatively low and does not allow for estimation of very
rich models, i.e. q and p cannot be chosen to be large. This is why estimation of a
universal model is infeasible.
4. Stylized facts
Before we present our model and the results of our formal econometric analysis, we
give some stylized facts about the relation between the soundness of the banking
system and the macroeconomy in this section.
The distress indicator shows a substantial degree of co-movement with most of the
macroeconomic variables (Figure 1).
It seems to be especially moving together with GDP growth, stock return, change of
government expenditure (GEX), and the exchange rate. This visual impression is
Figure 1.
Distress indicator and
macroeconomic variables
Distress indicator GDP growth (rhs) Inflation (rhs) Policy rate (rhs)
I
n
d
e
x
P
e
r
c
e
n
t
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.01
0.03
0.05
0.07
–2
0
2
4
6
Distress indicator Stock return (rhs) Exchange rate (rhs)
I
n
d
e
x
P
e
r
c
e
n
t
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.01
0.03
0.05
0.07
–60
–20
20
60
Distress indicator Government exp. (rhs) Government rev. (rhs)
I
n
d
e
x
P
e
r
c
e
n
t
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.01
0.03
0.05
0.07
–40
0
40
80
Note: Government revenues in 2000Q3 are affected by revenues from the auctioning of UMTS licences
JFEP
2,2
132
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
supported by an analysis of correlation coef?cients, which shows that the distress
indicator is substantially negatively correlated to those six variables (Appendix,
Table AI). There are differences, however, in the size of the correlation and the lag at
which a high correlation can be found.
The distress indicator shows a strong contemporaneous correlation with GDP
growth and stock returns, that is strong growth and high stock market returns go in
line with a reduction of the PD events in the banking sector. In contrast, the distress
indicator seems to lag behind the development of GEXs and, even more pronounced,
the exchange rate, i.e. increases of GEX and an appreciation of the Euro are followed by
a reduction of the PD events in the banking sector.
Hence, there seems to be indeed some interaction of the stance of the banking
system and macroeconomic factors. This in turn indicates that the identi?cation of
structural shocks is of particular importance. To analyze which shocks exactly drive
this interaction, we have to move to a formal econometric model.
5. Results
5.1 Micro-level estimates
On the micro-level, we estimate the impact of our bank-speci?c covariates and
macroeconomic variables on the PD using the model that is given by equation (1)[9].
5.1.1 Results for the full sample. In the baseline model, the estimates of the marginal
effects of bank-level and macro-level variables largely con?rm the results shown in
De Graeve et al. (2008). The explanatory power, as measured by the pseudo-R
2
is
11.45 percent.
Better capitalized banks, i.e. banks with a higher equity ratio and higher reserves,
have a lower PD. The coef?cients are negative and highly signi?cant. Higher customer
loans and broader off-balance sheet activities imply higher credit risk, and we would
expect a positive impact on the PD (Kick and Koetter, 2007). While we ?nd the
in?uence of off-balance sheet activities to be insigni?cant, this is indeed the case for
customer loans. The size of small and medium-sized banks signi?cantly reduces the
likelihood of a distress event. In addition, more pro?table banks, as measured by
the return on equity, have a lower PD. One could argue that banks with higher liquidity
are less likely to experience a distress event. On the other hand, liquidity is consistent
with a “signalling effect”. High liquidity signals a lack of interest-bearing investment
possibilities and thus low pro?tability. In fact, similar to Blank et al. (2009), we ?nd this
signaling effect to dominate, since liquidity enters positively signi?cant.
A positive macroeconomic environment, as measured by higher real GDP growth,
which is usually accompanied by rising prices, should reduce the PD. We ?nd indeed a
signi?cantly negative coef?cient for both, real GDP growth as well as in?ation for
most of the sub-samples. We would expect the interest rate to exert a positive in?uence
on the PD as it higher costs of re?nancing. However, in the baseline scenario, the
interest rate is insigni?cant.
The remaining models allowfor a more comprehensive impact of the macroeconomic
environment on the PD. The estimates show that adding additional macroeconomic
variables to this baseline speci?cation leaves the result for the bank-speci?c covariates
remarkably stable in general. Our results for the effects of a change in stock market
valuation showthat a rise of the DAXhas a signi?cant negative impact on the PD. Some
of the effect might be due to the signaling property of the DAX for business
The German
banking system
133
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
cycle movements. But there could also exist a direct effect since banks hold some of their
assets in equity which leads to re-valuations of the banks’ assets due to changes in stock
prices. This might – in case of an adverse shock – lead to an increasing distress
probability. The coef?cients for the other three macroeconomic variables remain
unchanged.
In our model to measure ?scal policy effects, we include the change of GRVs as well
as GEXs. In addition, we also include dummies to account for the extraordinary high
revenues through alienation of Universal Mobile Telecommunications System (UMTS)
licences in 2000. GRVs have a signi?cant positive effect on banks’ PD, whereas
expenditures are insigni?cant. The positive sign for revenues may be explained by
higher tax burden that might increase the likelihood of borrowers’ defaults. In addition,
we ?nd, somewhat unexpectedly, a negative coef?cient for the interest rate.
Next, we include the change of prices on the real estate market to analyze effects of
this high-volume market. Like increasing in?ation, rising prices in the housing sector
should reduce the PD. Indeed, we ?nd a negative signi?cant effect in both
speci?cations. Also, the interest rate now has the expected positive sign.
Finally, we want to analyze the effects of exchange rate movements because
roughly 50 percent of German banks’ foreign liabilities and 30-40 percent of banks’
foreign assets were denominated in foreign currency in the period under study. Thus,
an appreciation of the euro should have a signi?cant negative effect on banks’
cross-border assets and liabilities. However, since the ratio of foreign liabilities to
foreign assets ranges from 1.1 to 0.7, the impact of valuation changes stemming from
exchange rate ?uctuations and their effect on the PD is not clear cut. In fact, we ?nd the
marginal impact of changes in the NXR to be insigni?cant in our micro-level estimates.
5.1.2 Results for different distress categories. So far, we have not accounted for the
ordinal character of our endogenous variable. We check the robustness of our results
by splitting our sample according to the different categories of distress (De Graeve et al.,
2008)[10]. It is likely that the different distress events correlate differently with
macroeconomic shocks. One could expect for instance that the weaker distress events
as automatic signals or warnings by the ?nancial supervisor, as summarized in
categories I and II, are driven by both, bank-speci?c characteristics as well
macroeconomic factors, while severer events are caused solely by bank-speci?c factors.
In the baseline model without any additional macroeconomic variables, we indeed
?nd that – like most bank-speci?c covariates – GDP growth and in?ation enter
signi?cantly for distress categories I and II. In addition, the interest rate enters
signi?cantly positive in the sub-sample for distress category I. In contrast, there is no
impact from the macroeconomic sphere when considering events that comprise
interventions from the banking pillars head institution or the ?nancial supervisor into
the active business of a bank (distress category III). However, in the most severe
distress category, we ?nd a signi?cant impact of in?ation and, somewhat weaker, of
GDP growth. This indicates that the timing of restructuring mergers is in?uenced by
the macroeconomic environment (Blank et al., 2009).
We also check whether these results are in line with the speci?cations including
other macroeconomic factors. The most striking results are given for the model
including ?scal variables. For the distress category II, the results are very similar to the
full sample results. For distress categories I, both, GEXs and revenues are insigni?cant.
The only macroeconomic factor that drives the likelihood to observe an event assigned
JFEP
2,2
134
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
to distress category III is GEXs, which enters with a positive sign. One explanation
might be that the government could try to stimulate the economy when there is a
worsening of the business cycle. Hence, an increase in GEXs could be an indication for
a severe recession that increases the PD.
5.1.3 Results for different banking sectors. Next, we check the robustness of our
?ndings if we split up the sample according to Germany’s three banking pillars into
private commercial banks, savings banks, and cooperative banks. As the business
models of the different types of banks in Germany differ considerably, it is reasonable
to expect the different types of banks to be affected differently by certain
macroeconomic shocks. For example, cooperative banks may be less exposed to
international shocks, since these banks are mainly engaged on the domestic market.
In the baseline regressions without any additional macroeconomic variables, some of
the bank-speci?c variables as well the macroeconomic aggregates are insigni?cant. We
?nd the most stable results for cooperative banks, which drive the results for the full
sample, as these banks make up nearly 20,000 observations (out of roughly 28,000
observations in the full sample). However, the impact of the size of a bank nowswitches in
sign from being negative to positive. The sub-sample for savings banks comprises
approximately 6,000 observations. The only signi?cant in?uence from the
macroeconomic environment is given by the in?ation rate (INFL), which enters
negatively. In addition to in?ation, the interest rate is signi?cant with the expected
positive sign if we restrict the sample to private banks, which make up nearly 1,700
observations. Most of the banks-speci?c covariates turn out to be insigni?cant in this
speci?cation.
For the speci?cations with an enriched macroeconomic environment, the ?ndings
compared to the baseline regressions are more or less the same. However, some
differences are noteworthy. House prices have a negative signi?cant in?uence on the PD
in the sub-sample for cooperatives as well as in the sub-sample for savings banks. This
may be due to the fact that lending on mortgages is more important for these banking
sectors than for commercial banks. GEXs are positively signi?cant for the sub-sample
comprising savings banks. GRVs enter positively signi?cant in the sub-sample with
cooperative banks. However, the INFL now has a positive sign, thus increases the PD.
Although the in?uence of the nominal exchange rate is insigni?cant for the full sample
and for the different distress categories, we ?nd a negative signi?cant impact for both,
the sub-sample for the private banks and for the savings banks. We cannot ?nd a
signi?cant impact for cooperative banks, which are more concentrated on the national
economy and, thus, are less prone to valuation effects through exchange rate changes.
5.2 Responses to macroeconomic shocks in the combined framework
Having established the impact of the bank-speci?c and the macroeconomic variables
on the micro-level, we pursue and estimate the response of the PD in the VAR
framework described in Section 3.2. We do this for both, the pure macro VAR that does
not incorporate feedback effects from the macro-sphere to the ?nancial sector as well as
for the integrated model that allows for mutual macro-micro linkages. We also analyze
the response of the ?nancial stance to structural macroeconomic shocks in the various
sub-samples. Since the number of scenarios that we study is large, we report only a
collection of our results. As mentioned already in Section 3.2, we have to use different
versions of the extended model that include different sets of additional variables due to
The German
banking system
135
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
the low degrees of freedom. In Appendix, Table AII gives an overview over the
different shocks that we analyze and states which additional variables (analogously to
the micro-level estimations) we include in each case as well as which sign-restrictions
are used to identify the shocks[11].
5.2.1 Standard shocks in baseline model. The subsequent ?gures show annual
impulse responses to a one standard deviation shock. The solid line represents the
median of 500 simulations. As a benchmark, we replicate the ?ndings by De Graeve et al.
(2008) in the baseline model without any additional macroeconomic variables. To this
end, we analyze the response of the PD to a monetary policy shock. In Figure 2,
we show the impulse responses for the combined system.
A contractionary monetary policy, as given by a rise in the interest rate,
signi?cantly (on a 66 percent level) increases the aggregate PD from the ?rst year
onward, pointing to a potential con?ict of objectives between the ?nancial supervisor
and the European Central Bank. Note that like De Graeve et al. (2008) or Jacobson et al.
(2005), we ?nd that integrating the micro part of the model is very important for the
dynamics. Neglecting the impact of bank-speci?c variables leads to considerably
different responses for the PD. To check if these results are robust across different
sub-samples or whether we can observe some differences across sub-samples, we split
the sample and redo the impulse response analysis. We ?nd similar results for distress
category IV. In addition, the response of the PD is very pronounced when restricting
the sample to private banks. This may re?ect the fact that in contrast to savings banks
or cooperative banks those banks rely much more on re-?nancing via the capital
Figure 2.
