Hartmut Brinkmeyer Auth. Drivers of Bank Lending New Evidence From The Crisis

Schriften zum europischen
Management
Herausgegeben von/edited by
Roland Berger School of Strategy and Economics
Academic Network
Mnchen, Deutschland
Die Reihe wendet sich an Studenten sowie Praktiker und leistet wissenschaftliche
Beitrge zur konomischen Forschung im europischen Kontext.
This series is aimed at students and practitioners. It represents our academic contri-
butions to economic research in a European context.
Herausgegeben von/edited by
Roland Berger School of Strategy and Economics
Academic Network
Mnchen, Deutschland
Herausgeberrat/Editorial Council:
Prof. Dr. Thomas Bieger Prof. Dr. Kurt Reding
Universitt St. Gallen Universitt Kassel
Prof. Dr. Rolf Caspers () Prof. Dr. Dr. Karl-Ulrich Rudolph

European Business School Universitt Witten-Herdecke
Oestrich-Winkel
Prof. Dr. Klaus Spremann
Prof. Dr. Guido Eilenberger Universitt St. Gallen
Universitt Rostock
Prof. Dr. Dodo zu Knyphausen-Aufse
Prof. Dr. Dr. Werner Gocht () Technische Universitt Berlin
RWTH Aachen
Prof. Dr. Burkhard Schwenker
Prof. Dr. Karl-Werner Hansmann Roland Berger Strategy Consultants
Universitt Hamburg
Prof. Dr. Alfred Ktzle

Europa-Universitt Viadrina
Frankfurt/Oder
Hartmut Brinkmeyer
Drivers of Bank Lending

New Evidence from the Crisis
With a foreword by Prof. Dr. Christoph J. Brner
Hartmut Brinkmeyer
Dsseldorf, Germany
Dissertation Heinrich-Heine-Universitt Dsseldorf D61 / 2014
ISBN 978-3-658-07174-5 ISBN 978-3-658-07175-2 (eBook)

DOI 10.1007/978-3-658-07175-2
The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;

detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.
Library of Congress Control Number: 2014949154
Springer Gabler
Springer Fachmedien Wiesbaden 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole
or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical
way, and transmission or information storage and retrieval, electronic adaptation, compu-
ter software, or by similar or dissimilar methodology now known or hereafter developed.
Exempted from this legal reservation are brief excerpts in connection with reviews or schol-
arly analysis or material supplied specifically for the purpose of being entered and executed
on a computer system, for exclusive use by the purchaser of the work. Duplication of this
publication or parts thereof is permitted only under the provisions of the Copyright Law of
the Publishers location, in its current version, and permission for use must always be obtained
from Springer. Permissions for use may be obtained through RightsLink at the Copyright
Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date
of publication, neither the authors nor the editors nor the publisher can accept any legal re-
sponsibility for any errors or omissions that may be made. The publisher makes no warranty,
express or implied, with respect to the material contained herein.
Printed on acid-free paper
Springer Gabler is a brand of Springer DE.

Springer DE is part of Springer Science+Business Media.
www.springer-gabler.de
Foreword V
FOREWORD
The recent financial crisis hook the banking system to its very foundations. While the
most acute phase of the crisis seems to be over, very challenging questions remain
unanswered. In their capacity as financial intermediaries, banks both generate profits
and contribute to social welfare by taking risks. Yet when the crisis revealed that
there may be strong incentives for them to go too far, they were forced to reduce
their risky positions in a very short space of time. This in turn, however, may result in
less social welfare, particularly in the context of banks' lending business. Lending is
the most significant source of both income and risk for the banking sector, but it is
also the one outcome of financial intermediation that carries the greatest social
importance. A number of studies have already analyzed the lending behavior of
banks during the crisis. However, only a few studies examine the characteristics of
banks and how they influence the supply of bank loans. Evidence for European
banks in particular is very scant.
Hartmut Brinkmeyer's dissertation contributes to this field of research not only on a

general level, but also with respect to individual euro area countries. His analysis
provides a wealth of detailed results. One broad finding is that significant
relationships exist between lending and bank characteristics. In particular, the level
and nature of influence differs between countries and between times of crisis and
normal times. While great care must as always be taken when interpreting these
results, they clearly deliver a profound insight into the lending behavior of European
banks. The findings of the study are the fruit of a well-founded theoretical framework.
To develop hypotheses, the author applies a wide range of theoretical approaches to
the transmission of monetary policy, nevertheless focusing primarily on the new view
of the bank lending channel. This modern theoretical approach is tested against a
proprietary set of data. The econometrical design deploys a number of remarkably
innovative ideas. First, the author implements a bank-specific, self-chosen target
capital ratio in which the capital structure of a bank is driven not only by general
regulatory rules, but by internal considerations as well. This approach enables
management decisions to be introduced in a sophisticated and realistic way. Second,
the study adopts a very convincing approach to the disentanglement of loan supply
and loan demand.
VI Foreword
While some of the findings may line up with expectations, others are surprising
indeed. The study explicitly urges academic and practical discussion; and I am
convinced that it will have a place in the ongoing discussion of how banks acted in
the crisis. My hope is therefore that this dissertation receives the attention it deserves.
Dsseldorf, April 2014
Christoph J. Brner
Acknowledgements VII
ACKNOWLEDGEMENTS
An undertaking such as writing a dissertation is a great challenge. Thinking of some

hard times I had while working on it, I can definitely confirm that. At the same time,
however, it has also been a very satisfying task, because I had the opportunity to
work on a subject that I am truly interested in. This was one of the best sources of
inspiration and motivation I could possibly have and that gave me the stamina and
discipline required to pursue and successfully accomplish the present work.
All this would not have been possible without support of some people who I would
like to acknowledge here.
First and foremost, I would like to express my profound gratitude to my thesis

supervisor Professor Christoph J. Brner who accepted me as his doctoral candidate.
I very much appreciate his style, his constructive guidance and his stimulation which
made it easy for me to keep going. It has been a real pleasure to work under his
supervision. I also want to thank Professor Ulrike Neyer who agreed to take the role
as co-supervisor. Her unpretentious and enthusiastic nature is truly admirable.
My employer, Roland Berger Strategy Consultants, gave me the opportunity to take

time off for my dissertation. During that time I received support in many different ways.
Hence I would like to express my gratitude to the partner team of the CC Financial
Services for nominating me for the company's PhD program and to Christian Krys for
organizing it as well as for a great number of helpful pieces of advice. Thanks also to
my colleagues Dirk Thiele and Sleyman Ertan for their support in accessing the
required data.
Finally, I want to thank those whose contribution was less related to content but even
more valuable and special: my family and especially my wife, Anne. Writing a
dissertation is not always easy. Without your support, encouragement and
understanding this undertaking would not have been possible and I would have never
come this far.
Dsseldorf, June 2014 Hartmut Brinkmeyer

Table of Contents IX
TABLE OF CONTENTS
LIST OF FIGURES .................................................................................................. XIII
LIST OF TABLES .................................................................................................... XV
LIST OF NOTATIONS AND ABBREVIATIONS .................................................... XVII
1. Introduction ....................................................................................................... 1
1.1. Motivation ..................................................................................................... 1

1.2. Research questions and contribution ........................................................... 2
1.3. Scope and limitations ................................................................................... 4
1.4. Organization of the research ........................................................................ 5
2. Transmission channels of monetary policy .................................................... 7
2.1. The money view ........................................................................................... 7

2.2. The credit view ........................................................................................... 10
2.2.1. The balance sheet channel ................................................................... 11

2.2.2. The bank lending channel Overview .................................................. 13
3. The bank lending channel in detail ................................................................ 18
3.1. Structure and elements of the bank lending channel.................................. 18
3.1.1. Condition 1: The central bank must be able to affect the supply
scheme of bank loans ............................................................................ 22
3.1.1.1. Subcondition 1: No complete adjustment to adverse monetary

policy shocks by the sale of securities/liquid assets ........................ 23
3.1.1.2. Subcondition 2: No access to non-deposit forms of funding
without additional cost ..................................................................... 24
3.1.1.3. Subcondition 3: Banks must not be capital constrained ................. 26
3.1.2. Condition 2: Publicly issued debt and non-bank intermediated loans

must not be perfect substitutes for bank loans ....................................... 28
3.1.3. Condition 3: Prices must not adjust instantaneously ............................. 32
3.2. Conclusion ................................................................................................. 34

X Table of Contents
4. A new view: Implications of financial innovation for bank lending ............ 35
4.1. The bank lending channel revisited ............................................................ 35

4.2. Toward a conceptualization of the new view .............................................. 40
5. Bank lending against the background of the recent crises ......................... 44
5.1. The loss spiral ............................................................................................ 45

5.2. The margin spiral or leverage cycle............................................................ 50
5.3. Conclusion ................................................................................................. 52
6. Review of empirical evidence on bank lending and its implications .......... 55
6.1. Remarks on the difference between the US and the euro area .................. 56
6.2. Empirical evidence from the US ................................................................. 58
6.2.1. US evidence based on aggregate data ................................................. 58

6.2.2. US evidence based on data from individual banks ............................... 59
6.2.3. Conclusion ............................................................................................ 67
6.3. Evidence from the euro area ...................................................................... 68
6.3.1. Euro area evidence from before the crisis ............................................ 69
6.3.2. Euro area evidence in the wake of the crisis ......................................... 73
6.3.3. Conclusion ............................................................................................ 76
6.4. Implications of theoretical framework for interpretation of empirical

evidence....................................................................................................... 77
7. Empirical analysis approach ....................................................................... 82
7.1. Research hypotheses................................................................................. 82
7.1.1. General hypotheses .............................................................................. 83

7.1.2. Hypotheses involving the context of the recent crisis............................ 85
7.2. Overall empirical strategy and approach .................................................... 89

7.3. Empirical model .......................................................................................... 91
7.3.1. Derivation of a model of bank behavior ................................................. 91

7.3.2. Introduction of the empirical model ....................................................... 96
7.4. Data............................................................................................................ 98
7.4.1. Data sources ......................................................................................... 98

7.4.2. Target capital estimation ..................................................................... 110
Table of Contents XI
7.4.3. Special challenges .............................................................................. 113
7.4.3.1. Disentangling loan supply and loan demand ................................ 113

7.4.3.2. Determining the relevant crisis period .......................................... 115
7.4.4. Purging the data.................................................................................. 119
7.5. Estimation method.................................................................................... 121
8. Empirical analysis results ......................................................................... 129
8.1. Results for the euro area .......................................................................... 129
8.1.1. Descriptive statistics and correlations ................................................. 131

8.1.2. Baseline analysis ................................................................................ 134
8.1.2.1. Results of the standard specification ............................................ 134

8.1.2.2. Robustness checks ...................................................................... 139
8.1.2.3. Summary of main results and relationship to existing literature.... 150
8.1.3. Capital surplus .................................................................................... 153

8.1.4. Deposit overhang ................................................................................ 161

8.2. Results for major euro area countries ...................................................... 170
8.2.1. Composition of the country samples and tested specifications ........... 171
8.2.2. Main results for Germany .................................................................... 174
8.2.3. Main results for Italy ............................................................................ 178
8.2.4. Main results for France ....................................................................... 181
8.2.5. Main results for Spain ......................................................................... 184
8.2.6. Discussion of inter-country differences and differences between
countries and the euro area ................................................................. 187
8.3. Conclusion Main research hypotheses confirmed ................................. 191

XII Table of Contents
9. Final discussion and implications ............................................................... 194
9.1. Overall summary of results ....................................................................... 194

9.2. Theoretical contributions .......................................................................... 197
9.2.1. Contributions to research regarding the determinants of bank lending 197

9.2.2. Contributions to bank lending channel-related research ..................... 199
9.2.2.1. General contributions ................................................................... 199

9.2.2.2. Monetary policy indicator .............................................................. 200
9.2.2.3. Disentanglement of loan supply and loan demand ....................... 201
9.3. Implications for bank management........................................................... 202

9.4. Implications for monetary policymakers ................................................... 205
9.5. Implications for the discussion of banking supervision ............................. 206
9.6. Limitations and outlook............................................................................. 207
References ............................................................................................................ 211
Appendix ............................................................................................................... 222

List of Figures XIII
LIST OF FIGURES
Figure 1.1: Organization of the research .................................................................... 5

Figure 2.1: Mechanism behind the balance sheet channel of monetary policy
transmission ........................................................................................... 12
Figure 3.1: The three conditions for the existence of a bank lending channel .......... 19
Figure 3.2: Structure of conditions and subconditions of the bank lending channel . 20
Figure 3.3: The three subconditions of the first condition of the bank lending
channel .................................................................................................. 22
Figure 4.1: The traditional view and a critique of the traditional view........................ 37
Figure 4.2: The new view on the bank lending channel ............................................ 38
Figure 4.3: Determinants of cost or ease of access to alternative forms of funding.. 41
Figure 4.4: Breakdown of subconditions into economic concepts driving bank
lending ................................................................................................... 42
Figure 5.1: The US federal funds target rate, 2000-2012 ......................................... 46
Figure 5.2: House price development in the US, 2000-2012 .................................... 47
Figure 5.3: Catalysts to the loss and the margin spirals ........................................... 48
Figure 5.4: The loss spiral and the margin spiral/leverage cycle .............................. 49
Figure 6.1: Financial structure in the US compared to the euro area ....................... 57
Figure 6.2: Systematization of bank characteristics as drivers of bank lending ........ 79
Figure 6.3: Three stylized cases of capital ratios against the background of
regulatory requirements, self-imposed targets and their implications
for expansion of the loan portfolio .......................................................... 81
Figure 7.1: Overview of formulated hypotheses ....................................................... 89
Figure 7.2: Eonia and 3-month Euribor rates .......................................................... 107
Figure 7.3: 3-month Euribor-OIS spread ................................................................ 108
Figure 7.4: The impact of selected events on the evolution of the 3-month
Euribor-OIS spread during times of financial turmoil/crisis ................... 116
Figure 7.5: Overview of steps in purging and cleansing the data ........................... 119
Figure 7.6: Properties of the difference GMM estimator and their applicability ....... 125
Figure 8.1: Structure of the empirical estimations................................................... 130
XIV List of Figures
Figure 8.2: Tested hypotheses based on euro area sample ................................... 131
Figure 8.3: Tested hypotheses based on individual country samples..................... 170
Figure 8.4: Government debt as a percentage of national GDP ............................. 188
Figure 8.5: Government deficit or surplus as a percentage of national GDP .......... 189
Figure 8.6: Summary of the results of the hypothesis test ...................................... 191
List of Tables XV
LIST OF TABLES
Table 6.1: Selected empirical research from the US on bank lending and the bank
lending channel ....................................................................................... 60
Table 6.2: Selected empirical research from the euro area on bank lending and
the bank lending channel ........................................................................ 70
Table 7.1: Sources of the variables used ................................................................. 99
Table 7.2: Description and construction of variables used in the regression .......... 100
Table 8.1: Descriptive statistics for the euro area sample ...................................... 132
Table 8.2: Correlation matrix for the euro area ....................................................... 133
Table 8.3: Results of standard specification ........................................................... 137
Table 8.4: Integration of non-standard monetary policy measures ......................... 139
Table 8.5: Eonia as a monetary policy indicator ..................................................... 141
Table 8.6: The 3-month Euribor-OIS spread as a monetary policy indicator .......... 142
Table 8.7: Loan demand proxied by results of the ECB bank lending survey ......... 143
Table 8.8: Estimation with time fixed effects ........................................................... 146
Table 8.9: Capitalization measured in terms of tangible common equity over
tangible common assets ....................................................................... 147
Table 8.10: Estimation including the share of mark-to-market securities ................ 149
Table 8.11: Results of estimations including capital surplus only ........................... 155
Table 8.12: Results of estimations including capital surplus and capital ................ 157
Table 8.13: Capital surplus estimation with time fixed effects................................. 159
Table 8.14: Results of estimations including the deposit overhang variable ........... 164
Table 8.15: Estimation including a deposit overhang dummy variable ................... 167
Table 8.16: Deposit overhang dummy variable and time fixed effects .................... 168
Table 8.17: Baseline results for Germany .............................................................. 175
Table 8.18: Capital surplus and deposit overhang results for Germany ................. 176
Table 8.19: Baseline results for Italy ....................................................................... 178
Table 8.20: Capital surplus and deposit overhang results for Italy ......................... 179
Table 8.21: Baseline results for France .................................................................. 182
Table 8.22: Capital surplus and deposit overhang results for France ..................... 183
XVI List of Tables
Table 8.23: Baseline results for Spain .................................................................... 185

Table 8.24: Capital surplus and deposit overhang results for Spain ....................... 186
List of Notations and Abbreviations XVII
LIST OF NOTATIONS AND ABBREVIATIONSi
CBPP Covered bond purchase programme
Coeff. Coefficient
e.g. Exempli gratia; for example
ECB European Central Bank
EMU Economic and Monetary Union of the European Union
Eonia Euro OverNight Index Average
et al. et alii
et seq. et sequens; and the following
etc. et cetera
Euribor Euro Interbank Offered Rate
FED Federal Reserve System
FN Footnote
GDP Gross domestic product
GMM Generalized method of moments
HICP Harmonized Index of Consumer Prices
mon. pol. Monetary policy
n.a. Not applicable / not available
no. Number
obs. Observations
OIS Overnight indexed swap
OLS Ordinary least squares
Std. dev. Standard deviation
Std. error Standard error
UK United Kingdom
US United States
i
Does not contain the variables used in the empirical section (see table 7.2)
1.1 Motivation 1
1. Introduction
1.1. Motivation
The recent crisis has presented a major challenge to banks, monetary policymakers
and the stability of the financial system as whole. The collapse of the investment
bank Lehman Brothers marked the starting point of a protracted crisis period that
went through different aspects and phases (e.g. the subprime lending crisis, banking
crisis, global financial crisis and sovereign debt crisis). The latter phases are still
ongoing.
Banks and monetary policymakers were impacted by the crisis in important and
connected respects. The banks' granting of credit one of the most important
functions of banks in the economy was temporarily threatened by serious
disruptions. Since the lending business is the most significant source of income to
the banking sector, the inability to supply credit does not only endanger profitability
but, even worse, it also poses an existential threat to almost any bank. When faced
with the crisis it took banks great efforts to prevent the worst consequences.
The subsequent challenge for monetary policymakers was based on the fact that one
transmission channel of monetary policy impulses works through banks and impacts
the supply of loans. Accordingly, the observation that the banks' ability to supply
credit was threatened by the crisis has called the effectiveness of monetary policy
and the achievement of its ultimate goal, i.e. price stability, into question.
However, only few studies address the question of which bank characteristics affect
the supply of bank loans, especially during the recent crisis, and the available
empirical evidence is relatively weak. What is missing is a systematic review of the
crisis and its mechanics that focuses on the issue of bank lending.
Another gap in current research exists regarding the analysis of possible differences
in the impact of certain bank characteristics on the supply of bank loans between
individual euro area countries. The integration of European financial markets and the
introduction of the euro as a single currency seem to have concentrated scholars'
focus on the euro area as a whole. However, current discussion of whether key
interest rates are appropriate for all euro area countries alike and the observation
H. Brinkmeyer, Drivers of Bank Lending, Schriften zum europischen Management,

DOI 10.1007/978-3-658-07175-2_1, Springer Fachmedien Wiesbaden 2015
2 1 Introduction
that the crisis developed differently in different countries, while only ranking as
anecdotal evidence, nevertheless points to the real existence of institutional and
economic disparities that should not be neglected.
The present study addresses these gaps. It aims to deepen our knowledge of those
bank characteristics that impact bank lending and the mechanics that play a role in
this process, especially in light of the recent crisis.
1.2. Research questions and contribution
Generally speaking during the recent crisis banks were particularly affected toward
the beginning.1 Although this most threatening phase of the crisis is over and despite
its severity and significance, studies devoted to analyzing the crisis with respect to
banks and bank lending are still very much underrepresented in relevant literature.
Only a very small number of such studies is available so far. The present dissertation
seeks to address this issue.
Overall, this research undertaking focuses on the determinants of the supply of bank
loans in the euro area especially during the crisis, and on their implications. The
basic idea of the study can be summarized by four main research questions:
x Which bank characteristics have an effect on the supply of bank loans?

x How did the impact of bank characteristics on lending change during the crisis?
o Which bank characteristics gained or lost influence?
o Which bank characteristics had no impact on lending before the crisis
but did play a role during the crisis?
x What are the implications for bank management and banks' business models?
x What other implications are of relevance to monetary policymakers and the
debate on macroprudential supervision?
In this context, the euro area as a whole comes under scrutiny, but so too do the four
most important euro area countries (Germany, Italy, France and Spain). Examining
those bank characteristics that, according to existing literature, have been proven to
affect bank lending, putting them in the context of crisis and analyzing their
differential impact is one aspect of the present study. At the same time, it also
1
The exact definition of the crisis period relevant in this study is discussed in section 7.4.3.2.
1.2 Research questions and contribution 3
considers two other largely unexplored economic concepts. The first is whether a
bank has a capital surplus or deficit relative to a bank-specific, self-chosen target. As
argued below, there are good reasons to assume that banks target individual capital
ratios. This being the case, it is natural to look at the impact on bank lending when
such a target is missed or exceeded. This goes beyond the conventional analysis of
"pure" capital ratios. The second is whether a bank is characterized by an overhang
of insured retail deposits over the amount of loans (a "deposit overhang"). Any such
overhang should make it easier for banks to fund their loan portfolios and other
assets, which is an important aspect in context of banks' funding strategies.
Special attention is given to identifying the crisis period that is relevant in this context.
The term "crisis" covers different aspects and phases, not all of which are equally
important to all the research questions. The most relevant aspects and phases are
those in which the banks were most seriously affected.
Another important issue regards the correct disentanglement of loan supply and loan
demand (the "identification problem"). When certain events impact on factors that
influence loan supply and loan demand at the same time, it becomes hard to
distinguish whether the change in the observable loan volume on banks' balance
sheets should be ascribed to supply-side or demand-side factors. Hence, a thorough
identification strategy is chosen to ensure correct identification. In a novel approach,
an attempt is also made to make use of information on loan demand contained in
answers gathered in the euro area bank lending survey.
By answering the research questions, this study contributes to the existing literature
in several ways: First, it deepens our understanding of the role of bank
characteristics, especially under crisis conditions, and allows implications to be
derived for the management of banks. This knowledge can help managers to
organize banks in a way that is more resilient to adverse economic conditions.
Second, a comprehensive framework into which all bank characteristics can be
integrated is derived from literature on the bank lending channel. This framework can
be used to show how the crisis altered the way in which bank characteristics affected
the supply of loans a finding that can be explained by the debt-deflation mechanism
and liquidity spirals. The framework also reflects a new, up-to-date view of the bank
lending channel that has not previously been presented in literature. Third, although it
is not a focus of the study, the fact that measures taken by the ECB during the crisis
4 1 Introduction
to restore the banks' ability to grant credit are accounted for in the empirical
estimations also permits an assessment of the effectiveness of these measures. This
information is valuable because the ability of banks to supply loans is an important
cornerstone of economic activity and prosperity. Fourth, this study has implications
for the debate surrounding macroprudential supervision.
1.3. Scope and limitations
This study focuses on the role of certain bank characteristics and their significance
for the supply of bank loans. It is particularly interesting to note how the impact of
these bank characteristics changed during the recent crisis relative to "normal"
periods, and to explore the implications this has for bank management. To find an
empirical answer to this question, a new framework is presented that accounts for
developments regarding the integration of European financial markets and the field of
financial integration, but that also captures the impact of the crisis on the role which
bank characteristics play in the context of bank lending.
Although the framework derives from literature on the bank lending channel, this
dissertation is not explicitly devoted to confirming the existence of a bank lending
channel. Nor is it primarily geared to contributing to the debate on macroprudential
supervision which seeks to answer the question of how not only individual financial
institutions but the financial system as a whole can be made more resilient to crises.
While not focusing on these adjacent fields of research, the findings of this study
nevertheless certainly do have implications for monetary policymakers and for the
debate on macroprudential supervision, over and above their implications for bank
management.
The geographical focus of this study is on the euro area. In addition to analyzing the
euro area as a whole, it also studies the four most important individual euro area
countries (Germany, Italy, France and Spain). Since the number of banks per country
is too small outside the four named ones, analysis of every euro area country is
prevented by concerns about the validity of the results,.
The temporal focus is clearly on the recent banking crisis. In this context, it is
important to note the period that was chosen for investigation: The sample period
begins with stage three of European Economic and Monetary Union (EMU) in 1999
when the common monetary policy was introduced. It cannot be ruled out that this
1.4 Organization of the research 5
event marks a structural break in the way bank characteristics impact bank lending
so that an earlier begin of the sample might have biased the results.
This study is subject to a few limitations that are primarily of a technical nature and
are caused by the character of the data. These are discussed in section 9.6 and
allow to identify potential areas for future research.
1.4. Organization of the research
The research structure is shown in figure 1.1 and consists of five major blocks. The
first block gives an introduction to the topic, presents the research question and sets
the scope (chapter 1). It is complemented by an introduction to the transmission
channels of monetary policy in chapter 2 that lays the basis for the theoretical
framework in the second block.
Introduction and Theoretical Literature review & appli- Empirical analysis

I theoretical background II framework III cation of new framework IV approach and results
MAIN > General introduction > Detailed description of bank > Differences between the US > Approach
CONTENT > Research questions, lending channel and the euro area Research hypotheses
contribution and scope > Implications of financial > Empirical evidence from the Empirical model, data
innovation for theoretical US and estimation method
> Introduction to the framework
transmission channels of > Empirical evidence from the > Results
monetary policy as > Implications of the recent euro area
theoretical background crisis Results for the euro area
> Implications of theoretical Results for selected euro
framework for interpretation area countries
of empirical evidence
CHAPTER Chapter 1 and 2 Chapter 3,4 and 5 Chapter 6 Chapter 7 and 8
Final discussion, > Summary, theoretical contributions and implications

V implications & summary Chapter 9
> Limitations and outlook
Figure 1.1: Organization of the research
The second block starts with a detailed description of the bank lending channel
(chapter 3), complete with all its conditions and subconditions. No state-of-the-art
theoretical framework would be complete without an account of the implications that
developments in financial innovation have had for bank lending and the bank lending
channel (chapter 4). This block also describes the theoretical foundation for the idea
that certain bank characteristics have a different impact on lending during the crisis
than they do in "normal" periods (chapter 5).
The third block comprises a review of relevant literature and is divided into empirical
evidence obtained for the US and for the euro area (chapter 6). It concludes with
6 1 Introduction
implications that derive from the theoretical framework for the interpretation of the
available empirical evidence: By developing the theoretical framework further and by
linking its economic concepts to the bank characteristics reviewed in literature, it
represents a new systematization of bank characteristics in their ability to impact loan
supply.
The fourth block the biggest one explains the empirical approach (chapter 7) and
presents the results (chapter 8). Part of the approach includes explicit formulation of
the hypotheses to be tested, the data sources and data handling method, and an
outline of the estimation methodology. Chapter 8 presents all results for the euro
area as a whole and for the four countries analyzed individually.
The final block (chapter 9) concludes with a summary of the results, the contributions
they make to the body of research and the implications they have in practice, before
looking ahead to possibilities for further research.
Although the various blocks differ in length, each one plays an important role. The
first block spells out the author's motivation for choosing the topic and tackling this
research undertaking. It also guides the reader regarding what to expect from the
present study and provides an introduction to the subject matter. The second block
builds a theory to prepare the ground for empirical analysis. The literature review in
the third block identifies the research gap, which is addressed in the fourth block, the
empirical analysis. The fifth block formulates the implications for bank managers,
monetary policymakers and the discussion of banking supervision. It also elaborates
on the contributions this study makes to the existing literature body and suggests
possible directions for future research based on the findings.
2.1 The money view 7
2. Transmission channels of monetary policy
There is a consensus among economists that the instruments of monetary policy are
able to generate real effects at least in the short run. The exact mechanism is still
the subject of controversial debate. Some time has passed since Milton Friedman
concluded that long and variable lags are involved in transmitting monetary policy
impulses (Friedman (1960), p. 87); yet the controversy has remained.
To shed some light on the question how the transmission of monetary policy works,
this chapter introduces the topic and outlines the most important transmission
channels.
This discussion then lays the basis for a detailed review of the bank lending channel
in the next chapter (chapter 3). That is important, because the bank lending channel
is a key tenet of the theoretical framework that is necessary to explain the
determinants of banks lending reactions.
2.1. The money view
The most widely shared view on monetary policy transmission can be summarized
under the heading "the money view". 2 The most important representative of the
money view is the traditional interest-rate channel, which explains the effect of
monetary policy on aggregate spending through changes in interest rates.
This mechanism is based on two key assumptions. The first assumption is that the
central bank can affect the short-term nominal interest rate. This is doubtless the
case, as empirically supported by Mojon (2000), for example. Control over the short-
term nominal interest rate enables the central bank to influence both the short-term
and long-term real interest rates. In seeking to understand the transmission from
nominal to real short-term interest rates, the key concept is "price stickiness". Due to
2
The systematization of the money view presented herein follows Mishkin (2010), chapter 26.
Different approaches are taken by Bofinger (2001) or Jarchow (2003), for example.