Monetary policy shock
in the integrated VAR
GDP growth
Interest rate Distress frequency
Inflation rate
0.2
0.3 0.5
–0.5
–1
0
0.2
0.1
–0.1
0
–0.2
–0.4
–0.6
0 1 2 3 4
0
Note: Annual impulse responses to a one standard deviation shock. The solid lines
show the median impulse responses. The two dashed lines in each plot represent the
bounds of 66% confidence bands. For the sign restrictions used to identify each
shock see Appendix Table AII
1 2 3 4 0 1 2 3 4
0 1 2 3 4
0
0.1
–0.1
–0.2
–0.3
0
JFEP
2,2
136
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
market (as their deposit business is relatively small) and are, hence, directly affected by
a change of the yield curve that is usually implied by changes in the short-term interest
rate.
In addition to monetary policy shocks, the baseline model allows us to identify
aggregate demand and aggregate supply shocks. We model a supply shock by
restricting GDP growth to fall and in?ation to rise during the ?rst year. In the case of a
demand shock, we restrict GDP growth, in?ation, and the interest rate to fall on impact.
While we would expect the distress probability to rise in both recessionary
scenarios, we only ?nd the aggregate demand shock to have a signi?cant impact
(Figure 3), an aggregate supply shock is insigni?cant (Figure 4).
An analysis of sub-samples reveals that the distress probability of savings banks is
less affected by aggregate demand shocks – and the effect is only marginally
signi?cant. The distress probability of commercial banks is affected – but only for one
year after the shock hit the economy. The aggregate result seems to be solely driven by
the response of cooperative banks for which the reaction of the distress probability
looks like the one for the entire sample. Regarding the aggregate supply shocks, our
results show that they have no signi?cant impact for all of the sub-samples.
We now turn to the augmented models that allow us to identify a bunch of other
structural shocks in addition to the three basic macroeconomic shocks for which we
have presented results so far[12].
5.2.2 Equity price shocks. Next, we examine the consequences of an equity price
shock. Note that this shock can be understood as a combination of a immediate shock
to the banks’ balance sheets (because some of their assets are equity investments) and a
more precisely de?ned aggregate demand shock. The difference to the standard
Figure 3.
Aggregate demand shock
in the integrated VAR
0.2 0.05
0.4
0.2
–0.2
–0.4
–0.6
0
–0.05
–0.1
–0.15
–0.2
0
0.1
0.05
–0.05
–0.1
–0.15
0
0
Note: See Figure 2
1 2 3 4
0.1
–0.1
–0.2
–0.3
–0.4
–0.5
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate Distress frequency
The German
banking system
137
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
aggregate demand shock presented above is simply that we do not force aggregate
output to fall but set the sign-restriction in such a way that we specify a speci?c cause
for declining aggregate demand. As Figure 5 shows, we cannot ?nd any signi?cant
impact on the PD. This ?nding is neither reversed in the integrated model nor for any
of the sub-sample estimations.
5.2.3 Fiscal policy shocks. When we impose sign-restrictions to identify different
?scal policy shocks, the results depend to a high degree on how the ?scal policy shock
is designed. While a “pure government spending shock” (Caldara and Kamps, 2008)
does not signi?cantly affect the distress probability even in the integrated model, the
opposite is true for a “pure GRV shock” (Figure 6). A rise in GRVs leads to a signi?cant
increase of the distress probability after one year. An explanation is that an increased
tax burden leads to a higher insolvency rate among debtors that eventually causes the
frequency of distressed events to go up. A weaker, but still signi?cant impact can be
found for the sub-sample for cooperative banks as well as distress category II.
Instead of concentrating on just one side of the government budget, one could also
argue that what we will call a “contractionary ?scal shock” should be identi?ed by
decreasing GEXs and increasing GRVs. We ?nd that such an identi?cation scheme has
a pronounced in?uence on the distress probability in the integrated model (Figure 7).
The aggregate PD sharply and signi?cantly increases one period period after the
shock hits the economy. Again, we cannot detect such an impact in the pure macro
VAR. We further explore the role of the “contractionary ?scal shock” for ?nancial
stability and conduct the impulse response analysis for different sub-samples. For
distress category I and for cooperative banks, we qualitatively ?nd the same response
of the distress probability.
Figure 4.
Aggregate supply shock
in the integrated VAR
0.2
0.1 0.5
–0.5
0
0.05
–0.05
–0.1
–0.15
–0.2
–0.25
0
0.3
0.2
0.1
–0.1
–0.2
0
0.1
–0.1
–0.2
–0.3
–0.4
–0.5
0
0 1 2
GDP growth
Interest rate Distress frequency
Inflation rate
3 4
0
Note: See Figure 2
1 2 3 4 0 1 2 3 4
0 1 2 3 4
JFEP
2,2
138
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Figure 5.
Equity price shock in the
integrated VAR
0.4
0.2
–0.2
–0.4
0
0.2 5
0
–5
–10
0.1
–0.1
–0.2
0
0.4
0.2
–0.2
–0.4
0
0.2
0.1
–0.1
–0.2
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0
Note: See Figure 2
1 2 3 4
Interest rate
Distress frequency
Interest rate
Equity prices
GDP growth
Figure 6.
Pure GRV shock in the
integrated VAR
0.5 0.2
–0.2
0
0.2
10
–10
0
2
–2
0
–0.2
0
1
–1
0
–0.5
0
0 1 2 3 0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
4
0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Public expenditures Interest rate
Public revenues Distress frequency
Note: See Figure 2
The German
banking system
139
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
5.2.4 Real estate price shocks. When we impose the sign-restrictions to identify a real
estate price shock in the pure macro VAR, we cannot ?nd any signi?cant impact on the
PD (Figure 8)[13].
However, a real estate price shock signi?cantly increases the likelihood to observe a
distress event one period after the shock hits the economy if we shock the integrated
system (Figure 9). This signi?cant effect in the integrated model also holds if we
consider cooperative banks only, and, though somewhat weaker, if we restrict our
sample to private banks or to distress category II.
5.2.5 Exchange rate shocks. When we impose the sign-restrictions to identify an
exchange rate shock, we observe no signi?cant reaction of the distress probability –
neither in the pure macro VAR nor in the integrated model (Figure 10).
Also, the estimation based on the various sub-samples yields no signi?cant
response. The identi?cation of this shock is, however, a good example to demonstrate
the sensitivity of the results to the parameters chosen for estimation. We can for
instance impose the sign-restrictions for two instead of only one year. This might be
justi?ed by the fact that currency deprecations (appreciations) tend to occur in rather
persistent long movements. If we do so, it turns out that the distress probability rises,
though signi?cantly only two and three years after the shock hit the economy
(Figure 11).
It seems to be the case that banks get into trouble once the re-valuation of their
assets, that are denominated in foreign currency, causes write-offs due to a sustained
depreciation of the home currency.
Figure 7.
Contractionary ?scal
policy shock in the
integrated VAR
0.4
0.2
10 1.5
1
0.5
–0.5
0
5
–5
0
0.5
–0.5
–1
0
0.1
–0.1
–0.2
0
0.2
0.1
–0.1
–0.2
0
0.2
–0.2
–0.4
0
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate
Distress frequency Public revenues
Public expenditures
Note: See Figure 2
JFEP
2,2
140
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Figure 8.
Real estate price shock
in the pure macro VAR
0.5
0.2
–0.2
0
0.5
–0.5
0
0.5
–0.5
0
0.2
–0.2
0
–0.5
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate House prices
Distress frequency
Note: See Figure 2
Figure 9.
Real estate prices shock
in the integrated VAR
0.5
0.2
–0.2
0
0.5
–0.5
0
0.5
–0.5
0
0.2
–0.2
0
–0.5
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1
Note: See Figure 2
2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate House prices
Distress frequency
The German
banking system
141
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Figure 10.
Exchange rate shock
in the integrated VAR
0.4 0.3
0.2
0.1
–0.1
–1
–2
0
0
1 0.3
0.5
–0.5
0
0.2
0.1
–0.1
0
0.2
–0.2
–0.4
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth
Interest rate Nom. eff. exchange rate
Distress frequency
Inflation rate
Note: See Figure 2
Figure 11.
Exchange rate shock
in the integrated VAR
(longer restrictions)
0.6 0.3
0.2
0.1
–0.1
–1
–2
0
0
1 0.3
0.6
0.4
0.2
–0.2
0
0.2
0.1
–0.1
0
0.4
0
–0.2
0.2
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth
Interest rate Nom. eff. exchange rate
Distress frequency
Inflation rate
Notes: See Figure 2; in contrast to the other estimations we impose
sign-restrictions over two years instead of one year only
JFEP
2,2
142
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
6. Conclusion
In this paper, we have demonstrated how the model recently proposed by
De Graeve et al. (2008) can be enriched to allow for the analysis of impacts on the
banking system of a higher variety of structural macroeconomic shocks.
In sum, the empirical results show the following: ?rst, there is a close link between
macroeconomic developments and the stance of the banking sector. Second, monetary
policy shocks seem to be the most in?uential shocks for the development of the distress
indicator. Third, while also ?scal policy shocks and real estate price shocks have a
signi?cant impact on the distress indicator, the evidence is mixed for exchange rate
shocks. For equity price shocks we do not ?nd any impact. Fourth, for the identi?cation
of most shocks it is essential to work in the integrated model that combines micro and
macro evidences. Finally, we could identify some differences across sub-samples by
restricting the data for estimation to only those observations that corresponds to a
certain category of distress events or to certain types of banks. Most importantly, we
?nd that monetary policy shocks hit private banks much more severe than savings
banks or cooperative banks. In addition, we found that macroeconomic shocks tend to
increase the PDevents of category I or II more than they do for the more severe events of
categories III and IV. In general, however, the differences across sub-samples are smaller
than expected and do not show a systematic pattern.
These ?ndings lead to the following policy implications for ?nancial regulation
authorities; they show which developments should be closely monitored by them
because they increase the risk of ?nancial instability in the future. The results from the
baseline model showthat there might be a trade off between the proper monetary policy
stance and ?nancial stability; De Graeve et al. (2008) discuss potential intra-institutional
con?icts if a central bank is also involved in ?nancial regulation. The augmented models
showthat especially unexpected ?scal contraction can have strong impacts on ?nancial
stability; this suggests that the current stimulus packages around the world should be
removed slowly and according to a well-communicated time schedule. In addition, also
developments on the real estate market should be closely monitored because price drops
on that market can increase ?nancial stress directly via balance sheet effects and
indirectly via effects on aggregate demand. Finally, exchange rate movements should be
monitored; arguably, this should be done especially when judging the distress situation
of highly international connected banks with large assets positions denoted in foreign
currency. In sum, we conclude that ?nancial regulators should look at a broad range of
macroeconomic developments rather than focusing too narrowly on direct stability
measures like balance sheet data or current stress events. This is in analogy to monetary
policy, for which the experience of the recent years has shown that it is too short-sighted
to solely concentrate on the development of in?ation.
Our analysis suffers from the scarce data availability of distress events over the
time dimension. We expect most of the insigni?cant coef?cients to be due to the low
number of data points available for estimating the models, which impedes identifying
the true correlation structure of the data generating processes. In addition, the lack of
data makes an estimation of a “global” model, that includes all relevant variables
altogether, infeasible. However, it would be very interesting to address this issue in
future work with an extended dataset.