8 2 Transmission channels of monetary policy
factors such as menu costs and money illusion,3 the aggregate price level adjusts
slowly, with the result that an expansionary monetary policy shock lowers not only
the nominal but also the real short-term interest rate.4 The relationship between real
short-term interest rates and real long-term interest rates is established by a concept
called expectations theory and works as follows: In line with expectations theory, it
follows that real long-term interest rates are the average of expected future short-
term interest rates. For example, buying a bond with a maturity of one year, holding it
to maturity and then buying another bond with a maturity of one year should yield the
same expected return as a bond with a maturity of two years. Following the same
logic, different maturities can be regarded as substitutes for each other.
The second assumption of the money view is that investment and consumption
expenditures are sensitive to changes in the real interest rate. The more interest-rate
elastic both are, the greater is the impact of monetary policy stimulus. This is
especially plausible for long-term investments such as business fixed investment,
residential housing investment and consumer durable spending.
To sum up: Monetary policy makes use of its influence on short-term nominal interest
rates to affect long-term real interest rates. This precipitates a decline in the interest-
sensitive components of spending, especially those that are geared to long-term
considerations.
Three further prominent transmission mechanisms have also evolved under the
heading of the money view: first, the exchange rate mechanism, which is also
predicated on real interest-rate changes; second, the Tobin's q channel; and third,
the wealth mechanism (with the latter two both based on stock price values). All three
mechanisms are briefly sketched below.
The exchange rate mechanism of monetary policy transmission assumes that, if the
domestic real interest rate rises, domestic deposits will appear relatively more
3
Menu costs are the costs incurred in the change of prices. They include printing new price lists
(e.g. menus in restaurants), re-tagging items, updating systems and updating merchandise
material, etc. Money illusion refers to people's tendency to think of prices in nominal rather than in
real terms. Consequently, they do not adjust instantly to new real price levels (see Fisher (1928)
for the original reference to money illusion).
4
The discussion of the third condition required for the bank lending channel examines price
stickiness in more detail. See section 3.1.3.
2.1 The money view 9
attractive to investors than deposits held in a foreign currency. This leads to

increased demand for the domestic currency, and the domestic currency appreciates
relative to foreign currencies. Domestic goods thus become more expensive abroad,
while net exports are reduced. Since net exports are also a component of aggregate
spending, aggregate demand declines.
The Tobin's q channel (see Tobin (1969)) is based on the assumption that monetary
policy can affect the market valuation of a company's stocks. If monetary policy is
eased, more money flows into the stock markets, increasing the stock market value
of companies. The question is: Where is the connection to aggregate spending? The
necessary concept is the Tobin's q ratio, which is defined as the market value of an
enterprise divided by the replacement value of the enterprise's capital. Given a high
value for q, the market valuation of the company will exceed the replacement cost of
capital, thereby making it attractive to issue new stocks (equity) in order to finance
investments. An increase in investments also increases aggregate spending.
While the wealth mechanism (see Modigliani (1971)) is also based on stock prices,
this concept focuses on stocks as a component of private wealth. The basic premise
is that private individuals desire to smooth their periodic consumption 5 over time
depending not on current financial resources, but on lifetime financial resources.
Since stocks are a significant component of financial wealth, stock price movements
can affect lifetime financial resources and, hence, private consumption. The increase
in consumption positively affects aggregate spending.
These three mechanisms, together with the traditional interest-rate channel, add up
to the money view.
Criticism has been leveled at attempts to explain what are relatively substantial real
effects using the money view mechanism only. As pointed out above, the first
assumption is that the central bank can affect the short-term nominal interest rate.
The second assumption is that investment and consumption expenditures are
sensitive to changes in the real interest rate. Both assumptions have been
questioned mainly due to the fact that the relatively small impulses given by
monetary policy are not sufficient to explain the relatively large real effects as
explained, for example, by Bernanke and Gertler (1995) and Mishkin (1995). In
5
Consumption in this context excludes consumer durables expenditures.
particular, in order to explain the strong real effects using the direct interest-rate
channel only, the interest-rate changes effected by the central bank would have to be
much more pronounced than those that are observable, especially in light of the
relatively low interest-rate elasticity of investment. Furthermore, it is questionable
whether the empirically strong influence of the central bank on the demand for long-
lived assets can indeed be attributed to this channel: The power of the central bank
to influence long-term interest rates is observable, yes; but it is also limited according
to Bernanke and Gertler (1995).
These observations have led a number of authors, such as Gertler and Gilchrist
(1993), Bernanke and Gertler (1995), Cecchetti (1995), Hubbard (1995) and
Bernanke et al. (1996), to conclude that credit market imperfections too in addition
to the traditional interest-rate channel/money view must play a crucial role in
explaining the relatively large real effects stemming from relatively small monetary
policy impulses. This view, called the credit view of monetary transmission, is central
to the following section.
2.2. The credit view
As hinted at at the end of the last section, the beginnings of the credit view can be
traced back to some puzzling observations that could not be brought into line with the
conventional interest-rate channel view or money view of monetary transmission.
Most notably, the federal funds rate, over which the US Federal Reserve exercises
close control, is an overnight money market rate.6 One would therefore expect the
impact on (real) long-term interest rates to be relatively weak. Strikingly, however,
research on monetary policy has found a substantial impact of short-term rates on
aggregate demand, especially on long-lived assets such as housing or fixed business
equipment, which should in theory primarily be sensitive to long-term interest rates
(see e.g. Bernanke and Gertler (1995), Bernanke et al. (1996) and Peek and
Rosengren (2010)).
6
Reference is made here to the Federal Reserve and the federal funds rate simply because the
observations mentioned were first made in a US context. The credit view is not limited to the US,
of course, and its implications also hold for the euro area. Some differences between the US and
the euro area are discussed in the course of the available empirical evidence (section 6.1).
2.2 The credit view 11
The solution to the puzzle can be found in credit market imperfections, as myriad
papers emphasize (see Gertler and Gilchrist (1993), Cecchetti (1995), Bernanke and
Gertler (1995) and Bernanke et al. (1996), to name but a few). As pointed out by
Bernanke and Gertler (1989), the important insight is that deadweight losses occur
whenever there is an asymmetry in information between the borrower and the lender
relative to an equilibrium in a world of perfect information. These agency costs are
reflected in the difference between raising funds internally (e.g. through retained
earnings) as opposed to externally (e.g. by issuing equity or debt). The cost
differential is inversely related to the borrower's balance sheet position, especially the
borrower's net worth, and referred to as the external finance premium. The external
finance premium plays a role in both the balance sheet channel and the bank lending
channel, as will be outlined in the sections that follow.
According to Bernanke and Gertler (1995), the external finance premium reflects
three kinds of costs: first, the expected costs the lender has to bear for evaluating,
monitoring and administering the borrower; second, the costs of the typical "lemon's
premium" stemming from the fact that borrowers possess better information about
their financial position than lenders; and third, the expected costs associated with
moral hazard of the borrower.
When the European Economic and Monetary Union was established, the question of
the importance of the credit view attracted renewed attention in the euro area. As a
preliminary judgment of the ongoing debate it is fair to say this: In particular the
conclusion that the interest rate channel is not sufficient to explain the magnitude of
observable real effects also holds true for the euro area. Accordingly, there must be
some mechanism(s) at work that is (are) amplifying or complementing the
transmission of monetary policy impulses over and above what is explained by the
money view. These mechanisms are the balance sheet channel and the bank lending
channel.
2.2.1. The balance sheet channel
Central to the balance sheet channel is the concept of the external finance premium
which was introduced by Bernanke and Gertler (1989). The external finance premium
states that a borrower's cost of financing is inversely contingent on his financial
position, measured especially in terms of net worth but also in terms of liquidity and
current and future expected cash flows. The stronger the financial position of the
borrower, the more collateral he will be able to provide and the more he will be able
to bear his own losses. This fact strengthens the incentive not to act in a morally
hazardous way, but instead to do one's best to ensure favorable financial results,
because the borrower has more 'skin in the game' which he risks to lose (Bernanke
(2007)). This in turn makes investing in the borrower less risky, yielding more
favorable credit terms and lowering the overall cost of financing. Since it was first
introduced, the idea of an inverse relationship between a borrower's financial position
and the cost of credit has gained popularity. The concept has, for example, been
applied by Kiyotaki and Moore (1997), Bernanke and Gertler (1995), Bernanke et al.
(1999), Carlstrom and Fuerst (2001) and Iacoviello (2005), most notably in a
business cycle research context.
Unanticipated, adverse
Initial monetary policy shock
impulse (interest rate increase)
Direct effect Direct effect
Borrowing firm Customer 1 of borrowing firm
Worsened terms of floating-rate notes Customer

1 Worsened terms of2FRN
of borrowing
& ST debt firm
1
2 Declining asset and collateral values
or short-term debt Customer
1 Worsened terms of3FRN
of borrowing
& ST debt firm
Direct 2 Declining asset prices/values and 2 Declining asset and collateral values
1 Worsened terms of FRN & ST debt
result devalued collateral 2 Declining asset and collateral values
Financial position of borrower weakened Financial position of customers of borrowing

firm weakened
3 Declining demand by customers of Indirect effect

Indirect borrowing firm results in further decline of (from customers of borro-
result the financial position of borrowing firm wing firm to borrower)
Macro- Worsened conditions under which external finance becomes available (increase in
4
economic external finance premium) depresses interest-sensitive expenditures
result
Figure 2.1: Mechanism behind the balance sheet channel of monetary policy transmission
The research on business cycles referred to above used the concept of the external
finance premium to solve the puzzle of how relatively small, unanticipated monetary
changes can have substantial and persistent real effects: A change in interest rates
provoked by monetary policymakers an increase, say negatively impacts a

borrower's financial position both directly and indirectly (see figure 2.1).
Two distinct effects are immediately apparent (see Bernanke and Gertler (1995)).
First, to the extent that borrowers are financed using short-term (floating) debt, they
are directly affected by higher interest payments. This weakens their cash flows and
their overall financial positions and increases the external finance premium. Second,
a rise in short-term interest rates is usually accompanied by a decline in asset prices,
thereby diminishing the value of collateral and the creditworthiness of companies
through higher discount factors, again increasing the external finance premium.
Indirectly, there is also an effect on borrowing firms when customers of the borrowing
firms are themselves directly and negatively affected by an unanticipated monetary
policy shock, leading to a higher external finance premium for the customers of the
borrowing firm. As a result, the customer companies are likely to reduce spending on
goods or services from the borrower.
Ceteris paribus, this situation causes credit conditions to deteriorate and, ultimately,
leads to a higher cost of financing. This effect amplifies the traditional interest-rate
channel in the sense that the level of market interest rates which the borrower has to
pay is higher because the increased financing cost caused by the additional risk
premium places an even higher burden on spending and investment decisions,
thereby making even more marginal investment opportunities unprofitable. As a
consequence, aggregate spending, aggregate demand and real activity all slow
down.7
The balance sheet channel is one of two channels in the credit view. The following
section addresses the second of these two channels: the bank lending channel.
2.2.2. The bank lending channel Overview
In a study on the Great Depression of the early 1930s building on the work of
Friedman and Schwartz (1963), Bernanke (1983) suggests the existence of another
distinct mechanism by which monetary policy can have non-transitory real effects.
7
In the context of business cycle research, the fact "that endogenous procyclical movements in
borrower balance sheets can amplify and propagate business cycles" (Bernanke and Gertler
(1995)) is referred to as the 'financial accelerator'.
This idea was further developed by Bernanke and Blinder (1988) and explains how a
central bank's monetary policy can impact the supply of intermediated loans by
affecting banks' loanable funds. This leads to the bank lending channel, which
focuses on the role of banks in the propagation of monetary policy.
Despite the fact that non-bank financial intermediaries have gained importance in
recent decades and that firms today have access to alternative forms of funding via
public debt markets, banks still play the leading role in financing firms, especially in
the euro area. 8 The bank lending channel establishes the link between monetary
policy and banks and between banks, firms and real activity.
Since it examines the circumstances under which monetary policy can affect the
supply of bank lending, this is one of the key reference frameworks within the present
study. Before going into the details of the bank lending channel in the next chapter,
this section begins with a brief overview of the mechanism at work.
According to the traditional view of the bank lending channel, an increase in key
interest rates by the central bank worsens the terms on which banks can equip
themselves with reserves by the central bank. This has consequences for banks
because reserves are always linked to deposits: Banks are required to hold a certain
percentage of (insured) deposits as central bank reserves.9 As a result, the shortage
of or "drain" in reserves limits the banks' ability to create deposits and, at the same
time, to grant loans, because granting a loan means creating a deposit on the
account of the borrower. Therefore, reducing the availability of reserves impairs the
ability to provide loans.
There is another mechanism through which the central bank impacts the availability
of deposits as a funding source for banks. This mechanism works by affecting the
yields on deposits relative to other assets Disyatat (2011). It can be explained by the
motives for holding money.
8
Arguments substantiating the view of an intact bank lending channel despite the increased
significance of non-bank financial intermediaries and wholesale funding markets are outlined in
section 4.1.
9
The reserve requirement is 1% of all covered liabilities in the euro area (it was 2% until the end of
2011) and 10% in the US. It must be noted, though, that in the euro area the range of liabilities
covered by reserve requirements is larger than in the US. For more details on reserve
requirements, see Bofinger (2001), p. 343 et seq.
Excursus: motives for holding money
According to Keynes' famous disquisition (Keynes (1936), chapter 15), individuals

hold money as deposits for three reasons: for transaction purposes, out of a
precautionary motive and for speculative reasons.
First, the transaction motive reflects those parts of liquidity that are held in order to
make immediate purchases of goods or services (such as food, rent, electricity, etc.).
The desired money is mostly held in the form of demand deposits (rather than as
cash). The quantity of money designated for transaction purposes depends on
periodic income, but does not hinge on the level of interest rates.
Second, because individuals face expenses whose amount and probability of

occurrence is uncertain they hold money out of a precautionary motive to cover these
uncertain events (e.g. repairs, replacement purchases). Some textbooks group the
precautionary motive together with the transaction motive and do not regard the two
items separately. The reason for this is that the amount of money held for
unanticipated events stemming from a precautionary motive can be thought of mainly
as a function of the amount money spent on transactions and, therefore, ultimately as
a function of periodic income.
Third, money is held for speculative reasons. In order to illustrate the logic suppose
the simple case that economic agents have the choice between holding money in
form of non-interest bearing deposits and a non-maturing bond.10 The revenue from
the bond, i, arises immediately from the quotient of the annual interest payment, K,
and the purchase price, P: i = K/P. In addition, Keynes makes the assumption that
every individual makes some assessment of what is the "normal" level of interest
rates. The relation between the actual interest rates and the assessment regarding
their normal level can be used to reflect the incentive structure for the individual
(assuming constant interest payments): If the actual level of interest rates is below
the level that is considered to be normal, then rising interest rates and falling bond
prices will be expected in future. In this case, individuals hold money as speculative
accounts. Conversely, if the actual level of interest rates is above the level that is
considered to be normal, economic agents will expect falling interest rates
10
The simplifying assumption of a non-maturing bond is only made in order to be able to abstract
from repayment effects. Obviously, no conclusion is dependent on this assumption.
accompanied by rising bond prices. In this situation, purchasing a bond will be

profitable not only because of the bond price gain. Over and above this gain, a high
return is expected due to the low purchase price. Ceteris paribus, less money is
therefore held in the form of deposits owing to the increased opportunity cost faced
by individuals if they hold money in a non-interest bearing form as opposed to
interest-bearing bonds. Ultimately, the amount of money individuals hold as
speculative accounts correlates inversely to the level of interest rates.
To sum up: The amount of deposits available to banks is a function of (monetary

policy) interest rates due to the opportunity cost that individuals face as a result of the
speculative motive.
Coming back to the bank lending channel mechanism: In response to a shortfall in

funding on the liabilities side of the balance sheet, banks are forced to reduce assets
or to replace lost deposits with alternative forms of funding (liabilities).
Since adjusting the balance sheet solely on the asset side (e.g. by selling securities;
see detailed explanation below) is a suboptimal approach, banks will also adjust their
liabilities. This is where the concept of the external finance premium comes into play:
To the extent that lost deposits are replaced by alternative, uninsured forms of
funding (e.g. wholesale funding), banks, acting as borrowers in wholesale funding
markets, are exposed to the external finance premium which, in turn, depends on the
overall financial condition of the bank in question.
Generally speaking, because deposits are subject to deposit insurance schemes,

they are the only source of funding to banks that is not subject to asymmetric
information and moral hazard problems (see Stein (1998)). 11 Any other form of
uninsured funding potentially implies asymmetric information and moral hazard
issues, all of which involve an external funding premium.
11
Depending on how deposit insurance schemes are implemented in individual jurisdictions, the
contributions that banks have to make to the respective deposit insurance funds may be risk-
related (e.g. in the US since 1993). However, although this might constitute an incentive to reduce
risk from a bank's perspective (since the contribution to the insurance fund increases as the level
of risk increases), this has no impact on the customers' propensity to provide (insured) deposits.
Therefore, the issue of whether a bank's contribution to deposit insurance funds is risk-related or
not has no direct consequences for the significance of the asymmetric information problem.
Because external funding is always more expensive due to the funding premium that
investors demand, a bank will not be able to completely offset the reduction in its
deposits by other forms of funding without additional cost. By consequence, a bank
must reduce its assets in the same proportion as lost deposits cannot be replaced.
As explained in greater detail in the next chapter, the optimal approach is to always
hold a certain fraction of assets in the form of liquid assets; this is what makes banks
reluctant to simply sell off liquid securities. Hence, banks also have to reduce the
volume of loans they grant. To the extent that borrowers that need money from banks
cannot replace bank loans without additional cost with credit from other sources, this
impacts aggregate spending.
In a nutshell, the bank lending channel describes how the central bank can influence
the real economy by exerting an influence on the supply of intermediated loans via
the availability deposits.
Following on from this brief introduction to the main transmission channels of

monetary policy, the next chapter focuses specifically on the bank lending channel.
This channel is central to the current research undertaking, as it is a key tenet of the
theoretical framework that explains lending reactions by banks.
18 3 The bank lending channel in detail
3. The bank lending channel in detail
The bank lending channel focuses on the transmission of monetary policy actions via
banks and bank lending and is, therefore, central to the purpose of this study. The
concept of the bank lending channel is a key cornerstone of the theoretical
framework which is used as a basis to explain the determinants of bank lending in
the empirical section of this study. It is therefore necessary to review in detail both
the mechanism itself and the conditions under which the bank lending channel is
active.
The following sections focus on the theoretical foundations of the bank lending
channel: first, by laying down the mechanism of the bank lending channel, and
second, by discussing in detail the conditions and subconditions that must be fulfilled
if it is to function.
In order to briefly pre-structure the further course of argumentation it should be noted

that the following sections address what can be called the "traditional view" on the
bank lending channel. With regard to certain aspects it is indicated to develop it
further to what this study calls the "new view". Without touching the main essence of
the bank lending channel the new view provides an enhanced interpretation against
the background of today's operational frameworks of major central banks and recent
developments in financial markets. The new view is dealt with in a separate chapter
(chapter 4).
3.1. Structure and elements of the bank lending channel
The first element of the mechanism describes how the central bank influences the
supply of bank loans by controlling both reserves and the ability to create deposits.
The second element describes firms' dependence on bank-intermediated loans. Both
elements are reflected in the first two conditions that must be met to create an active
bank lending channel. On top of these conditions, there is also the general condition
that prices must not adjust instantaneously subsequent to monetary policy changes.
To establish a comprehensive understanding of the bank lending channel, it makes
sense to introduce the conditions required for its distinct existence as a separate

3.1 Structure and elements of the bank lending channel 19
transmission channel step by step (see figure 3.1 for a brief description of all three
conditions).
Condition 1 Condition 2 Condition 3

Central bank must be able to Publicly issued debt and non-bank Prices must not adjust
affect supply scheme of intermediated loans must not be instantaneously (monetary
bank loans perfect substitutes for bank loans policy must not be neutral)
> Banking sector as a whole > It must not be the case that > Frictionless price adjustments
must not be able to or willing to firms are able to offset the would imply that policy rate
completely insulate lending decline in the supply of bank changes immediately translate
portfolio from monetary policy loans completely, i.e. without into price adjustments of an
Explanation shocks, either by incurring additional cost, by equal proportion
switching from deposits to borrowing more directly from > This must not be the case
other forms of funding or household sector in public
selling securities (liquid markets
assets)
Figure 3.1: The three conditions for the existence of a bank lending channel
One of the most famous early descriptions of the lending channel and its conditions
was penned by Bernanke and Blinder (1988). Another illustrative formulation can be
found in Kashyap and Stein (1994). Essentially, the bank lending channel requires
three conditions: The first condition states that central banks must be able to affect
the supply of bank loans. The second condition is that publicly issued debt and non-
bank intermediated loans must not be perfect substitutes for bank loans. The third
condition is that prices must not be adjusted instantaneously subsequent to monetary
policy changes, resulting in the notion, as touched on earlier, that monetary policy is
not neutral. The first two conditions can be broken down further into subconditions
(see figure 3.2 for an overview of the structure of conditions and subconditions.
These subconditions are phrased in such a way that their fulfillment warrants the
fulfillment of the governing condition.) 12
12
As presented here, the order of the first two conditions is swapped compared to the common
approach found in literature, as this order better captures the sequence of steps in the bank
lending channel mechanism from the central bank via banks to the real economy. A further
modification to the customary manner of presentation concerns the subconditions under
condition 2: The usual approach is to discuss whether bank loans can be substituted by non-bank
intermediated loans in connection with condition 1 (e.g. Bernanke and Blinder (1988) and Kashyap
and Stein (1994)). However, the author believes that this issue slots more naturally into the
discussion in connection with condition 2: Whether a firm resorts to bonds or non-bank
intermediated loans is effectively the same thing insofar as, in both cases, bank loans are avoided.
For this reason, the wording of the conditions too differs slightly compared to the references stated.
As regards the first condition, the basic requirement is that the central bank must be
able to affect the supply of bank loans. Let us consider a tightening of monetary
policy, in the course of which the central bank increases key interest rates. For banks,
this immediately means that the terms under which central bank reserves are
available deteriorate. This is important insofar as banks are required to hold a certain
proportion of deposits as reserves. As a consequence, banks are therefore limited in
their ability to create new (reservable) deposits.13
The bank lending channel conditions
Condition 1 Condition 2 Condition 3
Publicly issued debt and non- Prices must not adjust

Central bank must be able
bank intermediated loans instantaneously
to affect supply scheme
must not be perfect substitutes (monetary policy
of bank loans
for bank loans must not be neutral)
Subcondition 1 Subcondition 1
No complete adjustment to adverse Firms must not be able to
monetary policy shocks by saleI) of completely replace bank loans by
securities/ liquid assets publicly issued debt (e.g. bonds)
Subcondition 2 Subcondition 2
Banks must not have access to Firms must not be able to
non-deposit forms of funding completely replace bank loans by
without additional cost non-bank intermediated loans
Subcondition 3
Banks must not be capital

constrained
I) The opposite holds in case of expansionary monetary policy shocks: Banks must not react to interest rate cuts completely by buying securities/ liquid assets
Figure 3.2: Structure of conditions and subconditions of the bank lending channel
Banks can react to this reduction in deposits as a funding source in several ways
(see Peek and Rosengren (2010), p. 261, for example). In principle, they can make
the necessary adjustments to rectify this imbalance on either the asset side or the
liabilities side of the balance sheet. Attempts to replace deposits by alternative forms
of funding (on the liabilities side) involve an external finance premium, because this
kind of funding is, unlike insured deposits, subject to issues of asymmetric
information. This makes the replacement of deposits by other forms of funding costly.
13
Note that the granting of a credit is associated with the creation of a deposit on the account of the
beneficiary which is subject to reserve requirements.
On the asset side, banks can either reduce loans or securities or both to countervail
the drained deposits. Since the optimal approach is to always hold a certain portion
of assets in the form of deposits, banks will never perform the adjustment entirely by
selling securities, as is discussed below. In other words, banks will offset part of the
drain on deposits by reducing lending by cutting back on their loan supply.14
This mechanism is what gives the central bank the power to influence the supply of
bank loans. This first condition and the subconditions are discussed below in greater
detail.
A second condition of the bank lending channel is that publicly issued debt (e.g.
bonds) and loans by non-bank financial intermediaries must not be perfect
substitutes for bank loans, at least for some firms. In other words, firms must not be
able to circumvent a decline in bank loans by resorting to other sources of funding.
As hinted at by the formulation of the condition itself, there are two subconditions:
The first subcondition is the imperfect substitutability of bank loans and publicly
issued debt. The second is the imperfect substitutability of bank loans and loans by
non-bank financial intermediaries. If both subconditions hold, then the reduced
supply of bank loans will carry over to the real economy. This second condition and
the subconditions are also discussed in greater detail below.
The third condition of the bank lending channel relates to imperfect price adjustments,
in the sense that prices are sticky in the short run. Price rigidities are what enable
monetary policy to achieve real effects. Without this condition, monetary policy
shocks and the changes in reserves associated with them would immediately affect
prices and would not have any real effects. Monetary policy would then be pointless.
This is thus a general condition that is necessary for any channel of monetary policy
transmission.
Since an overwhelming body of literature is devoted to imperfect price adjustments,

and since this issue is not unique to the bank lending channel, the section on this
third condition concentrates on the basic mechanism and certain historical aspects of
price rigidities.
14
The theoretical derivation of this proposition is commented on below in the course of the
discussion of subcondition 1 of condition 1 (section 3.1.1.1).
To sum up: The theoretical mechanism of the bank lending channel establishes two
main links. The first is from monetary policy actions to banks and bank lending. This
link is reflected in the condition that the central bank must be able to affect the supply
scheme of bank loans. The second link is the connection between bank lending and
the real economy. This connection is captured by the condition that, at least for some
firms, publicly issued debt (e.g. bonds) and non-bank intermediated loans must not
be perfect substitutes for bank loans.
Only if neither of these two links is interrupted and if both conditions hold will the
mechanism of the bank lending channel apply in practice. In addition, prices must not
adjust instantaneously to changes in the monetary environment. Without this general
precondition monetary policy would not exert any influence on the real economy,
irrespective of the transmission channel considered.
The section below further elaborates on the conditions and subconditions that are
required for the mechanism to take effect.
3.1.1. Condition 1: The central bank must be able to affect the supply scheme of
bank loans
Subcondition 1 Subcondition 2 Subcondition 3

No complete adjustment to ad- No access to non-deposit
Banks must not be capital
verse monetary policy shocks by forms of funding without
constrained
saleI) of securities/ liquid assets additional cost
Necessary con- > Banks must not fully compensate > Banks must not be able to com- > Banks must not fail to expand
dition for central the drain in deposits resulting pensate the drain in deposits by lending due to too low capital
banks being able to from a monetary policy tightening other non-insured forms of funding ratios (induced by regulator or
affect loan supply by selling securities without additional cost self-induced )
> Banks constantly face a certain > Investors face an asymmetric > Under the Basel Accords and
probability of random deposit information and adverse selection under national legislation banks
withdrawals (particularly after a problem when lending uninsured need to hold a certain fraction of
change of monetary policy rates) funds to banks loans as capital
> In order not to be forced to > This requires a premium that > Banks for which such a capital
Theoretical liquidate loans on short notice in banks have to pay which is requirement is binding cannot
substantiation face of random withdrawals banks associated with the underlying expand their loan supply in case
hold a certain amount of securities banks' credit risk of monetary policy eases (interest
Is the condition > The optimal amount is subject to > By this, non-deposit forms of rate decrease)
likely to hold? the banks' risk appetite funding are imperfect substitutes > The same logic applies when a
> Hence banks will not completely for deposits self-chosen capital target is
compensate a drain in deposits by > The same logic applies to equity missed
selling securities since this would capital which is another potential
reduce liquidity below the amount form of funding
chosen to be optimal
I) The opposite holds in case of expansionary monetary policy shocks: banks must not react to interest rate cuts completely by buying securities/ liquid assets
Figure 3.3: The three subconditions of the first condition of the bank lending channel
The first condition is the most important one in the context of this study, because it
concerns itself with how banks respond to monetary policy shocks in terms of their
lending. It can be broken down into the three subconditions summarized together
with the underlying theoretical mechanism in figure 3.3. Fulfillment of all three of
these subconditions is tantamount to fulfillment of the first condition.
3.1.1.1. Subcondition 1: No complete adjustment to adverse monetary policy

shocks by the sale of securities/liquid assets
The first subcondition is related to the asset side of the balance sheet and to banks'
holdings of liquid assets (e.g. securities). The requirement is that banks do not fully
compensate for the drain in deposits when monetary policy is tightened by selling
liquid assets. If banks did adjust on the asset side solely by selling liquid assets, the
loan portfolio and, hence, loan supply would not be affected at all. The question is
therefore: Is it theoretically plausible for banks to adjust not simply by selling liquid
assets, but, instead, to also adjust via the more profitable loan portfolio, thereby
incurring lost profits?
Tobin (1982) formulates a simple model for profit-maximizing banks that have to
choose between loans and other relatively illiquid investments on the one hand and
liquid assets on the other hand, not knowing beforehand the volume of deposits that
will be available (stochastic flow of deposits). A volume of deposits lower than
expected requires a bank either to sell (liquid or illiquid) assets or to borrow from
alternative sources of funding. Both instantaneously trying to reduce illiquid assets
such as loans (see Blinder (1984)) and borrowing from alternative funding sources
(see King (1986)) are typically more costly options than reducing holdings of
comparatively liquid assets.
Douglas and Rajan (2001) point in the same direction: Acting as a lender, a bank will
not be able to redeem the present value of a loan by selling it when it needs to
generate liquidity, but will instead lose money.
Therefore, in order to protect against instantaneous loan reductions becoming

necessary because of random deposit outflows, banks will always hold a certain
portion of liquid assets. The exact portion is determined by a bank's expectation
regarding the available deposit volume and its appetite for or aversion to risk
regarding the volatility of deposit outflows. 15 As a consequence and this is the

important point a bank will respond to a deposit outflow by also curtailing lending
not in the short-term but in the medium term in order to maintain the desired ratio of
deposits to liquid assets. It does this to ensure that it does not directly have to reduce
loans or agreed credit lines for existing bank customers.16
The same logic applies when monetary policy is eased: In this case, banks will react
to the excess in available deposits not only by increasing their lending, but also by
holding an additional share of liquid assets.
Do these theoretical considerations fit actual economic circumstances? Kashyap and

Stein (1994 and 2000) provide data on the share of securities held (as a percentage
of total assets) by large, medium-sized and small banks.17 For all three size classes,
the data shows consistent patterns of portfolio composition that are in line with
expectations: Large banks' share of securities is much smaller than that of small
banks. Since small banks are more vulnerable to random large-scale deposit
withdrawals (e.g. for diversification purposes), they have to hold a larger share of
securities than large banks. The persistency of this pattern suggests that these kinds
of portfolio compositions are not the expression of randomly chosen levels of liquidity.
3.1.1.2. Subcondition 2: No access to non-deposit forms of funding without

additional cost
The second subcondition is related to the substitutability of deposits by alternative

forms of funding. It states that banks must not be able to fully compensate for the
deposit outflow by other forms of external funding, e.g. on wholesale funding markets.
In other words: Alternative forms of funding must not be perfect substitutes for
deposits from a bank's perspective in order for this subcondition to be valid. While
15
In fact, as also pointed out by Tobin (1982), the desire of banks to hold liquid assets is an example
of the precautionary motive for holding money in accordance with Keynes (see section 2.2.2).
16
As explained by Bernanke and Blinder (1992), loans are quasi-contractual agreements that cannot
easily be drawn on to reduce assets. This is reflected in the response pattern to monetary policy
shocks: Banks' short-term reaction to reduced deposits is to sell securities. In the medium to long
term, a reduction in loans is the dominant effect.
17
Large banks are those in the 1st percentile of total assets in their sample. Medium-sized banks are
those from the 99th through the 75th percentile. Small banks are those below the 75th percentile.
Romer and Romer (1990) stated that banks do, indeed, have the capability to easily
replace drained deposits by alternative funding, important aspects of this view have
been criticized. As pointed out by Kashyap and Stein (1994), the idea of Romer and
Romer is posited on the assumption that demand in wholesale funding markets is
perfectly elastic. This means that, following a drain in deposits when monetary policy
is tightened, banks can issue as large a volume of open-market liabilities as needed
or desired without the volume affecting their costs at all. This is not very likely to be
realistic, for the following reasons.
Unlike deposits, wholesale funding is not insured. Wholesale funding includes the
(secured or unsecured) borrowing in money markets and the issuance of short-term
debt, commercial papers, for example, but also comprises longer-term debt such as
bonds and covered bonds (see e.g. van Rixtel and Gasperini (2013)). The fact that it
is not insured means that lenders (to banks) must concern themselves with the
creditworthiness or riskiness of the borrowing bank. In this context, the concept of the
external finance premium, first introduced by Bernanke and Gertler (1989), plays a
key role. The external finance premium is the spread between the cost of raising
funds externally and the opportunity cost of raising them internally. Due to standard
asymmetric information and moral hazard problems and the resulting costs that
lenders have to bear to monitor the borrower's actions, the external finance premium
is always positive. Moreover, the premium that outside investors demand for the
funding they provide should be a negative function of the borrower's financial position,
as Carlstrom and Fuerst (2001) demonstrate by building on a model of Kiyotaki and
Moore (1997). The financial position can be thought of as proxied by net worth,
liquidity and current and future cash flows (see Bernanke (2007)). For "healthy"
borrowers there is more at stake, which makes it less likely that they will act in a way
that causes the lender to lose money. It thus becomes clear that the amount of
external financing is, generally speaking, limited by the borrower's financial health, i.e.
by the value of his net worth or collateral. As a consequence, the cost of external
financing is indeed a function of its volume, which Kashyap and Stein (1994) show
formally in a simple partial-equilibrium model. 18 While these arguments are more
closely geared to debt financing, the same logic obviously also applies to equity
18
The model of Kashyap and Stein (1994) also captures the argument why banks hold a certain
fraction of deposits as liquid securities and is, therefore, also insightful for this reason.
capital. Because equity is junior compared to debt capital, the premium that equity
investors demand should be even higher. Another problem that arises when raising
new equity is stated by Myers and Majluf (1984), who point out the special
asymmetric information problems between old and new shareholders that may
prevent the issue of new equity (debt overhang). In short, potential new
shareholders face the threat that their capital is used to service debt which has been
accumulated during the leadership of the old shareholders. This overhang of "old"
debt which becomes a burden on new capital will make potential new shareholders
reluctant to invest if they are unsure regarding the exact financial position of
investment target.
Why is this subcondition necessary for the bank lending channel to be active? Or, to
put that another way, what would happen if alternative forms of funding were perfect
substitutes for deposits? It is not difficult to see that, if alternative funding were a
perfect substitute, adjusting for lost deposits could simply be performed at no cost on
the liabilities side of the balance sheet. Since the cost of capital are not perfectly
inelastic for a bank when the assumption of perfect substitutability of insured deposits
and uninsured debt is abandoned, this has implications for the asset side. In
connection with the first subcondition, which states (for the reasons given above) that
adjustment for drained deposits is not only performed by selling deposits, a bank will
curtail those lending opportunities that are, relatively speaking, the most unprofitable
ones. Similarly, a bank that offers loan rates which reflect the higher cost of capital
cannot expect to maintain the same volume of loans as before the drain in deposits.
To sum up: The key point concerning the second subcondition is not that a bank
must be unable to replace the lost deposits. For the second subcondition to hold, it is
only necessary that demand for a bank's wholesale liabilities must not be perfectly
elastic, so that they are not perfect substitutes for lost deposits from a bank's
perspective. This marks a departure from the famous Modigliani-Miller theorem
(Modigliani and Miller (1958)), according to which the capital structure does not make
any difference to the value of a firm.
3.1.1.3. Subcondition 3: Banks must not be capital constrained
The third subcondition is related to the capital position of a bank. Under the accords
of Basel I (which was introduced in 1993), its successor Basel II (in force since 2007)
and the different national legislations, banks are required to maintain a certain equity
capital ratio. Failure to do so can have serious consequences for the institution
concerned and is ultimately threatened by the takeover of control or forced closure by
the authorities.
The third subcondition requires that banks must not be capital constraint, i.e. that
their ability to engage in additional lending should not be restricted because of too
low capital ratios.
As stated in the context of the first subcondition (section 3.1.1.1), banks do not adjust
for a drain in deposits following a monetary tightening only by selling securities. In the
event of an adverse monetary policy shock, a bank will also adjust by curtailing
lending in the medium term in order to maintain the desired ratio of liquid assets to
deposits. The exact ratio that each bank can decide on at discretion depends on its
expectations concerning future deposit availability and deposit flows.
There is one exception, in the event of which a bank will adjust to the reduction in
deposits solely by selling securities: if it is bound by capital constraints. Under these
circumstances, the bank already holds more securities than are considered optimal
based on its expectations and risk appetite with regard to coverage of possible future
deposit outflows. This is because the bank does not have the capital it would need to
accommodate higher loan volumes. Hence, a bank in such a situation will handle
adjustment solely by selling securities, as long as the proportion of liquid assets is
still above the optimum.
In the opposite case, i.e. where monetary policy is relaxed, banks that are capital
constrained cannot expand their lending, but will instead invest the additional
available deposits in liquid assets.
Raising new equity is not an option, because it is even more costly than debt capital,
as already discussed in the context of the second subcondition (section 3.1.1.2).
It is important to note that capital constraints need not necessarily be related to a