The German
banking system
143
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Notes
1. See International Monetary Fund (2008) for more information on those programs.
2. Note that the macroeconomic variables do not vary across the different banks. Hence, we
cannot include time ?xed effects in the regressions. As a consequence, the marginal effects of
the macroeconomic variables potentially re?ect to some degree the effects of unobserved
macroeconomic factors.
3. De Graeve et al. (2008) note that this feature makes the coef?cients state dependent and
allows for experiments that analyzes the system’s behavior for different levels of the bank
speci?c variables. While we do not pursue this issue in this paper and set P
MF
equal to its
value obtained for the sample averages of X
it
and Z
t
, an example of such an analysis can be
found in the paper by De Graeve et al. (2008).
4. Initially, it has been proposed to obtain these restrictions by choosing A
0
to be a Cholesky
factorization of S, implying a recursive ordering of the variables as in Sims (1986). This
method has been questioned on various grounds (Sims, 1992; Grilli and Roubini, 1996;
Christiano et al., 1998). Although, alternative identi?cation schemes like for instance the
approach proposed by Blanchard and Quah (1989) have been introduced in the past, none of
them has remained without criticism (Fernald, 2007).
5. For the speci?cs of this Bayesian estimation strategy, we refer to Appendix B in Uhlig
(2005).
6. De Graeve et al. (2008) note that in fact for some standard identifying restrictions like for
instance for a monetary policy shock, “the restrictions [in the sign-restriction approach] nest
the recursive (or Choleski) response.”
7. A more detailed discussion of the construction can be found in Kick and Koetter (2007).
8. Three subset databases of the distress database (measures, incidents, distressed mergers),
which contain information about exact dates, are used to allocate the distress events to
quarters to construct this series on a quarterly frequency; this is done to match the highest
frequency available for the macroeconomic variables described below.
9. The detailed results are collected in an appendix and available from the authors upon
request.
10. Alternatively, we could also adopt the method employed by Kick and Koetter (2007) and
estimate an ordered logit. However, we stick to this more intuitive approach.
11. In setting the identi?cation restrictions, we mainly follow various approaches presented in the
literature – among others by Uhlig (2005), Caldara and Kamps (2008), or Fratzscher et al. (2007).
12. Note that we will do not attribute shares of the variation of the distress probability to speci?c
shocks, since we do not identify all shocks in one model; consequently we cannot perform a
forecast error decomposition that takes into account all structural shocks.
13. Note that this shock can be understood as a more precisely de?ned aggregate demand shock.
The difference to the standard aggregate demand shock presented above is simply that we
do not force aggregate output to fall but set the sign-restriction in such a way that we specify
a speci?c cause for declining aggregate demand.
References
Allen, L. and Saunders, A. (2004), “Incorporating systemic in?uences into risk measurements:
survey of the literature”, Journal of Financial Services Research, Vol. 26 No. 2, pp. 161-92.
Bernanke, B.S. and Gertler, M. (1989), “Agency cost, net worth, and business ?uctuations”,
American Economic Review, Vol. 79 No. 1, pp. 14-31.
JFEP
2,2
144
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Bernanke, B.S., Gertler, M. and Gilchrist, S. (1999), “Financial accelerator in a quantitative
business cycle framework”, in Taylor, J.B. and Woodford, M. (Eds), Handbook of
Macroeconomics, Vol. 1, Elsevier, Amsterdam.
Blanchard, O.J. and Quah, D. (1989), “The dynamic effects of aggregate demand and supply
disturbances”, American Economic Review, Vol. 79 No. 4, pp. 655-73.
Blank, S., Buch, C.M. and Neugebauer, K. (2009), “Shocks at large banks and banking sector distress:
the banking granular residual”, Journal of Financial Stability, Vol. 5 No. 4, pp. 353-73.
Caldara, D. and Kamps, C. (2008), “What are the effects of ?scal policy shocks? A VAR-based
comparative analysis”, Working Paper Series 877, European Central Bank, Frankfurt am
Main, March.
Christiano, L., Eichenbaum, M.S. and Evans, C.L. (1998), “Monetary policy shocks: what have we
learned and to what end?”, Working Paper 6400, National Bureau for Economic Research,
Cambridge, MA, February.
Committee on the Global Financial System (2001), “A survey of stress tests and current practice
at major ?nancial institutions”, Report, Bank for International Settlements, Basel, April.
De Graeve, F. and Koetter, M. (2007a), Shock to the System: Real Shocks, Financial Shocks and
Financial Distress, Mimeo, University of Groningen, Groningen.
De Graeve, F. and Koetter, M. (2007b), The Role of Balance Sheet Adjustments in Transmitting
Monetary Policy to Financial Distress, Mimeo, University of Groningen, Groningen.
De Graeve, F., Kick, T. and Koetter, M. (2008), “Monetary policy and ?nancial (in)stability:
an integrated micro-macro approach”, Journal of Financial Stability, Vol. 4 No. 3, pp. 205-31.
Dovern, J., Meier, C.-P. and Vilmeier, J. (2008), “How resilient is the German banking system to
macroeconomic shocks?”, Kiel Working Papers 1419, Kiel Institute for the World
Economy, Kiel, May.
European Central Bank (2005), Financial Stability Report, European Central Bank, Frankfurt am
Main.
European Central Bank (2006), Financial Stability Review, European Central Bank, Frankfurt am
Main.
Fernald, J.G. (2007), “Trend breaks, long run restrictions, and contractionary technology
improvements”, Working Paper 2005-21, Federal Reserve Bank of San Francisco,
San Francisco, CA, June.
Fratzscher, M., Juvenal, L. and Sarno, L. (2007), “Asset prices, exchange rates and the current
account”, Working Paper 790, European Central Bank, Frankfurt am Main, August.
Gavin, M. and Hausmann, R. (1996), “The roots of banking crises: the macroeconomic context”,
Working Paper Series 318, Inter-American Development Bank, Washington, DC, January.
Goodhart, C.A., Sunirand, P. and Tsomocos, D. (2004), “A model to analyse ?nancial fragility:
applications”, Journal of Financial Stability, Vol. 1 No. 1, pp. 1-30.
Goodhart, C.A., Sunirand, P. and Tsomocos, D. (2006), “A model to analyse ?nancial fragility”,
Economic Theory, Vol. 27 No. 1, pp. 107-42.
Grilli, V.U. and Roubini, N. (1996), “Liquidity models in open economies: theory and empirical
evidence”, European Economic Review, Vol. 40 Nos 3-5, pp. 847-59.
Hosmer, D.W. and Lemeshow, S. (2000), Applied Logistic Regression, 2nd ed., Wiley, NewYork, NY.
International Monetary Fund (2008), “Financial Sector Assessment Program (FSAP)”, available
at: www.imf.org/external/NP/fsap/fsap.asp (accessed July 28, 2009).
Jacobson, T., Linde´, J. and Roszbach, K. (2005), “Exploring interactions between real activity and
the ?nancial stance”, Journal of Financial Stability, Vol. 1 No. 3, pp. 308-41.
The German
banking system
145
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Kick, T. and Koetter, M. (2007), “Slippery slopes of stress: ordered failure events in German
banking”, Journal of Financial Stability, Vol. 3 No. 2, pp. 132-48.
King, T.B., Nuxoll, D. and Yeager, T.J. (2006), “Are the causes of bank distress changing? Can
researchers keep up?”, Federal Reserve Bank of St Louis Review, Vol. 88 No. 1, pp. 57-80.
Kiyotaki, N. and Moore, J. (1997), “Credit cycles”, Journal of Political Economy, Vol. 105 No. 2,
pp. 211-48.
Sims, C.A. (1986), “Are forecasting models usable for policy analysis?”, Quarterly Review, Winter,
pp. 2-16.
Sims, C.A. (1992), “Interpreting the macroeconomic time series facts: the effects of monetary
policy”, European Economic Review, Vol. 36 No. 5, pp. 975-1011.
Sorge, M. (2004), “Stress-testing ?nancial systems: an overview of current methodologies”, BIS
Working Papers 165, Bank for International Settlements, Basel, December.
Uhlig, H. (2005), “What are the effects of monetary policy? Results from an agnostic identi?cation
procedure”, Journal of Monetary Economics, Vol. 52 No. 2, pp. 381-419.
Appendix
Correlation of the distress indicator with. . . . . .At lag 0 . . .At lag 1 . . .At lag 2 . . .At lag 4
GDP growth 20.37
* *
20.29 20.11 0.18
In?ation 0.11 0.04 20.09 20.41
* *
Policy rate 20.11 20.03 0.06 0.25
De?ator for construction 20.37 20.43
*
20.46
* *
20.42
*
Stock returns 20.50
* * *
20.49
* * *
20.48
* * *
20.02
Exchange rate 0.11 20.02 20.15 20.48
* * *
GEX 20.16 20.29
*
20.31
*
20.42
* *
GRVs 20.08 20.09 20.11 0.06
Notes: Signi?cance at
*
10,
* *
5, and
* * *
1 percent level, respectively; correlations are calculated over
the sample 1994Q1-2004Q3; for GRVs we excluded those observations that are distorted by the
revenues from the auction of UMTS licences
Table AI.
Cross-correlations
JFEP
2,2
146
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
M
o
n
e
t
a
r
y
p
o
l
i
c
y
s
h
o
c
k
A
g
g
r
e
g
a
t
e
s
u
p
p
l
y
s
h
o
c
k
A
g
g
r
e
g
a
t
e
d
e
m
a
n
d
s
h
o
c
k
E
q
u
i
t
y
p
r
i
c
e
s
h
o
c
k
G
E
X
s
h
o
c
k
G
R
V
s
h
o
c
k
C
o
n
t
r
a
c
t
i
o
n
a
r
y
?
s
c
a
l
s
h
o
c
k
R
e
a
l
e
s
t
a
t
e
p
r
i
c
e
s
h
o
c
k
E
x
c
h
a
n
g
e
r
a
t
e
s
h
o
c
k
D
G
D
P
2
2
2
?
?
?
?
?
?
I
N
F
L
2
þ
2
?
?
?
?
2
þ
I
R
A
T
E
þ
?
2
2
?
?
?
2
þ
P
D
?
?
?
?
?
?
?
?
?
A
d
d
i
t
i
o
n
a
l
v
a
r
i
a
b
l
e
s
D
D
A
X
X
X
X
2
X
2
X
X
X
D
G
E
X
X
X
X
X
2
?
2
X
X
D
G
R
V
X
X
X
X
?
þ
þ
X
X
D
C
O
P
X
X
X
X
X
X
2
2
X
D
N
X
R
X
X
X
X
X
X
2
X
2
N
o
t
e
s
:
A
“
2
”
i
n
d
i
c
a
t
e
s
t
h
a
t
w
e
i
m
p
o
s
e
a
d
e
c
r
e
a
s
i
n
g
o
f
t
h
e
c
o
r
r
e
s
p
o
n
d
i
n
g
v
a
r
i
a
b
l
e
t
o
i
d
e
n
t
i
f
y
t
h
e
s
h
o
c
k
,
l
i
k
e
w
i
s
e
a
“
þ
”
i
n
d
i
c
a
t
e
s
t
h
a
t
w
e
i
m
p
o
s
e
a
n
i
n
c
r
e
a
s
e
,
a
n
d
a
“
?