regulatory requirement, as the subcondition also applies to self-imposed capital
constraints (see Berrospide and Edge (2010)). The fact that a bank might operate at
the limit of a self-imposed capital constraint, even though it is not restricted by
regulatory requirements, is one of the hypotheses in section 7.1. It is incorporated in
the empirical analysis following the methodology originally developed by Hancock
and Wilcox (1993 and 1994) for the application in context of regulatory capital
constraints.
All three subconditions together ensure that the first necessary condition for the bank
lending channel that the central bank must be able to affect the supply scheme of
bank loans holds true. The first subcondition requires that lost deposits must not
simply be compensated for by selling liquid assets. The second subcondition
presupposes that banks cannot compensate for lost deposits by alternative forms of
(uninsured) funding, including both external finance and equity. The third
subcondition makes clear that the lending channel only works in the event that a
bank is not capital constraint either for regulatory reasons or due to a self-imposed
capital ratio.
3.1.2. Condition 2: Publicly issued debt and non-bank intermediated loans must not
be perfect substitutes for bank loans
The condition that publicly issued debt and non-bank intermediated loans must not
be perfect substitutes for bank loans is the second requirement for the bank lending
channel. It establishes the link between the banking sector and firms and the real
economy. Without this condition, there would be no guarantee that a monetary shock
affecting the supply of bank loans filters down into the real economy. Or, to put that
another way: If the condition did not hold, this would mean that firms could, at no
additional cost, fully compensate for the supply in bank loans by borrowing from other
sources, thereby escaping the influence of monetary policy.
As the formulation of the second condition implies, a firm has, in principle, two
possibilities to circumvent a decline in the supply of bank loans: It can either take
loans from non-bank financial intermediaries or it can rely on publicly issued debt,
especially in the form of bonds or commercial papers. Thus, in order for the condition
to hold, two questions need to be answered. First, why is it difficult for firms to
replace bank loans with loans from non-bank financial intermediaries? And second,
why are bank loans and publicly issued debt not perfect substitutes (at least for some
firms)?
To begin answering the first question, it is helpful to clarify why firms are at all
dependent on financial intermediaries. Three concepts play a role here: first,
transaction costs; second, risk diversification and liquidity transformation; third, the
cost of (gathering) information.
Without financial intermediaries, both borrowers and lenders would need to handle all
the details associated with a loan contract themselves, e.g. finding a lawyer to set up
a proper contract between the two parties, formulating all the clauses, etc. Especially
for small-scale borrowers and lenders, this would be enormously costly. Financial
intermediaries such as banks can drastically lower this kind of transaction costs due
to their economies of scale.
The second aspect is risk diversification. An investor not big enough to diversify in its
own right can avoid "putting all its eggs in one basket" by giving money to a financial
intermediary big enough to allow for proper diversification of investments and
investment risk. Without the intermediary, the funds would not be available to
borrowers because the investor lacked the opportunity to diversify risk. Closely linked
to the idea that a bank offers a service to both investors and deposit holders without
which funds would not be available is the function of liquidity transformation, which
banks perform at the same time. Banks are able to transform relatively illiquid assets,
such as loans, into liquid deposits, thereby insuring investors who want to consume
whenever they need to ('liquidity on demand'; see Kashyap et al. (2002)). Both ideas,
the one of risk sharing and of liquidity transformation, have been shaped by Diamond
and Dybvig (1983).
The third aspect regards the cost of gathering information. Asymmetric information
between borrower and lender is a natural consequence of the fact that the borrower
usually has better information regarding the risk and reward profile of an investment
and, therefore, regarding the probability of repayment than the lender who is asked to
grant a loan. This problem is relevant before the loan contract is closed, i.e. when the
lender seeks to avoid selecting the most risky borrowers (adverse selection), and
after the loan contract is closed, i.e. when the lender needs to monitor and control the
borrower's actions to prevent the borrower from acting in a manner contrary to the
interests of the lender (moral hazard).
Based on the work of Jaffee and Russell (1976), Stiglitz and Weiss (1981) presents a
theoretical model of asymmetric information which highlights their significance in the
context of financial intermediation that leads to conventional moral hazard
problems. 19 Taking this argument further, Blinder (1987) notes that, for the same
reason, specialized institutions exist that gather information about borrowers. These
institutions can realize economies of scale with respect to the acquisition of
information and monitoring borrowers' actions.
Although the above discussion makes the case for financial intermediaries in general,
it still does not answer the question why firms depend on banks in particular.
By assessing risk before granting a loan and by monitoring borrowers' actions,

lenders acquire a monopoly of information with respect to their borrowers. This
monopoly results in lock-in effects, as theoretically underpinned by Sharpe (1990)
and Rajan (1992). Ultimately, from a firm's perspective, it is not easily possible to
switch lenders without incurring additional costs even though credit markets are
competitive. The same argument, viewed from the perspective of a competing lender,
states that it is hard to step into an existing and established lender-borrower
relationship.
In addition, Nakamura (1984) makes a case for economies of scope stemming from
the fact that banks already acquire information by managing a firm's deposits. This is
clearly an advantage over non-bank financial corporations that do not manage
deposits, but that only lend money.20
Moreover, competing lenders who try to lure away borrowers by offering lower rates
could well be countered by an equivalent or even better offering from the current
bank in case of a "good" borrower. Only in case of a "bad", i.e. riskier borrower do
competitors have a chance to win the client. 21 Obviously, this adverse selection
problem can lead to an undesired, risky portfolio (see Sharpe (1990)). This kind of
19
Both Stiglitz and Weiss (1981) and Jaffee and Russell (1976) discuss the impact of monetary
policy on the supply of lending in light of credit rationing. As Kashyap and Stein (1994) points out,
even though shifts in the supply of bank loans may involve credit rationing to some degree, the
bank lending channel is not predicated on whether there is credit rationing or not.
20
This is empirically plausible against the backdrop of models that calculate a score on the basis of
account movements and account balances over time. This score enters a borrower's risk
assessment. Today, these models are used by all major commercial banks.
21
This view is an abstraction of the concept of standard unit costs. Theoretically, it is also possible
for a competitor to offer lower loan rates than the existing lender for a "good" borrower due to
lower standard unit costs. This is also plausible in practice.
problem can be reduced if firms provide more transparent information; this should
apply especially to large firms that may have also been rated by a rating agency.
These arguments show how gathering information can lead to a lock-in between the
borrower and the lender. Although particularly large firms may not be subject to this
effect, and although one could argue that it only applies in cases where the existing
borrower is a bank (as newly founded firms may, theoretically, resort to non-bank
financial intermediaries right from the beginning), at least small and medium-sized
firms will not find it easy to switch to non-bank financial intermediaries.
As a matter of fact, comparison of the relative sizes of bank credit and non-bank
credit, as performed by Kashyap and Stein (1994), shows that non-bank credit has
gained in importance to some extent in recent decades. However, it also shows that
bank credit is still by far the most important source of firms' funding, emphasizing the
significance of the banking sector in the process of financial intermediation. This
finding is supported by Mayer (1990), who, in a comparison of major OECD countries,
shows that firms in all the countries surveyed obtain funds primarily through financial
intermediaries, despite existing differences between countries. The work of Fiore and
Uhlig (2005) points in the same direction. Fiore and Uhlig (2005) finds a high
proportion of bank loans in the euro area in particular (as opposed to the US), and
attributes this to the high efficiency of euro-area banks in acquiring information on
borrowers.22
To sum up: The answer to the first question why it is difficult for firms to replace
bank loans with loans from non-bank financial intermediaries (the probable future
gain in importance of non-bank financial intermediaries notwithstanding) is that, up
to now, the importance of banks as intermediaries is firmly established, particularly in
the euro area. It is therefore not to be expected that a shift in the supply of loans
following the tightening of monetary policy can be completely offset by non-bank
financial intermediaries. 23
Comparison with the US leads to the second question: why bank loans and publicly
issued debt are not perfect substitutes. While the LM curve in the traditional IS/LM
22
See section 6.1 for figures on the significance of bank lending in the euro area and the US.
23
There are good reasons to believe that also non-bank financial intermediaries are impacted by
changes in monetary policy. See chapter 4 for a discussion.
model implicitly assumes that they are indeed perfect substitutes, there are a number
of reasons to believe that they are not (see e.g. Peek and Rosengren (2010)). First,
the fixed costs associated with issuing a bond are likely to be too large for small firms
and relatively small ticket sizes. Second, smaller firms are not transparent enough
with respect to information they provide or sufficiently creditworthy, which prevents
their direct access to the bond market. Third, one could argue that these constraints
do not apply to large firms that enjoy an adequate credit rating from a rating agency
and, therefore, could exploit direct access to public credit markets. However, the
issuance of commercial papers typically entails making use of bank services and,
more importantly, bank balance sheets. The fact that banks usually acquire a share
of the issued volume to send a positive signal about the borrower's creditworthiness
once again establishes firms' dependence on banks and leaves firms bound by the
bank's ability to grant loans along with all the sensitivities to monetary policy
explained under the first condition of the bank lending channel.
The answers to both questions thus suggest that publicly issued debt and non-bank
intermediated loans are not perfect substitutes for bank loans, at least for some firms.
This satisfies the second condition that is necessary for the bank lending channel to
be active.
3.1.3. Condition 3: Prices must not adjust instantaneously
The third condition reflects a general requirement without which monetary policy
would not have any influence on the real economy, irrespective of the channel of
transmission under consideration. In such a situation, monetary expansion would
lead to immediate price increases and would therefore not have any real effect.
Given that the issue of imperfect price adjustment is dealt with by an enormous body
of literature, is empirically well established and is an element of any macroeconomic
textbook, this section focuses solely on the most important building blocks of its
microfoundations.
The most important concept to which price rigidities can be attributed is what is
known as the menu cost. The menu cost is the cost that firms necessarily incur when
changing prices. Cost items range from printing new price lists (e.g. a restaurant's
menu) and re-tagging items to updating systems and merchandise material. From a
business perspective, price changes are only beneficial if the associated costs are
outweighed by the expected revenues, resulting in stepwise (rather than continuous)

price adjustments.
Long-term contracts such as tenancy and lease contracts are an additional reason
for price rigidities. In these contracts, prices are fixed for the entire duration of the
contract without the possibility of adjustments.
The labor market is a further example of an area where price adjustments do not
happen instantaneously. To a significant extent, nominal wages are settled in
collective wage agreements and fixed for some period in advance. More frequent
wage adjustments would involve prohibitively high negotiation costs.
What all these examples have in common is the idea that price rigidities exist
because frequent revisions of prices and contracts entail transaction costs that move
the parties involved away from their profit or utility maximum.24
This third condition completes the set of sufficient conditions that must hold in order
for the bank lending channel to apply. This concludes the traditional view on the bank
lending channel.
24
As a historical reference to price rigidities: The role of price and wage rigidities is captured by the
Phillips curve, pioneered by Phillips (1958). The key element is that nominal wages are a function
of unemployment or, more generally, of the cyclical state of the economy. The link between
unemployment and the inflation rate was formulated by Samuelson and Solow (1960) in terms of
the modified Phillip curve, which suggests an exploitable trade-off between unemployment and
inflation. The flaw in the modified Phillips curve was that it disregarded the crucial role of
expectations. Since it is the real wage and not the nominal wage that is important for workers, this
omission was accounted for by including inflation expectations as a factor in the rationale of wage
negotiations. The theoretical foundation was delivered by Phelps (1967) and Friedman (1968) in
the form of the expectations-augmented Phillips curve. It must be noted that the relationship
between unemployment and inflation in the expectations-augmented Phillips curve exists only if
there is a difference between the actual and expected rates of inflation. This can only be the case
in the short run: In the long run, inflation expectations and inflation will be identical, and no money
illusion exists. This gives rise to two implications: First, the Phillips-curve trade-off cannot be
exploited by policymakers. Second, analysis of inflation on basis of the Phillips curve involves a
relatively short time horizon.
3.2. Conclusion
This chapter has introduced the theoretical foundations of the bank lending channel
with its conditions and subconditions. It has been outlined that three conditions need
to be fulfilled for an active bank lending channel. The first condition of the mechanism
describes how the central bank influences the supply of bank loans. The second
condition describes firms' dependence on bank-intermediated loans. The third
condition, which is a general one applying to all transmission channels of monetary
policy, stipulates that prices must not adjust instantaneously subsequent to monetary
policy changes.
As this study focuses on the determinants of the supply of bank loans in different
economic circumstances, it is apparent that the first condition is of primary interest for
the further reasoning since this condition directly addresses banks and their supply of
loans in the transmission process.
The way in which the whole mechanism of the bank lending channel and especially
the first condition has been presented above can be called the "traditional view".
What makes it traditional? Typical of this view is the emphasis on deposits being
affected by monetary policy via control over reserves.
However, according to today's operational frameworks of major central banks

reserves are not controlled directly. Furthermore, developments with regard to
deregulation and financial innovation in recent decades have important implications
for bank lending. These facts challenge the bank lending channel in its traditional
notion presented above.
Consequently, in the next chapter, it is argued that the traditional view must be
reconsidered: On the one hand a comprehensive perspective on bank lending and
the bank lending channel must account for the actual functioning and operational
frameworks of central banks (such as the ECB or the FED). On the other hand it
needs to account for developments in the financial sector.
However, this in no way undermines the importance of the bank lending channel as a
concept to explain the lending reaction of banks, the essence of which is that the
central bank influences the supply of bank loans. Why this is the case is central to
the following chapter.
4.1 The bank lending channel revisited 35
4. A new view: Implications of financial innovation for bank lending
The previous chapter has laid the important theoretical foundations of the bank
lending channel by presenting the underlying mechanism in its traditional
interpretation. The first section of this chapter addresses the question why a new
view on the bank lending channel is necessary not only as a result of developments
in the field of financial innovation but also against the background of today's
operational frameworks of major central banks.
On basis of this, the new view is conceptualized in the second section of this chapter.
The new view on the bank lending channel and its conceptualization is central to the
reasoning of this study because it is a necessary condition for the argumentation in
the subsequent chapter 5: The argumentation why bank lending has been different
against the background of the recent crisis.
4.1. The bank lending channel revisited
So far, the arguments have followed the traditional vein of the bank lending channel,
which relies heavily on central banks control over the level of deposits via reserve
requirements and the money multiplier. Because deposits are regarded as the most
dominant form of funding, a deposit outflow impacts on the ability of banks to grant
new loans, as discussed in detail in section 3.1.1.
Especially with regard to the European Central Bank's operational framework, within
which the ECB provides any amount of reserves demanded by the financial system
given that the deposited collateral is of a satisfactory quality and thus compensates
for deposit outflows it seems questionable whether the traditional view is (still) a valid
representation of one of the monetary policy transmission mechanisms used in
modern financial systems, or at least in the euro area and the US (see e.g. Marques-
Ibanez (2009), Borio and Disyatat (2010) and especially Disyatat (2011)).
This is not to say that the transmission of monetary policy via banks is less effective
or even muted. It rather appears that developments in recent decades, namely those
summarized under the heading of financial innovation, have acted as a game
changer for the banking industry and now imply a new view on the mechanism of the

36 4 A new view: Implications of financial innovation for bank lending
bank lending channel. Whereas banks previously had to resort chiefly to deposits as
the main source of funding, the development of wholesale funding markets has given
banks access to funding opportunities that were not available on such a large scale
some 20 years ago.25 This makes banks probably more dependent than ever on risk
appraisals of investors and on the external finance premium (see Altunbas et al.
(2010), Disyatat (2011) or Borio and Zhu (2012)). Uninsured funds can no longer be
regarded as merely a marginal funding source, but have instead emerged as an
important pillar of overall funding strategy in many institutions. One consequence of
this is that the characteristics that influence a banks risk position and the investors
risk perception of a bank determine to a significant extent both a banks ability to fund
loans through market sources of funding and its reaction to monetary policy.26
To avoid any misconceptions regarding the new view as opposed to the traditional
view: This new view does not come in complete contrast to the traditional bank
lending channel literature. Neither does it mean that there is no longer any role for
deposits in this interpretation of the bank lending channel, nor does traditional
literature neglect the fact that risk premiums influence a banks ability to access
uninsured funding.
There are two new aspects to the new interpretation. First, due to the dimension to
which wholesale funding markets have grown and due to financial innovation
(including securitization and credit derivatives), the problems associated with
asymmetric information and moral hazard have also increased in significance.
Second, the new view explicitly acknowledges that, even if a bank is not dependent
at all on (insured) deposits, it is all the more subject to an external finance premium,
which originates in the informational asymmetry between lender and borrower. In
short, in accounting for the developments of financial markets over recent decades,
the focus has shifted to the increased importance of funding via wholesale markets.
25
Wholesale funding can be distinguished according to maturities into short-term and long-term
funding. Short-term wholesale funding comprises unsecured and secured money market
borrowings and the issuance of short-term debt (commercial paper, for example). Long-term
wholesale funding includes bonds and covered bonds.
26
One very telling example of the perceived riskiness of banks as a determinant for their ability to
access uninsured sources of funding is the drying-out of the interbank capital markets following the
collapse of Lehman Brothers: Uncertain about other institutions' levels of exposure, banks simply
ceased to lend to each other (see van Rixtel and Gasperini (2013), for example).
The cost of raising funds on these markets is dependent on an institution's perceived

riskiness and is affected by monetary policy.
TRADITIONAL VIEW: CRITIQUE TO THE

Mechanism focuses on central banks' control over loans via deposits and reserves TRADITIONAL VIEW
1. Reserve requirements must be met by all banks CB supplies any amount of reserves
> Reserve requirements are calculated according to: as demanded by the system
All eligible liabilities (esp. deposits) > The volume of reserves provided varies
only with the requirements of the banking
system (and not with the desire of the
x Reserve ratio central bank to steer the level of loans)
= Reserve requirements Banks regularly report their amounts
of reserve-eligible liabilities to the
central bank
2. Central bank influences volume and terms of available reserves Thus, the central bank knows how
> Central bank exercises short-term controlI) over much reserves are required by the
volume of reserves made available system at any time
terms of available reserves This volume of required deposits is
always supplied by the central bank
3. Thereby, the CB affects deposit creation and the granting of loans > As a matter of principle, the central bank
> As a result, with the control over terms / volume of available reserves the central bank accommodates any level of loans and
can influence the volume of deposits available to fund loans and the ability to grant loans deposits by supplying the required
> For example: volume of reserves
CB increases key interest rates (monetary policy tightening) > This applies to any fractional-reserve
banking system (e.g. ECB, FED)
Consequently, reserves turn more expensive and the total volume of reserves in the
system declines
With the decline of reserves the volume of deposits that can be maintained decreases Thus, no direct link exists from
(see calculation of res. requirements) reserves to bank lending
A decrease in the volume of deposits limits a bank's ability to grant new loans because > A direct, mechanical link from the level of
the granting of a loan involves the creation of a equally large deposit reserves to bank lending, as postulated
according to the traditional view of the
bank lending channel, does not exist
A monetary tightening/easing leads to a shortage/surplus of liquidity (deposits)
I) The central bank also has control over the reserve ratio and the range of liabilities covered by reserve requirements. However, these are not subject to regular,
short-term changes.
Figure 4.1: The traditional view and a critique of the traditional view
The traditional view of the bank lending channel mechanism and the critique of it in
light of the developments mentioned above are shown in figure 4.1.
Before attempting to conceptualize this new view on the bank lending channel, it
makes sense to clarify how monetary policy influences bank lending in light of the
new view (see figure 4.2). The mechanism works through changes in the funding
conditions for banks that are induced by monetary policy changes. It is then reflected
in the cost of loans for borrowers: With the change of key interest rates, the central
bank induces disproportionate changes in the funding conditions for banks. How
does this happen?
According to Disyatat (2011), tighter monetary policy, for example, has a negative
impact on banks cash flow, net interest margins and asset valuations. These factors
are all reflected, ceteris paribus, in negative changes to banks capital positions.
Lower capital ratios signal higher risks to providers of uninsured funding, for which
they demand a higher external finance premium. Furthermore, the widespread use of
mark-to-market accounting exposes banks to higher interest-rate risks which, in the

same way, trigger demand for a higher risk premium (see Adrian and Shin (2008a).
Moreover, increased reliance on short-term funding adds further to banks'
augmented sensitivity to changes in the monetary policy rate.
NEW VIEW:
Mechanism focuses on central banks' power to affect terms of uninsured funding
Monetary policy changes lead to disproportionate changes in availability of funding
> Over the last decades the integration of financial markets and financial innovation have led to
the development of deep and liquid wholesale funding markets
> Nowadays, many banks and the banking system as a whole rely on (uninsured) wholesale
funding as an important funding source (besides insured deposits)
> Monetary policy is able to disproportionately affect the terms under which this (uninsured)
wholesale funding is available
A feature of markets for uninsured funding is that there is informational asymmetry
between the lenders and banks as borrowers
Lenders demand a risk premium which depends, broadly speaking, on the perceived
financial position/health
Banks with weaker financial positions are unlikely to meet their repayment obligations,
because:
- An increase of monetary policy rates negatively affects cash flows, net interest margins
and the valuation of assets
- Low level of interest rates increase banks' propensities for risk taking by aggravating the
search for yield-problem
Thus, in case of a monetary tightening the problem of asymmetric information is
aggravated most severely for those banks whose characteristics (e.g. capital position etc.)
signal relatively weak financial positions
> Consequently, the bank lending channel works predominantly through the impact of
monetary policy on the availability of uninsured sources of funding
> This mechanism operates without any recourse to reserves affecting deposits
Figure 4.2: The new view on the bank lending channel
A further rationale is worth considering that also sees risk as an important factor,
albeit from a slightly different angle: According to Borio and Zhu (2012), a central
bank directly impacts banks incentives to take risks. Low levels of interest rates have
the potential to increase a banks tolerance for risk by aggravating the search for
yield problem.27 Riskier operations will, however, translate into a higher risk premium
demanded on wholesale funding markets.
These examples show how monetary policy can trigger variations in the perceived
riskiness of banks. The magnitude of the variation depends on banks
characteristics.28 The result is that changes in the terms at which banks can borrow
in funding markets are passed on to borrowers of loans.29
Interestingly, this new view on the mechanism of how monetary policy affects the
supply of bank loans does not hinge on a drain in reserves. It reconciles the fact that
insured deposits are not a limiting factor to the funding of banks assets. As noted
earlier, a deposit is created when a bank grants a loan because it credits the amount
of the loan to the debtors account as a deposit. Hence, deposits are not a constraint,
but they are endogenously created in the financial sector as a function of demand for
loans (see Disyatat (2011)). Reserves are then provided by the central bank
depending on the amount of deposits and the reserve ratio. This is the reversal of
what is claimed according to the traditional view.
Another interesting consideration is that non-bank financial intermediaries that do not

fund themselves with insured deposits at all but that have to rely completely on
market funding are likewise sensitive to the mechanism triggered by monetary policy
changes. Their cost of funding too is dependent on their perceived riskiness. Via the
same mechanism, monetary policy also influences the balance sheets, cash flows
and capital of non-bank financial intermediaries, which determines the probability that
27
Borio and Zhu (2012) coin the notion that monetary policy has an impact on the riskiness of an
institution's risk-taking channel. Although the role of risk is fully acknowledged (and proven
empirically; see e.g. Jimenez et al. (2008)), the question addressed in the present study is what
factors impact the lending behavior of banks and their lending response to monetary policy. For
this reason, the above notation of Borio and Zhu is not followed here. (The authors themselves do
not claim it as a distinct channel.) Instead, risk is incorporated into the systematization of bank
characteristics stemming from the conditions and subconditions of the bank lending channel (see
figure 4.4 and figure 6.2).
28
For a formalization of this idea, see the model proposed in Disyatat (2011), adapted from Disyatat
(2004).
29
It is important to note that this is not a reflection of the traditional interest-rate channel. The
transmission of monetary policy through the interest rate channel leads to a proportional increase
of costs of credit over the risk-free rate (see Disyatat (2011)).
they will be able to pay back borrowed funds. The implication is therefore that non-
bank financial intermediaries are subject to monetary policy changes and to the bank
lending channel in just the same way as normal" financial intermediaries, and that
the determinants of their lending reaction are closely comparable to those of banks.
Claims that the bank lending channel may have become muted or have seen its
significance reduced in light of the growing importance of non-bank financial
intermediaries may, then, have been premature (see Bernanke (2007)).
4.2. Toward a conceptualization of the new view
As shown in figure 4.1, criticism has been leveled at the traditional view of the bank
lending channel, which focuses on the central bank's influence on deposits as a
funding source via its ability to control the volume of available reserves. The
argument is that, in a fractional-reserve banking system, central banks can supply
any amount of reserves as demanded by the system. However, the bank lending
channel is unlikely to have ceased to exist. In light of developments such as
deregulation and financial innovation, the channel continues to work through the
central banks' impact on the availability of uninsured funding in public markets. Banks
whose financial position is relatively weak are more severely impacted by monetary
policy tightening, for example, because the problem of asymmetric information is
amplified for these banks, in particular due to the negative impact on cash flows, net
interest margins and asset quality.
Taking account of these new aspects of the bank lending channel and their
theoretical substantiation as outlined in section 4.1, this leads to a new approach to
conceptualizing the bank lending channel, its conditions and subconditions.
Before going into the details, one important theoretical step must first be
acknowledged. There is no question that the total cost of alternative funding for a
bank is determined not only by the risk premium, which is the cost per unit of
borrowed funds: It also depends on the amount or volume of borrowed funds. This
relationship is illustrated in figure 4.3. Figure 4.3 also previews the determinants of
both volume and (the risk component of) cost per unit of alternative, uninsured
4.2 Toward a conceptualization of the new view 41
funding (see figure 4.4 for more details).30 The key insight, from a theoretical point of
view, is that all characteristics that either impact on the volume or on the unit cost of
uninsured funding determine bank lending.
Cost per unitI) of

alternative forms
of funding
Total cost of Cost per unitI) determined by

alternative forms > Transparency
of funding > Business risk
> Risk of moral hazard
Volume/amount of
alternative forms of funding
Volume/amount determined by
> Dependence on alternative Total cost of alternative
(uninsured) forms of funding forms of funding
> Access to (insured) retail deposits
I) The focus here is on the factors driving the risk component of the cost per unit. Therefore, the risk-free interest rate is not listed.
Figure 4.3: Determinants of cost or ease of access to alternative forms of funding
This leads to a enhanced conceptualization of the bank lending channel including the
economic concepts determining bank lending. Figure 4.4 presents a breakdown of
the subconditions of the relevant first condition of the bank lending channel into the
related economic concepts.
On the left-hand side of the figure, all three subconditions of condition 1 are arranged
one below the other. Moving one column to the right, the economic concepts that
serve to operationalize the various subconditions are depicted. For the second
subcondition, these economic concepts are split into concepts that impact the cost
per unit of alternative forms of funding and concepts that impact the volume/amount
of alternative forms of funding.
30
These are the concepts on which the determinants are based that are used to explain banks'
lending behavior in the empirical analysis in chapter 7. See also section 6.4 for the further
translation and specification of the concepts toward individual bank characteristics.
Let us take an example: One of the economic concepts behind the second
subcondition is "business risk". "Business risk" has an impact on the cost per unit of
funding: The less risky the bank, the more favorable will be the terms on which
external funds can be raised.
Subconditions Economic/theoretic concept

(operationalizing the subcondition)
Subcondition 1
Liquidity
Holdings of liquid (as buffer against random deposit outflows)
assets
Transparency
Subcondition 2 Business risk

Cost per unit
of alt. forms (focus on perception
(Ease of) regarding riskiness
Access to of funding
of a bank's operations)
alternative forms
of funding
(determines total
cost of alternative
forms of funding) Risk of moral hazard
(focus on risk of bank acting
in a morally hazardous way)
Volume/
amount of Degree of dependence
alt. forms of insured funding
funding
Subcondition 3 Facing a binding capital constraint

No capital
constraint (capital ratio below certain threshold)
Figure 4.4: Breakdown of subconditions into economic concepts driving bank lending
Already previewing the following discussion, this way of depicting a framework for the
breakdown of the subconditions has two main advantages regarding the further
course of analysis in the next chapters.
First, after the review of empirical evidence in chapter 6, an attempt can be made to
connect each of the bank characteristics that authors have identified to have an
impact on bank lending to the economic concepts shown in figure 4.4 and thereby
4.2 Toward a conceptualization of the new view 43
integrating them into overall framework. This is useful as a way to transfer the
subconditions into proxy variables used later in the econometric analysis. As such, it
highlights structural relationships between bank characteristics and the conditions
and subconditions. This exercise is done in section 6.4.
Second, it provides a framework into which each bank characteristic that is relevant
must be meaningfully integrated, thereby serving as an initial plausibility check for
any future developments in this field.
44 5 Bank lending against the background of the recent crises
5. Bank lending against the background of the recent crises
The focus so far has been on changes in bank lending in the context of the theory of
the bank landing channel. The bank lending channel and the conceptualization
formulated in chapter 4 lay the foundation for the subsequent empirical analysis.
However, this context must be extended by one important aspect: While initial and by
no means exhaustive attempts have been made to investigate the empirical influence
of the recent crisis on the factors that determine bank lending, the current literature
lacks any framework that is able to theoretically account for a differential impact of
bank characteristics on lending during periods of crisis. This chapter closes the gap
by linking bank lending to two phenomena that were observable during the recent
crisis in the financial sector and that acted as an amplification mechanism. Taken
together, these two concepts form the basis, from a theoretical perspective, to
substantiate the view that some bank characteristics played a different role in their
influence on bank lending during the recent crises compared to normal times. The
first concept is the loss spiral and the second is the margin spiral or leverage cycle.
These two negative feedback loops reveal externalities that can be seen as having
significantly contributed to the propagation of the crisis in the financial system and as
having promoted contagion among financial investors such as banks, insurance
companies and others. In particular, while it is desirable from the perspective of an
individual institution to sell assets in order to restore adequate capital ratios in
response to initial losses to a banks equity, the same course of action might not be
desirable from the perspective of the stability of the financial system as a whole. The
forced and uncoordinated sale of assets (deleveraging) when a pronounced boom
period turns negative can put the entire financial system under a level of stress that
leads to economy-wide contraction in lending and economic activity.
This important issue is prominent in the ongoing discussion of macroprudential

supervision. Macroprudential supervision aims to develop and implement measures

5.1 The loss spiral 45
to guard financial stability against the risk of contagious effects and the build-up of
structural imbalances.31
In the sections that follow, the general mechanics of the loss spiral and the margin
spiral/leverage cycle are presented and linked to the lending of financial institutions.
Implications are then drawn with regard to bank characteristics that impact lending in
a crisis context.
5.1. The loss spiral
The loss spiral is one of the two concepts which give reason to believe that some of
the bank characteristics that determine the credit supply behave differently during
times of crisis than at normal (i.e. non-crisis) times.
To see the connection, it is helpful to first understand the general mechanics. At the
heart of the liquidity spiral is the famous debt-deflation mechanism pioneered by
Fisher (1933), which helps to explain business cycles against the background of the
Great Depression. 32 However, mapping this idea onto the financial sector and
applying it to financial crises, especially to the most recent one, is a fruitful exercise.
Without going into the details of the origins of the recent crisis and only in order to
sketch a few important features, it is well established in literature that the US Federal
Reserve created an environment of low key interest rates in response to events at
beginning of the last decade in order to counteract the ensuing recession (see figure
5.1). This policy left the economy awash with liquidity which, in connection with other
factors, such as development of credit derivatives and securitization markets,
governmental subsidies for mortgages, etc., created a real estate boom. As long as
real estate prices were rising (see figure 5.2) and debtors were able to meet their
repayment obligations, this boom also drove up banks net worth or equity. To the
31
Macroprudential concepts are believed to be necessary to prevent individual, utility-maximizing
financial firms from deviating from the actions a social planner would induce them to engage in, i.e.
to avoid the social costs that result from a sudden, uncoordinated shrinkage of balance sheets. For
a recent survey of literature on macroprudential policy, its goals and tools, see Galati and
Moessner (2012).
32
For a more recent analytical underpinning of business cycles, see Kiyotaki and Moore (1997) and
Bernanke et al. (1999), for example.
extent that financial intermediaries balance sheets had to be marked to market, e.g.
due to their holdings of credit derivatives, net worth was impacted on a daily basis.
Target Federal Funds Rate

7
6
Period of falling
interest rates
5
4
Percent
3
0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Year
Source: U.S. Federal Reserve
Figure 5.1: The US federal funds target rate, 2000-2012
As Adrian and Shin (2010b) argue, this development encouraged banks that were
keen to find favorable investment opportunities to lend to borrowers of questionable
creditworthiness. At the time, this was not a major concern due to the fact that rising
housing prices guaranteed sufficient collateral.
When the Federal Reserve began to gradually increase target interest rates from the
second half of 2004 onward and the economic outlook began to deteriorate, real
estate prices started to decline by the end of 2006 (figure 5.2). This set the debt-
deflation mechanism of the loss spiral in motion.33
The start of the actual spiral is marked by initial losses from which banks suffered as
house owners began to struggle to repay their loans. Banks in such a situation are
faced with a funding constraint in the sense the losses wipe out equity and increase
the leverage. Increased leverage ratios exacerbate the asymmetric information
problem between banks as borrowers and the providers of (uninsured) funds. The
33
The focus here is more on the mechanism of the loss spiral rather than the sequence of events
during the recent crisis. The latter topic is already amply dealt with by almost innumerable
accounts that cover the crisis from many different angles.
borrower is more risky and, therefore, the problem is the more severe the higher an
institution's leverage.
Case-Shiller Home Price IndexI)

(composite 20 index, seasonally adjusted)
210
200
190
180
170
160
Index
150
values
140
130
120
110
100
0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Year
I) The Case-Shiller Home Price Index is one of the leading house price indices in the U.S. The composite 20 index is calculated from sales of single-family house
prices from 20 major metropolitan areas. The methodology has been developed by Karl E. Case, Robert J. Shiller und Allan Weiss. Today, it is compiled by Standard
& Poor's. For details on the methodology refer to Standard & Poor's (2009).
Source: Standard & Poor's
Figure 5.2: House price development in the US, 2000-2012
In response to the increased asymmetric information problem and to perceived

riskiness, banks try to push leverage ratios down to earlier levels by selling assets to
avoid being cut off from funding liquidity. 34, 35 Where not enough short-term assets
are available, even long-term assets have to be sold at short notice, which becomes
more likely the greater the maturity mismatch is on the banks balance sheet.
Shleifer and Vishny (1992) cite an example from the real economy to point out that
the forced sale of an asset in times of recession can lead to significant price
34
The assumption of constant leverage ratios may be questioned. More light on that issue is shed in
the next section on margin spirals in which the assumption is relaxed in favor of the even more
severe one of procyclical leverage ratios.
35
There different approaches to theoretically model the funding constraint stemming from too high
leverages which all amount to the same result. For example, Bianchi and Mendoza (2011) impose
that agents may not borrow more than given by a certain fraction of the market value of their
collateral assets.
discounts (fire sales) precisely because potential buyers to whom the asset has the
highest value are typically those who are from the same industry and tend to be
affected by the same distress as the seller. This leads to an equilibrium in which the
asset is sold to a buyer at a price below the value in best use (see also Geanakoplos
(2010)).
The same pattern can be applied to securities in financial markets. Securities that are
liquidated under fire-sale conditions are sold at a discount because those market
participants who are most knowledgeable about the underlying asset and the risks
associated with it are distressed as well. Only investors with deep pockets may have
the capacity to buy these securities, but only at a discount due to a lack of knowledge
and experience in this field. To the extent that securities are marked to market, the
result is an immediate equity loss and a need to sell further assets.
Factor acting as a catalyst for spiral Explanation
> The higher the leverage the greater the need to sell assets after initial equity
loss to maintain leverage ratio (or to meet increased margin requirements)
> Institution hit by initial equity
> This increases the probability of asset sales with significant differences
capital loss is highly leveraged
between proceeds and the book value
> The result is further losses in the form of write-downs
> Greater risk to suffer from dry-up in funding liquidity (funding liquidity risk)
when rolling over short-term debt in case investors withdraw funds in light of
> Institution's maturity mismatch is gloomy economic outlook
largeI)
> In case of matching maturities no such problem exists because a bank can
simply stop to renew maturing assets e.g. loans if funding becomes difficult
> Equity is sensitive to asset price movements in case of mark-to-market

> Institution's equity capital that securities in the trading book
serves as margin in leveraged
> Securities marked to market hit by negative price shock demand an
positions is sensitive to asset
immediate reaction on the asset side of the balance sheet (liquidation)
price movements
> That, in turn, increases the risk of liquidations under fire-sale conditions
I) Maturity mismatch in terms of short-term uninsured debt (not deposits!) serves as funding for longer-term assets
Figure 5.3: Catalysts to the loss and the margin spirals
The fact that securities are marked to market has another important effect, producing
the following externality: Not only the bank selling the asset has to write down the
difference between the book value and the price of the sale, but so too do all holders
of the same asset. This results in further (mark-to-market) equity losses at all affected
institutions which, in turn, triggers a vicious cycle: the aggravated asymmetric
information problem, the need to sell assets even at fire-sale prices to keep leverage
ratios from rising too much and thus cutting the bank off from funding liquidity, the
depreciation of assets, additional equity losses, and so on.
The factors acting as catalysts to the loss spiral are high leverage, a large maturity
mismatch and the fact that an institutions equity is subject to marked-to-market asset
price movements as shown in figure 5.3 (see also European Central Bank (2010a), p.
139). The overall mechanism of the loss spiral is illustrated in figure 5.4.
Needless to say, if a bank finds itself in a position where it has to liquidate assets at
fire-sale prices, there is no room to expand lending because of the need to shrink the
asset side of the balance sheet.
Asset fire
sales
Initial equity loss Funding problemsI) Asset depreciation
Further equity
losses on held
positions
Decreased leverage
= Loss spiral tolerance/
higher margin
= Margin spiral/leverage cycle requirements
I) Funding problems: Institutions with need for funding encounter difficulties to raise short-term funding due to an asymmetric information problem
Source: Adaption from European Central Bank (2010), p. 141, and Brunnermeier (2009a)
Figure 5.4: The loss spiral and the margin spiral/leverage cycle
Against the background of the crisis, it is interesting to see that, while the friction was
originally sparked off in the market for subprime mortgages, it also spilled over to
other asset classes, as is apparent from the breakdown of the jumbo mortgage
markets, the markets for asset-backed commercial papers and those for
collateralized debt obligations, for example (see Greenlaw et al. (2008)). This effect
had already been anticipated by Shleifer and Vishny (1992) in an example of fire
sales of real assets.
The loss spiral is not the only mechanism that contributed to the propagation of the
recent financial crisis. The second, closely related, mechanism is the margin spiral or
leverage cycle.
5.2. The margin spiral or leverage cycle
A sibling of the loss spiral and equally involved in propagating the crisis and
promoting its contagious effects is a concept to which Brunnermeier (2009a) refers to
as the margin spiral, while Geanakoplos (2010) refers to it as the leverage cycle. It
can be seen as a mechanism that enhances the sequence described in the concept
of the loss spiral.
One assumption made in the case of the loss spiral is that banks react to initial
losses in their equity by selling assets in order to keep their leverage ratio constant.
In the case of the margin spiral, this assumption is abandoned in favor of a more
realistic, but also more severe one: As the two cited studies argue, leverage is
cyclical, with leverage ratios increasing in boom times and declining in recessions a
picture that is empirically supported by Adrian and Shin (2008b and 2010a) and
Hanson et al. (2011). In short, the margin spiral centers the cyclicality of the leverage
ratios of financial intermediaries.36
As in the case of the loss spiral, when an economic downturn is perceptible, initial
losses to a banks equity force a bank to sell assets. Given that, in a recession,
borrowers have stronger incentives to act in a morally hazardous way, lenders of
uninsured funds tend to have a reduced tolerance for leverage at such times.
The reduced tolerance for leverage is not only motivated by lenders seeking to avoid
being negatively affected by moral hazard but also by the following, according to
Brunnermeier and Pedersen (2009): The first unexpected asset price decreases
could be interpreted by providers of uninsured debt as warning signals, and the
higher volatility in the markets may lead to higher margin requirements. For
borrowers, this is particularly harmful when resorting to short-term debt. This is
36
Leverage ratios for households show exactly the opposite pattern: Households do not tend to
adjust their balance sheets in response to asset price increases. For example, on aggregate,
house owners who have raised a mortgage for which the house serves as collateral usually do not
take on more debt when house prices rise (see Greenlaw et al. (2008)).
5.2 The margin spiral or leverage cycle 51
because short-term debt has to be rolled over more frequently so that changes in
margin requirements take effect immediately.
To meet the lenders reduced requirements, borrowers must liquidate assets on a

larger scale (as compared to the loss spiral with constant leverage ratios), causing
even more losses due to fire sales stemming from the considerably greater need for
depreciation. This is anticipated by borrowers, who demand even more collateral or,
on the other hand, tolerate even lower leverage ratios. This aggravates the funding
problems over and above the loss spiral which is at work at the same time. The
margin spiral/loss cycle is depicted in figure 5.4.
The pressure on banks to sell assets in economic crises may further be aggravated
by risk-management procedures that are used in banks resulting in even more
pronounced procyclical leverage ratios. Brunnermeier (2009b) and Greenlaw et al.
(2008) state this by pointing to the countercyclical character of conventional value-at-
risk measures: In recessions, volatility tends to be higher than in boom phases,
leading to higher possible losses in accordance with the value-at-risk methodology.
The assumption is that a bank can only hold an amount of assets whose value at risk
(or possible loss) must correspond to the equity capital that is available to absorb
possible losses. This means that, in a recession, a given equity capital endowment
only allows for a smaller balance sheet size or less leverage due to higher value-at-
risk estimates. In a boom phase, the opposite is the case. Since banks manage their
balance sheets on the basis of value-at-risk estimates, their leverage correlates
positively to the business cycle.37
A possible question is why banks do not, in anticipation of this recurring pattern, raise
sufficient equity capital and build up adequate buffers before a crisis.38
Clearly, once a crisis is underway and a bank is struggling to repay its debt, it might
already be too late to do so because of the debt-overhang problem identified by
Myers (1977): New equity that a bank raises to avoid having to reduce assets would
then simply be siphoned off by more senior credit tranches.
37
Note that to lenders of uninsured funds higher leverages are tolerable in boom times primarily due
to the rising prices of those assets that serve as collateral. The opposite is the case in a recession.
38
This question is also valid in the context of the loss spiral but especially indicated against the
background of cyclical leverage ratios.
An explanation offered by Stein (2012) is that banks prefer short-term debt over
equity capital (and over long-term debt) because it is cheaper.39 The risk of default to
investors is much lower and, consequently, the price of this kind of funding is lower,
which leads to excessive bank leverage ratios. The problem with this excessive
leverage is that banks do not internalize all the associated costs when they have
problems rolling over their short-term debt in a recession and are thus forced to
liquidate assets under fire-sale conditions. As already pointed out, the decline in
value of the sold assets triggers write-downs at all banks that hold the same assets.
To conclude this chapter, it is worth briefly examining why the margin spiral in the
recent financial crisis was worse than previous spirals. Geanakoplos (2010) believes
that the recent crisis was characterized by several elements that have not occurred in
previous spirals. The most important one is that leverage ratios were higher than
ever before, leading to balance sheet adjustments on an unprecedented scale. In
addition, there were actually two mutually reinforcing margin spirals: one in the
housing market and a second one in the market for mortgage securities. Moreover,
Geanakoplos (2010) claims that, with the development of credit default swaps (CDS),
investors have been able to bet against a given underlying asset, thereby further
pushing down asset prices.
What conclusions can therefore be drawn from this discussion? And what
implications do the two spirals have for bank lending?
5.3. Conclusion
The loss spiral and the margin spiral/leverage cycle are both phenomena that occur
at times of crisis. They become problematic for the financial system as a whole in an
economic downturn due to the negative externalities they produce. The speed and
severity of a crisis can be thought of as being related to the factors that promote
these spirals.
Their mechanics suggest that certain bank characteristics or properties of banks

balance sheets affect bank lending differently at normal times compared to times of
crisis. In the course of the review of empirical evidence and especially by deriving
39
This assumption is a deviation from the Modigliani-Miller theorem (Modigliani and Miller (1958)),
the capital structure irrelevance principle.
5.3 Conclusion 53
implications from the new view on the bank lending channel for the interpretation of
available empirical results in the next chapter (chapter 6), it will become evident how
the bank characteristics that play a special role during crisis periods fit in the
framework which has been already developed in section 4.2.
Precisely those characteristics or balance sheet properties that set the spirals in
motion and that are to be held accountable for the gravity of the course the crisis
takes are of primary interest when comparing normal to crisis periods in the empirical
analysis of this study. So which characteristics are they?
First, it has become clear that high leverage is an important factor. While high
leverage does not pose a problem in itself during normal periods, it becomes harmful
to a bank during a crisis because initial equity losses involve asset liquidations that
amounting to values equal to the loss multiplied by the leverage.
Second, while a maturity mismatch can already be a sign of generally elevated

funding risk even at normal times, this risk materializes in full force during a crisis.
When investors withdraw their funds and funding markets start to dry out, rolling over
short-term debt becomes much more difficult. This increases the risk that a bank will
have to liquidate even long-term assets at short notice and leaves no room to expand
lending activities quite the contrary. In addition, even outside of a real banking
crisis, an inverted yield curve 40 means that assets with longer maturities must be
financed using more expensive short-term debt. This is an additional source of risk
stemming from maturity mismatches.
Third, the share of mark-to-market securities increases a bank's exposure to asset

price volatility. In an economic downturn, the increased volatility of the assets held
leads directly to equity losses. Moreover, the sale of a security by a bank that is also
held by other banks causes the other banks to write down the value as well. This is
part of the negative externality described above.41
40
An inverted yield curve shows higher interest rates for short maturities than for longer maturities.
41
The vulnerability to asset price depreciations of banks balance sheets characterized by a
significant volume of mark-to-market securities on their trading books drove many banks to
reclassify all eligible assets and transfer them to the held-to-maturity portfolio during the recent
crisis, in accordance with amendments to IAS 39 and IFRS 7 (see International Accounting
Standards Board (2008)).
Fourth, while counting as short-term (but insured) debt, a large proportion of funding
in the form of insured retail deposits diminishes dependency on wholesale funding
markets. Again, when liquidity on these markets evaporates as the recession
worsens, a bank that funds itself to a smaller extent by means of short-term
wholesale debt is more shielded by these adverse conditions.
These considerations suggest that there are, indeed, bank characteristics that play a
different role during a crisis compared to normal times. Empirical verification of this
claim is one of the central concerns of this study and is part of the empirical analysis
in chapter 7.
All bank characteristics identified in this chapter will be incorporated in the framework
shown in figure 4.4 which is further developed in section 6.4. This implies that the
framework is also capable of capturing the effect of the crisis on bank lending. Before
that, available empirical evidence is reviewed in the next chapter, on basis of which
the exact research gaps are derived.
6.1 Remarks on the difference between the US and the euro area 55
6. Review of empirical evidence on bank lending and its

implications
This chapter deals with the empirical evidence that has been produced so far
regarding bank lending and the bank lending channel. Furthermore it focuses on
implications for the interpretation of the empirical evidence in light of the new view on
the bank lending channel introduced in chapter 4.
To validate the entire bank lending channel mechanism from central banks via banks
to the real economy, it would be necessary to provide evidence substantiating at
least the first and the second conditions.42 However, a complete proof of the bank
lending channel is not within the scope of the present study. The aim of this research
undertaking is to identify the driving factors behind the expansion or contraction of
bank lending. The empirical approach necessary to do so is closely related to an
analysis of the validity of condition 1. Accordingly, the review of relevant empirical
literature is confined to the scope of the first condition of the bank lending channel.43
The specific research questions and the hypotheses to be tested in the empirical
section emerge from gaps in this literature.
There are various dimensions along which the literature can be organized, the most
important of which are the country dimension, aggregate versus bank-specific data,
and a chronological order. Since the first empirical tests were carried out using US
data, starting with evidence from the US best captures the development and
advancements that have been made in the past two-and-a-half decades.
The results of early bank lending channel tests are based on aggregate data. They
have been deemed unsatisfactory for the reasons outlined below. In the course of
econometric advancements toward panel estimation techniques, the characteristics
of individual (US) banks came to be the center of attention.
42
The validity of the third condition prices must not adjust instantaneously and monetary policy
must not be neutral is taken as given.
43
A detailed treatment of empirical work on the second condition would be too far off focus and is,
therefore, not justified.

56 6 Review of empirical evidence on bank lending and its implications
In order to capture the progress made in understanding the drivers of bank lending
and the bank lending channel, the following sections first reconstruct the academic
debate in the US based on aggregate and, subsequently, bank-specific data.
Following on from the US case, the empirical evidence produced in Europe is
presented with a focus on the euro area, again based on bank-specific data.44 The
empirical literature review also includes an overview of existing contributions that
already deal with the most recent financial crisis.
The chapter concludes with focusing on implications that can be derived from the
empirical literature against the background of the new view on the bank lending
channel outlined in chapter 4. An attempt is made to integrate all bank characteristics
that have been identified to have an impact on the supply of bank loans into the
framework that has been presented in figure 4.4. By applying this framework which
leads to a systematization of bank characteristics driving bank lending, the basis is
built not only for the formulation of hypotheses to test and but also for the empirical
analysis as a whole.
Before reviewing the empirical literature and discussing how the available evidence
can be reconciled with the new view on the bank lending channel, attention is drawn
to some institutional differences between the US and the euro area whose
implications for bank lending should be borne in mind.
6.1. Remarks on the difference between the US and the euro area
It is a well-known fact that there are substantial differences in the financial structures
of different countries and, in particular, between the United States and countries
belonging to the euro area. As shown in figure 6.1, enterprises in the euro area rely
much more heavily on bank financing than their counterparts in the US. On the
contrary, US companies avail themselves of the financial markets to a much greater
extent, as reflected in the higher share of issued debt securities and the higher stock
market capitalization (measured as shares of GDP) compared to the euro area.
Does this mean that the bank lending channel is relevant in the euro area but
irrelevant in the US? Care must be exercised in deriving a priori implications from this
44
When researchers started to focus their attention on European countries, analyses based on
techniques involving aggregate data were already outdated and are, therefore not available.
6.1 Remarks on the difference between the US and the euro area 57
pattern for the importance of the bank lending channel. One could be tempted to
conclude that the economic impact of monetary policy on bank lending and the
overall economy must be stronger in a country or region in which the bulk of
corporate debt is reflected in assets on banks balance sheets, as opposed to a
country in which a relatively low share of corporate debt involves bank lending. On
the contrary, however, not only the markets for debt that are tapped by corporations
are more highly developed: The same also goes for the debt markets that are tapped
by banks. Indeed, by aggravating the potential adverse selection and moral hazard
problems merely through the higher volume of wholesale funding, banks might
actually be even more sensitive to changes in monetary policy (as discussed in
chapter 4), thereby amplifying the function of the bank lending channel.
Bank loans to corporate sector Debt securities issued by corporate Stock market capitalization
[in percentage of GDP, 2012] sector [in percentage of GDP, 2012] [in percentage of GDP, 2012]
47,9% 119,0%
41,0%
51,7%
18,7%
10,7%
US Euro US Euro US Euro

area area area
Source: US FED; ECB; Eurostat; Bank for International Settlements; The World Bank; International Federation of Stock Exchanges; own calculations
Figure 6.1: Financial structure in the US compared to the euro area
In the end, when judging economic significance, there are many reasons to believe
that the role of the individual bank characteristics must be examined carefully in the
context of the financial structure of a given economy, and also in the context of the
prevailing economic conditions (e.g. during or outside a crisis period) before a
meaningful conclusion can be reached. For this reason, most of the following
sections on empirical evidence are devoted to the impact of bank characteristics.
6.2. Empirical evidence from the US
6.2.1. US evidence based on aggregate data
Beginning with evidence from the US, a first promising empirical attempt to establish
the claim that monetary policy works in part through its effect on the availability of
bank loans, over and above the traditional interest-rate channel, is presented by
Bernanke and Blinder (1992) on the basis of aggregate data. 45 Building on their
earlier work (Bernanke and Blinder (1988)), they show that loans respond to changes
in monetary policy. But as they admit themselves, this finding is also consistent with
the interpretation that the decline in bank loans merely reflects falling demand for
credit caused by demand factors.
An alternative approach by Kashyap et al. (1993) investigates the composition of

firms' external finance to identify a bank lending channel. These authors show that, in
response to monetary tightening, the share of commercial papers rises while the
share of bank loans declines. If this decline in bank loans were merely reflecting
depressed economic conditions, one would expect commercial papers to decline by
a comparable proportion, leaving the overall composition unchanged. Since this is
not the fact, the empirical evidence gives rise to believe that the supply of bank loans
is reacting to monetary policy.
The evidence presented by Ludvigson (1998) is based on automobile loans and goes
along the same lines. The finding is that, after a contractionary monetary shock, the
ratio of banks' automobile loans falls relative to the sum of bank and non-bank
automobile loans, which indicates a decline in the supply of consumer loans by
banks.
One objection to results obtained in the manner adopted by Kashyap et al. (1993),
put forward by Oliner and Rudebusch (1996b), is that the observed change in the mix
of bank loans and commercial papers is not necessarily due to a decline in the
supply of bank loans. They suggest that the rising share of commercial papers could
also be attributed to heterogeneous demand for credit: Large firms account for the
bulk of commercial papers, while small firms usually do not have access to the
45
A list of selected empirical research from the US on bank lending and the bank lending channel is
provided in table 6.1
6.2 Empirical evidence from the US 59
commercial paper markets. Now, if small firms are hit harder than large firms by a
demand shock probably because small firms are more vulnerable to the business
cycle the composition of bank loans and commercial papers will change without a
decline in the supply of bank loans.
6.2.2. US evidence based on data from individual banks
Advances in panel estimation techniques made it possible to focus on the

characteristics of individual banks and their role in the transmission process of
monetary policy. Specifically, it is now possible to analyze the extent to which
different bank characteristics account for the varying cross-sectional responses of
bank lending to monetary policy shocks.
Most studies in this vein are primarily geared to exploiting the differences between
banks in the way they access alternative forms of funding. These contributions
explore those characteristics that are assumed to affect the magnitude of the
asymmetric information and moral hazard problems (and, ultimately, the external
finance premium) that is involved when banks are themselves borrowers of
uninsured sources of funding. In terms of conditions of the bank lending channel this
connection is stipulated in the second subcondition of the condition 1.
Other contributions focus on the analysis of the relationship between the holdings of
liquid assets and bank lending which is the theoretical background of the first
subcondition.
Studies arguing that the ability to expand lending is affected when banks are subject
to binding self-imposed or regulatory capital constraints are related to the third
subcondition of condition 1.
In the discussion that follows it has adhered to the numerical sequence of the
subconditions (subcondition 1, subcondition 2, subcondition 3). This style is preferred
to ensure congruence with the theoretical foundations of the bank lending channel as
discussed in sections 3.1.1 and 4.2, and to make the present study easier to read.
Geographical Sample No. of banks and

Paper focus period no. of observations (Main) Data source Main findings
Bernanke / Lown US 1981- Based on aggregate data FED: Shortage of capital responsible for slowdown of
1991 1991 call reports lending
Bernanke / US 1957- Based on aggregate data DRI Loans respond to monetary policy change
Blinder 1992 1989
Hall US 1998- 11,507 banks (in '91) FED: Introduction of Basel I accord caused slowdown
1993 1992 No. of obs.: n.a. call reports in lending
Kashyap et al. US 1964- Based on aggregate data FED: Mix of company financing consisting of bank
1993 1989 flow-of-funds loans and commercial paper (CP) changes
publication toward CP after monetary policy change
Berger / Udell US 1979- No. of banks: n.a. FED: Lending was reduced in reaction to Basel I
1994 1992 >600,000 obs. call reports accord to reduce funding costs and not due to a
(unbalanced) direct capital constraint
Kashyap / Stein US 1976- 14,280 Banks (in '84) FED: Small banks reduce loan portfolio more sharply
1995 1992 3,360 obs. (balanced) call reports in reaction to monetary policy rate increase
Peek / Rosengren US 1976- No. of banks: n.a. FED: Capital-restricted banks reduce lending more
1995a 1994 No. of obs.: n.a. call reports significantly then non-restricted banks
Peek / Rosengren US 1989- 150 banks FED: Banks for which capital constraint becomes
1995b (New England) 1992 No. of obs.: n.a. call reports binding due to regulatory enforcement actions
reduce loans faster in response to low capital
ratios than unbound banks
Peek / Rosengren US 1990- 407 banks FED: Better capitalized banks shrink lending by less
1995c (New England) 1991 No. of obs.: n.a. call reports after hit by negative capital shock
Peek / Rosengren US 1989- n.a. (only US branches of FED: Capital ratio of Japanese parent of US branch
1997 1995 Japanese banks) call reports has significant impact on lending
370 obs. in the US
Ludvigson US 1965- Based on aggregate data FED: G.19 & G.20 Monetary tightening results in relative decline of
1998 1994 statistical release bank loans used for automobile finance
Kashyap / Stein US 1976- 11,206 banks (in '93) FED: Banks with little holdings of liquid assets are
2000 1993 961,530 obs. call reports most responsive to monetary policy changes
(unbalanced) (esp. small banks)
Kishan / Opiela US 1980- 13,042 banks FED: (Small) banks with higher capital ratios are less
2000 1995 No. of obs.: n.a. call reports sensitive to monetary policy changes
Campello US 1981- No. of banks: n.a. FED: Small banks that are affiliated to a multibank
2002 1997 547,390 obs. call reports holding react less to monetary policy tightening
(unbalanced)
Kishan / Opiela US 1980- No. of banks: n.a. FED: Banks with low capital ratios react more
2006 1999 No. of obs.: n.a. call reports sensitive to monetary policy tightening than to
monetary policy expansion
Holod / Peek US 1986- No. of banks: n.a. FED: Loans portfolio of publicly traded banks are less
2007 2005 675,546 obs. call reports and sensitive to monetary policy tightening
(unbalanced) FR Y-9C
Berrospide / Edge US 1992- 140 banks (holdings) FED: Consolidated Modest effect of low capitalization on credit
2010 2008 11,099 obs. (unbalanced) Fin. Statement for supply; sizable impact of scarce liquidity
Holding Companies
(FR Y-9C)
Table 6.1: Selected empirical research from the US on bank lending and the bank lending channel
Holdings of liquid assets: Turning to evidence regarding the question whether

holdings of liquid assets have an impact on bank lending as this is the essence of
subcondition 1 leads us to a study composed by Kashyap and Stein (2000). These

authors test the notion that banks with large holdings of liquid assets should be better
able to shield their loan portfolio from monetary policy shocks because, instead of
adjusting their balance sheet by resorting to external funding (i.e. on the liabilities
side), they can respond to the effects of monetary policy by selling securities (i.e. on
the asset side).46 Based on a large sample covering every insured commercial bank
in the US over more than 15 years, they find that the more liquidity a bank holds
(measured as the ratio of securities to total assets), the lower is its sensitivity to
monetary policy shocks an effect they mainly attribute to small banks. Berrospide
and Edge (2010) also report a positive relation between securities holdings and bank
lending for bank holding companies.
Size (as a sign of transparency and also correlated with degree of diversification of
business/income sources and professionalism): Historically, the first idea when
seeking to identify those bank characteristics that impact lending was to test the size
of a bank in terms of its balance sheet, usually measured in total assets. Bigger
banks are thought to be more professional and more diversified in terms of income
sources, business lines etc. resulting in a lower risk. Furthermore, bigger banks
should have more advanced controlling and reporting procedures in place, leading to
a higher degree of transparency for investors. This results in lower risk premiums,
more favorable conditions at which the banks can borrow and lend, and a bigger loan
portfolio. Kashyap and Stein (1995) support this empirically by finding that smaller
banks reduce their loan portfolio to a greater extent in response to an increase of the
federal funds rate.
Stock exchange listing: Another characteristic which, like size, is linked to the
transparency of a bank from the investor's perspective (see figure 4.4 and also figure
6.2) is whether a bank is listed on a stock exchange. Listed companies must fulfill a
number of disclosure requirements stated in stock exchange market regulations, and
this distinguishes them from non-listed companies. The result is that listed companies
make more information available to the public, as Healy and Palepu (2001) report in
46
A theoretical model which incorporates the impact of liquid asset holdings on the sensitivity of
monetary policy changes has been developed by Stein (1998).
a survey of empirical disclosure literature. This enhanced transparency should result

in a lower external finance premium and, therefore, in better access to uninsured
external funding, which is exactly the hypothesis that Holod and Peek (2007) test.
According to them, whether a bank is publicly traded or not is a better, i.e. more
direct indicator of transparency than size, which is only an indirect measure. Their
results show that listed banks are better able to raise large time deposits than
otherwise similar non-listed banks. As a consequence, the loan portfolios of listed
banks are, indeed, less sensitive to monetary policy shocks than those of non-listed
banks, as their results indicate even after controlling for balance sheet size and
capitalization (capital-to-assets ratio; not risk weighted). Moreover, Holod and Peek
(2007) conclude that the explanatory power of the size of a bank as indicating the
ease of access to alternative forms of funding is dominated by whether or not it is
listed.
Affiliation to a bigger bank network: One interesting idea is raised by Campello