”
i
n
d
i
c
a
t
e
s
t
h
a
t
w
e
d
o
n
o
t
i
m
p
o
s
e
a
n
y
r
e
s
t
r
i
c
t
i
o
n
s
,
“
X
”
d
e
n
o
t
e
s
t
h
a
t
t
h
e
c
o
r
r
e
s
p
o
n
d
i
n
g
v
a
r
i
a
b
l
e
i
s
n
o
t
i
n
c
l
u
d
e
d
i
n
t
h
e
m
o
d
e
l
u
s
e
d
t
o
i
d
e
n
t
i
f
y
t
h
e
s
h
o
c
k
;
I
R
A
T
E
,
m
a
i
n
p
o
l
i
c
y
r
a
t
e
;
C
O
P
,
c
o
n
s
t
r
u
c
t
i
o
n
p
r
i
c
e
s
Table AII.
Analyzed shocks and
corresponding models
The German
banking system
147
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
About the authors
Sven Blank is a Research Assistant at the chair for “International Macroeconomics and Finance”
at the Eberhard-Karls Universita¨t Tu¨bingen. His research focus is on the link between ?nancial
markets and the macroeconomy as well as on international ?nance. Sven Blank is the
corresponding author and can be contacted at: [email protected]
Jonas Dovern is Founder and Director of Kiel Economics Research & Forecasting GmbH &
Co. KG, an enterprise that offers forecasting products and services as well as economic expertise
to the private and public sectors. Before, he was Researcher at the Kiel Institute for the World
Economy (IfW) where he was working for both the research area on monetary policy and the
business cycle forecasting center as an econometrician. Jonas Dovern studied econometrics at
Maastricht University and holds a PhD in economics from Kiel University. His research activities
cover different topics in macroeconomics, monetary policy, and econometric forecasting.
JFEP
2,2
148
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This article has been cited by:
1. Jarmo Pesola. 2011. Joint effect of financial fragility and macroeconomic shocks on bank loan losses:
Evidence from Europe. Journal of Banking & Finance 35, 3134-3144. [CrossRef]
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_308635051.pdf
The paper is to understand how the financial system is influenced by macroeconomic
shocks and how the financial stance, in turn, feeds back into the macroeconomic environment is key
for policy makers. The most recent financial crisis has demonstrated the need for a deeper
understanding of these interdependencies. The purpose of this paper is to analyze what
macroeconomic shocks affect the soundness of the German banking system
Journal of Financial Economic Policy
What macroeconomic shocks affect the German banking system?: Analysis in an
integrated micro-macro model
Sven Blank J onas Dovern
Article information:
To cite this document:
Sven Blank J onas Dovern, (2010),"What macroeconomic shocks affect the German banking system?",
J ournal of Financial Economic Policy, Vol. 2 Iss 2 pp. 126 - 148
Permanent link to this document:http://dx.doi.org/10.1108/17576381011070193
Downloaded on: 24 January 2016, At: 21:38 (PT)
References: this document contains references to 30 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 777 times since 2010*
Users who downloaded this article also downloaded:
Dilip K. Das, (2010),"Financial globalization: a macroeconomic angle", J ournal of Financial Economic
Policy, Vol. 2 Iss 4 pp. 307-325http://dx.doi.org/10.1108/17576381011100847
Donald J . Mullineaux, Mark K. Pyles, (2010),"Bank marketing investments and bank performance", J ournal
of Financial Economic Policy, Vol. 2 Iss 4 pp. 326-345http://dx.doi.org/10.1108/17576381011100856
Subrata Sarkar, Rudra Sensarma, (2010),"Partial privatization and bank performance:
evidence from India", J ournal of Financial Economic Policy, Vol. 2 Iss 4 pp. 276-306 http://
dx.doi.org/10.1108/17576381011100838
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
What macroeconomic shocks
affect the German banking
system?
Analysis in an integrated micro-macro model
Sven Blank
Economics Department, University of Tuebingen, Tuebingen, Germany, and
Jonas Dovern
Kiel Economics Research & Forecasting GmbH & Co. KG, Kiel, Germany
Abstract
Purpose – The paper is to understand how the ?nancial system is in?uenced by macroeconomic
shocks and how the ?nancial stance, in turn, feeds back into the macroeconomic environment is key
for policy makers. The most recent ?nancial crisis has demonstrated the need for a deeper
understanding of these interdependencies. The purpose of this paper is to analyze what
macroeconomic shocks affect the soundness of the German banking system.
Design/methodology/approach – The paper draws on a micro-macro stress-testing framework for
the German banking system in which macroeconomic and bank-speci?c data are used to identify the
effects of various shocks in a structural vector autoregressive model, which includes main
macroeconomic variables and an indicator of stress in the banking system. To this end, the
sign-restriction approach is applied.
Findings – First, it is found that there is a close link between macroeconomic developments and the
stance of the banking sector. Second, monetary policy shocks are the most in?uential shocks for
distress in the banking sector. Third, ?scal policy shocks and real estate price shocks have a
signi?cant impact on the distress indicator, while evidence is mixed for the exchange rate. Fourth, for
the identi?cation of most shocks it is essential to work in the integrated model that combines the
micro- and the macro-sphere.
Originality/value – The paper analyzes various shock scenarios in an integrated micro-macro
framework that takes the mutual relationship between the ?nancial stance and the macroeconomic
environment into account.
Keywords Banking, Monetary policy, Germany, Macroeconomics, Risk analysis
Paper type Research paper
1. Introduction
We analyze what macroeconomic shocks affect the soundness of the German banking
system and how this, in turn, feeds back into the macroeconomic environment.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1757-6385.htm
JEL classi?cation – C32, E44, G21
Financial support by “Stiftung Geld und Wa¨hrung” is highly appreciated by both authors.
The authors would like to thank the Banking Supervision Department of the Deutsche
Bundesbank for its hospitality and for access to its bank-level data; also Claudia Buch, Ferre De
Graeve, Thomas Kick, Michael Koetter, and Thomas Laubach for very helpful comments. Any
errors are of course, the responsibility of the authors. The views presented in this paper re?ect
the authors’ opinion, and do not necessarily coincide with those of the Deutsche Bundesbank.
JFEP
2,2
126
Journal of Financial Economic Policy
Vol. 2 No. 2, 2010
pp. 126-148
qEmerald Group Publishing Limited
1757-6385
DOI 10.1108/17576381011070193
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The recent turmoil on the international ?nancial markets has shown very clearly that
continuously monitoring the banking system and quantifying the risks to which banks
are exposed is of utmost importance to investors and policy makers. Over the last two
decades, stress testing at the level of ?nancial institutions has become more and more
important in addition to traditional approaches like value-at-risk (Committee on the
Global Financial System, 2001). However, stress testing from a macroeconomic
perspective has attracted attention not until more recent years (Sorge, 2004).
In this paper, we demonstrate how the recently introduced approach by
De Graeve et al. (2008) can be extended to allow for identi?cation of a richer variety of
macroeconomic shocks. Their approach combines the micro-sphere (banks) and
macro-sphere (entire economy) into an integrated model framework for analyzing the
sensitivity of the banking sector to macroeconomic shocks. The two appealing features
of this approach are, on the one hand, that it makes use of both macro- and bank-speci?c
data and, on the other hand, that it allows for contemporaneous feedback effects between
the macro- and micro-level in both directions. Both features are also advantageous when
analyzing shocks different to the ones presented in De Graeve et al. (2008).
Motivated by the ?nancial crises in emerging markets in the late 1990s and the
increasing worldwide integration of ?nancial markets, central banks, and international
institutions took lead in augmenting the micro-perspective at the individual bank by a
macro-perspective that addresses overall ?nancial stability. Even though the weakness
of individual banks possibly will be the trigger to larger crises, it is mostly the
deterioration of macroeconomic environment that makes the single bank fail and may
cause chain reactions in a tightened surrounding (Gavin and Hausmann, 1996). Major
crises in the ?nancial system, therefore, cannot simply be dispatched as a result of
failures in single institutions. It is the interaction between the ?nancial system and the
macroeconomy that drives the dynamics. Increasingly, central banks and international
organizations study this interaction to assess the resilience of the ?nancial system –
especially the banking sector – to extreme but plausible shocks to its operational
environment (European Central Bank, 2006). The most extensive appliance of
macroeconomic stress testing so far has been accomplished by the International
Monetary Fund as part of its Financial System Assessment Programs[1].
Existing evidence on feedback effects between the real economy and the ?nancial
sector suggests that the link between the two parts of the De Graeve et al. (2008) model
is indeed very important (Goodhart et al., 2004, 2006). On the one hand, the well being
of the banking sector – as measured by various balance sheet items – can be affected
by macroeconomic shocks (Dovern et al., 2008). On the other hand, transmission
channels are also working in the opposite direction; the two most important concepts
are the bank lending channel and the ?nancial accelerator effect. The former concerns
the intermediation role of banks in transmitting changes in the monetary policy stance
into the real economy (Bernanke and Gertler, 1989). The second effect refers to the fact
that frictions in the banking sector can amplify business cycle ?uctuations (Kiyotaki
and Moore, 1997; Bernanke et al., 1999; Allen and Saunders, 2004). While we do not
identify the contribution of each channel to the shock transmission in our empirical
analysis below, the integrated micro-macro framework in principle allows both
channels to be prevalent.
The basic model by De Graeve et al. (2008) has been extended recently along several
dimensions. De Graeve and Koetter (2007b) expand the model to allow for endogenous
The German
banking system
127
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
balance sheet adjustments in the banking sector in response to macroeconomic shocks
by making the bank speci?c variables in the baseline model react endogenously to
variations in the other variables. They show that only the share of customer loans
reacts signi?cantly to monetary policy shocks while the other bank speci?c variables
are not affected. De Graeve and Koetter (2007a) propose an identi?cation scheme that
can be used to identify an exogenous ?nancial distress shock in the extended model.
They show that while the effects of monetary policy and aggregate supply shocks
remain the same, those effects attributed to aggregate demand shocks in the original
model are likely to actually re?ect the in?uence of ?nancial distress shocks.
The original model is extended by Blank et al. (2009) along a different dimension. The
authors explore whether shocks originating at large banks affect the probability of
distress (PD) of smaller banks. To this end, they construct a measure of idiosyncratic
shocks at large banks and include this measure into the integrated model proposed by De
Graeve et al. (2008). They conclude that positive shocks at the large banks reduce the PD
for smaller banks.
In this paper, we extend the model along a third dimension. While the contributions
mentioned so far made an attempt to introduce more realistic features into the micro part
of the integrated model, we enlarge the macro part of the model. By doing so, we are able
to identify additional macroeconomic structural shocks which might be of interest in
stress testing the banking system.
Our results support the following conclusions. First, the results show a close link
between macroeconomic developments and the stance of the banking sector. We ?nd
some differences across sub-samples by restricting the data for estimation to only those
observations that corresponds to a certain category of distress events or to certain types
of banks. There is a larger impact of macroeconomic variables for weaker distress
events. In addition, the results are more robust for cooperative banks. Second, monetary
policy shocks are the most in?uential shocks for the development of the distress
indicator. Third, while also ?scal policy shocks and real estate price shocks have a
signi?cant impact onthe distress indicator. However, the evidence is mixedfor exchange
rate shocks. For equity price shocks, we do not ?nd any impact. Fourth, for the
identi?cation of most shocks it is essential to work in the integrated model that combines
micro and macro evidences. In sum, we conclude based on these ?ndings that ?nancial
regulators should look at a broad range of macroeconomic developments rather than
focusing too narrowly on direct stability measures like balance sheet data or current
stress events.
The remainder of the paper is structured as follows. In Section 2, we present the model
framework in which we analyze the stability of the banking sector. We show how the
reduced formintegrated micro-macro vector autoregressive (VAR) model is extended to
allowfor a richer set of structural shocks and howwe identify these shocks. In Section 3,
we brie?y outline the different datasets, which are used in the empirical analysis. In
Section 4, we showsome basic correlation statistics to demonstrate howour indicator of
?nancial distress moves in line with several macroeconomic variables. In Section 5, we
present the empirical results of our analysis. We present results for the microeconomic
model as well as for the combined approach for the full sample and various sub-samples
of our data. Section 6 concludes the paper.
JFEP
2,2
128
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
2. The model
2.1 The basic model
The basic model is presented in De Graeve et al. (2008). Their extension of an approach
proposed by Jacobson et al. (2005) combines macroeconomic and individual data on
single banks in one integrated model framework that allows for feedback effects
between micro and macro data in both directions. The microeconometric part of the
model is given by a binary model that links the PD for a bank to bank speci?c
covariates and macroeconomic variables. The macroeconometric part is modeled as a
VAR model including the most important macroeconomic variables. As shown below,
the two parts are combined to yield an integrated VAR model that inherits all features
described so far.
The key equation at the micro-level explains a bank’s PD as a function of k
1
bank
speci?c covariates, collected in vector X
it
, and k
2
macroeconomic variables, collected in
vector Z
t
:
PD
it
¼
expðb
1
X
it21
þb
2
Z
t21
Þ
1 þ expðb
1
X
it21
þb
2
Z
t21
Þ
ð1Þ
This equation is estimated using yearly data and a pooled logit model[2]. Since we
expect the explanatory variables to have a delayed impact on the probability of a
distress event of a bank, we adopt the speci?cation proposed by De Graeve et al. (2008)
and choose a lag length of one year which is suggested by conventional lag selection
criteria.
The macro part of the model can be written as:
Z
t
¼ P
MM
Z
t21
þP
FM
PD
a
t21
þu
t:
ð2Þ
It models the dynamics of the macroeconomy as an autoregressive process of Z
t
. As an
additional explanatory variable, we include the aggregate probability of bank distress,
PD
a
t
, which is constructed as in De Graeve et al. (2008) and measured by the frequency
of distressed events across all banks in the sample.
The VAR approach is used for three reasons. First, VARs usually perform relatively
well in describing the data generating process of macroeconomic variables. Second, the
general form of the model is not based on any structural assumptions on the way
?nancial distress and the macroeconomy interact. A consensus on the effectiveness of
different transmission channels has not yet emerged (European Central Bank, 2005). It
is therefore appealing to resort to a model that needs as little a priori theorizing as
possible. Finally, the structure of the model allows us to introduce feedback effects in a
very convenient way in the next step.
Note that so far the model does not incorporate the feedback effects between the
micro- and the macro-sphere, since PD
a
t
is assumed to be exogenously determined. To
obtain the integrated model, we need to augment the macroeconometric model by one
equation originating from the micro part that describes the evolution of the PD:
Z
t
PD
a
t
" #
¼
P
MM
P
MF
!
Z
t21
þ
P
FM
P
FF
!
PD
a
t21
þ1
t
ð3Þ
The German
banking system
129
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The elasticity of the PD with respect to the macroeconomic variables is given by the
outcome from the microeconometric part of the model:
P
MF
;
›PðPD
it
¼ 1jZ
t21
; X
it21
Þ
›Z
t
¼
expð
^
b
1
X
it21
þ
^
b
2
Z
t21
Þ
½1 þ expð
^
b
1
X
it21
þ
^
b
2
Z
t21
Þ?
2
^
b
2
ð4Þ
Note that by imposing the estimated effects from the microeconometric model, the
bank speci?c variables retain an important role in the integrated model although they
are assumed to be exogenous at this stage. This is because P
MF
depends on the level of
each of the variables in the micro part of the model[3].
2.2 Extending the basic model
In this subsection, we enhance the basic model by De Graeve et al. (2008) to allow for a
wider set of macroeconomic shocks. We can write an augmented version of the model
presented in equation (3) as:
Y
t
¼ PY
t21
þGW
t
þ1
t
ð5Þ
where Y
t
¼ ½Z
t
PD
a
t
y
1
y
2
. . .y
p
?
0
includes the endogenous variables of the basic model
together with p additional endogenous variables, W
t
¼ ½w
1
w
2
. . .w
q
? denotes a vector
of q exogenous variables, and P and G are properly dimensioned coef?cient matrices.
By extending the model with the appropriate additional exogenous and endogenous
variables, it becomes feasible to identify more speci?c fundamental shocks – such as
?scal shocks (Caldara and Kamps, 2008), or exchange rate and asset price shocks
(Fratzscher et al., 2007) – than in the basic model where the variation could only be
attributed to monetary policy shocks, aggregate supply shocks, or aggregate demand
shocks (De Graeve et al., 2008).
2.3 Identi?cation of structural shocks
Starting from the complete reduced form n-dimensional VAR given in equation (5),
we are interested in the responses of the variables in Y
t
to various structural shocks. To
this end, the vector of prediction errors u
t
has to be translated into a vector of
economically meaningful structural innovations. The essential assumption in this
context is that these structural innovations are orthogonal to each other. Consequently,
identi?cation amounts to providing enough restrictions to uniquely solve for a
decomposition of the covariance matrix of the reduced form VAR: S ¼ A
0
A
0
0
. This
de?nes a one-to-one mapping from the vector of orthogonal structural shocks v
t
to the
reduced-form residuals, u
t
¼ A
0
v
t
. Because of the orthogonality assumption, and the
symmetry of S, only n(n 2 1)/2 restrictions need to be imposed to pin down A
0
.
Like De Graeve et al. (2008), we follow an approach recently proposed by Uhlig
(2005), which achieves identi?cation of the VAR by imposing sign-restrictions on the
impulse responses of a set of variables[4]. Identi?cation of the model is achieved by a
simulation approach[5].
This so-called “pure sign-restriction approach” has one main advantage.
Sign-restrictions are relative mild identifying assumptions compared to conventional
approaches like identi?cation through recursive ordering of the variables or long-run
zero restrictions. The obtained results are, for instance, insensitive to the speci?c
decomposition of S or the ordering of the variables in the VAR[6]. Thus, we do not
JFEP
2,2
130
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
have to take any stance on the question to which particular shocks the banking sector
corresponds contemporaneously. Instead, we let the data speak about the effects of the
shocks on the banking sector.
One major disadvantage of this approach is that the number of restrictions has to be
suf?ciently large to identify the impact of a shock. Hence, in larger systems, the
number of necessary identifying restrictions can grow rapidly. In this study, however,
the simulated models are suf?ciently small to dismantle this concern.
3. Data
3.1 Data sources
3.1.1 Distress indicators. For the estimation of the micro part of the model, we use data
on distress events among German banks between 1994 and 2004 from the annual
distress database of the Deutsche Bundesbank. These data are con?dential and are
available on the premises of the Deutsche Bundesbank only. The construction of the
distress events follows the classi?cation proposed by Kick and Koetter (2007) and used
by De Graeve et al. (2008)[7]. The data includes the following four categories of distress
events. The weakest type of distress (“distress category I”) comprises mandatory
announcements by individual banks to the supervisory authority like a drop of annual
operational pro?ts or liable capital by more than 25 percent. The second category
(“distress category II”) captures of?cial warnings by the German Financial Supervisory
Authority (Bundesanstalt fu¨r Finanzdienstleistungsaufsicht, BaFin). A more severe
sign of banking distress (“distress category III”) are direct interventions into the
ongoing business of a bank by the BaFin, like restrictions to lending or deposit taking.
In addition, this category also comprises capital injections by the insurance scheme of
the respective banking sector. Finally, the worst distress category (“distress category
IV”) comprises all closures of banks and restructuring mergers.
On the aggregate level, the PD is constructed as the ratio of distress events in the
banking sector to the total number of banks in the sample. It measures the
unconditional frequency of distress events across the banking sector[8].
3.1.2 Other bank-speci?c variables. Information on individual bank balance sheets
comes from con?dential data collected by the Deutsche Bundesbank. Since the number
of bank-speci?c covariates is potentially very large, we use a selection of variables that
have been identi?ed by De Graeve et al. (2008) following an approach by Hosmer and
Lemeshow (2000). The selection procedure is oriented at the so-called capitalization,
asset quality, management, earnings, and liquidity taxonomy (King et al., 2006). Details
on the selection process can be found in the paper by De Graeve et al. (2008). The
bank-speci?c variables which are taken into account, are the equity ratio, total
reserves, customer loans, off-balance sheet activities, size, return on equity, and
liquidity. Consequently, the vectors X
it
contains eight variables (including a constant).
3.1.3 Macroeconomic variables. The set of variables representing the macroeconomy
consist of the three standard variables, namely gross domestic product (GDP) growth,
consumer price in?ation, and the short-term interest rate; consequently, the vector Z
t
has dimension three. In the different augmented models that we show in the next
section, we include the German Stock Market Index (DAX-index), government revenues
(GRVs) and expenditures, a price de?ator for construction in the housing sector (which
serves as a proxy for real estate prices), and the nominal effective exchange
The German
banking system
131
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
rate (NXR),respectively. The data – except for the exchange rate and the interest rate –
have been seasonally adjusted prior to differencing.
3.2 Data availability
The sample period is restricted by the availability of bank speci?c data on distress
events and the banks’ characteristics. In its currently available form, these data cover
the period from 1994 to 2004. We use annual data to estimate the model in equation (1),
since the properties of the data on the bank speci?c covariates do not allow to switch to
a higher frequency. Owing to the panel structure of the data with a fairly high
cross-section dimension, estimation is feasible with effectively ten years of data.
The sample for the macroeconomic variables is chosen correspondingly. To enable
us to estimate versions of the extended model given in equation (5), we use quarterly
data and transform the estimates subsequently to conform to the microeconometric
part. Still, the sample size is relatively low and does not allow for estimation of very
rich models, i.e. q and p cannot be chosen to be large. This is why estimation of a
universal model is infeasible.
4. Stylized facts
Before we present our model and the results of our formal econometric analysis, we
give some stylized facts about the relation between the soundness of the banking
system and the macroeconomy in this section.
The distress indicator shows a substantial degree of co-movement with most of the
macroeconomic variables (Figure 1).
It seems to be especially moving together with GDP growth, stock return, change of
government expenditure (GEX), and the exchange rate. This visual impression is
Figure 1.
Distress indicator and
macroeconomic variables
Distress indicator GDP growth (rhs) Inflation (rhs) Policy rate (rhs)
I
n
d
e
x
P
e
r
c
e
n
t
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.01
0.03
0.05
0.07
–2
0
2
4
6
Distress indicator Stock return (rhs) Exchange rate (rhs)
I
n
d
e
x
P
e
r
c
e
n
t
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.01
0.03
0.05
0.07
–60
–20
20
60
Distress indicator Government exp. (rhs) Government rev. (rhs)
I
n
d
e
x
P
e
r
c
e
n
t
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.01
0.03
0.05
0.07
–40
0
40
80
Note: Government revenues in 2000Q3 are affected by revenues from the auctioning of UMTS licences
JFEP
2,2
132
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
supported by an analysis of correlation coef?cients, which shows that the distress
indicator is substantially negatively correlated to those six variables (Appendix,
Table AI). There are differences, however, in the size of the correlation and the lag at
which a high correlation can be found.
The distress indicator shows a strong contemporaneous correlation with GDP
growth and stock returns, that is strong growth and high stock market returns go in
line with a reduction of the PD events in the banking sector. In contrast, the distress
indicator seems to lag behind the development of GEXs and, even more pronounced,
the exchange rate, i.e. increases of GEX and an appreciation of the Euro are followed by
a reduction of the PD events in the banking sector.