(2002). According to this author, banks that are affiliated to a multibank holding
conglomerate should have better access to funding due to internal capital markets.
Supporting his hypothesis, he finds evidence that small banks affiliated to a bank
network are better able to shield their loan portfolio from adverse monetary policy
shocks than small but unaffiliated banks. This mechanism is supported by Ashcraft
(2006), who finds that negative lending responses to adverse monetary policy shocks
are mitigated in the case of banks that are affiliated to multibank holding companies.
Capitalization (as a signal of a bank's riskiness to providers of debt capital): In this

context, there are two sides to capitalization. First, it reduces the incentive to take
undue risks that run counter to the interests of the lender (moral hazard). Second, the
amount of capital is an indicator of the buffer that is available to absorb losses. Both
aspects of capitalization add to the perceived riskiness of a bank.47 The effect of an
47
In this context, capitalization is not related to the notion of whether a bank is facing a binding self-
imposed or regulatory-induced capital constraint. The empirical evidence for the latter issue is
discussed in this section below. Distinguishing between the different notions of capitalization has
relevant theoretical and practical implications (see the end of section 6.4 for a theoretical
discussion).
institution's capitalization in combination with its size has been examined by Kishan
and Opiela (2000). According to the authors, the smallest and least capitalized banks
are the ones most sensitive to monetary policy changes. They counter the possible
refutation that this finding is also in line with the balance sheet channel by drawing on
delinquency rates: An active balance sheet channel implies an inverse relationship
between the quality of bank borrowers (measured in terms of delinquency rates) and
capital. Since this inverse relationship does not hold, Kishan and Opiela suggest that
the bank lending channel is at work.
In an extension of their earlier study, Kishan and Opiela (2006) examine whether
capital is different in its effect on bank lending in an expansionary environment as
opposed to a contractionary monetary policy environment. First, in line with the
preceding empirical literature, they find asymmetric loan responses to monetary
policy shocks between well capitalized and sparsely capitalized banks. The
distinction made between the two directions of monetary policy reveals that sparsely
capitalized banks react more strongly to a contractionary monetary policy than well
capitalized banks. In addition, the former do not expand lending to the same degree
as the latter in response to an expansionary monetary policy. The authors ascribe
this effect to the binding application of the Basel capital requirements to small banks.
The reaction of well capitalized banks is inverse: They do not react strongly to
negative monetary policy shocks, but expand lending (especially mortgage lending)
in response to expansionary monetary policy shocks. These patterns are only valid in
the post-Basel era, whereas the minimum capital requirements in place in the pre-
Basel era were not consistently enforced.
A slightly different picture is drawn by Berrospide and Edge (2010). Although their
results also suggest a positive relationship between capital and loan growth, the
effect is small in quantitative terms. Their principal interpretation is that banks do not
actively manage their assets (and especially loans) as a function of their capital
endowment. Against the backdrop of any real-life banking experience, this argument
is barely conceivable and contrasts with that of Berger et al. (2008), for example, who
show that banks do indeed actively manage their capital ratios, including both capital
and (risk-weighted) assets.48 A different picture is also drawn by Gropp and Heider
(2010), who find that banks seek to comply with self-chosen capital targets.
A more recent contribution is a study by Carlson et al. (2013). In order to largely rule
out demand effects on loan growth, they match banks of comparable size in the
same geographical area and analyze the extent to which differences in capital
endowment can explain loan growth differentials. They find that there is a positive
relationship between capital and loan growth, but only during the recent financial
crisis and not in the period before, which is in contrast to findings of other scholars.
Their conclusion is that, at normal times, banks have little difficulty managing their
capital by retaining earnings or issuing equity capital, for example. In times of
turmoil, however, the depressed economic situation impacts earnings negatively and
it is also much more difficult to raise new equity on the markets.
There is a huge body of empirical literature dealing with the argument that banks
constrained by binding (regulatory or self-imposed) capital requirements
(subcondition 3) play a lesser part in the transmission of monetary policy shocks. To
a significant extent, this is to be ascribed to the attention scholars have given to the
discussion of the Basel (I, II and, most recently, III) frameworks.
Before examining the details of the findings obtained regarding this argument, one
distinction must be noted: Capital in this sense is not used in the notion of reducing
risk for a lender who provides uninsured funding.49 In this case, the focus is rather on
capital as a way to determine whether or not a bank is operating under capital
constraints.
Whether a constraint is binding for regulatory or self-imposed reasons is irrelevant

since, like regulatory requirements, self-imposed constraints too can be considered
as not arbitrarily variable, at least in the short run. This is plausible, because the
change of a self-imposed capital ratio will not be without consequences for the
48
A second interpretation of the authors is that the measure of capital they use is not informative for
assessing the capital position of banks. This interpretation is also not satisfying as it questions
their entire approach in one of its key points.
49
Capitalization as an indicator of a bank's riskiness in the perception of providers of debt capital is
discussed in this section above.
riskiness perceived by potential lenders and, therefore, for a bank's access to

uninsured funding.
The question whether a bank is working under a binding constraint demands a yes or
no answer: Either yes, a bank is operating under capital constraints, or no, it is not. If
it is, the bank lending channel should be muted, because even if enough funding
(debt) is available it lacks the required equity capital.
Beginning with studies that focus particularly on capital constraints in the wake of the
implementation of the Basel I capital rules and the US credit crunch of the early
1990s, one of the first contributions was made by Bernanke and Lown (1991). They
find evidence for the claim that a shortage of equity capital is responsible for the
slowdown in lending during the credit crunch in the early 1990s.
Analyzing the same period, Hall (1993) finds that the Basel I accord that took effect in
1989 was accompanied by a decline in lending. His conclusion is that this was due to
the incentives the Basel accord provided for banks: Since loans, especially to
businesses, demand large amounts of capital relative to other investments, banks
decide to constrain lending in favor of alternative investments, thereby contributing to
the credit crunch.
Berger and Udell (1994) come to a similar conclusion, although they prefer
alternative hypotheses that are equally consistent with the data, e.g. a voluntary
reduction of leverage in order to reduce the cost of funding, rather than a reduction
directly induced by regulation.
Peek and Rosengren (1995b) relate the credit crunch explicitly to the regulatory
enforcement actions to which some banks have been subject. When the Basel I
capital rules were implemented in the US and supplemented by additional regulatory
standards to account for risks not covered by the Basel I framework (e.g. interest-rate
risk), banks in New England were the first to be faced with regulatory enforcement
actions.50 Since their capital ratios were too low under the new regulatory standards,
they had to significantly shrink their loan portfolios. In light of this observation, Peek
and Rosengren (1995b) conclude a direct link between regulatory actions and bank
lending. This view is supported by Furfine (2001), who draws this conclusion based
50
Banks in New England were the first because they were the first to sign formal accords with the
regulatory authorities.
on a structural dynamic model using a representative bank whose implications are

subsequently tested with real US data.
Turning to evidence that abstracts from the US credit crunch in the early 1990s, two
further studies are to be mentioned. Although the first is not explicitly related to
literature on the bank lending channel, Peek and Rosengren (1995c) first show that,
theoretically, banks subject to capital constraints react to negative shocks to their
capital by reducing assets to a greater extent than unconstrained banks. This effect
applies particularly to loan portfolios. The authors then show empirically that banks
with lower capital-to-asset ratios behave in line with the predictions derived from their
theoretical model: After controlling for demand factors, particularly by using a cross-
section of banks in New England that are assumed to face similar economic demand
shocks, the better capitalized banks shrink their lending to a lesser degree after
being hit by a negative capital shock.
Extending their analysis explicitly to the responses of capital-constrained and

unconstrained banks to monetary policy shocks, Peek and Rosengren (1995a) report
at least slight evidence for the assumption that unconstrained banks' lending
behaves in line with predictions based on the concept of the bank lending channel.
In an attempt to isolate loan supply effects and to avoid the objection that the decline
in the observable loan volume can be ascribed to demand factors, two studies take a
different path to identify the impact of capital. Peek and Rosengren (2000) and
especially Peek and Rosengren (1997) use the Japanese banking crisis of the early
1990s as a natural experiment. In the late 1980s, Japanese stock markets suffered
from significant declines. Due to the fact that large volumes of the stock of Japanese
corporations are traditionally held by Japanese banks, these adverse stock market
developments translated into losses that weakened the capital positions of Japanese
banks. At the same time, the new Basel (I) Accord was introduced, resulting in
stricter requirements for capital adequacy. These events put considerable pressure
on the banks, with the result that they primarily reduced assets outside Japan,
namely in the US, in order to restore acceptable ratios as measured against self-
imposed or regulatory targets. Since the shock to the Japanese stock markets did not
correlate to demand conditions in the US, the lending response of Japanese banks in
the US can be regarded as reflecting supply factors only. As a result, Peek and
Rosengren (1997) report that a decline of one percentage point in the risk-based
capital ratio of US branches/agencies of Japanese banks led to a 6% decline in loans

during this period.
6.2.3. Conclusion
What, then, is the bottom line with regard to the determinants of bank lending in the
US? Taking all factors into account, there is ample empirical evidence to support the
hypothesis that the central bank is able to affect the supply scheme of loans. This
suggests that the differences in the increase or decrease of banks loan portfolios
can be attributed to banks characteristics.
Specifically, more liquid, more transparent banks (reflected in the size of a bank and
the fact whether or not it is listed) and banks that are affiliated to a network are better
able to shield their loan portfolios from monetary policy shocks, even though the
details of interdependencies relating to size, listing and affiliation do not seem to be
clear (e.g. under what precise conditions does size cease to be relevant in explaining
lending?). Moreover, the fact that banks subject to binding capital constraints react
by cutting down on lending is also documented. However, the various studies that
mainly focus on the regulatory constraint also neglect the impact of self-imposed
capital constraints to a large extent. Capital (as an indicator of the riskiness of a bank
to providers of debt) has likewise been found to be an influential factor: The better a
bank's capital endowment, the lower is its sensitivity to monetary policy shocks. This
finding suggests that capital could be a means to alleviate problems associated with
asymmetric information and moral hazard, although there is some evidence to
suggest that the impact of capital might not be constant over time.
6.3. Evidence from the euro area
This section examines those contributions that have a geographical focus on

European countries. Special emphasis is given to literature that reviews the recent
financial crisis. 51, 52
Toward the beginning of stage three of the European Economic and Monetary Union
and the introduction of a common monetary policy, scholars addressed the question
whether there is an active bank lending channel in each of the individual European
countries and in the countries forming the euro area as a whole.
On the one hand, a bank lending channel might be particularly effective owing to the
accepted fact that banks in the euro area are, generally speaking, more important to
the provision of debt than their counterparts in the US. On the other hand, it must be
admitted that, in line with the new view on the bank lending channel, the role of
deposits no longer receives the same emphasis (see section 4.1): The ECB always
provides the amount of liquidity that is required by the banks in the system to meet
their reserve requirements, subject to the condition that collateral of an accepted
quality can be provided.53
This fact which, at first sight, challenges the central banks ability to affect the supply
scheme of bank loans (condition 1), has recently begun to attract attention. The
question is whether there is still an empirically identifiable effect on the conditions
under which banks are able to raise funds. (The theoretical foundation of the belief
that there might well be an effect is discussed in chapter 4).
51
It should be noted that, in this section, the structure used to present evidence from the euro area
differs from that based on data from the US in section 6.2. First, there is no section on evidence
based on aggregate data. This is due to the fact that, when researchers began to focus their
attention on European countries, analyses based on techniques involving aggregate data were
already outdated. For this reason, no such evidence exists for the euro area. Second, the
evidence from the euro area is subdivided into a pre-crisis section and a section that takes the
effects of the crisis into account. The reason is that, since the recent crisis period is of particular
interest to this research undertaking, the literature already available merits special attention.
52
A list of selected empirical research from the euro area on bank lending and the bank lending
channel is provided in table 6.2.
53
For details of the operational framework of the ECB, see Bofinger (2001), p. 300 et seq. and
European Central Bank (2011b), especially chapter 4.2, p. 96 et seq.
6.3 Evidence from the euro area 69
Evidence about whether the central bank (ECB) is able to affect the supply of bank
loans in the individual countries of the euro area dating from the beginning of the first
decade of this century is mixed. Evidence for supply effects has been found in some
countries, but not in others.
Over time, more data on the consolidated euro area has become available. At the
same time, the deregulation of the financial sector has advanced and the significance
of financial innovation has grown into new dimensions.54 These events again fueled
both the discussion of the bank lending channel in general and the question whether
the ECB is able to affect the supply scheme of bank loans in particular and also
improved the chances of arriving at a conclusive answer.
Analysis of the interaction of these developments with the recent financial crisis and
the implications for bank lending has only yielded preliminary findings but is, at the
present time, far from complete.
6.3.1. Euro area evidence from before the crisis
The earliest studies of the euro area made several important observations. Apart
from the fact that, compared to the US, more corporate debt takes the form of bank
loans rather than bonds (see figure 6.1), the banking system in many large European
countries is characterized by major bank networks (see e.g. Ehrmann et al. (2003)).
Taken as an indication that informational asymmetries play a lesser role, this is
responsible for the finding that the size factor (total assets of a bank) is insignificant
in many cases that use data on European banks, as Ehrmann et al. (2003) and
Ehrmann and Worms (2004) report. How is this finding to be understood? The
studies mentioned (along with many subsequent studies) attribute the absence of a
size effect to the fact that a large proportion of small banks belong to a bigger bank
network. In the event of funding difficulties that make it harder to roll over debt, intra-
network flows of liquidity supply funds to the institutions affected to avoid reductions
in their loan portfolios, as exemplarily shown for Germany by Ehrmann and Worms
(2001). In addition, European small banks are especially strong in relationship
lending and might therefore be particularly reluctant to constrain lending, preferring
54
A study by Weber et al. (2009) suggests that the time period analyzed may indeed make a
difference. They discovered structural breaks in euro area data in 1996 and 1999.
instead to sell liquid assets. These factors suggest that size is not a decisive factor
for differences in the response of banks to monetary policy changes.
Geographical Sample No. of banks and

Paper focus period no. of observations (Main) Data source Main findings
Ehrmann / Worms Germany 1992- 3,665 banks Deutsche Bundes- Existence of bank networks important to
2001 & 2004 1998 No. of obs.: n.a. bank: balance sheet explain lending reaction of banks to monetary
statistics policy changes
Worms Germany 1992- 2,659 banks Deutsche Bundes- Loan portfolios of banks with large short-term
2001 1998 58,374 obs. (unbalanced) bank: balance sheet interbank deposits react more sensitive to
statistics monetary policy changes
Altunbas et al. All euro area 1991- 9,991 obs. (unbalanced) Bankscope Findings vary depending on country; refer to
2002 & 2004 countries 1999 in total text for main findings
Angeloni et al. All euro area 1971- 18 to 3,207 banks ECB: Findings vary depending on country; refer to
2003 countries 2000 No. of obs.: n.a. area-wide model text and FN 55 for main findings
Ehrmann et al Euro area '92-'99 4,425 banks Bankscope Lending of less liquid banks react more strongly
2003 France '93-'00 496 banks Eurosystem to monetary policy changes;
Germany '93-'98 3,281 banks dataset; respective Size and capital are not relevant for lending
Italy '86-'98 785 banks national banks reaction
Spain '91-'98 264 banks supervisory reports
Gambacorta/ Italy 1992- 556 banks Bank of Italy: Capital in excess of regulatory minimum
Mistrulli 2001 17,792 obs. (unbalanced) supervisory repots determines lending reaction to monetary policy
2004 changes
Gambacorta Italy 1986- 759 banks Bank of Italy: More liquid and better capitalized banks
2005 1998 35,678 obs. (unbalanced) supervisory reports maintain lending after monetary policy
tightening
Altunbas et al. Euro area 1999- 2,947 banks Bankscope Banks that are active in securitization are less
2009 2005 15,403 obs. (unbalanced) sensitive to monetary policy changes
Gambacorta / Euro area 1985- Based on aggregate data ECB Monetary tightening has larger impact on
Rossi 2005 lending than monetary easing
2010
Albertazzi / Italy 2008- ~ 500 banks Bank of Italy: Well-capitalized and liquid banks managed best
Marchetti 2010 2009 ~ 19,000 obs. credit register to maintain loan supply during crisis
Bonaccorsi / Italy 2007- No. of banks: n.a. Bank of Italy: There are only significant effects of capitali-
Sette 2008 No. of obs.: n.a. credit register zation on bank lending if cap is interacted with
2010 liquidity, securitization activity, return on
assets and degree of interbank funding
Gambacorta / Euro area, 1999- 1,008 banks Bloomberg Well-capitalized banks that are active in
Marques-Ibanez UK, US 2009 30,920 obs. (unbalanced) securitization, shield lending better against
2011 monetary policy changes;
Central banks' non-standard measures helped
to maintain loan supply during crisis
Brei et al. Euro area, 1995- 108 banks Bankscope Central banks' rescue measures helped to
2013 UK, US 2010 1,615 obs. (unbalanced) sustain bank loan supply during crisis
Table 6.2: Selected empirical research from the euro area on bank lending and the bank lending channel
Interestingly, Ehrmann and Worms (2004) are able to show for Germany that, for
those banks that are not affiliated to a bank network, size does matter. In this respect,
the affiliation dominates the impact of size. Moreover, for affiliated banks, short-term
interbank deposits in particular determine the lending reaction of an institution. This is
also supported by Worms (2001). These findings further suggest that, within a bank
network, liquidity can be channeled effectively from head or affiliated institutions

toward smaller members.
In an attempt to shed further light on the role of liquidity, Ehrmann et al. (2003) and
Ehrmann and Worms (2004) find evidence for the hypothesis that high levels of liquid
assets entail a certain robustness against monetary policy shocks in the most
important European economies. In addition, using a sample for Italian banks,
Gambacorta (2005) points out that small banks on average hold more liquid assets
(as a percentage of total assets), and that small banks reduce the share of liquid
assets to a greater extent than bigger institutions do in response to monetary
tightening. This can be seen as supporting the idea that especially small institutions,
which rely heavily on relationship lending, like to insure themselves against adverse
monetary policy shocks or random deposit withdrawals by holding a relatively high
share of assets in liquid forms.
Comparing the individual European countries, Angeloni et al. (2003) find evidence for
loan supply effects in response to monetary policy in Germany, France and Italy. In
all these cases, they identify liquidity as playing an important role in explaining the
difference in lending reactions between banks.55
As regards the capitalization of a bank as a determinant of its lending reaction to

monetary policy, the evidence for European countries in the period before the recent
crisis is not fully conclusive. Ehrmann et al. (2003) do not find any statistically
significant effects of capital on bank lending. The authors suspect that this might be
due to the historically low number of bank failures compared to the US, which may be
indicative of a generally lower level of informational asymmetries in the euro area.
On the contrary, Altunbas et al. (2004) find, if banks in France, Germany and Italy are
ordered by groups of capitalization, the group of banks with the lowest capitalization
is the one most sensitive to the tightening of monetary policy, especially in France
55
Loan supply effects are also reported for Greece, the Netherlands and Portugal. The selection of
countries mentioned above represents the most important economies. Of the major European
economies, only Spain does not seem to exhibit shifts in the loan supply after a monetary policy
shock. Bank lending channel effects are limited in Austria and doubtful in Finland.
and Italy. In Germany, banks seem to be able to avoid the contraction of their loan
supply by reducing securities holdings and interbank borrowings factors that
suggest a relatively weak bank lending channel. Comparable country patterns have
been observed by Altunbas et al. (2002), who conclude that the impact of monetary
policy on the supply of bank loans is strongest in undercapitalized and small
European countries.
Less ambiguous results are obtained by emphasizing a different aspect of

capitalization, namely the amount that a bank holds in excess of the regulatory
requirement. Using excess capital, Gambacorta and Mistrulli (2004) and Gambacorta
(2005) show for Italy that an ample (excess) equity capital position provides better
protection for the loan portfolio. This suggests that excess capital ensures better
access to uninsured sources of funding due to a reduced risk of the bank acting in a
morally hazardous way. This view is also consistent with the implications of the bank
lending channel.
The relatively mixed indications with respect to the existence of a bank lending
channel and the role of certain bank characteristics in the countries that form the
euro area may be due to the time period selected, i.e. the fact that the samples used
in the literature cited so far are either individual country samples or synthetic
European ones formed by adding all the individual country samples together.
Developments attributed to the European Monetary Union (such as deeper financial
integration) and other trends observable in recent years (such as the shift toward a
stronger market-based financial system) are not incorporated.
Asymmetric effects of monetary policy: A contribution to a new strand of literature,

allowing for a more recent, "true" euro area sample, has been made by Gambacorta
and Rossi (2010). They are, first, able to detect loan supply effects and, second, able
to show that the effect on lending is larger in case of monetary tightening as opposed
to the relaxation of monetary policy. The authors ascribe this pattern to differential
reactions in investment and self-financing opportunities vis--vis monetary policy
regimes, which they take as an indication of a broad credit channel.
Against the backdrop of developments in the financial sector, one interesting

question is the extent to which changes in banks business models play a role in
influencing bank lending.
A study presented by Altunbas et al. (2009) finds that the lending of those banks that
are more heavily engaged in securitization activities is less sensitive to changes in
monetary policy. 56 Due to equity capital relief and the provision of liquidity,
securitization allows banks to expand their supply of loans, generally making banks
more flexible in the way they respond to monetary policy shocks. Altunbas et al.
(2009) also point out that securitization partially explains the weakened impact of the
size factor when large amounts of assets are transferred off the balance sheet. For
the bank lending channel discussion, this means that securitization may weaken the
effectiveness of monetary policy (see also Loutskina and Strahan (2009)).
The other side of the coin is that securitization could create incentives not to screen
and monitor borrowers as diligently as under the originate and hold business model,
as reported in a study of the US by Keys et al. (2010). Theoretically, this has the
ability to aggravate the problem of asymmetric information between lender and
borrower.57
What are the implications during a crisis period if incentives to screen and monitor
borrowers for bank lending are probably affected?
6.3.2. Euro area evidence in the wake of the crisis
The last point mentioned that the degree of securitization could have an impact on
the quality of screening and monitoring of borrowers is the reason, from a
theoretical perspective, why securitization may have a different effect on lending
during or outside of crisis periods. At times of crisis when the markets for
securitization are distressed and tend to dry out, the lost opportunity to repackage
and sell loans and the need to hold them to maturity may result in a loan portfolio
56
For a more detailed overview of securitization instruments and their implications for financial
markets, see Marques-Ibanez and Scheicher (2010).
57
For the US, Dell'Ariccia et al. (2012) show that lending standards declined most in areas with lively
securitization activities, which also suggests an exacerbated asymmetric information problem.
characterized by an increased risk and deteriorating opportunities to raise uninsured

funds. Because securitization can no longer serve to absorb monetary policy shocks,
this should magnify the effects on bank lending during a crisis. This reasoning is
empirically supported by Gambacorta and Marques-Ibanez (2011), who find that,
during crisis periods, the difference in the lending responses to monetary policy
shocks between banks that actively pursue securitization and all other banks is
smaller than at normal times.
To date, the analysis carried out by Gambacorta and Marques-Ibanez (2011) is one
of the very few that is explicitly devoted to the impact the recent financial crisis had
on how bank characteristics drive lending and how the different characteristics
interact with changes in monetary policy. It must be noted however, that it is not
exclusively geared to the euro area as it also comprises banks from the US and the
UK.
Their results support the view that structural changes did indeed occur during the
crisis. In addition to their findings on securitization activities, these authors' results
suggest that banks that are weakly capitalized, more dependent on market funding
and have a higher share of non-interest sources of income tended to restrict their
loan supply more strongly during the recent crisis. The particularly strong in-crisis
effect of an ample equity capital endowment found by Gambacorta and Marques-
Ibanez (2011) can be attributed to the flight-to-quality behavior of risk-sensitive
providers of debt capital. This supports our theoretical considerations regarding the
role of those bank characteristics that have an impact on perceived riskiness.
According to the authors, the observation that lending is dependent on the

importance of market funding to an individual institution is to be ascribed to the
following pattern: Banks that rely heavily on deposit funding could make use of the
opportunity not to lower their deposit interest rates in line with monetary interest rates,
but to try to maintain an attractive level to avoid losing deposit funding. An alternative
and more obvious interpretation is that banks with a high share of market funding
have greater difficulty rolling over their debt.
The outcome of greater lending sensitivity toward a crisis among those banks whose
income is characterized by a large proportion of non-interest income can be
explained by the higher volatility of related business (especially investment banking
business). During times of financial turmoil, the relevant markets feature a low level
of activity with fewer opportunities for banks to generate income, whereas the cost
base does not adjust flexibly to the new business situation. This makes banks riskier
for investors, undermining their ability to raise uninsured sources of funding.
An additional finding of Gambacorta and Marques-Ibanez (2011), though not geared

to bank characteristics, is that the non-standard measures taken by central banks like
the ECB significantly helped to prevent the supply of bank credit from drying up.
An analysis of bank rescue measures (provision of extended deposit insurance,

capital injections, debt guarantees and asset relief programs) has been carried out by
Brei et al. (2013). The authors find that capital injections are effective in
strengthening the loan supply only in cases where the capital ratio rises above a
certain critical threshold. In accordance with Bhattacharya and Nyborg (2010) they
argue that beyond this threshold banks are relieved from the debt-overhang problem.
Interestingly, they do not consider bank-specific, self-chosen capital targets as they
will be discussed in detail in the present study below. In general, the impact of capital
on lending is found to be higher during times of turmoil than at normal times.
A few additional observations have been made with regard to the recent crisis period,
although their informative value and universal applicability are probably restricted by
the limited geographical scope. Based on data obtained from the Italian credit
register, Albertazzi and Marchetti (2010) finds that, during the crisis,58 those banks
with low capital ratios in particular cut back their credit supply to the greatest degree,
while relatively liquid Italian banks managed to maintain their supply fairly well.
Extending the empirical model to analyze the effect of a banks size (in terms of total
assets) or its affiliation to a banking network, however, they find only relatively minor
effects that are dominated by the impact of capitalization. However, the largest group
of banks in their sample had to reduce lending by more than the smaller banks.
Also referring to Italian credit register data, the contribution by Bonaccorsi di Patti
and Sette (2012) establish no direct effect of the level of bank capital on lending.
When capital is regressed on loans, they cannot detect any statistically significant
effect. However, where capital interacts with other bank characteristics (liquidity,
securitization, level of interbank funding and the return on assets), the authors find
that these characteristics are affected by the level of capital a bank had at the outset
58
The exact crisis period examined comprises September 2008 until March 2009.
of the crisis. The influence of securitization and the return on assets from lending is
reduced for well capitalized banks.
These results especially the finding that the (stand-alone, non-interacted) level of
capital has no significant impact on lending contrasts with the results of Albertazzi
and Marchetti (2010), although this paper also use data on the Italian credit register
and, at least to a large extent, both papers analyze overlapping periods of the crisis.
It can thus be inferred that the exact role of capital in impacting bank lending during a
crisis is not yet fully understood.
6.3.3. Conclusion
What is the bottom line with regard to the factors that impact bank lending in the euro
area? Generally speaking, the evidence for a bank lending channel in Europe based
on synthetically aggregated country samples dating from the period before stage
three of European Monetary Union took effect is relatively heterogeneous. In some
countries, evidence in support of a bank lending channel or at least in support of
the first condition is stronger than in other countries.
Apart from geographical differences, with respect to the individual bank

characteristics it can be noted much the same as in the US that
interdependencies exist between the size of an institution and whether it is affiliated
to a bank network. While some studies have not found any size effect at all, at least
small institutions affiliated to networks do not seem to be at a disadvantage regarding
their lending reaction to monetary policy. It is not entirely clear whether this pattern
also applies to euro area countries for the period after stage three of European
Monetary Union.
In line with the theoretical framework for the bank lending channel, liquidity can help
to shield especially small banks loan portfolios from adverse shocks at normal times.
As regards capitalization, the distinction between its two main roles on the one
hand capital as a buffer against losses and to mitigate problems of asymmetric
information, and on the other hand capital resulting in a binding constraint if a bank
falls short of either regulatory or self-imposed targets/standards is widely
neglected. 59 Furthermore, the largely identical implications following from a capital
59
This issue is further elaborated on in the next section (section 6.4).
6.4 Implications of theoretical framework for interpretation of empirical evidence 77
constraint that is binding either for regulatory or self-imposed reasons are widely
disregarded. Probably owing to the inappropriate handling of capitalization, some
scholars have produced results of (statistically and economically) insignificant
relevance for an institution's capital endowment even during a crisis, while other
scholars' results highlight the opposite.
Essentially, the crisis period and its impact on the characteristics that drive bank
lending is still underrepresented in literature. There are reasons to believe that not all
patterns have yet been recognized.
In this context, a closer look at rescue measures taken by national governments

during the crisis would also appear expedient. Although initial evidence supports the
view of a generally positive impact on lending, it is not clear whether there are any
differences in the way the various kinds of rescue measures (e.g. debt guarantees,
capital injections, asset purchase/insurance) affect aggregate lending and the lending
of individual banks.
Moreover, the monetary policy indicator used hitherto in the euro area is the
overnight (Eonia) interest rate. During the crisis, because interbank markets dried out
and the overnight rate rose sharply, the ECB adopted the full allotment policy,
which equipped all institutions with the amount of liquidity they demanded. This
expansion of central bank liquidity together with low levels of liquidity in interbank
markets is responsible for the fact that the overnight rate may not have been fully
representative of banks funding conditions. Funding conditions during the crisis were
probably better captured by the Euribor-OIS spread (reflecting the risk of default
when banks lend to each other60).
6.4. Implications of theoretical framework for interpretation of empirical

evidence
Before proceeding to the empirical part in the next chapter, it should be recognized
how the bank characteristics reviewed in the previous sections can be integrated into
the theoretical framework in the spirit of the new view on the bank lending channel
60
For more details of the Libor-OIS spread, to which the same logic applies as to the Euribor-OIS
spread, see Thornton (2009).
presented in section 4.2. The result is a systematization of bank characteristics that

is unique to the bank lending channel literature.
With the development of the understanding of the bank lending channel and
econometric advancements toward panel estimation techniques a number of bank
characteristics have been found to influence banks' lending behavior as outlined in
the previous sections on empirical evidence. All the reviewed bank characteristics
can now be integrated into the framework reflecting the new view on the bank lending
channel. As stated in chapter 4, the new view focuses on the significance of
uninsured funding and on the bank characteristics that determine its costs and de-
emphasizes the role of reserves for the loan supply in reaction to changes of
monetary policy rates.
Figure 4.4 has already presented a novel framework which links the conditions and
subconditions of the bank lending channel to economic concepts. Figure 6.2,
showing a systematic outline of bank characteristics, is an enhancement of figure 4.4:
The economic concepts that have been introduced in figure 4.4 can be further
operationalized by formulating examples or proxies for these concepts. To give an
example: In the case of "business risk", examples for representative concepts are a
bank's "capitalization" or its "share of non-interest income". Banks that have more
capital have a stronger buffer against losses and are, thus, less risky from an
investors point of view. A high share of non-interest income may be interpreted as a
sign that earnings will be more volatile during the business cycle, since investment
banking earnings in particular account for non-interest income. In an economic
downturn, these earnings usually decrease by more than traditional interest income.
These representative concepts are more narrowly defined than the corresponding
economic concepts. They bridge the gap to the proxy variables that are used in
empirical analysis. Each representative concept can be thought of as a proxy that is
part of the econometric model which is introduced in section 7.3.
The depiction of the links between conditions and bank characteristics in figure 6.2 is
an attempt to lend a more structural emphasis to the discussion of which bank
characteristics may or may not play a role in affecting bank lending within the
framework of the bank lending channel.
Economic/theoretic concept Representation of concept

(operationalizing the subcondition)
Subcondition 1 Liquidity Share of liquid assets

Holdings of (as buffer against random deposit outflows)
liquid assets
Transparency Size
Listed/rated
Business risk Capitalization

(focus on perception
Subcondition 2 regarding riskiness Degree of maturity mismatch (proxied by
of a bank's operations) share of short-term funding)
(Ease of) Cost per
Access to unit of alt.
Share of non-interest income
alternative forms of
forms of funding
Share of marked-to-market securities
funding
Size (correlated with degree of diversification
(determines of income sources and professionalism)
total cost of
alternative Risk of moral hazard Capitalization
forms of (focus on risk of bank acting
funding) in a morally hazardous way)
Degree of dependence Share of wholesale/market funding

Volume/ insured funding
amount of Share of deposit funding (negative relation)
alt. forms
of funding Affiliation to bank network (providing access
to intra-group funding; negative relation)
Subcondition 3 Facing a binding capital constraint Shortfall relative to self-imposed capital ratio
No capital (capital ratio below certain threshold)
constraint Shortfall relative to regulatory capital ratio
Explicitly accounted for in the empirical analysis Not accounted for in the empirical analysis
Figure 6.2: Systematization of bank characteristics as drivers of bank lending
One particular matter should be noted: As can be seen in figure 6.2, "capitalization"
appears in two different roles (to which it has already been referred to in the literature
review sections above). In the first role, capital is to be interpreted as providing a
buffer against losses (related to "business risk") and as alleviating the potential
conflict of interest between the providers of equity and the providers of debt (related
to "risk of moral hazard"). As a result, a larger amount of capital causes lenders to
demand a lower risk premium. Although two aspects of capital can be distinguished
its first role, both aspects affect the unit cost of alternative forms of funding.
In the second role, the level of a banks capital endowment is assessed against the
background of regulatory or self-imposed requirements. Failing to comply with
regulatory requirements can result in serious action by the regulator, including the
takeover of managerial authority and even the forced closure of a financial institution.
In this sense, there is a "yes or no" answer to the question of whether a bank is under
constraints: Either a bank is impeded to supply loans by a capital constraint or it is
not. If it is, it will not be able to expand lending in response to a relaxed monetary
policy, nor will it participate in propagating the monetary policy impulse. The more
constrained banks there are in an economy, the more muted the bank lending
channel tends to be.
It is important to bear in mind that not only a regulatory capital ratio but also a self-
imposed capital target can have the same constraining effect in the event that the
target ratio is undercut. As reported by Gropp and Heider (2010), almost all
institutions target a certain magnitude of capital endowment over and above
regulatory requirements which they judge to be adequate. Berger et al. (2008)
confirm that this magnitude is indeed well above the regulatory minimum. It is chosen
by banks at their own discretion for the reason that the gains arising from a lower risk
premium outweigh the higher cost of equity capital. In particular, failure to meet
market standards could result in serious difficulties for rollover debt.61 This being the
case, a self-imposed constraint too shares the relevant "yes or no" implications of
situations in which banks are subject to regulatory constraints.
It should be kept in mind that this kind of self-imposed constraint is adopted with a
view to reducing the demanded risk premium or the unit cost of alternative funding. In
this regard, both roles of capital are interrelated and connected. One implication of
this interrelationship is the realization that even if a bank's capital is not constrained
for regulatory reasons (which can be deduced from regulatory capital ratios), it can
nevertheless be constrained by a self-imposed capital ratio stemming from
considerations regarding the unit cost of alternative funding (see figure 6.3).
61
As Hanson et al. (2011) show for the four largest US banks, even in middle of the crisis (Q1 2010)
these institutions maintained capital ratios well above the regulatory minimum, i.e. ratios that were
not binding. This is an indication of the significance of self-imposed capital ratios and ratios that
are demanded by the market.
The bottom line is that the question whether the expansion of a bank's loan supply is
constrained by a self-imposed or a regulatory capital ratio can be answered
independently of the question whether a bank is struggling to expand its lending due
to adverse funding conditions. This distinction must be borne in mind especially when
conducting empirical analyses.
Three stylized cases of capital ratios Adherence to Adherence to self- Implications for expansion of
regulatory minimum imposed target bank lending
capital requirement capital ratio
> Expansion of loan portfolio
Case possible
1
.
> No expansion due to too low

capital as measured by self-
Case
2
.
imposed target
> Expansion possible from reg.
perspective
> No room for expansion of loan
Case portfolio
3
.
Regulatory minimum Self-imposed

capital ratio target capital ratio
Both the regulatory requirement and the self-imposed target need to be met for an expansion of the loan portfolio
Figure 6.3: Three stylized cases of capital ratios against the background of regulatory requirements, self-
imposed targets and their implications for expansion of the loan portfolio
Consequently, a hypothesis accounting for such a capital constraint is formulated in

the following chapter. Additionally, the following chapter addresses the research gaps
and ideas that are stated in section 6.3.3 as a result of the review of the literature
body dealing with the euro area.
82 7 Empirical analysis approach
7. Empirical analysis approach
7.1. Research hypotheses
As discussed in the previous chapters, there are still a number of gaps in the
academic debate regarding analyses of those factors that influence bank lending
which, up to now, have mainly been studied within the bank lending channel
framework.
The gaps are related to the distinction between two different roles of capital (capital
that mitigates the asymmetric information problem/buffers against losses and capital
that leads to constraints when it falls short of self-imposed or regulatory capital ratios),
related to an incomplete understanding of financial innovation and business models,
which have only started to be explored very recently, and in particular to several
aspects of the financial crisis, with its implications for bank lending (e.g. maturity
mismatches, capital and the success of (unconventional) measures taken by the
ECB).
Within the scope of this study these gaps are addressed in the following by
formulating testable hypotheses. Hypotheses are classified according to whether
they apply generally (i.e. also at normal times), or whether the rationale behind them
necessarily involves a crisis context. Special attention is paid particularly to the latter
aspect: disentangling the recent financial crisis and clarifying the question of which
bank characteristics and other factors impact bank lending in times of financial
turmoil. An overview of all hypotheses is given by figure 7.1 at the end of
section 7.1.2
A further aspect of this study is to analyze whether the following hypotheses are valid
not only in the euro area as a whole but also in the individual countries of the euro
area. Accordingly, the hypotheses are tested with reference to the euro area but also
with a focus on the four most important euro area countries in the next chapter
(chapter 8).