Hence, there seems to be indeed some interaction of the stance of the banking
system and macroeconomic factors. This in turn indicates that the identi?cation of
structural shocks is of particular importance. To analyze which shocks exactly drive
this interaction, we have to move to a formal econometric model.
5. Results
5.1 Micro-level estimates
On the micro-level, we estimate the impact of our bank-speci?c covariates and
macroeconomic variables on the PD using the model that is given by equation (1)[9].
5.1.1 Results for the full sample. In the baseline model, the estimates of the marginal
effects of bank-level and macro-level variables largely con?rm the results shown in
De Graeve et al. (2008). The explanatory power, as measured by the pseudo-R
2
is
11.45 percent.
Better capitalized banks, i.e. banks with a higher equity ratio and higher reserves,
have a lower PD. The coef?cients are negative and highly signi?cant. Higher customer
loans and broader off-balance sheet activities imply higher credit risk, and we would
expect a positive impact on the PD (Kick and Koetter, 2007). While we ?nd the
in?uence of off-balance sheet activities to be insigni?cant, this is indeed the case for
customer loans. The size of small and medium-sized banks signi?cantly reduces the
likelihood of a distress event. In addition, more pro?table banks, as measured by
the return on equity, have a lower PD. One could argue that banks with higher liquidity
are less likely to experience a distress event. On the other hand, liquidity is consistent
with a “signalling effect”. High liquidity signals a lack of interest-bearing investment
possibilities and thus low pro?tability. In fact, similar to Blank et al. (2009), we ?nd this
signaling effect to dominate, since liquidity enters positively signi?cant.
A positive macroeconomic environment, as measured by higher real GDP growth,
which is usually accompanied by rising prices, should reduce the PD. We ?nd indeed a
signi?cantly negative coef?cient for both, real GDP growth as well as in?ation for
most of the sub-samples. We would expect the interest rate to exert a positive in?uence
on the PD as it higher costs of re?nancing. However, in the baseline scenario, the
interest rate is insigni?cant.
The remaining models allowfor a more comprehensive impact of the macroeconomic
environment on the PD. The estimates show that adding additional macroeconomic
variables to this baseline speci?cation leaves the result for the bank-speci?c covariates
remarkably stable in general. Our results for the effects of a change in stock market
valuation showthat a rise of the DAXhas a signi?cant negative impact on the PD. Some
of the effect might be due to the signaling property of the DAX for business
The German
banking system
133
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
cycle movements. But there could also exist a direct effect since banks hold some of their
assets in equity which leads to re-valuations of the banks’ assets due to changes in stock
prices. This might – in case of an adverse shock – lead to an increasing distress
probability. The coef?cients for the other three macroeconomic variables remain
unchanged.
In our model to measure ?scal policy effects, we include the change of GRVs as well
as GEXs. In addition, we also include dummies to account for the extraordinary high
revenues through alienation of Universal Mobile Telecommunications System (UMTS)
licences in 2000. GRVs have a signi?cant positive effect on banks’ PD, whereas
expenditures are insigni?cant. The positive sign for revenues may be explained by
higher tax burden that might increase the likelihood of borrowers’ defaults. In addition,
we ?nd, somewhat unexpectedly, a negative coef?cient for the interest rate.
Next, we include the change of prices on the real estate market to analyze effects of
this high-volume market. Like increasing in?ation, rising prices in the housing sector
should reduce the PD. Indeed, we ?nd a negative signi?cant effect in both
speci?cations. Also, the interest rate now has the expected positive sign.
Finally, we want to analyze the effects of exchange rate movements because
roughly 50 percent of German banks’ foreign liabilities and 30-40 percent of banks’
foreign assets were denominated in foreign currency in the period under study. Thus,
an appreciation of the euro should have a signi?cant negative effect on banks’
cross-border assets and liabilities. However, since the ratio of foreign liabilities to
foreign assets ranges from 1.1 to 0.7, the impact of valuation changes stemming from
exchange rate ?uctuations and their effect on the PD is not clear cut. In fact, we ?nd the
marginal impact of changes in the NXR to be insigni?cant in our micro-level estimates.
5.1.2 Results for different distress categories. So far, we have not accounted for the
ordinal character of our endogenous variable. We check the robustness of our results
by splitting our sample according to the different categories of distress (De Graeve et al.,
2008)[10]. It is likely that the different distress events correlate differently with
macroeconomic shocks. One could expect for instance that the weaker distress events
as automatic signals or warnings by the ?nancial supervisor, as summarized in
categories I and II, are driven by both, bank-speci?c characteristics as well
macroeconomic factors, while severer events are caused solely by bank-speci?c factors.
In the baseline model without any additional macroeconomic variables, we indeed
?nd that – like most bank-speci?c covariates – GDP growth and in?ation enter
signi?cantly for distress categories I and II. In addition, the interest rate enters
signi?cantly positive in the sub-sample for distress category I. In contrast, there is no
impact from the macroeconomic sphere when considering events that comprise
interventions from the banking pillars head institution or the ?nancial supervisor into
the active business of a bank (distress category III). However, in the most severe
distress category, we ?nd a signi?cant impact of in?ation and, somewhat weaker, of
GDP growth. This indicates that the timing of restructuring mergers is in?uenced by
the macroeconomic environment (Blank et al., 2009).
We also check whether these results are in line with the speci?cations including
other macroeconomic factors. The most striking results are given for the model
including ?scal variables. For the distress category II, the results are very similar to the
full sample results. For distress categories I, both, GEXs and revenues are insigni?cant.
The only macroeconomic factor that drives the likelihood to observe an event assigned
JFEP
2,2
134
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
to distress category III is GEXs, which enters with a positive sign. One explanation
might be that the government could try to stimulate the economy when there is a
worsening of the business cycle. Hence, an increase in GEXs could be an indication for
a severe recession that increases the PD.
5.1.3 Results for different banking sectors. Next, we check the robustness of our
?ndings if we split up the sample according to Germany’s three banking pillars into
private commercial banks, savings banks, and cooperative banks. As the business
models of the different types of banks in Germany differ considerably, it is reasonable
to expect the different types of banks to be affected differently by certain
macroeconomic shocks. For example, cooperative banks may be less exposed to
international shocks, since these banks are mainly engaged on the domestic market.
In the baseline regressions without any additional macroeconomic variables, some of
the bank-speci?c variables as well the macroeconomic aggregates are insigni?cant. We
?nd the most stable results for cooperative banks, which drive the results for the full
sample, as these banks make up nearly 20,000 observations (out of roughly 28,000
observations in the full sample). However, the impact of the size of a bank nowswitches in
sign from being negative to positive. The sub-sample for savings banks comprises
approximately 6,000 observations. The only signi?cant in?uence from the
macroeconomic environment is given by the in?ation rate (INFL), which enters
negatively. In addition to in?ation, the interest rate is signi?cant with the expected
positive sign if we restrict the sample to private banks, which make up nearly 1,700
observations. Most of the banks-speci?c covariates turn out to be insigni?cant in this
speci?cation.
For the speci?cations with an enriched macroeconomic environment, the ?ndings
compared to the baseline regressions are more or less the same. However, some
differences are noteworthy. House prices have a negative signi?cant in?uence on the PD
in the sub-sample for cooperatives as well as in the sub-sample for savings banks. This
may be due to the fact that lending on mortgages is more important for these banking
sectors than for commercial banks. GEXs are positively signi?cant for the sub-sample
comprising savings banks. GRVs enter positively signi?cant in the sub-sample with
cooperative banks. However, the INFL now has a positive sign, thus increases the PD.
Although the in?uence of the nominal exchange rate is insigni?cant for the full sample
and for the different distress categories, we ?nd a negative signi?cant impact for both,
the sub-sample for the private banks and for the savings banks. We cannot ?nd a
signi?cant impact for cooperative banks, which are more concentrated on the national
economy and, thus, are less prone to valuation effects through exchange rate changes.
5.2 Responses to macroeconomic shocks in the combined framework
Having established the impact of the bank-speci?c and the macroeconomic variables
on the micro-level, we pursue and estimate the response of the PD in the VAR
framework described in Section 3.2. We do this for both, the pure macro VAR that does
not incorporate feedback effects from the macro-sphere to the ?nancial sector as well as
for the integrated model that allows for mutual macro-micro linkages. We also analyze
the response of the ?nancial stance to structural macroeconomic shocks in the various
sub-samples. Since the number of scenarios that we study is large, we report only a
collection of our results. As mentioned already in Section 3.2, we have to use different
versions of the extended model that include different sets of additional variables due to
The German
banking system
135
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
the low degrees of freedom. In Appendix, Table AII gives an overview over the
different shocks that we analyze and states which additional variables (analogously to
the micro-level estimations) we include in each case as well as which sign-restrictions
are used to identify the shocks[11].
5.2.1 Standard shocks in baseline model. The subsequent ?gures show annual
impulse responses to a one standard deviation shock. The solid line represents the
median of 500 simulations. As a benchmark, we replicate the ?ndings by De Graeve et al.
(2008) in the baseline model without any additional macroeconomic variables. To this
end, we analyze the response of the PD to a monetary policy shock. In Figure 2,
we show the impulse responses for the combined system.
A contractionary monetary policy, as given by a rise in the interest rate,
signi?cantly (on a 66 percent level) increases the aggregate PD from the ?rst year
onward, pointing to a potential con?ict of objectives between the ?nancial supervisor
and the European Central Bank. Note that like De Graeve et al. (2008) or Jacobson et al.
(2005), we ?nd that integrating the micro part of the model is very important for the
dynamics. Neglecting the impact of bank-speci?c variables leads to considerably
different responses for the PD. To check if these results are robust across different
sub-samples or whether we can observe some differences across sub-samples, we split
the sample and redo the impulse response analysis. We ?nd similar results for distress
category IV. In addition, the response of the PD is very pronounced when restricting
the sample to private banks. This may re?ect the fact that in contrast to savings banks
or cooperative banks those banks rely much more on re-?nancing via the capital
Figure 2.
Monetary policy shock
in the integrated VAR
GDP growth
Interest rate Distress frequency
Inflation rate
0.2
0.3 0.5
–0.5
–1
0
0.2
0.1
–0.1
0
–0.2
–0.4
–0.6
0 1 2 3 4
0
Note: Annual impulse responses to a one standard deviation shock. The solid lines
show the median impulse responses. The two dashed lines in each plot represent the
bounds of 66% confidence bands. For the sign restrictions used to identify each
shock see Appendix Table AII
1 2 3 4 0 1 2 3 4
0 1 2 3 4
0
0.1
–0.1
–0.2
–0.3
0
JFEP
2,2
136
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
market (as their deposit business is relatively small) and are, hence, directly affected by
a change of the yield curve that is usually implied by changes in the short-term interest
rate.
In addition to monetary policy shocks, the baseline model allows us to identify
aggregate demand and aggregate supply shocks. We model a supply shock by
restricting GDP growth to fall and in?ation to rise during the ?rst year. In the case of a
demand shock, we restrict GDP growth, in?ation, and the interest rate to fall on impact.
While we would expect the distress probability to rise in both recessionary
scenarios, we only ?nd the aggregate demand shock to have a signi?cant impact
(Figure 3), an aggregate supply shock is insigni?cant (Figure 4).
An analysis of sub-samples reveals that the distress probability of savings banks is
less affected by aggregate demand shocks – and the effect is only marginally
signi?cant. The distress probability of commercial banks is affected – but only for one
year after the shock hit the economy. The aggregate result seems to be solely driven by
the response of cooperative banks for which the reaction of the distress probability
looks like the one for the entire sample. Regarding the aggregate supply shocks, our
results show that they have no signi?cant impact for all of the sub-samples.