7.1 Research hypotheses 83
7.1.1. General hypotheses
Existing literature does not satisfactorily account for a bank's capital endowment. The
usual approach is to include pure capital or capital ratios in regressions. Where this
is the case, the intention is to account for capital in its role in providing a buffer
against losses, or in attenuating the asymmetric information problem. What is missing
is a perspective on capital that takes into account whether a bank is operating below
its self-imposed target capital ratio (or even the regulatory capital ratio; see figure
6.3).
Besides the extensive body of literature dealing with the capital structures of non-
financial firms and their determinants (see Brner et al. (2010), among many others),
there are also contributions that explicitly study the capital structure of banks. For
example, Gropp and Heider (2010) estimate banks target capital ratios, an approach
also pursued by Flannery and Rangan (2006) and Lemmon et al. (2008) in the
context of non-financial firms. Their findings are that banks' capital ratios are best
explained by time-invariant, bank fixed effects and that they converge to form bank-
specific, time-invariant targets.62
The first hypothesis derives from this insight:
H1: A shortfall in capital relative to a targeted ratio leads to a reduction of the

loan portfolio.
The idea is to test whether the fact that a bank is close to the limits of defined capital
constraints has an effect on its lending. It must be noted that this hypothesis claims
that the relationship applies both at normal times and during a crisis. When a bank is
short of capital or is already operating close to the limit of its self-imposed capital
ratio, this should have an effect on its ability or willingness to provide loans. When a
self-chosen capital target is missed, this should imply that a bank is operating under
a constraint and will not expand its lending. This effect is assumed to be in addition to
widely reviewed conventional measures of capitalization. The latter take only "pure"
62
This is consistent with practical experience of how banks deal with the amount of capital available.
In practice, virtually every bank targets a certain capital ratio which it deems adequate and
opportune given all relevant circumstances, i.e. the returns expected by providers of equity, the
requirements of providers of debt, the cost of debt, market disciplinary forces, etc. This determines
to a large extent the amount of risk-weighted assets a bank is willing to hold.
capital ratios into account and are motivated by the connection between capital and
access to uninsured sources of funding (which makes perfect sense).
To clarify the point: The use of a measure of capital surplus (or shortfall) does not
mean that "pure" capital ratios have no influence on lending. However, the distinction
between the two roles of capital marks an important extension to the well-
documented impact of pure capital or capital ratios.63
Next, the total cost of alternative (uninsured, non-deposit) forms of funding can be
divided into two aspects: the volume and the cost per unit a relationship that is
illustrated in figure 4.3. A precondition for the notion that certain bank characteristics
affect the cost per unit is that a bank does raise funding on wholesale funding
markets. The bank characteristics that impact the cost per unit affect lending only if a
bank taps wholesale funding markets at all. The concept of the deposit overhang
serves as to proxy if a bank does so or not. Accordingly, the hypotheses are:
H2: For banks that do not face a deposit overhang, lending depends on
H2a: the share of uninsured short-term funding.
H2b: the share of non-interest income.
H2c: the share of securities that must be marked to market.
As illustrated in figure 4.4 and figure 6.2, the hypotheses stated above reflect those
factors that have an impact on how risky a bank is from an investors point of view.64
The link to bank lending is substantiated by the following rationale: The more difficult
it is to obtain funding, the more likely it is that the loan portfolio will be reduced. To
date, these factors have not been scrutinized in a coherent manner. By explicitly
subjecting them to hypothesis tests, it becomes possible to account for differences in
business models as well as changes in banks funding practices.
63
Consequently, "pure" capital measures are indeed part of the estimated empirical equations in
chapter 8, but they are not subject to explicit hypothesis tests.
64
The characteristics size and capitalization are also associated with the cost per unit, according to
figure 6.2. Due to their multiples role, however, it is highly likely that they have an impact on
lending not only if a bank has an overhang in deposits but also "stand-alone". This is tested in the
course of the empirical analysis. (Results confirm this presumption.)
More precisely, the higher the share of short-term funding, the greater the risk that a
bank will have difficulty rolling over its debt when the economy dips into recession.
This could involve the problems associated with fire sales and the beginnings of
vicious circles as discussed in chapter 5.
Furthermore, non-interest income is known to be more volatile (see Gambacorta and

Marques-Ibanez (2011), for example). While a bank with a large proportion of non-
interest income might therefore be more profitable at normal times,65 this might also
be an indicator of higher risk to investors.
The rationale for the share of mark-to-market securities is as follows: The bigger their
share, the more vulnerable a bank will be to shocks such as increases in monetary
policy interest rates. The resultant higher discount factors lead to immediate write-
offs, thereby impairing the bank's financial position and increasing the external
finance premium.
All three patterns reflected in hypotheses H2a, H2b and H2c should generally be
present, irrespective of the economic circumstances, i.e. whether or not a crisis
currently prevails.
7.1.2. Hypotheses involving the context of the recent crisis
In an attempt to isolate the impact of the crisis on bank lending, the first hypothesis
can be derived directly from the theories outlined in chapter 5. In line with the theory
of the loss spiral and the margin spiral, high leverage increases the liquidation of
assets under fire-sale conditions and aggravates adverse effects. In a crisis,
providers of uninsured funding should therefore be more concerned with banks'
leverage ratios. Since leverage is the reciprocal of the capital-to-asset ratio, the
hypothesis is this:
H3: During a crisis, the capital ratio has a more positive impact on lending
than it does at normal times.
65
This is true with regard to profitability in terms of the return on equity, for example, because the
business that generates non-interest income does not require the provision of any capital.
In addition, the theory of the loss spiral and the margin spiral suggests that, in an
economic downturn, the inability to roll over short-term liabilities forces banks to
reduce assets under fire-sale conditions and/or to liquidate longer-term assets at
considerable discounts. Thus:
H4: During a crisis, the impact of a pronounced maturity mismatch (evidenced

by a large proportion of short-term funding)66 on lending is lower (in terms
of the value of the coefficient) compared to normal times.
Staying with the problem of raising funding during a crisis, this hypothesis serves as
the motivation for further hypotheses. Another factor that determines the unit cost of
uninsured funding is the share of non-interest income (see figure 4.4 and figure 6.2).
This is because the business from which non-interest income is generated is more
vulnerable to crisis conditions than interest income. A reduction in this income source
increases the risk that a bank may not be able to meet its repayment obligations.
Hence:
H5: During a crisis, the impact of a large proportion of non-interest income

on lending is lower (in terms of the value of the coefficient) compared to
normal times.
A further bank characteristic that is suspected of influencing the cost per unit of
uninsured funding is size. The literature review in section 2 shows that, in many
studies, the pure size of a bank has not been found to impact lending at normal times.
In a crisis, however, size could become more of an issue: Size is known to have
been used as a proxy for transparency (see figure 4.4 and figure 6.2). During a crisis,
however, providers of uninsured debt might also perceive size as a proxy for the
professionalism of a bank in general e.g. regarding its crisis management or its
ability to take action to counter adverse developments and as a proxy for a more
favorable long-term outlook, e.g. due to the fact that larger institutions usually have
more diversified sources of income. The resultant hypothesis is:
66
Recall that the average maturity of loans is much higher than the maturity of the funding that is
called "short-term". Hence, a higher share of short-term funding usually implies a stronger
mismatch between the maturities on the asset and the liabilities side ("maturity mismatch").
H6: During a crisis, the size of a bank has a more positive impact on lending
than it does at normal times.
The next hypothesis also derives directly from chapter 5 and focuses on the share of
securities that have to be marked to market. In an economic downturn, if a bank
holds marked-to-market securities that are sold by other banks under fire-sale
conditions it also has to write down the respective asset. A high proportion of
marked-to-market securities thus increases the risk for write-downs and makes the
bank more vulnerable and risky. Write-downs are immediate losses and lower the
capitalization of a bank. Even if a bank does not encounter write-downs the
heightened risk should negatively impact the cost per unit of alternative funding and,
ultimately the ability to grant credit. The seventh hypothesis is therefore:
H7: During a crisis, the impact of a large proportion of marked-to-market

securities on lending is lower (in terms of the value of the coefficient)
compared to normal times.
Furthermore, the more a bank relies on (insured) deposits as a funding source, the
more it is independent of wholesale markets and the turmoil that was observable
during the recent crisis. Thanks to deposit insurance schemes, depositors are not
faced with an asymmetric information issue vis--vis deposit-taking banks and are
thus less concerned about draining their funds. This leads to the following hypothesis:
H8: During a crisis, a large proportion of deposit funding has a positive impact
on lending.
Building on this idea, a higher (than average) proportion of deposits is not the only
factor that should have a positive impact on the loan portfolio. An overhang of
deposits over the amount of loans should also be beneficial, since it implies that a
bank does not depend (to a very great extent 67 ) on wholesale markets to raise
funding. Especially during a crisis, a deposit overhang should provide relief for banks'
funding problems. Since the literature reviewed shows no evidence that the test of a
deposit overhang has been carried out, it will be interesting to see whether or not the
data confirms the following hypothesis:
H9: During a crisis, a deposit overhang has a positive impact on lending.
The difference of hypothesis H9 and hypotheses H2a to H2c all somehow related
to a deposit overhang is as follows: Hypotheses H2a to H2c are supposed to apply
generally, including normal periods. The (lack of a) deposit overhang is used to proxy
whether a bank needs to tap wholesale funding markets. When it needs wholesale
funding, then factors that make a bank appear riskier from an investor's perspective
should have an impact on funding costs and thereby also on loan rates and
quantities. Only for those banks that do not face a deposit overhang a relationship is
supposed between lending on the one hand and the share of short-term funding, of
non-interest income and of securities that need to be marked to market on the other
hand. Important to note is that in the end it is not the fact that a bank needs
wholesale funding at all that affects the loan supply of this bank. It is the dependence
on wholesale funding in connection with characteristics that make a bank appear
riskier.
By contrast Hypothesis H9 is meant to directly address the impact of the need to tap
wholesale funding markets during a crisis. The need of doing so is also proxied by a
lack of a deposit overhang. The important difference is that during a crisis, according
to hypothesis H9, the fact that a bank needs wholesale funding is sufficient in itself to
suppose that a bank has to curtail lending due to the difficult funding market situation.
All the above hypotheses (see figure 7.1) derive from gaps in existing literature and
from the rigorous application of the theoretical framework or concepts that address
the lending behavior of banks. The following sections deal with the introduction of the
67
Even with a deposit overhang, a bank could still rely on wholesale funding markets, e.g. because it
has to fund liquid assets that banks hold for precautionary (and other) reasons. See section
3.1.1.1 for the underlying rationale. As discussed in context with hypotheses 2, nevertheless,
(in)dependence of wholesale funding markets is proxied by a deposit overhang.
7.2 Overall empirical strategy and approach 89
empirical framework, including the model, data and estimation method that are
chosen to address these gaps and to test the stated hypotheses.
Hypotheses
A shortfall in capital relative to a targeted ratio leads to a reduction

H1
of the loan portfolio
General For banks that do not face a deposit overhang, lending depends on
hypo-
theses H2 > H2a: The share of uninsured short-term funding
> H2b: The share of non-interest income
> H2c: The share of securities that must be marked to market
During a crisis, the capital ratio has a more positive impact on
H3 lending than it does at normal times
During a crisis, the impact of a pronounced maturity mismatch

(evidenced by a large proportion of short-term funding) on lending
H4 is lower (in terms of the value of the coefficient) compared to
normal times
During a crisis, the impact of a large proportion of non-interest
Hypo- H5 income on lending is lower (in terms of the value of the coefficient)
theses compared to normal times
involving During a crisis, the size of a bank has a more positive impact on
crisis H6
lending than it does at normal times
context
During a crisis, the impact of a large proportion of marked-to-
H7 market securities on lending is lower (in terms of the value of the
coefficient) compared to normal times
During a crisis, a large proportion of deposit funding has a positive
H8
impact on lending
During a crisis, a deposit overhang has a positive impact on
H9
lending
Figure 7.1: Overview of formulated hypotheses
7.2. Overall empirical strategy and approach
To test the above hypotheses, the key identification strategy is to relate cross-
sectional differences in banks' characteristics to changes in the supply of bank loans,
with a special focus on comparing bank lending at "normal" times to the lending
practice adopted during the recent crisis. Doing so involves special challenges that
are explicitly addressed below. The first is the need to draw a clear distinction
between loan demand and loan supply. The problem here is that only the volume of
loans on banks' balance sheets can be observed, without knowing a priori whether it
is driven by supply or demand factors. The second challenge is to identify the criteria
based on which the period of the crisis of relevance to this study can be defined. This
is important to determine start, ending and duration of the crisis.
Before addressing these challenges, the first step is to derive an empirical model
based on the work of Peek and Rosengren (1995a) and Kishan and Opiela (2000).
The purpose of this exercise is to formalize the arguments set out especially in
chapters 4 and 5, and to add some mathematical intuition. After that, the model is
adapted to the empirical requirements of this study.
The data used has both a cross-sectional and a time series dimension, which yields
a panel structure. The panel is estimated using the generalized methods of moments
(GMM) methodology, or, more precisely, difference GMM. This estimator uses
instrumental variables, as it is able to provide a solution to the endogeneity problem
(explanatory variables are not strictly exogenous) which might otherwise compromise
the results if not treated properly. Furthermore, difference GMM estimator is designed
for panels comprising many cross-sectional observations and relatively few time
periods, for capturing individual fixed effects, and for dynamic dependent variables
that depend on past values as is the case here.
This section is organized as follows: First, an econometric model is developed on the

basis of a model of bank behavior, providing important insights into the relationship
between bank characteristics and lending. Second, building on this model in
conjunction with the hypotheses, the data required to feed the model is presented
and an overview of the data sources is given. This includes a description of the
purging/correction steps involved in preparing the data for use in the estimations.
Special attention is given to construction of the capital surplus variable, the use of
which in the context of bank lending in the euro area represents an academic
innovation. A separate subsection is devoted to the special challenges involved in
disentangling loan supply and loan demand and determining the duration of the crisis
period. Finally, the motivation for using this estimation methodology is explained. The
estimation methodology is an answer to all the challenges that emerge out of the
structure and characteristics of the data and is therefore, discussed at the end of the
chapter.
Together, these elements lay the foundation for empirical analysis, the results of
which are presented in the next chapter (chapter 8).
7.3 Empirical model 91
7.3. Empirical model
7.3.1. Derivation of a model of bank behavior
The model on whose implications the empirical model is based is a version of the
one presented in Peek and Rosengren (1995a and 1995c). It is especially suitable
because, unlike other models in the context of bank lending and the bank lending
channel, it explicitly specifies a capital constraint. Although the motivation for
constraints in the present study is geared to a self-chosen capital target rather than
strictly to regulatory requirements, as in the original version, the mechanics are very
similar.68
The starting point is a balance sheet constraint which states that the sum of all
assets must equal the sum of all liabilities. Each bank holds loans, Li, securities, Si
and reserves, Ri, as assets. On the liabilities side there is (equity) capital, Ki, insured
deposits, Di, and uninsured deposits, UDi:
. (1)
Insured deposits are assumed to be inversely related to a market interest rate, e.g.
the monetary policy rate (iMP). The inverse relationship is motivated by opportunity
cost considerations: In light of relatively high market interest rates that are
determined by the monetary policy rate, individuals shift from deposits into other
asset classes that yield returns market-linked rates. As a result, it is the less
attractive to hold money as non interest-bearing deposits the higher the market level
of interest rates:69
. (2)
Uninsured deposits are assumed to be the marginal source of funding. The market
for these funds is characterized by imperfect competitiveness in the sense that a
bank has the power to attract additional funds by raising the interest rate it is willing
to pay (iUD) above the market level (
):

. (3)
68
The following development of the model is based on Peek and Rosengren (1995a). A similar
approach has been taken by Kishan and Opiela (2000), among others.
69
This point is further elaborated on in the excursus on motives for holding money in section 2.2.2.
By raising interest rates on uninsured deposits above market rates, a bank not only
profits from inflows of funds from the clients of other banks, but also from inflows of
financial instruments that are close substitutes. The magnitude of this effect is
captured by b1i, where the index i indicates that this effect is bank-specific. In
particular, b1i is assumed to be a function of individual bank characteristics, which is
an important feature of the model:
(4)
The variable xik contains all bank-individual characteristics k that are postulated to
have an impact on the supply of bank loans and to be sensitive to the economic
situation (normal times versus periods of crisis). These characteristics were
discussed in chapters 4, 5 and 6 and are shown in figure 6.2.
On the asset side, banks hold reserves as a certain percentage (c) of insured
deposits exactly as demanded by the central bank, which means that they do not
hold any additional reserves. This reflects a fractional reserve system such as the
one run, for example, by the ECB:70
. (5)
In addition, banks hold a certain fraction of deposits as securities to protect

themselves against random and large deposit outflows. As discussed in section
3.1.1.1, if they did not hold securities, banks would, when confronted with large
unexpected deposit withdrawals, be forced to terminate loan contracts prior to
maturity to avoid being unable to replace the lost deposits with other forms of funding,
which would normally not be possible without frictions. The premature termination of
loan contracts is assumed to be relatively costly. The exact fraction of deposits held
in the form of securities (d1) is the result of a bank's expectations about the available
deposit volume and its risk appetite or aversion regarding the volatility of deposit
outflows. Securities as a fraction of deposits are held net of reserves. This yields:
. (6)
Both the market for uninsured funding and the loan market experience imperfect
competition. If a bank demands an individual interest rate for loans (iiL) that is below
70
One simplifying assumption made is that reserves are non-interest-bearing. In the euro area,
however, required reserves are actually remunerated at the (marginal) rate of the main refinancing
operations (see European Central Bank (2011b), section 4.3, p.102).
the market interest rate ( ), it will be able to increase the volume of loans on its
balance sheet:
(7)
The market rates on loans, securities and uninsured deposits are modeled to be
dependent on the monetary policy rate. In the interests of simplicity, it is assumed
that a change in the monetary policy rate translates into an equal reaction () by
each of the market rates for loans, securities and uninsured deposits:71
, (8)
, (9)

. (10)
A further feature of the model is the specification of a capital constraint. In the version
of Peek and Rosengren (1995a), the capital constraint is motivated by minimum
regulatory requirements. However, as discussed in sections 3.1.1.3 and 4.2, a
constraint can also be self-imposed in the sense that a bank targets an individual
equity capital ratio as a result of its judgment regarding what it believes to be its
optimal capital structure.72, 73
Irrespective of its motivation, the capital constraint can be thought of as binding if the
capital ratio falls below a certain (regulatory-imposed or self-imposed) fraction (i) of
total assets:
. (11)
71
This assumption is only made to simplify the algebra. It is not critical in any sense for the further
analysis.
72
Again, this implies an objection to the Modigliani-Miller theorem (see Modigliani and Miller (1958)),
according to which the capital structure is under certain conditions irrelevant to the value of a
firm.
73
The explicit modeling of an empirical target capital ratio for an individual bank including its
derivation follows in section 7.4.2.
Reference is made to a fraction of total assets and not to a fraction of a subsample of

assets e.g. risk-weighted assets in the interests of simplicity. All relevant
conclusions remain unchanged.74
It is assumed that banks aim to maximize their profits, . Since the model abstracts
from loan losses, overhead costs and fee income, the individual profit maximization
function is given by the sum of interest earnings on loans (iiL) and securities ( )75
less interest expenses on insured deposits (iD) and uninsured deposits (iUD):
. (12)
The profit function (12) can then be maximized using the capital constraint as a side
condition in a Lagrangian equation. The Lagrangian function, G, yields:
. (13)
Equations (1) to (3) and (5) to (10) can be used to eliminate Si Ri UDi Di iiL iD and the
three market interest rates, , and
, from equations (11), (12) and (13).
At this point, the usual approach in literature on the bank lending channel the
approach adopted by Peek and Rosengren (1995a) and Kishan and Opiela (2000),
for example76 is to analyze the effect of monetary policy changes on loans and on
other variables, such as the amount of uninsured deposits. Mathematically speaking,
to this end, the Lagrangian function, G, is maximized with respect to loans and the
first-order conditions are used to solve for Li. Testable hypotheses are then obtained
by taking the loan equation with respect to the monetary policy variable.77 It is then
judged whether changes to the monetary policy variable are expected to have a
positive or negative sign and impact on loans.
74
Interestingly, the "pure" ratio of capital to total assets has now (once again) been included in the
requirements stated in the Basel III accord in the form of the leverage ratio, although the exact
specification is still to be settled at a later point in time (see Basel Committee on Banking
Supervision (2012)).
75
Securities are multiplied by the market rate, because banks are assumed to be price takers in the
market for securities.
76
The latter authors use a similar model to the former, but without incorporating a capital constraint.
77
This process can also be employed to obtain values for other variables of interest, e.g. the amount
of uninsured deposits.
In this study, the impact of monetary policy on bank lending is of secondary

relevance. The focus here is more on the question of which bank characteristics drive
bank lending at normal times and during crises. The monetary policy indicator is
therefore used as a control variable, whereas the primary focus is on bank
characteristics that have an impact on b1i in equation (4).
Despite this slight difference of focus, it is still useful to exploit the predictions of the
model regarding the impact of capital on banks' supply of loans both when capital is
subject to constraints and when it is not.
In the event that a bank is subject to capital constraints, applying the approach
outlined above maximizing (13) with respect to loans, using the first-order
conditions to solve for Li and then taking the derivative of the loan equation with
respect to Ki yields:

. (14)

Equation (14) must be positive, because i is the equity capital ratio that a bank must
not undercut for regulatory or self-imposed reasons. This means that, for a bank
subject to capital constraints, a decrease in capital is associated with a decline in the
supply of loans. The magnitude of the effect is the reciprocal value of the equity
capital ratio and is, therefore, greater than one.
This result can easily be reconciled to economic intuition: When a bank holds only an
amount of capital which exactly matches the desired quantity and, then, suffers
capital losses, it obviously needs to reduce its assets. 78 , 79 As securities do not
change by construction, the capital loss translates into a reduction of loans. Because
a bank needs, or has decided, to provide capital equal to the amount of loans times
the capital ratio, the consequence of a loss of one unit of capital is a loss in the
capacity of accommodating loans in the magnitude of the reciprocal value of the
capital ratio. It should be noted that, in the case of a bank subject to capital
78
It is abstracted from the possibility of raising additional funds on the liabilities side of the balance
sheet. See section 3.1.1.1 for a deeper discussion of this issue.
79
As a reminder: Capital in this case is not a bank characteristic that acts as a buffer against
possible losses from an investor's perspective. This motive is captured in b1i (see equation (4)).
Here, capital is used to decide whether its quantity satisfies a self-imposed or regulatory target.
constraints, the model does not predict any influence of bank characteristics on the
supply of loans, ceteris paribus.
In the case where no capital constraints apply, the picture is different when the
capital ratio is already above the targeted threshold:

. (15)

Again, equation (15) is greater than zero as in the case where constraints apply. The
effect of capital on loans is positive. However, in equation (15), the impact depends
on e1 and b1i. While e1 captures the sensitivity of the loan demand to a rise in the
interest rate on loans above the market rate and is not of any particular interest, b1i
reflects the sensitivity of uninsured deposits to changes in the rates offered by banks.
As stated in equation (4), b1i is a function of bank characteristics.
Hence, the conclusion for the case where no constraints apply is that the response of
the loan supply depends on certain bank characteristics. Which of these, then,
deserve consideration? The characteristics postulated to have an impact on the
supply of loans are the ones shown in figure 6.2 and discussed in chapters 4, 5 and 6.
These bank characteristics are explicitly incorporated in the empirical model in the
section that follows.
7.3.2. Introduction of the empirical model
Having derived the model and established the link to bank characteristics, this
section presents the model that is used for the hypotheses tests.
The empirical model that is actually estimated has been used in a comparable
fashion in a number of publications.80 This implies that its empirical specification is
well established and has been approved by many successful predecessors.
The model is designed to capture the impact of certain bank characteristics on the
supply of bank loans. It is modified to allow for a structural change in the period of the
80
Among many others there are Kashyap and Stein (2000), Kishan and Opiela (2000) and Kishan
and Opiela (2006), Ehrmann et al. (2003), Gambacorta and Mistrulli (2004), Gambacorta (2005),
Ashcraft (2006), Gambacorta and Marques-Ibanez (2011), and Altunbas et al. (2009).
recent financial crisis. In its baseline specification, it is expressed by the following

equation:81

(16)
with , , , and , where N is the number of

banks in the sample, T denotes the final year, J stands for the number of lags and K
denotes the number of different bank characteristics.
In equation (16), the growth rate of the natural logarithm of bank loans, ,
is regressed on a vector of fixed effect ai, 82 on lagged values of the dependent
variable, and on k different bank characteristics represented by xk that are
interacted with a crisis dummy, C, that assumes the value one in the period from
2008 to 2009 and the value zero at all other times. It is also regressed on bank
characteristics that are interacted with a non-crisis dummy, NC, that takes the value
of zero in the period from 2008 to 2009 and one at all other times.83 It must be noted
that this does not lead to the introduction of perfect collinearity (a dummy variable
trap).84 As outlined in detail in the section on the estimation methodology (section
7.5), this is because all variables are transformed by applying orthogonal deviations
in order to remove bank-specific fixed effects. Comparison of the non-interacted
parameters with the parameters of the interaction terms (of the crisis dummy with the
bank characteristics) and inspection of statistical significance in each case is
intended to reveal any difference in the impact of the economic situation (normal
times versus crises) on the effect of the different bank characteristics on bank loans.
Among all the bank characteristics in the standard specification are variables such as
the size of an institution, a measure of the amount of liquid assets, the share of
funding via customer deposits, a measure of the capitalization of a bank and other
81
Additional specifications (augmentations of the baseline specification, specifications for sub-
samples, robustness checks etc.) are reported further below together with the results.
82
A motivation of for this fixed effect and the approach to dealing with it, especially against the
background of a lagged endogenous variable which is also present here, is given in the section on
the estimation method (section 7.5).
83
The exact definition of the crisis period relevant in this context is motivated in section 7.4.3.2.
84
See also the explanation in footnote 120.
factors. Precisely which variables are used, how they are constructed and from which
sources they originate is presented in detail in the data section (section 7.4.) and
shown in table 7.1 and table 7.2. Anticipating objections that may be justified with
regard to the problem that certain right-hand-side variables (apart from loans) might
not be entirely endogenous, all bank characteristics enter the estimated equation with
one lag. An additional measure to fix the endogeneity issue is to apply the
generalized method of moments methodology which is described in detail in
section 7.5.
Proceeding with the remaining explanatory variables, the growth rate in the natural
logarithm of bank loans is regressed on changes in the monetary policy indicator,
MP, on variables to control for loan demand effects, y, such as nominal GDP, on the
inflation rate and on a more direct measure of loan demand provided by the euro
area bank lending survey.85 Since the only figure that can be observed is the amount
of loans on banks' balance sheets, which is the outcome of the settlement of both
loan supply and loan demand, it is crucial to disentangle the two. Otherwise, changes
in the amount of loans observed might be attributed to the supply side while the loan
demand side is the actual driver.86 The error term is given by .
It is assumed that a linear relationship exists between bank loans and the parameters
of the explanatory variables.
7.4. Data
7.4.1. Data sources
The model introduced above is fed with data from different sources (see table 7.1).
The data spans the period from 1999 to 2011 and is based on an annual frequency.
The year 1999 was chosen because it marks the beginning of stage three of
European Economic and Monetary Union when a common monetary policy under the
authority of the ECB took effect and the euro became a real currency. It has been
85
Details of the construction of the variable based on the euro area bank lending survey are given in
the next section (section 7.4.1).
86
See section 7.4.3.1 for a detailed discussion of the disentanglement of loan supply and loan
demand effects.
7.4 Data 99
argued (in chapter 4) that three events taken together the advancing integration of
financial markets, developments in the field of financial innovation and the concurrent
launch of the common European monetary policy may have changed the
relationship between bank characteristics and lending, especially with respect to
other funding opportunities that have become available. To avoid any exposure to
structural breaks caused by these events, it therefore seems reasonable to choose
1999 as the beginning of the sample. An overview of the variables used, and the
construction/transformation in each case is given in table 7.2.87
Variable Period Source

Loans 1999-2011 Bankscope
Size (total assets) 1999-2011 Bankscope
Capital: Equity capital ratio 1999-2011 Bankscope
Capital: Tangible common equity ratio 1999-2011 Bankscope
Capital: Tier 1 ratio 1999-2011 Bankscope
Bank Capital surplus 1999-2011 Own calculations based on Bankcope data
charac- Liquidity 1999-2011 Bankscope
teristics Deposit funding ratio 1999-2011 Bankscope
Deposit overhang 1999-2011 Own calculations based on Bankcope data
Dummy deposit overhang 1999-2011 Own calculations based on Bankcope data
Dummy no deposit overhang 1999-2011 Own calculations based on Bankcope data
Short-term funding ratio 1999-2011 Bankscope
Share of non-interest income 1999-2011 Bankscope
Share of mark-to-market securities 1999-2011 Bankscope
EURIBOR 1999-2011 ECB (originally from Thompson Reuters)

Monetary
EONIA 1999-2011 ECB
policy
variables EURIBOR-OIS spread 1999-2011 Own calculations based on ECB data
Non-standard measures 1999-2011 Own calculations based on ECB / Eurostat data
GDP (euro area) 1999-2011 Eurostat