We now turn to the augmented models that allow us to identify a bunch of other
structural shocks in addition to the three basic macroeconomic shocks for which we
have presented results so far[12].
5.2.2 Equity price shocks. Next, we examine the consequences of an equity price
shock. Note that this shock can be understood as a combination of a immediate shock
to the banks’ balance sheets (because some of their assets are equity investments) and a
more precisely de?ned aggregate demand shock. The difference to the standard
Figure 3.
Aggregate demand shock
in the integrated VAR
0.2 0.05
0.4
0.2
–0.2
–0.4
–0.6
0
–0.05
–0.1
–0.15
–0.2
0
0.1
0.05
–0.05
–0.1
–0.15
0
0
Note: See Figure 2
1 2 3 4
0.1
–0.1
–0.2
–0.3
–0.4
–0.5
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate Distress frequency
The German
banking system
137
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
aggregate demand shock presented above is simply that we do not force aggregate
output to fall but set the sign-restriction in such a way that we specify a speci?c cause
for declining aggregate demand. As Figure 5 shows, we cannot ?nd any signi?cant
impact on the PD. This ?nding is neither reversed in the integrated model nor for any
of the sub-sample estimations.
5.2.3 Fiscal policy shocks. When we impose sign-restrictions to identify different
?scal policy shocks, the results depend to a high degree on how the ?scal policy shock
is designed. While a “pure government spending shock” (Caldara and Kamps, 2008)
does not signi?cantly affect the distress probability even in the integrated model, the
opposite is true for a “pure GRV shock” (Figure 6). A rise in GRVs leads to a signi?cant
increase of the distress probability after one year. An explanation is that an increased
tax burden leads to a higher insolvency rate among debtors that eventually causes the
frequency of distressed events to go up. A weaker, but still signi?cant impact can be
found for the sub-sample for cooperative banks as well as distress category II.
Instead of concentrating on just one side of the government budget, one could also
argue that what we will call a “contractionary ?scal shock” should be identi?ed by
decreasing GEXs and increasing GRVs. We ?nd that such an identi?cation scheme has
a pronounced in?uence on the distress probability in the integrated model (Figure 7).
The aggregate PD sharply and signi?cantly increases one period period after the
shock hits the economy. Again, we cannot detect such an impact in the pure macro
VAR. We further explore the role of the “contractionary ?scal shock” for ?nancial
stability and conduct the impulse response analysis for different sub-samples. For
distress category I and for cooperative banks, we qualitatively ?nd the same response
of the distress probability.
Figure 4.
Aggregate supply shock
in the integrated VAR
0.2
0.1 0.5
–0.5
0
0.05
–0.05
–0.1
–0.15
–0.2
–0.25
0
0.3
0.2
0.1
–0.1
–0.2
0
0.1
–0.1
–0.2
–0.3
–0.4
–0.5
0
0 1 2
GDP growth
Interest rate Distress frequency
Inflation rate
3 4
0
Note: See Figure 2
1 2 3 4 0 1 2 3 4
0 1 2 3 4
JFEP
2,2
138
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Figure 5.
Equity price shock in the
integrated VAR
0.4
0.2
–0.2
–0.4
0
0.2 5
0
–5
–10
0.1
–0.1
–0.2
0
0.4
0.2
–0.2
–0.4
0
0.2
0.1
–0.1
–0.2
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0
Note: See Figure 2
1 2 3 4
Interest rate
Distress frequency
Interest rate
Equity prices
GDP growth
Figure 6.
Pure GRV shock in the
integrated VAR
0.5 0.2
–0.2
0
0.2
10
–10
0
2
–2
0
–0.2
0
1
–1
0
–0.5
0
0 1 2 3 0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
4
0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Public expenditures Interest rate
Public revenues Distress frequency
Note: See Figure 2
The German
banking system
139
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
5.2.4 Real estate price shocks. When we impose the sign-restrictions to identify a real
estate price shock in the pure macro VAR, we cannot ?nd any signi?cant impact on the
PD (Figure 8)[13].
However, a real estate price shock signi?cantly increases the likelihood to observe a
distress event one period after the shock hits the economy if we shock the integrated
system (Figure 9). This signi?cant effect in the integrated model also holds if we
consider cooperative banks only, and, though somewhat weaker, if we restrict our
sample to private banks or to distress category II.
5.2.5 Exchange rate shocks. When we impose the sign-restrictions to identify an
exchange rate shock, we observe no signi?cant reaction of the distress probability –
neither in the pure macro VAR nor in the integrated model (Figure 10).
Also, the estimation based on the various sub-samples yields no signi?cant
response. The identi?cation of this shock is, however, a good example to demonstrate
the sensitivity of the results to the parameters chosen for estimation. We can for
instance impose the sign-restrictions for two instead of only one year. This might be
justi?ed by the fact that currency deprecations (appreciations) tend to occur in rather
persistent long movements. If we do so, it turns out that the distress probability rises,
though signi?cantly only two and three years after the shock hit the economy
(Figure 11).
It seems to be the case that banks get into trouble once the re-valuation of their
assets, that are denominated in foreign currency, causes write-offs due to a sustained
depreciation of the home currency.
Figure 7.
Contractionary ?scal
policy shock in the
integrated VAR
0.4
0.2
10 1.5
1
0.5
–0.5
0
5
–5
0
0.5
–0.5
–1
0
0.1
–0.1
–0.2
0
0.2
0.1
–0.1
–0.2
0
0.2
–0.2
–0.4
0
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate
Distress frequency Public revenues
Public expenditures
Note: See Figure 2
JFEP
2,2
140
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Figure 8.
Real estate price shock
in the pure macro VAR
0.5
0.2
–0.2
0
0.5
–0.5
0
0.5
–0.5
0
0.2
–0.2
0
–0.5
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate House prices
Distress frequency
Note: See Figure 2
Figure 9.
Real estate prices shock
in the integrated VAR
0.5
0.2
–0.2
0
0.5
–0.5
0
0.5
–0.5
0
0.2
–0.2
0
–0.5
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1
Note: See Figure 2
2 3 4
0 1 2 3 4
GDP growth Inflation rate
Interest rate House prices
Distress frequency
The German
banking system
141
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Figure 10.
Exchange rate shock
in the integrated VAR
0.4 0.3
0.2
0.1
–0.1
–1
–2
0
0
1 0.3
0.5
–0.5
0
0.2
0.1
–0.1
0
0.2
–0.2
–0.4
0
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth
Interest rate Nom. eff. exchange rate
Distress frequency
Inflation rate
Note: See Figure 2
Figure 11.
Exchange rate shock
in the integrated VAR
(longer restrictions)
0.6 0.3
0.2
0.1
–0.1
–1
–2
0
0
1 0.3
0.6
0.4
0.2
–0.2
0
0.2
0.1
–0.1
0
0.4
0
–0.2
0.2
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4 0 1 2 3 4
0 1 2 3 4
GDP growth
Interest rate Nom. eff. exchange rate
Distress frequency
Inflation rate
Notes: See Figure 2; in contrast to the other estimations we impose
sign-restrictions over two years instead of one year only
JFEP
2,2
142
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
6. Conclusion
In this paper, we have demonstrated how the model recently proposed by
De Graeve et al. (2008) can be enriched to allow for the analysis of impacts on the
banking system of a higher variety of structural macroeconomic shocks.
In sum, the empirical results show the following: ?rst, there is a close link between
macroeconomic developments and the stance of the banking sector. Second, monetary
policy shocks seem to be the most in?uential shocks for the development of the distress
indicator. Third, while also ?scal policy shocks and real estate price shocks have a
signi?cant impact on the distress indicator, the evidence is mixed for exchange rate
shocks. For equity price shocks we do not ?nd any impact. Fourth, for the identi?cation
of most shocks it is essential to work in the integrated model that combines micro and
macro evidences. Finally, we could identify some differences across sub-samples by
restricting the data for estimation to only those observations that corresponds to a
certain category of distress events or to certain types of banks. Most importantly, we
?nd that monetary policy shocks hit private banks much more severe than savings
banks or cooperative banks. In addition, we found that macroeconomic shocks tend to
increase the PDevents of category I or II more than they do for the more severe events of
categories III and IV. In general, however, the differences across sub-samples are smaller
than expected and do not show a systematic pattern.
These ?ndings lead to the following policy implications for ?nancial regulation
authorities; they show which developments should be closely monitored by them
because they increase the risk of ?nancial instability in the future. The results from the
baseline model showthat there might be a trade off between the proper monetary policy
stance and ?nancial stability; De Graeve et al. (2008) discuss potential intra-institutional
con?icts if a central bank is also involved in ?nancial regulation. The augmented models
showthat especially unexpected ?scal contraction can have strong impacts on ?nancial
stability; this suggests that the current stimulus packages around the world should be
removed slowly and according to a well-communicated time schedule. In addition, also
developments on the real estate market should be closely monitored because price drops
on that market can increase ?nancial stress directly via balance sheet effects and
indirectly via effects on aggregate demand. Finally, exchange rate movements should be
monitored; arguably, this should be done especially when judging the distress situation
of highly international connected banks with large assets positions denoted in foreign
currency. In sum, we conclude that ?nancial regulators should look at a broad range of
macroeconomic developments rather than focusing too narrowly on direct stability
measures like balance sheet data or current stress events. This is in analogy to monetary
policy, for which the experience of the recent years has shown that it is too short-sighted
to solely concentrate on the development of in?ation.
Our analysis suffers from the scarce data availability of distress events over the
time dimension. We expect most of the insigni?cant coef?cients to be due to the low
number of data points available for estimating the models, which impedes identifying
the true correlation structure of the data generating processes. In addition, the lack of
data makes an estimation of a “global” model, that includes all relevant variables
altogether, infeasible. However, it would be very interesting to address this issue in
future work with an extended dataset.
The German
banking system
143
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Notes
1. See International Monetary Fund (2008) for more information on those programs.
2. Note that the macroeconomic variables do not vary across the different banks. Hence, we
cannot include time ?xed effects in the regressions. As a consequence, the marginal effects of
the macroeconomic variables potentially re?ect to some degree the effects of unobserved
macroeconomic factors.
3. De Graeve et al. (2008) note that this feature makes the coef?cients state dependent and
allows for experiments that analyzes the system’s behavior for different levels of the bank
speci?c variables. While we do not pursue this issue in this paper and set P
MF
equal to its
value obtained for the sample averages of X
it
and Z
t
, an example of such an analysis can be
found in the paper by De Graeve et al. (2008).
4. Initially, it has been proposed to obtain these restrictions by choosing A
0
to be a Cholesky
factorization of S, implying a recursive ordering of the variables as in Sims (1986). This
method has been questioned on various grounds (Sims, 1992; Grilli and Roubini, 1996;
Christiano et al., 1998). Although, alternative identi?cation schemes like for instance the
approach proposed by Blanchard and Quah (1989) have been introduced in the past, none of
them has remained without criticism (Fernald, 2007).
5. For the speci?cs of this Bayesian estimation strategy, we refer to Appendix B in Uhlig
(2005).
6. De Graeve et al. (2008) note that in fact for some standard identifying restrictions like for
instance for a monetary policy shock, “the restrictions [in the sign-restriction approach] nest
the recursive (or Choleski) response.”
7. A more detailed discussion of the construction can be found in Kick and Koetter (2007).
8. Three subset databases of the distress database (measures, incidents, distressed mergers),
which contain information about exact dates, are used to allocate the distress events to
quarters to construct this series on a quarterly frequency; this is done to match the highest
frequency available for the macroeconomic variables described below.