GDP deflator (euro area) 1999-2011 Eurostat
GDP (Germany) 1999-2011 Eurostat
GDP deflator (Germany) 1999-2011 Eurostat
Macro- GDP (Italy) 1999-2011 Eurostat
economic GDP deflator (Italy) 1999-2011 Eurostat
variables GDP (France) 1999-2011 Eurostat
GDP deflator (France) 1999-2011 Eurostat
GDP (Spain) 1999-2011 Eurostat
GDP deflator (Spain) 1999-2011 Eurostat
Lending survey 2003-2011 ECB
Other Crisis dummy 1999-2011 Own calculations

variables Non-crisis dummy 1999-2011 Own calculations
Table 7.1: Sources of the variables used
87
All variables have been tested for unit roots. Each time tests suggested the presence of a unit-root
in the original times series, the respective variable was first-differenced.
Bank characteristics
All the bank characteristics used in the model which make up the bulk of the data are
taken from Bankscope. Bankscope, a commercial database maintained by the
Bureau van Dijk in cooperation with Fitch Solutions, provides standardized data at
the micro-level on banks' financial statements. It is the most comprehensive
cdatabase in terms of geographical coverage (the number of countries) and coverage
within a country (the number of banks within a country), and is widely used by
academic and non-academic researchers for studies and policymaking. There is no
better publicly or commercially available database of banks' financial statements
Variable Construction/transformation Symbol (for transformed

(short title) (as used in estimation) variable as used in estimation)
Loans First differences of natural logarithm of loans; normalized with LOG(LOANS)

respect to average
Size (total assets) Natural logarithm of total assets; normalized w.r.t. average SIZE
Capital: Equity capital ratio Total equity over total assets; normalized w.r.t. average CAP
Capital: Tangible common equity Tangible common equity over tangible assets; normalized w.r.t. CAP_TCE_TCA
ratio average
Capital: Tier 1 ratio Tier 1 regulatory capital ratio; normalized w.r.t. average TIER_ONE
Capital surplus Capital surplus variable constructed as outlined in section 7.4.1 CAPSUR
Liquidity Cash and due from banks over total assets; normalized w.r.t. LIQ
average
Bank Deposit funding ratio Total customer deposits over total assets; normalized w.r.t. DEP
charac- average
teristics Deposit overhang Loans over total customer deposits minus one OVERHANG
Dummy deposit overhang Dummy variable: 1 if total customer deposits exceed volume of DUMMY_OVERHANG
loans exceeds, 0 otherwise
Dummy no deposit overhang Dummy variable: 0 if total customer deposits exceed volume of DUMMY_NO_OVERHANG
loans exceeds, 1 otherwise
Short-term funding ratio Deposits plus short term funding minus total customer STF
deposits over total assets; normalized w.r.t. average
Share of non-interest income Total non-interest operating income over the sum of net interest NII
income and total non-interest operating income; normalized
w.r.t. average
Share of mark-to-market securities Available for sale securities over total assets; normalized w.r.t. AFS
average
EURIBOR 3 month Euro Interbank Offered Rate EURIBOR

Monetary EONIA Euro OverNight Index Average EONIA
policy EURIBOR-OIS spread 3 month EURIBOR-OIS spread EURIBOR_OIS
variables Non-standard measures Assets on ECB's balance sheet over euro area GDP; first (NSM)
differences
GDP (euro area) First differences of natural logarithm of first difference of euro LOG(GDP)
area's nominal GDP
GDP deflator (euro area) GDP deflator for the euro area GDP_DEFLATOR
GDP (Germany) First differences of natural logarithm of first difference of LOG(GDP)
Germany's nominal GDP
GDP deflator (Germany) GDP deflator for Germany GDP_DEFLATOR
GDP (Italy) First differences of natural logarithm of first difference of Italy's LOG(GDP)
nominal GDP
Macro-
GDP deflator (Italy) GDP deflator for Italy GDP_DEFLATOR
economic
GDP (France) First differences of natural logarithm of first difference of LOG(GDP)
variables
France's nominal GDP
GDP deflator (France) GDP deflator for France GDP_DEFLATOR
GDP (Spain) First differences of natural logarithm of first difference of Spain's LOG(GDP)
nominal GDP
GDP deflator (Spain) GDP deflator for Spain GDP_DEFLATOR
Lending survey Weighted net percentage (tightened-eased) of overall answers (LEND_SURV)
to question 4 (demand for loans or credit lines to enterprises) of
bank lending survey; first differences
Other Crisis dummy Dummy variable: 1 for the years 2008-2009; 0 otherwise CRISIS
variables Non-crisis dummy Dummy variable: 0 for the years 2008-2009; 1 otherwise NC
Table 7.2: Description and construction of variables used in the regression

7.4 Data 101
The annual frequency of the data corresponds to the frequency with which
Bankscope provides it. As an alternative, one could employ data taken from
Bloomberg, a US-based financial data provider. However, since Bloomberg data only
comprises listed companies, using it would be a major drawback in terms of coverage
of the banking population. Most likely, this would result in a sampling bias
overweighting large banks (in terms of assets). This is because listed banks are
bigger on average than non-listed companies.
In addition, several studies compare quarterly and annual data with respect to the
conclusions that follow from the different frequencies. In a contribution based on US
Call reports,88 Ashcraft (2006) reports similar results for both frequencies. In a study
of Italian banks, Gambacorta (2005) resorts to both data from non-publicly available
Italian supervisory reports and Bankscope data. Their approach has been to pick
those banks that are part of the Bankscope sample. They then look up the data for
these banks in Italian supervisory reports and transform this data into a quarterly
frequency and into an annual frequency. Comparison of the results based on this
procedure yields no substantial differences between the two frequencies.
This leads to the conclusion that the annual frequency is sufficient, and that data with
a quarterly frequency does not seem to contain additional relevant information in this
context. There is therefore no justification for the disadvantages that come with
employing quarterly data taken from the available sources, i.e. the limited coverage
and a sampling bias toward large banks.
The sample includes banks from all countries that have been part of stage three of
the European Economic and Monetary Union right from the beginning in 1999. These
countries are Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg,
the Netherlands, Portugal and Spain. In addition, banks from Greece (which joined
the euro area in 2001) are also part of the sample. Countries that have joined the
euro area since then are not included.
With regard to the bank characteristics included in the model, loans, which are
measured using the natural logarithm of gross loans, are the dependent variable. In
addition, loans in lags are used as explanatory variables.
88
In the US, banks are required to give answers as demanded by the Consolidated Report of
Condition and Income (in short: Call report) every quarter. The Call reports are administered by
the Federal Financial Institutions Examination Council (FFIEC). This data is publically available.
One of the characteristics included in xk in equation (16) is the size of a bank, which
is the natural logarithm of its total assets.
To capture the capitalization of a bank, the use of various measures creates the
opportunity to perform robustness checks regarding the different concepts of
capitalization. The standard capitalization variable is constructed by taking total
equity over total assets. Total equity includes common equity, non-controlling interest,
revaluation reserves for securities, foreign exchange instruments and fixed assets, as
well as other accumulated comprehensive income.
A second measure of capitalization is tangible common equity divided by tangible

assets. Construction involves taking total equity and total assets and subtracting
goodwill, other intangibles and deferred tax from both. This is a relatively
conservative measure of loss-absorbing capacity. It is especially interesting to a
bank's investors, because it does not include the preferred stocks with which a
number of banks were bailed out by national or federal state governments or other
government institutions during the financial crisis.89
A third measure of capitalization is the Tier 1 capital ratio, which is geared to

regulatory capital requirements and may, therefore, not be of primarily relevance for
providers of uninsured funding. In addition, it is a relatively new concept (although it
was first defined in the Basel I framework as far back as 1988), which is probably
why many observations on this item are missing especially from the earlier years of
the sample. It is nevertheless useful as a robustness check.
Another capital-related variable is the capital surplus or shortfall relative to a bank-

specific target. The exact methodology of the construction of the capital surplus
variable is presented in the following section (section 7.4.2). Basically, the
construction involves two main steps: First, the bank-specific capital target is
estimated, and then the actual capital endowment is subtracted from this target,
producing either a surplus or a shortfall. As stated, a detailed explanation and
discussion follows in the next section.
The measure of the liquidity of a bank is a variable made up of cash and receivables
from banks over total assets.
89
See e.g. Fratianni and Marchionne (2010) for an account of interventions or Congleton (2009) for a
more general discussion of public interventions.
7.4 Data 103
To examine the funding situation of a bank, two variables are used that are geared to
the share of insured and uninsured funding. Insured funding is captured by the share
of customer deposits and expressed as total customer deposits including current,
savings and term deposits divided by total assets. Uninsured funding is measured
by the sum of all deposits and short-term funding, from which total customer deposits
are subtracted. The result is divided by total assets.
The extent to which a bank's securities are subject to market movements is given by
the volume of available-for-sale securities over total assets. This is believed to hold
true especially in crisis periods, making banks particularly vulnerable at such times.
A further variable quantifies the share of non-interest income as opposed to income

from interest-bearing business. It is constructed by taking total non-interest operating
income over the sum of net interest income and total non-interest operating income.
All bank characteristics, apart from the capital surplus variable, are normalized with
respect to their averages across all banks, which yields indicators that add up to zero
across all observations. The resultant parameters may be interpreted directly as the
loan response of an average bank. This is the approach taken by e.g. Ehrmann et al.
(2003), Gambacorta (2005), Altunbas et al. (2009) and many others.
Besides bank characteristics, the model also includes macroeconomic and monetary
policy variables. Data on these variables also covers the period from 1999 to 2011.
Like the data on bank characteristics, the frequency is annual and the data
comprises the same countries, i.e. those that constituted the euro area in 1999 and
complemented by Greece, which joined in 2001.
Macroeconomic variables
Among the macroeconomic variables included in the model to control for loan
demand factors influencing the observable loan volumes on banks' balance sheets,90
there is nominal GDP and inflation, which is proxied by the GDP deflator and,
alternatively, by the harmonized index of consumer prices (HICP). The data is
provided by Eurostat, the statistical office of the European Union.
90
The question of how to disentangle loan supply and loan demand factors are thoroughly
considered in section 7.4.3.1.
A new and innovative approach in this study is the utilization of data from the so
called euro area bank lending survey. As part of this survey, senior loan officers at
banks are asked quantitative questions on past and expected future developments in
the loan markets four times a year. The data is collected by the ECB. A detailed
discussion is provided in Berg et al. (2005).
The survey was announced in 2002 (see European Central Bank (21 November,
2002)) and first introduced by the ECB in 2003 to obtain further information about
credit markets and the business cycle, and to assess the effects of monetary policy
on the credit-related aspect of the transmission mechanism. The survey is addressed
to about 90 senior loan officers in all euro area countries. Generally speaking, the
questions distinguish between enterprises and households and relate to both past
and expected future developments. One section covers the supply of loans and is
geared to credit conditions (e.g. interest rate levels, collateral requirements,
maturities) and credit standards (e.g. the factors that influence the tightening/easing
of loan granting). In the other section, which includes the questions that are of
interest to this study, respondents are asked whether demand for loans from different
debtors, for different purposes and with different maturities has increased or
decreased in recent months.
The data on the bank lending survey is of good quality in the sense that the answers
are a reliable predictor of real GDP growth, as supported by the findings in De Bondt
et al. (2010). In addition, the ECB, which carries out the survey, has the opportunity
to cross-check the answers against the "hard" information it receives from banks as
part of a regular and compulsory reporting program. This further validates the
reliability of the survey.
For the purpose of this study, the answers to questions on the demand for loans
have been evaluated. In particular, the weighted net percentage is calculated from
the five possible answers to the question how the demand for loans and credit lines
(to enterprises) has changed over the past three months over and above normal
seasonal fluctuations.91 To calculate an index for the euro area, all answers from loan
officers in the individual countries are aggregated in accordance with the share of
91
Answers are assigned values ranging from -2 (demand "decreased considerably") to +2 (demand
"increased considerably").
7.4 Data 105
national lending aggregates in the sum of euro area lending aggregates (see Berg et
al. (2005)).
Because annual data is used in this study, the results for the four quarters the
answers refer to a three-month period in each case are aggregated to deliver a
single observation.92 This time series is then used in the regression to control for loan
demand effects that impact the loan volume.
To date, only very few studies have tried to make use of the euro area bank lending
survey; and those that have done so all investigate different questions. The study by
Ciccarelli et al. (2010) relies exclusively on the answers to the bank lending survey to
evaluate how monetary policy shocks influence loan demand and loan supply (albeit
without using data on bank characteristics). Focusing on those Italian banks that take
part in the survey, Del Giovane et al. (2011) use the survey data to cross-check the
developments condensed from banks' balance sheets. Maddaloni and Peydro (2011)
reveal that monetary policy rates affect credit standards, as reflected in the answers
to the survey questions. With the exception of these studies, however, the lending
survey has been largely ignored by academic papers. In particular, it has not been
used to capture loan demand in a euro area context.
Although one loses observations at the beginning of the sample as the survey only
started in 2003, it is still a useful instrument to complement other macroeconomic
variables.
The monetary policy variable and non-standard monetary policy measures
Different approaches are adopted with regard to the monetary policy variable, whose
purpose is to capture the stance of monetary policy. The monetary policy stance is a
measure of whether a central bank's monetary policy contributes to the economic,
financial and monetary developments in a way that is in accordance with its goals
which, in the case of the ECB and the euro area, is ultimately the goal of achieving
92
To be absolutely precise: The results published, say, in April refer to the preceding three months
(January to March). Therefore, the results for April, July and October of year t0 and January of t1
are aggregated to form one observation for the year t0.
price stability.93 A measure of the stance is included in the empirical model because
the fact of whether it is accommodative, neutral or restrictive should be closely
reflected in the banks' funding conditions and their lending reactions.
There seems to be no generally accepted measure.94 However, most studies dealing

with bank lending and the bank lending channel in a European context use a short-
term interbank rate to proxy banks' funding conditions. In this study, three different
approaches are adopted to ensure the robustness of the results with regard to
variations in the measurement of the monetary policy stance.
Since the ECB sets key interest rates with the aim of steering short-term money
market interest rates (see European Central Bank (2011b), p. 93), it seems
reasonable to use these rates as monetary policy rates. Two variables of this kind
are the 3-month Euribor (Euro InterBank Offered Rate) and the Eonia rate (Euro
OverNight Index Average), both of which are rates for relatively short-term time
horizons at which banks in the euro area are willing to lend unsecured funds to each
other. As shown in figure 7.2, the 3-month Euribor shows somewhat more fluctuation
from mid-2007 onward, which is likely be due to other disruptions and is probably not
a reflection of changes in the stance of monetary policy. 95 This finding might be
considered as a cautious argument in favor of the Eonia rate as the superior
measure.
However, strong demand for liquidity in connection with the dysfunction of interbank
markets during the crisis led to sharp increases in both the overnight and 3-month
interbank rates. As stated in the discussion of Gambacorta and Marques-Ibanez
(2011), these rates thus do not appropriately represent banks' funding conditions
during large parts of the crisis, especially against the background of the drying up of
interbank markets and the sharp associated rate hikes that have been observable
since the summer 2007 (see figure 7.2).
93
For details of the monetary policy stance of the ECB during the recent crisis, see European
Central Bank (2010b), p. 63 et seq.
94
For an overview of different approaches, see e.g. Oliner and Rudebusch (1996a) or the survey
provided in Khan and Qayyum (2007).
95
Reasons for these fluctuations included tensions on the interbank markets. This issue is further
discussed in section 7.4.3.2.
7.4 Data 107
In practice, a meaningful indicator of banks' funding conditions should abstract from

the banks' tendency to hoard liquidity during the crisis. This was the case when
tensions in the financial markets in general and the interbank market in particular
began to unfold because banks feared they might not be able to raise the amount
necessary to meet reserve requirements in the latter. The resultant shortage of
liquidity contributed to a sharp rise in interbank interest rates (liquidity premium),
obviously in connection with increased risk premiums. To better capture the risk
associated with lending by banks to each other, it is worthwhile considering the
Euribor-OIS spread.
EONIA and 3-month EURIBOR rates

6
EONIA/OIS
3-month EURIBOR
5
Percent 3
0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Year
Source: European Central Bank (original data provider for 3-month EURIBOR: Thompson Reuters)
Figure 7.2: Eonia and 3-month Euribor rates
The OIS (overnight index swap) rate is the rate on a derivative contract in the course
of which the geometric average of overnight interest rates is exchanged against the
term OIS rate over the maturity of the contract. The difference is settled in cash. In
this respect the OIS rate is a reflection of financial markets' expectations regarding
the overnight rate (Eonia in the euro area) over the maturity of the contract (see
Thornton (2009)). Because no exchange of principal is involved, there is very little
default risk associated with OIS contracts.
The Euribor indicates the rate at which banks are willing to lend to each over the term
of the Euribor contract. It includes both their expectations about how interest rates
will evolve over the term of the contract and the risk that is associated with lending to
other banks.
Taking the spread between the 3-month Euribor and the OIS excludes or effectively
subtracts these expectations about the development of interest rates (that are
included in the OIS). This exercise thus yields a measure of the risk that is involved
when banks lend to other banks.96 This should be a much better reflection of the
funding conditions than "pure" overnight or Euribor rates (see figure 7.3 for the
evolution of the Euribor-OIS spread).
3-month EURIBOR to EONIA/OIS spread

175
150
125
100
Basis points
75
50
25
0
2005 2006 2007 2008 2009 2010 2011 2012
Year
Source: European Central Bank (original data provider for 3-month EURIBOR: Thompson Reuters); own calculations
Figure 7.3: 3-month Euribor-OIS spread
Therefore, one innovation in this study compared to others dealing with the period of
the financial crisis is that the 3-month Euribor-OIS spread is used as an indicator of
banks' funding conditions in addition to the Eonia rate and the 3-month Euribor.
The data on the 3-month Euribor is taken from the European Central Bank. Eonia
data is also taken from the ECB, while the original data provider is Thompson
Reuters. The Euribor-OIS spread is a derivative of the two other variables and is
based on the author's own calculations.
96
Putting it simply, the idea can be expressed as: Euribor (expectations + risk) - OIS (expectations
only) = Euribor-OIS (risk).
7.4 Data 109
A special category of data reflects the implementation of non-standard monetary

policy measures. In response to the shortage of liquidity in the wake of dysfunctional
interbank markets, the European Central Bank took a number of unconventional
steps. The ECB responded, first, by injecting additional liquidity through large-scale
overnight fine-tuning and additional longer-term refinancing operations. Then, from
September 2008 onward, its most important measure was adopting the fixed-rate full
allotment policy and relaxing collateral requirements (see Trichet (2010), European
Central Bank (2009) and Lenza et al. (2010)97).
These measures had a significant alleviating impact on interbank rates and thus
influenced banks' funding conditions during the time of financial turmoil. Restoring the
functioning of the short-term funding markets was indeed the primary goal of the
ECB's unconventional measures (see Trichet (2010) and Gonzlez-Pramo (2011)).
To account for the impact of these measures on bank lending, this study followed the
approach adopted by Gambacorta and Marques-Ibanez (2011) and Brei et al. (2013).
They form the quotient of the central bank's assets in relation to nominal GDP and
use it in the estimation as a proxy for the ECB's non-standard measures. This
measure is especially designed to capture the growth of the ECB's balance sheet
caused by the extra liquidity provided.
From a theoretical point of view, however, it is not completely convincing to let the
ratio of the central bank's assets to nominal GDP influence lending directly (see
discussion of Gambacorta and Marques-Ibanez (2011)). As stated above, the
unconventional measures sought to restore the short-term interbank markets to
proper working order. Consequently, these measures may be already reflected in the
monetary policy variables (the Eonia, Euribor or Euribor-OIS spread, respectively)
which are in any case included in the estimated equation. To shed light on the
question of whether this is actually the case is one of the results the use of the non-
standard-measures variable is expected to give.
97
Other measures taken by the ECB from September 2008 include the extension of the maturity of
refinancing operations (LTROs), the provision of liquidity in US dollars, and the purchase of euro-
denominated covered bonds. For a review see Trichet (2010), Lenza et al. (2010), European
Central Bank (2009) or Mercier (2009).
To construct the quotient of assets in relation to nominal GDP, the total assets of the
Eurosystem as provided by the ECB, are taken. Nominal GDP, as stated above,
originates from Eurostat.
7.4.2. Target capital estimation
Using a variable that measures a capital surplus or shortfall, as briefly touched on in

the previous section, is a relative novelty in the context of literature on bank lending.
This approach involves estimating bank-specific capital targets.
Considerations regarding target capital ratios can be traced back to the capital
structure puzzle put forward by Myers (1984). The basic question is how firms
choose their capital structures, which, in itself implies a departure from the capital
irrelevance proposition formulated by Modigliani and Miller (1958). This question has
produced a very large body of literature in the field of corporate finance; and without
revisiting this discussion in detail, some studies are worth noting in the context of this
study. One of them is Flannery and Rangan (2006), who conclude that non-financial
firms do target capital ratios. After a target has been missed firms do not fully but
partially adjust in the subsequent period. Lemmon et al. (2008) add to this finding by
observing that, despite the explanatory power of certain firm characteristics, the
majority of the variation in non-financial firms' leverage ratios can be explained by
time-invariant fixed effects. As debatable as this result may be, a related idea applied
to banks has recently started to gain popularity.
The general idea that banks, too, follow target ratios and that this has an impact on
their lending behavior was first expressed by Hancock and Wilcox (1994). In this and
a subsequent study (Hancock and Wilcox (1998)), they employ variables for capital
shortfalls or surpluses in a bank lending context. However, the shortfalls are
measured relative to regulatory standards, not self-imposed ones. The authors report
a significant lending reaction on the part of banks in response to shortfalls relative to
unweighted capital ratio standards.
Borrowing from corporate finance literature and following the methodology already
applied to the capital structures of non-financial firms, Gropp and Heider (2010) find
that banks, like non-financial corporations, target individual capital ratios. These can
largely be explained by bank-fixed effects, which are supplemented by bank
characteristics as further explanatory variables. The individual nature of the capital
7.4 Data 111
targets implies that there is no common capital target, such as the regulatory
minimum capital ratios stated in the various Basel accords. The same line of thought
is pursued by Berrospide and Edge (2010) and Francis and Osborne (2009), who
model banks' target capital ratios as a function of a fixed effect and of certain bank
characteristics.
All three authors who concern themselves with banks' capital structures (Francis and
Osborne, Berrospide and Edge, Gropp and Heider) confirm that, where capital
targets are missed, a partial adjustment process can be observed. According to this
process, only a part of the gap between actual and targeted capital endowment, but
not the entire gap, is closed in the subsequent period.
How exactly does the process of estimating the capital target ratio and of
constructing the capital surplus/shortfall variable work? The process involves two
stages: The first stage is to estimate bank-specific target capital ratios. In the second
stage, estimated target capital ratios are put into relation with actual capital ratios. A
positive deviation stands for a capital surplus, while a negative deviation represents a
shortfall.
For the first stage, the equity capital target for each bank i at time t, cap*it, is modeled
as dependent on a constant i (representing the individual fixed effect), the lagged
capital ratio, capit1, and a vector of N different bank characteristics:

. (17)
The variables included in Xn broadly follow those that are used in the cited adjacent
literature: They are size, the share of non-interest income, the share of short-term
funding and the share of liquid assets. All these variables are constructed as
explained in section 7.4.1.
The size of a bank, measured in logs of total assets, is believed to have an impact on
the targeted capital ratio because, as stated above, size can be perceived as a proxy
for the professionalism and transparency of a bank. This should have an impact on
the equity capital buffer demanded by the providers of external finance 98 and,
therefore, on the target ratio that a bank deems adequate, optimal or "right".
98
Or, vice versa, given a certain equity capital endowment the external finance premium should be
lower for bigger banks than for smaller banks.
A bank with a larger share of non-interest income might prefer to operate at higher
levels of capital, because this form of income is more volatile than interest income.
A higher share of short-term funding can be interpreted as a reflection of a rather

risky business or funding model, since it is typically associated with significant
maturity mismatches between the assets and the liabilities side of the balance sheet.
Accordingly, banks characterized by large proportions of short-term funding will
probably tend to have more capital.
The thinking behind the share of liquid assets is that liquid assets act as a buffer to
shield the bank from unforeseen funding problems. If raising funding is difficult for
whatever reason, a bank can then respond by selling securities at short notice
without being forced to reduce other, less liquid assets at higher cost, or even at a
loss. Relatively liquid banks might therefore be comfortable with a lower level of
capital.
Obviously, equation (17) cannot be estimated since values for the target ratios, cap*it,
are missing. Hence, a further step is necessary: The assumption is that banks
partially adjust in line with their capital targets after a deviation. This is expressed as

, (18)
where capit is the actual capital ratio at time t, is the speed of adjustment and it is
an error term. The missing step is now to substitute (17) into (18). Rearranging this
yields:

, (19)
Equation (19) implies that all banks adjust toward their targets at the same speed.
Flannery and Rangan (2006) once modeled as a function of firm characteristics, but
reported no significant improvements by doing so. It further implied that the long-run
impact of the bank characteristics in Xn on capital are given by the estimated
parameters, n, divided by .
Equation (19) can now be estimated with OLS. The coefficients obtained are plugged
into equation (17). Together with the respective representations of the variables of
the individual banks, this yields individual and time-variant capital targets for each
bank in the sample.
All that remains to be done to construct the surplus variable is to calculate the
difference between capit and cap*it for each bank in every period. The difference is
7.4 Data 113
then divided by the capital target. This is necessary to make the surplus or shortfall
proportional to the target ratio. Otherwise, a shortfall of, say, one percentage point
would be equally "bad" irrespective of whether the target ratio is five per cent or 25
per cent.99
Constructed in this way, this variable is used as an additional bank characteristic in

the empirical model (equation (16)).
7.4.3. Special challenges
Having presented the sources of the data and the construction of the variables in
detail in the sections above, the present study now addresses to two further data-
related topics. Both represent special challenges and therefore merit separate
treatment. The first issue concerns itself with the question how it is possible to
effectively control for factors that drive loan demand. This is necessary because the
amount of loans observed on banks' balance sheets is merely the outcome of the
settlement of supply and demand. The second issue deals with the length of the
crisis period and the line of argument based on which the crisis period relevant to this
research enterprise can be correctly identified.
7.4.3.1. Disentangling loan supply and loan demand
One well-known problem in the context of the bank lending channel and a major
empirical challenge when trying to identify the impact of bank characteristics on
lending is the disentanglement of supply and demand factors. Theoretically, an
observed change in the volume of loans recorded on banks' balance sheets could be
attributed either to shifts in the supply or in the demand of loans or to a combination
of the two. The correlation between loan demand and loan supply or, more precisely,
between demand and certain factors that have an impact on the supply of loans, can
work through variations in the course of the business cycle and follows two main
patterns (see e.g. the discussion of Berrospide and Edge (2010).
According to the first pattern, deteriorating investment opportunities and subdued

economic prospects in an economic downturn lead to a decline in demand for loans
99
In the first case the deviation from the target would be 20%, in the second case it would be only
4%.
from potential borrowers. At the same time, in the downturn credit defaults require
banks to write off loans, which has a negative impact on the bank's capital position
with the known adverse consequences for lending. Moreover, insofar as the
economic downturn translates into a decline in the value of securities, the liquidity
positions of banks too are affected. In addition, both aspects lead to a shrinkage of
the balance sheet, such that the size of a bank as a factor potentially impacting the
supply of loans also declines.
Deposits serve as an additional example (see Loutskina and Strahan (2009)): A large
amount of insured deposits available to a bank is supposed to have a positive impact
on the supply of loans. However, strong demand for loans, e.g. in a boom phase,
might cause banks to intensify their efforts to attract insured retail deposits. This
could induce a correlation between demand for loans and deposits which, when
regressed, suggests that deposits drive the loan supply (because both observed
loans and deposits rise), when in fact deposits might equally have been driven by
loan demand. These examples reveal a first pattern from which a business-cycle
related correlation between supply and demand can arise.
According to the second pattern, the quality of borrowers declines in an economic

downturn, making them more risky. Simultaneously, a downward shift in the quality of
the borrower pool results in higher risk weights and necessitates higher equity capital
holdings than before for the same credit portfolio, such that less capital is available to
expand the credit supply. This effect thus has two implications: One is a direct
decrease in the (risk-weighted) capital ratio and the other is that the bank will come
closer to being bound by regulatory or self-imposed capital constraints. This is the
second example of a correlation between demand-related and supply-related factors
that have an impact on the observable amount of loans. Since, in fact, capital does
not drive loans in this example, but both capital and loans are driven by deteriorating
economic conditions this correlation will lead to conclude that the effect of capital on
loans is stronger than it really is if one fails to account for this factor.
In addressing these issues, the most common approach is to use macroeconomic

variables such as GDP and inflation-related variables in the regressions. The GDP
variable is an outright measure of aggregate demand and has a direct, demand-
driven interpretation regarding loan growth. The inflation variable reflects the positive
impact of price level changes on nominal loan growth.
7.4 Data 115
The standard assumption is that these variables effectively capture the demand
effects on the observed volume of loans, such that the partial effects of bank
characteristics on bank loans that remain have a causal interpretation rather than
representing mere correlations.
In addition, this study follows a new and innovative approach in disentangling loan
supply from loan demand by using data taken from the euro area bank lending
survey, as explained in section 7.4.1. In particular, answers to the question of how
demand for loans and credit lines (to enterprises) has changed over the past three
months above and beyond normal seasonal fluctuations are evaluated and included
as a supplementary variable intended to capture the demand for credit.
The demand-related responses to the bank lending survey may be superior to the
use of GDP measures, because they are directly geared to loan demand. In addition,
it is theoretically possible that a causality between GDP and loans might exist that
works in the opposite direction: A shock to the supply of loans (a "credit crunch"), for
example, could cause GDP to contract. From a theoretical perspective, there is
therefore much to suggest that using data obtained from the euro area bank lending
survey is a fruitful option indeed.
7.4.3.2. Determining the relevant crisis period
One central concern of this study is to clarify the role of those bank characteristics
that drive bank lending and, especially, to identify their impact during the recent crisis.
Although the term "crisis" has already been used several times in this study, a
definition has not yet been provided. According to conventional wisdom one can
distinguish between different stages or emphasize different aspects of the crisis
(subprime lending crisis, banking crisis, global financial crisis, sovereign debt crisis,
etc.). Some of these stages or aspects of the crisis are still ongoing.
This section serves to answer the question which aspect of the crisis is relevant in
the context of bank lending and what is the distinctive feature that characterizes this
aspect. On the basis of this definition, the length of the crisis period is determined,
which in turn determines the crisis dummy to be used in the empirical model.
The starting point for these considerations is to ask in what respect or by what events
banks have been affected in the conduct of their business. The first event to be
mentioned is the financial turmoil that started in August 2007. Around this date,
overnight money market rates rose to higher levels and were accompanied by
greater volatility in the money market (see figure 7.4). The ECB responded by
providing overnight credit of EUR 95 billion at the main refinancing rate in exchange
for collateral (see e.g. Trichet (2010)). This measure helped to relieve some of the
tension on the money market, although spreads100 and the volatility of money market
rates, especially at longer maturities, remained significantly higher than before.
3-month EURIBOR-OIS spread during times of financial turmoil/crisis
Relevant crisis period
200
August 2007 15 Sep. 2008 6 July 2009 30 June 2010
Begin of tensions Lehman collapse Start of CBPP End of CBPP
175
15 Oct. 2008
150 Begin of fixed
rate full Dec. 2009
allotment Phase-out of
125
non-standard
measures
Basis 100 announced
points
75
50
25
0
Okt. 07
Okt. 08
Okt. 09
Okt. 10
Apr. 07
Apr. 08
Apr. 09
Apr. 10
Juli 07
Juli 08
Juli 09
Juli 10
Jan. 11
Jan. 07
Jan. 08
Jan. 09
Jan. 10
Apr. 11
3-month EURIBOR-OIS spread Month/Year