9. The detailed results are collected in an appendix and available from the authors upon
request.
10. Alternatively, we could also adopt the method employed by Kick and Koetter (2007) and
estimate an ordered logit. However, we stick to this more intuitive approach.
11. In setting the identi?cation restrictions, we mainly follow various approaches presented in the
literature – among others by Uhlig (2005), Caldara and Kamps (2008), or Fratzscher et al. (2007).
12. Note that we will do not attribute shares of the variation of the distress probability to speci?c
shocks, since we do not identify all shocks in one model; consequently we cannot perform a
forecast error decomposition that takes into account all structural shocks.
13. Note that this shock can be understood as a more precisely de?ned aggregate demand shock.
The difference to the standard aggregate demand shock presented above is simply that we
do not force aggregate output to fall but set the sign-restriction in such a way that we specify
a speci?c cause for declining aggregate demand.
References
Allen, L. and Saunders, A. (2004), “Incorporating systemic in?uences into risk measurements:
survey of the literature”, Journal of Financial Services Research, Vol. 26 No. 2, pp. 161-92.
Bernanke, B.S. and Gertler, M. (1989), “Agency cost, net worth, and business ?uctuations”,
American Economic Review, Vol. 79 No. 1, pp. 14-31.
JFEP
2,2
144
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Bernanke, B.S., Gertler, M. and Gilchrist, S. (1999), “Financial accelerator in a quantitative
business cycle framework”, in Taylor, J.B. and Woodford, M. (Eds), Handbook of
Macroeconomics, Vol. 1, Elsevier, Amsterdam.
Blanchard, O.J. and Quah, D. (1989), “The dynamic effects of aggregate demand and supply
disturbances”, American Economic Review, Vol. 79 No. 4, pp. 655-73.
Blank, S., Buch, C.M. and Neugebauer, K. (2009), “Shocks at large banks and banking sector distress:
the banking granular residual”, Journal of Financial Stability, Vol. 5 No. 4, pp. 353-73.
Caldara, D. and Kamps, C. (2008), “What are the effects of ?scal policy shocks? A VAR-based
comparative analysis”, Working Paper Series 877, European Central Bank, Frankfurt am
Main, March.
Christiano, L., Eichenbaum, M.S. and Evans, C.L. (1998), “Monetary policy shocks: what have we
learned and to what end?”, Working Paper 6400, National Bureau for Economic Research,
Cambridge, MA, February.
Committee on the Global Financial System (2001), “A survey of stress tests and current practice
at major ?nancial institutions”, Report, Bank for International Settlements, Basel, April.
De Graeve, F. and Koetter, M. (2007a), Shock to the System: Real Shocks, Financial Shocks and
Financial Distress, Mimeo, University of Groningen, Groningen.
De Graeve, F. and Koetter, M. (2007b), The Role of Balance Sheet Adjustments in Transmitting
Monetary Policy to Financial Distress, Mimeo, University of Groningen, Groningen.
De Graeve, F., Kick, T. and Koetter, M. (2008), “Monetary policy and ?nancial (in)stability:
an integrated micro-macro approach”, Journal of Financial Stability, Vol. 4 No. 3, pp. 205-31.
Dovern, J., Meier, C.-P. and Vilmeier, J. (2008), “How resilient is the German banking system to
macroeconomic shocks?”, Kiel Working Papers 1419, Kiel Institute for the World
Economy, Kiel, May.
European Central Bank (2005), Financial Stability Report, European Central Bank, Frankfurt am
Main.
European Central Bank (2006), Financial Stability Review, European Central Bank, Frankfurt am
Main.
Fernald, J.G. (2007), “Trend breaks, long run restrictions, and contractionary technology
improvements”, Working Paper 2005-21, Federal Reserve Bank of San Francisco,
San Francisco, CA, June.
Fratzscher, M., Juvenal, L. and Sarno, L. (2007), “Asset prices, exchange rates and the current
account”, Working Paper 790, European Central Bank, Frankfurt am Main, August.
Gavin, M. and Hausmann, R. (1996), “The roots of banking crises: the macroeconomic context”,
Working Paper Series 318, Inter-American Development Bank, Washington, DC, January.
Goodhart, C.A., Sunirand, P. and Tsomocos, D. (2004), “A model to analyse ?nancial fragility:
applications”, Journal of Financial Stability, Vol. 1 No. 1, pp. 1-30.
Goodhart, C.A., Sunirand, P. and Tsomocos, D. (2006), “A model to analyse ?nancial fragility”,
Economic Theory, Vol. 27 No. 1, pp. 107-42.
Grilli, V.U. and Roubini, N. (1996), “Liquidity models in open economies: theory and empirical
evidence”, European Economic Review, Vol. 40 Nos 3-5, pp. 847-59.
Hosmer, D.W. and Lemeshow, S. (2000), Applied Logistic Regression, 2nd ed., Wiley, NewYork, NY.
International Monetary Fund (2008), “Financial Sector Assessment Program (FSAP)”, available
at: www.imf.org/external/NP/fsap/fsap.asp (accessed July 28, 2009).
Jacobson, T., Linde´, J. and Roszbach, K. (2005), “Exploring interactions between real activity and
the ?nancial stance”, Journal of Financial Stability, Vol. 1 No. 3, pp. 308-41.
The German
banking system
145
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Kick, T. and Koetter, M. (2007), “Slippery slopes of stress: ordered failure events in German
banking”, Journal of Financial Stability, Vol. 3 No. 2, pp. 132-48.
King, T.B., Nuxoll, D. and Yeager, T.J. (2006), “Are the causes of bank distress changing? Can
researchers keep up?”, Federal Reserve Bank of St Louis Review, Vol. 88 No. 1, pp. 57-80.
Kiyotaki, N. and Moore, J. (1997), “Credit cycles”, Journal of Political Economy, Vol. 105 No. 2,
pp. 211-48.
Sims, C.A. (1986), “Are forecasting models usable for policy analysis?”, Quarterly Review, Winter,
pp. 2-16.
Sims, C.A. (1992), “Interpreting the macroeconomic time series facts: the effects of monetary
policy”, European Economic Review, Vol. 36 No. 5, pp. 975-1011.
Sorge, M. (2004), “Stress-testing ?nancial systems: an overview of current methodologies”, BIS
Working Papers 165, Bank for International Settlements, Basel, December.
Uhlig, H. (2005), “What are the effects of monetary policy? Results from an agnostic identi?cation
procedure”, Journal of Monetary Economics, Vol. 52 No. 2, pp. 381-419.
Appendix
Correlation of the distress indicator with. . . . . .At lag 0 . . .At lag 1 . . .At lag 2 . . .At lag 4
GDP growth 20.37
* *
20.29 20.11 0.18
In?ation 0.11 0.04 20.09 20.41
* *
Policy rate 20.11 20.03 0.06 0.25
De?ator for construction 20.37 20.43
*
20.46
* *
20.42
*
Stock returns 20.50
* * *
20.49
* * *
20.48
* * *
20.02
Exchange rate 0.11 20.02 20.15 20.48
* * *
GEX 20.16 20.29
*
20.31
*
20.42
* *
GRVs 20.08 20.09 20.11 0.06
Notes: Signi?cance at
*
10,
* *
5, and
* * *
1 percent level, respectively; correlations are calculated over
the sample 1994Q1-2004Q3; for GRVs we excluded those observations that are distorted by the
revenues from the auction of UMTS licences
Table AI.
Cross-correlations
JFEP
2,2
146
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
M
o
n
e
t
a
r
y
p
o
l
i
c
y
s
h
o
c
k
A
g
g
r
e
g
a
t
e
s
u
p
p
l
y
s
h
o
c
k
A
g
g
r
e
g
a
t
e
d
e
m
a
n
d
s
h
o
c
k
E
q
u
i
t
y
p
r
i
c
e
s
h
o
c
k
G
E
X
s
h
o
c
k
G
R
V
s
h
o
c
k
C
o
n
t
r
a
c
t
i
o
n
a
r
y
?
s
c
a
l
s
h
o
c
k
R
e
a
l
e
s
t
a
t
e
p
r
i
c
e
s
h
o
c
k
E
x
c
h
a
n
g
e
r
a
t
e
s
h
o
c
k
D
G
D
P
2
2
2
?
?
?
?
?
?
I
N
F
L
2
þ
2
?
?
?
?
2
þ
I
R
A
T
E
þ
?
2
2
?
?
?
2
þ
P
D
?
?
?
?
?
?
?
?
?
A
d
d
i
t
i
o
n
a
l
v
a
r
i
a
b
l
e
s
D
D
A
X
X
X
X
2
X
2
X
X
X
D
G
E
X
X
X
X
X
2
?
2
X
X
D
G
R
V
X
X
X
X
?
þ
þ
X
X
D
C
O
P
X
X
X
X
X
X
2
2
X
D
N
X
R
X
X
X
X
X
X
2
X
2
N
o
t
e
s
:
A
“
2
”
i
n
d
i
c
a
t
e
s
t
h
a
t
w
e
i
m
p
o
s
e
a
d
e
c
r
e
a
s
i
n
g
o
f
t
h
e
c
o
r
r
e
s
p
o
n
d
i
n
g
v
a
r
i
a
b
l
e
t
o
i
d
e
n
t
i
f
y
t
h
e
s
h
o
c
k
,
l
i
k
e
w
i
s
e
a
“
þ
”
i
n
d
i
c
a
t
e
s
t
h
a
t
w
e
i
m
p
o
s
e
a
n
i
n
c
r
e
a
s
e
,
a
n
d
a
“
?
”
i
n
d
i
c
a
t
e
s
t
h
a
t
w
e
d
o
n
o
t
i
m
p
o
s
e
a
n
y
r
e
s
t
r
i
c
t
i
o
n
s
,
“
X
”
d
e
n
o
t
e
s
t
h
a
t
t
h
e
c
o
r
r
e
s
p
o
n
d
i
n
g
v
a
r
i
a
b
l
e
i
s
n
o
t
i
n
c
l
u
d
e
d
i
n
t
h
e
m
o
d
e
l
u
s
e
d
t
o
i
d
e
n
t
i
f
y
t
h
e
s
h
o
c
k
;
I
R
A
T
E
,
m
a
i
n
p
o
l
i
c
y
r
a
t
e
;
C
O
P
,
c
o
n
s
t
r
u
c
t
i
o
n
p
r
i
c
e
s
Table AII.
Analyzed shocks and
corresponding models
The German
banking system
147
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
About the authors
Sven Blank is a Research Assistant at the chair for “International Macroeconomics and Finance”
at the Eberhard-Karls Universita¨t Tu¨bingen. His research focus is on the link between ?nancial
markets and the macroeconomy as well as on international ?nance. Sven Blank is the
corresponding author and can be contacted at: [email protected]
Jonas Dovern is Founder and Director of Kiel Economics Research & Forecasting GmbH &
Co. KG, an enterprise that offers forecasting products and services as well as economic expertise
to the private and public sectors. Before, he was Researcher at the Kiel Institute for the World
Economy (IfW) where he was working for both the research area on monetary policy and the
business cycle forecasting center as an econometrician. Jonas Dovern studied econometrics at
Maastricht University and holds a PhD in economics from Kiel University. His research activities
cover different topics in macroeconomics, monetary policy, and econometric forecasting.
JFEP
2,2
148
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This article has been cited by:
1. Jarmo Pesola. 2011. Joint effect of financial fragility and macroeconomic shocks on bank loan losses:
Evidence from Europe. Journal of Banking & Finance 35, 3134-3144. [CrossRef]
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
3
8
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_308635051.pdf