Source: European Central Bank (original data provider for 3-month EURIBOR: Thompson Reuters); own calculations
Figure 7.4: The impact of selected events on the evolution of the 3-month Euribor-OIS spread during
times of financial turmoil/crisis
To understand how this negatively affected the funding conditions for banks, one
must be mindful of the role of the money market (market for reserves/monetary base)
in this context: The monetary base that the ECB provides corresponds to the
aggregate need of the euro area banking system. It is determined on the one hand
by the reserve ratio, which is the share of deposits (subject to reserve requirements)
100
For example, the 3-month Euribor-OIS spread and the spreads between the overnight index swap
rate and Euribor rates of other maturities. These spreads can be interpreted as a measure of
counterparty risk. See section 7.4.1 for details.
7.4 Data 117
that have to be held in the form of reserves with the ECB.101 On the other hand, the
need for central bank liquidity is determined by the volume of deposits itself which, in
turn, depends on the amount of loans that the banking system grants on aggregate to
non-financial corporations. 102 Since banks regularly report all relevant loan and
deposit positions to the ECB, it knows exactly how much liquidity it must provide to
the banking system on aggregate. A necessary condition for the functioning of the
interbank market is that banks actively lend to each other so that the liquidity the
central bank provides is effectively exchanged among the participants.
In the event that some banks' access to the required liquidity via interbank markets is
compromised as was the case in the period starting in August 2007 this is
associated with a rise in money market rates which, in the period in question,
reflected some banks' desire to hoard liquidity. These rates are then translated into
rates on, first, short-term and, then, longer-term loans and are reflected in the price of
external funding. This has an impact on banks' overall funding costs and their ability
to grant credit.
When Lehman Brothers collapsed on 15 September 2008, the tension immediately

escalated into a financial and economic crisis. Money market rates increased sharply
from already high levels (see figure 7.4) and exacerbated the funding problems of
banks caused by a breakdown of the interbank money markets. The ECB responded
by taking non-standard measures in order to restore the functioning of the money
markets as a key element in the transmission of monetary policy. These measures
included the unlimited provision of central bank money at a fixed rate (full-allotment
policy) and expansion of the scope of accepted securities to serve as collateral.103
While the ECB's deposit facility had been hardly used until then, a significant
increase in the use of the deposit facility was observable in the wake of the full
allotment policy and long-term tenders (see European Central Bank (2011a)),
101
For details of the fractional reserve system, see European Central Bank (2011b), p. 101 et seq. or
Bofinger (2001), p. 343 et seq.
102
Note that the link between a loan and a deposit is given by the fact that the granting of a loan is
associated with the creation of a deposit of equal value to the account of the beneficiary. This
deposit is then subject to reserve requirements.
103
An overview of additional measures taken by the ECB can be found e.g in Trichet (2010) or in
European Central Bank (2011a).
highlighting the mood of uncertainty and the failure of the money market to effectively
distribute liquidity among the participants. Another important element to improve the
funding conditions of banks was the "covered bond purchase programme" (CBPP),
which was announced in May 2009 and included outright purchases of covered
bonds in issue during the 12-month period beginning in July 2009 (see European
Central Bank (2011a)).104 The launch of this program further underlines the funding
problems that banks were facing due to the fact that the money market was no longer
working properly.
Together with a reduction in the key interest rates, the measures taken by the ECB
succeeded in reducing money market rates to normal levels105 and improving banks'
funding conditions. Because non-standard measures such as those undertaken in
the preceding months were no longer necessary, the ECB was able to announce the
gradual phase-out of these measures in December 2009.
In a nutshell, the dysfunction of the money market in conjunction with higher rates
translated into higher funding costs for banks even though the bank characteristics
had not necessarily changed. In the course of these developments, negative
feedback loops such as those described in chapter 5 were set in motion, with all the
externalities that then significantly contributed to the propagation of the crisis within
the financial system. This is the key rationale for the hypothesis that, during the crisis,
some bank characteristics played a different role in determining bank lending.
It follows that the period of unusually high money market rates shown in figure 7.4 is
the crisis period of relevance to this study. It spans the years 2008 and 2009. The
crisis dummy used in the estimation of equation (16) is constructed accordingly.106
104
For an (ECB in-house) assessment of the CBPP see Beirne et al. (2011).
105
"Normal" levels can probably be described as not being influenced by liquidity shortages and as
characterized by the absence of unusual high volatility.
106
This choice is in line with other studies that explicitly consider the crisis period, e.g. Gambacorta
and Marques-Ibanez (2011). Based on quarterly data, the latter define the relevant crisis period as
extending from the third quarter of 2007 to the fourth quarter of 2009.
7.4 Data 119
7.4.4. Purging the data
This section describes in detail the steps necessary to prepare the data for empirical
analysis and without which meaningful results would not be possible. The steps
involve excluding banks that have only marginal lending business, correcting for
mergers and acquisitions, eliminating banks with missing values for capitalization,
deposits and liquid assets, and removing outliers. An overview of the steps is given in
figure 7.5.
Exclusion of banks that have only marginal lending business
In line with e.g. Favero et al. (1999) and to avoid distortions in the results stemming
from banks for which lending is only a marginal business, all institutions whose ratio
of loans to total assets is less than five per cent are eliminated. 107 This group of
banks consists mainly of specialized institutions such as guarantee banks, trust
companies, securities banks etc., for which it can be assumed that loan business
follows patterns that differ from those under scrutiny in this study those that operate
significant client loan business.
Rationale
Rationale Approach/procedure
1 > Exclusion of banks with a marginal lending > Eliminiation of banks with a share of loans over total
business (mainly securities banks, guarantee banks, assets smaller than 5%
trust companies)
2 > Correction for mergers & acquisitions > Elimination of banks who show an asset growth of greater
than 67% in at least one year of the sample period
3 > Elimination of banks with missing values for > Elimination of banks without any values for total
capitalization otherwise no meaningful analysis equity/total assets
possible (capitalization key driver of lending)
4 > Elimination of banks with missing values for > Elimination of banks without any values for customer
deposits and liquidity otherwise no meaningful deposits and cash and due from banks over total assets
analysis (e.g. also concerning short-term funding,
deposit overhang) possible
5 > Elimination of outliers for > Elimination of banks with annual growth rates for the
Loans respective variables within the 1st or above the 99th
Total equity over total assets percentile in at least one year of the sample period
Cash and due from banks over total assets
Figure 7.5: Overview of steps in purging and cleansing the data
107
The five per cent threshold is calculated as an average of the available values over the sample
period (1999-2011).
Elimination of banks with no reported values for capital, deposits and liquidity
A number of the hypotheses to be tested (see section 7.1) are directly geared to
capital, deposit and liquidity variables. Banks that do not report any values for these
variables cannot contribute at all to exploitation of the differences in these variables
for the purpose of explaining lending since, given all that we know from the available
literature, these variables have considerable explanatory power regarding the lending
behavior of banks. Worse still, including these banks could contribute to biased
estimators, because the time-series variation in loans for these banks might in part
be attributed to other variables for which observations are available that are loosely
correlated with capitalization, deposit or liquidity variables. Hence, in order to obviate
the omitted variable problem, these banks are completely eliminated from the sample.
Treatment of mergers and acquisitions
Another issue regards the treatment of mergers and acquisitions. Several

approaches are pursued in the relevant literature. The most important of these are
the complete exclusion of banks that were involved in a merger or acquisition during
the sample period, and the synthetic aggregation of merging banks for the years
before the merger actually took place. The latter approach includes the assumption
that the merger in question took place in t0, i.e. at the beginning of the sample period.
In a study in which the robustness of results was explicitly tested with regard to
alternative ways of handling mergers, Worms (2001) finds only minor qualitative and
quantitative differences. Moreover, even where a study does not control for mergers
at all, the broad picture does not change more than marginally. Similar results have
been obtained by Kishan and Opiela (2000) and Ehrmann et al. (2003). This
suggests that the treatment of mergers and acquisitions is of little relevance to the
results.
However, to account for mergers in this study, the general approach adopted by
Favero et al. (1999), Worms (2001) and Alper et al. (2012), for example, was applied.
Accordingly, banks that are affected by mergers are excluded from the sample. For
7.5 Estimation method 121
this purpose, all banks whose total assets grow by more than 67% in any of the years
in the sample period are excluded.108
Elimination of outliers
An important issue with outliers which are probably attributable to measurement

errors, input errors or one-time events, for example is that they have the potential
to significantly bias estimation results in a panel regression. This holds true
especially where the first-difference estimator method is used in the context of
dynamic panels (see Wooldridge (2002), chapter 11, Griliches and Hausman (1986)
and Solon (1985)). It is therefore advisable to eliminate these outliers.109 The risk of
removing regular observations with the method used to detect outliers is justifiable
in light of the danger of obtaining biased results (see e.g. Worms (2001)).
Therefore, following the general approach of Worms (2001), Ehrmann et al. (2003)
and Gambacorta (2005), for example, an observation is defined as an outlier if its first
difference is in the top or bottom percentile of the given distribution. This applies to
capitalization, loans and liquidity as the most relevant variables with respect to the
hypotheses tested in this study. Every bank that has at least one outlier is removed
from the sample entirely.
7.5. Estimation method
Having introduced the model to be estimated and the data to be used, the question of
the appropriate estimation method arises. This section motivates the choice
especially of the generalized method of moments (GMM) methodology and certain
features of the estimation, including a discussion of the reasons why these methods
are advantageous in the context of the current research undertaking.
The starting point is the observation that the model estimated in equation (16) has a
lagged endogenous variable as regressor and a time-invariant but bank-specific
108
An alternative approach along the same lines is to refer to a database of mergers and acquisitions,
and then to manually identify and eliminate all relevant banks. Unfortunately, no such database
was available.
109
The first-difference estimator method and the GMM methodology are discussed in detail in
connection with the estimation method in section 7.5.
factor (ai). This factor is often called fixed effect because it does not change over time.
Why is it likely to encounter a fixed effect, i.e. why is incorporation of this factor
justified?
In the context of the determinants of bank lending, there may be unobserved factors
that could, for example, emanate from business stances such as risk aversion,
managerial/employer quality, the organizational structure of a bank and/or the
"quality" of its processes. Hence, one could conceivably encounter a situation in
which different levels of loans exist even if all observed variables take the same
values. Ultimately, therefore, one cannot rule out the possibility that (observable or
unobserved) factors exist which also impact bank lending.
Because the fixed effect is unobserved, it would normally be covered by the error
term, along with the time-variant, idiosyncratic error. However, a problem now arises
in connection with the presence of the lagged endogenous explanatory variable: As
Nickell (1981) shows, the lagged endogenous regressor is correlated with the fixed-
effect proportion in the error term, which violates one of the key conditions under
which OLS produces unbiased estimators. Using OLS thus results in the famous
"dynamic panel bias".
However, the panel structure of the data offers several solutions to deal with fixed
effects. In addition to the alternative methods dummy variable regression, time-
demeaning/within transformation or random effects model, all of which are
inappropriate for specific reasons,110 two popular alternatives are the method of
first-differencing and orthogonal transformation. They are designed to capture the
110
In a dummy variable regression, one dummy is included for every "individual", with the
disadvantage that this regression quickly becomes unmanageable when the number of cross-
sections is large, as it is in this case (see Wooldridge (2002), p. 272 et seq.).
The procedure of time-demeaning, also known as within transformation, involves writing the
equation to be estimated in averages of the time for each cross section. In the averaged equation,
the fixed effect is also present. If one subtracts the averaged equation from the original one, the
fixed effect drops out (see Wooldridge (2002), p. 265 et seq.). Unfortunately, as Nickell (1981) and
Bond (2002) confirm, this approach addresses the fixed effect but does not eliminate the dynamic
panel bias.
The random effects model requires that the unobserved effect be uncorrelated to all explanatory
variables in all periods (see Wooldridge (2002), p. 272 et seq.). This assumption is hard to justify
in this case.
fixed-effect part of the information otherwise contained in the error, while the time-
variant (idiosyncratic) components remain in the error term.111
How does first-differencing work? Without going into too much detail regarding
formalization the basic idea is easy to explain: The equation to be estimated is
expressed as representing a certain point in time (for example at t = ). In the next
step, the equation is expressed as representing the next available point in time (e.g. t
= ). By subtracting the equation at t = from the equation at t = , the intercept that
captures the individual effect missed by observed variables is canceled out since it is
time-invariant. Thus, by subtracting the observations of all explanatory variables at t =
from the relevant previous ones, one obtains a first-differenced equation in which
the individual unobserved effect drops out. According to Roodman (2009b), the
weakness of the method of first-differencing is that, when a data point is missing at a
certain point in time, one loses not just one observation but also the adjacent
observation due to the differencing procedure .
This shortcoming is cured by the orthogonal deviations procedure (Arellano and

Bover (1995)), which involves subtracting the average of all future available
observations of a variable from the contemporaneous observation of this variable. In
doing so, one only loses the last observation instead of many more in the event of
data gaps. Given a fully balanced panel in which no observations are missing, both
approaches yield the same result.
The orthogonal deviations procedure is generally used in connection with the GMM
estimator (see explanation further below), whereas the method of first-differencing is
also applied in when estimating with OLS. Why not applying first-differences and
perform the estimation using OLS? Besides the issue of data losses, this leads to
another problem: Estimation with OLS results in an unbiased and consistent
estimators only under certain conditions. The crucial assumption is the exogeneity of
the explanatory variables. When employing first-differencing, the least strict
formulation of the exogeneity condition amounts to , t = ,,T, which
means that the difference in the error term is uncorrelated to the bank-specific
111
See Wooldridge (2002), p. 248 et seq. for a general treatment of unobserved effects panel data
models, and also for the assumptions underlying difference GMM estimations.
explanatory variables (see Wooldridge (2002), p. 317 et seq.). Unfortunately, the

exogeneity requirement is likely to be violated.
The nature of the model estimated in equation (16) and the relationships between the
variables therein suggest that the problem of endogenous explanatory variables
could well be relevant. When certain bank characteristics are strongly correlated to
bank loans, it is not easy to clearly determine whether only these bank characteristics
drive loans, or whether the volume of loans might also drive certain bank
characteristics. To give an example: While the economic theory referred to in
previous chapters suggests that the capitalization of bank has a positive impact on
bank lending for a number of reasons i.e. allowing easier access to uninsured
sources of funding by mitigating the moral hazard problem associated with
asymmetric information or acting as a buffer against possible losses an increase in
the volume of loans also demands an increase in capital. Thus, the direction of
causality cannot readily be established in every case.
Another obvious and harmful example is the size of a bank measured in terms of
total assets or some derivative thereof. According to the standard hypothesis, size
matters for the supply of bank credit because it is assumed to reduce informational
friction, especially when a bank tries to tap alternative sources of funding. Large
banks thus benefit from more favorable funding conditions which they can pass on to
borrowers, and which should be reflected in the size of the loan portfolio. The thing is
that, when a bank grants credit, the amount of assets increases by exactly the
amount of the loan at the same moment. This shows that causality works in the
opposite direction as well a direction that logically derives from the mere fact that
loans are a subset of total assets. Hence, it is not in the least surprising that a
statistical connection between size and loans can easily be found. However,
interpreting this observation as being in line with the hypothesis stated above might
be premature if one does not take into account that size measured in terms of total
assets (or in logs or differences of total assets) is endogenous to the amount of loans.
Unfortunately, it appears that not all scholars have paid particular attention to this
issue.
To sum up: The issue with fixed effects can be resolved using the first-differencing
method. However, the problem of endogenous explanatory variables means that the
OLS estimator cannot be used as it produces biased and inconsistent estimators.

This speaks for the orthogonal deviations procedure.
When the data has a panel structure the GMM framework is especially suitable.112
The GMM estimator proposed by Arellano and Bond (1991), called difference GMM
estimator,113 is particularly popular in the context of bank lending and bank lending
literature. It has certain properties (see figure 7.6) that do not only help to tackle the
issue with fixed effects but also that of the endogeneity of explanatory variables.
Properties of difference GMM Rationale for applicability for this study
1 > Especially suitable for panels with small T > Sample period comprises years 1999-2011 9
and large N > More than 3.000 banks enter estimations (after purging)
> Power of GMM increases with growing N
compared with alternatives (e.g. direct bias
corrected estimates by Hansen (2001))
2 > Linear functional relationship > Loans depend linearly on explanatory variables 9
3 > Dynamic left-hand-side variable that > Lags of loans used as explanatory variable 9
depends on its own past realizations
4 > Presence of individual fixed effects > Presence of fixed individual effects cannot be ruled out 9
> Emanating from managerial/employer quality, organiza-
tional structure, "quality" of processes or risk aversion
5 > Presence of endogenous explanatory > Strict exogeneity cannot be warranted for all explanatory 9
variables variables
6 > Only available instruments are internal/ > Other, external and valid instruments not available 9
based on lags of instrumented variable
7 > Heteroskedasticity and autocorrelation may > Here, idiosyncratic disturbances (those apart from fixed 9
be present within individuals but not across effects) may have individual-specific patterns of
them heteroskedasticity and serial correlation
> Not likely that these patterns are present across individuals,
especially if it is controlled for demand factors
Source: Figure according to Roodman (2009b)
Figure 7.6: Properties of the difference GMM estimator and their applicability
112
See e.g. Newey and McFadden (1994) and Harris and Matyas (1999) for a general treatment of
the GMM estimator, or Hansen (1982) for the original source.
113
It is called difference GMM, because in its original form the authors used the first-differencing
approach (not the orthogonal deviations method). However, the difference GMM estimator allows
for both methods to remove individual fixed effects first differencing and the orthogonal
deviations method.
As one element to counter the endogeneity issue the GMM methodology involves the
instrumental variable procedure. 114 It produces consistent estimates, provided that
two conditions hold true: The first is that the instrumental variable used by the
endogenous explanatory variable must be uncorrelated to the error term or,
equivalently, be exogenous. Formally, it can be expressed as , where z
is the instrumental variable and is the error. This makes perfect sense: Without this
condition, the same problem of endogeneity which, as stated above, causes the OLS
estimator to be unsuitable would exist and nothing would be won.
The second condition is that the instrumental variable must be correlated with the
endogenous explanatory variable. Formally, this can be expressed as ,
where x stands for an explanatory variable that is suspected of being exogenous.
Because the condition demands some form of relationship between the endogenous
explanatory variable and the instrument, it is referred to as the instrumental
relevance.
To estimate the parameters consistently, at least as many instrumental variables as

endogenous explanatory variables are required. Otherwise, the equation remains
unidentified. If there are more instruments available than required, the equation is
said to be overidentified. This implies that there are more moment conditions
available than parameters to be estimated, which allows the additional moment
conditions to be used in a test for overidentifying restrictions. The J-test also known
as the Sargan or Hansen test is a specification test that sheds light on the validity
of instruments (see Newey and McFadden (1994) for formalization and a detailed
derivation). If the J-statistic suggests rejecting the null hypothesis of valid instruments,
the instrumental variables are poor.
According to Arellano and Bond (1991), another property of the difference GMM
estimator is that it is especially suitable in situations where only internal instruments
are available, meaning that all instruments are based on lags of the instrumented
variables, which is the case in the present study. With regard to individual fixed
114
As already stated, a second element in tackling the endogeneity issue, alongside application of the
GMM methodology, is to use bank characteristics as explanatory variables that are all lagged once.
The basic idea in the case of size, for example is that lagged values of size are not influenced
by current changes in the amount of loans, such that the coefficients on the lags are more suitable
to be interpreted in line with the hypothesis that size impacts lending.
effects, lagged dependent variables and endogenous explanatory variables, their

procedure produces unbiased and consistent estimators provided that the errors in
the model are not subject to serial correlation of order two.
This methodology has been further developed by Arellano and Bover (1995) and
Blundell and Bond (1998) toward what is referred to as the system GMM estimator.115,
116
System GMM is motivated by one weakness of difference GMM. It is relative
inefficient in the presence of weak instruments. To overcome this shortcoming, they
suggest making use of the information contained in the levels of the instruments that
disappear by transforming the variables into first differences. Consequently, they
advise to take lags in levels as instruments for those right-hand side variables that
are suspected of not being strictly exogenous. The authors report significant
performance gains by exploiting the additional moment conditions. The power of this
approach in a dynamic panel environment is further underscored by the results of
Behr (2003).
Bun and Windmeijer (2010) show that the system GMM is particularly advantageous
when the underlying time series are persistent. As becomes clear in the course of the
discussion of the results of the empirical analysis, however, this is not the case here.
Accordingly, this study follows the lead given by Arellano and Bond (1991) and uses
the difference GMM estimator.
One crucial point (for difference GMM and system GMM) is that invalid instruments
can induce a substantial bias for the estimators. The validity of the instruments used
is regularly tested with the Sargan test. If the respective J-statistics show satisfactory
results, researchers usually accept the validity of the instruments. However, an
important caveat in this context, as rigorously set out by Roodman (2009a), is the
weakness of such tests when the number of instruments grows large. This can easily
be the case if one uses all available lags as instruments, such that the number of
instruments explodes as the sample period increases. The consequence is that the
test suggests that the instruments are valid on the basis of which, together with
115
The original reference is Arellano and Bover (1995) while Blundell and Bond (1998) have made
explicit the assumptions under which the new-style instruments (see below) are valid. This has
resulted in the nomination of Blundell and Bond as the authors of system GMM.
116
The properties presented in figure 7.6 also apply to system GMM.
other necessary assumptions, one concludes that the estimated coefficients are
unbiased although in fact they are not.
To avoid this pitfall, this study follows Roodman (2009a), who suggests limiting the
number of instruments by only using certain lags, e.g. the first lag for all bank
characteristics (apart from loans) instead of all available lags. This limitation restores
the power of tests for the validity of the instruments. In practical terms, this means
that the endogenous explanatory variables are instrumented by the third and fourth
lag of loans in levels. In addition, the first lag of the bank characteristics and the
interaction terms of the bank characteristics with the crisis dummy used in the
equation serve as instruments. 117 The macro variable(s) and the monetary policy
variable are regarded as completely exogenous and instrumented by themselves.
To sum up: The validity of the instruments, in addition to the absence of serial
correlation of order two in the errors, shows that the GMM estimator is consistent,
efficient and asymptotically normally distributed.
For the present study the instruments are chosen in accordance with Arellano and
Bond (1991). The validity of the (limited number of) instruments is tested with the
Sargan test. The respective J-statistic and the number of instruments used are
reported together with the results. Moreover, because of gaps in the data, orthogonal
deviation is preferred over first-differencing. The estimations are carried out within the
difference GMM framework. It is the estimator which is the one most widely used and
best established in the relevant literature a further reason that argues for it.
117
The exact choice of instruments is reported together with the corresponding estimation results,
since it is sensitive to the exact specification of the estimated equation.
8.1 Results for the euro area 129
8. Empirical analysis results
Having clarified the author's approach to empirical analysis in the previous chapter,
the scene is set to dedicate this chapter to the results as one important centerpiece
of the entire study.
The chapter on results can be divided into two parts: The first part deals with the
results for the euro area as a whole. For the second part, subsamples of the four
major euro area countries Germany, Italy, France and Spain are constructed.
These subsamples are analyzed separately.
To make it easy to follow the results, this chapter has a clear structure: The results
for the euro area as a whole include the descriptive statistics and correlations as well
as the results of the empirical analysis (see figure 8.1 for the structure of the
empirical estimations).
The empirical analysis comprises the baseline analysis consisting of a standard

specification and a considerable number of robustness checks. This is followed by
two groups of analyses, the first of which focuses on the capital surplus and the
second of which focuses on the deposit overhang. Again, each analysis consists of a
standard specification and various robustness checks.
That is followed by a presentation of the results for the four individual countries. To
allow for comparison, a standardized, unvarying approach to the estimations is
chosen: Three types of estimations are carried out for each country: The first type is
the baseline model. The second is designed to test the impact of a capital surplus or
overhang, and the third focuses on deposit overhangs. Each type includes several
specifications to check for the robustness of the results.
8.1. Results for the euro area
The results of the baseline analysis are presented first. These results already reflect
answers to many of the hypotheses to be tested. After that, two groups of analyses
the first one focused on a capital surplus/shortfall relative to a bank-specific target
and the second one focused on the impact of a deposit overhang complement the
analysis and yield complete picture regarding derived hypotheses. A summary of

130 8 Empirical analysis results
tested hypotheses which are based on the euro area sample is given in figure 8.2
below.
Tested specifications for euro area (as a whole) Tested specifications for respective country samples
Euro area Germany Italy France Spain
Base- Standard specification Base- (I) Standard specification

line line
analysis Robust- Euribor + NSM analysis
ness Monetary Robustness check (II) Euribor-OIS spread + NSM
checks policy Eonia + NSM
indicator
Euribor-OIS-spread + NSM (III) Time-fixed effects
ECB Bank lending survey
Capital (IV) Standard specification
Time-fixed effects surplus
Capitalization (tangible common equity)
Robustness check (V)Time-fixed effects
Mark-to-market securities
Capital Standard specification Deposit (VI) Standard specification

surplus over-
Robust- Capital surplus and "plain" capital hang
ness Robustness check (VII)Time-fixed effects
checks Time-fixed effects
Deposit Standard specification

over-
hang Robust- Deposit overhang dummy variable
ness
checks Time-fixed effects
Figure 8.1: Structure of the empirical estimations
For all three groups of estimations, a generally similar approach is chosen: The first
estimation is given by a standard specification. This is followed by a number of
robustness checks to determine whether the results stand up to changes in the
relevant specifications. One important type of robustness check is the estimation
using time fixed effects, which involves the addition of time dummies for each year of
the estimation period. This design is intended to capture the full cross-sectional
impact of the explanatory variables.
The results are contrasted with those produced by studies of the euro area, where
available. Where relevant studies of the euro area are not available, also studies that
focus on the US are consulted.118
118
In this case it has to be borne differences in results can be due to generally different economic,
institutional and financial circumstances.
Tested using
Hypotheses euro area
sample
H1
hypo-

H4
is lower (in terms of the value of the coefficient) compared to
normal times
crisis H6
context
H8
impact on lending

H9
lending
Hypothesis tested Hypothesis not tested (refer to text for further remarks)
Figure 8.2: Tested hypotheses based on euro area sample
8.1.1. Descriptive statistics and correlations
An overview of descriptive statistics on euro area data and the correlation matrix are
given in table 8.1 and table 8.2.
Descriptive statistics are notable with respect to the number of observations of

certain variables: Data on securities that need to be marked to market (AFS) is
scarce compared to other bank characteristics. This poses limitations to AFS-related
estimations as is discussed further below. The lending survey variable (LEND_SURV)
is the second variable for which the number of observations is relatively low. This
reflects that the survey was not started by the ECB before 2003. In addition, all bank
characteristics are normalized with respect to their respective averages (apart from
capital surplus and deposit overhang as described in section 7.4). This results in a
zero mean. A further implication of the normalization procedure is that minimal values
are negative which as such has no economic interpretation but is technical or

mathematical result. Negative values for the median are sign of a distribution which is
skewed toward large values. Since the distances from the median values to zero are
relatively small in terms of standard deviations this is not considered to have a
relevant impact on the results.
EUROAREA
Variablename/symbol Numberof Mean Median Minimum Maximum Std.Dev.
observations
Dependentvariable
LOG(LOANS) 20,313 0.046 0.032 0.642 0.693 0.088
Independentvariables
Bankcharacteristics
CAP 23,799 0.000 0.014 0.076 0.924 0.062
CAP_TCE_TCA 23,799 0.000 0.013 0.076 0.925 0.061
CAPSUR 23,799 0.001 0.010 0.950 0.210 0.065
SIZE 22,967 0.000 0.151 7.511 8.052 1.579
DEP 22,750 0.000 0.044 0.659 0.337 0.180
OVERHANG 22,744 0.205 0.112 1.000 33.077 0.777
LIQ 22,846 0.000 0.000 0.018 0.375 0.013
STF 22,750 0.000 0.028 0.168 0.805 0.145
NII 22,834 0.000 0.006 2.957 2.405 0.148
AFS 8,485 0.000 0.040 0.069 0.830 0.092
Monetarypolicyvariables
EURIBOR 35,100 3.022 3.078 0.814 4.644 1.245
EONIA 30,420 2.583 2.736 0.438 4.387 1.276
EURIBOR_OIS 30,420 0.263 0.245 0.121 0.774 0.242
(NSM) 30,420 0.013 0.005 0.011 0.072 0.024
Macroeconomicvariables
LOG(GDP) 32,760 0.032 0.036 0.035 0.052 0.021
GDP_DEFLATOR 35,100 1.580 1.800 0.300 2.500 0.715
(LEND_SURV) 18,720 0.813 6.125 133.500 73.500 68.590
Table 8.1: Descriptive statistics for the euro area sample
The correlation matrix for the euro area sample reveals some interesting patterns.
The clearly negative correlation between size and the capital ratio of a bank means
that on average smaller banks have a higher capital ratio. This is consistent with the
conjecture that shareholders of large institutes demand higher returns on their
invested equity and that a higher leverage is a possibility to enhance profitability. It is
also consistent with the notion that bigger institutions are more robust to business
cycle fluctuations and, therefore, need less capital to reduce the asymmetric
information and moral hazard problem toward providers of uninsured funds.
CorrelationmatrixEuroarea
LOG(LOANS) CAP CAP_TCE_TCA CAPSUR SIZE DEP OVERHANG DUMMY_OVERHANG LIQ STF NII AFS EURIBOR EONIA EURIBOR_OIS (NSM) LOG(GDP) GDP_DEFLATOR (LEND_SURV) CRISIS
LOG(LOANS) 1.000

CAP 0.048 1.000
0.000
CAP_TCE_TCA 0.042 0.997 1.000

0.000 0.000
CAPSUR 0.074 0.005 0.008 1.000

0.000 0.422 0.253
SIZE 0.052 0.203 0.214 0.147 1.000
0.000 0.000 0.000 0.000
DEP 0.124 0.311 0.302 0.058 0.353 1.000

0.000 0.000 0.000 0.000 0.000
OVERHANG 0.111 0.109 0.109 0.015 0.144 0.519 1.000

0.000 0.000 0.000 0.031 0.000 0.000
8.1 Results for the euro area
DUMMY_OVERHANG 0.159 0.173 0.166 0.034 0.232 0.673 0.485 1.000

0.000 0.000 0.000 0.000 0.000 0.000 0.000
LIQ 0.150 0.103 0.098 0.018 0.072 0.331 0.180 0.258 1.000
0.000 0.000 0.000 0.010 0.000 0.000 0.000 0.000
STF 0.024 0.137 0.135 0.011 0.349 0.619 0.332 0.363 0.104 1.000
Table 8.2: Correlation matrix for the euro area

0.001 0.000 0.000 0.108 0.000 0.000 0.000 0.000 0.000
NII 0.014 0.131 0.121 0.026 0.066 0.154 0.041 0.051 0.045 0.140 1.000
0.046 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
AFS 0.148 0.304 0.299 0.018 0.151 0.137 0.048 0.102 0.191 0.288 0.098 1.000
0.000 0.000 0.000 0.098 0.000 0.000 0.000 0.000 0.000 0.000 0.000
EURIBOR 0.067 0.047 0.045 0.031 0.029 0.008 0.014 0.015 0.052 0.029 0.003 0.105 1.000
0.000 0.000 0.000 0.000 0.000 0.266 0.052 0.033 0.000 0.000 0.643 0.000
EONIA 0.061 0.063 0.060 0.034 0.030 0.003 0.015 0.013 0.064 0.043 0.007 0.128 0.982 1.000
0.000 0.000 0.000 0.000 0.000 0.718 0.027 0.071 0.000 0.000 0.347 0.000 0.000
EURIBOR_OIS 0.030 0.089 0.085 0.022 0.005 0.028 0.009 0.012 0.059 0.071 0.051 0.123 0.034 0.223 1.000
0.000 0.000 0.000 0.001 0.508 0.000 0.179 0.082 0.000 0.000 0.000 0.000 0.000 0.000
(NSM) 0.035 0.073 0.071 0.040 0.007 0.023 0.001 0.018 0.028 0.064 0.041 0.050 0.052 0.078 0.677 1.000
0.000 0.000 0.000 0.000 0.306 0.001 0.893 0.011 0.000 0.000 0.000 0.000 0.000 0.000 0.000
LOG(GDP) 0.071 0.045 0.042 0.015 0.026 0.014 0.017 0.009 0.029 0.029 0.010 0.171 0.518 0.582 0.404 0.068 1.000
0.000 0.000 0.000 0.021 0.000 0.053 0.017 0.206 0.000 0.000 0.141 0.000 0.000 0.000 0.000 0.000
GDP_DEFLATOR 0.015 0.054 0.051 0.034 0.018 0.008 0.007 0.000 0.066 0.028 0.021 0.172 0.633 0.719 0.534 0.128 0.469 1.000
0.033 0.000 0.000 0.000 0.011 0.271 0.296 0.966 0.000 0.000 0.003 0.000 0.000 0.000 0.000 0.000 0.000
(LEND_SURV) 0.055 0.005 0.006 0.005 0.006 0.015 0.006 0.013 0.029 0.005 0.018 0.048 0.832 0.763 0.581 0.742 0.333 0.562 1.000
0.000 0.491 0.423 0.567 0.510 0.081 0.506 0.116 0.001 0.584 0.035 0.000 0.000 0.000 0.000 0.000 0.000 0.000
CRISIS 0.027 0.042 0.039 0.006 0.011 0.023 0.014 0.001 0.025 0.041 0.011 0.146 0.029 0.098 0.670 0.160 0.764 0.257 0.246 1.000
0.000 0.000 0.000 0.339 0.101 0.001 0.042 0.920 0.000 0.000 0.116 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Thesampleperiodgoesfrom1999to2011.P valuesinitalics.Pairwisesamples(pairwisemissingvaluesdeletion).
133
In this picture fits also that there is a positive correlation between size and the
amount of short-term funding (+0.349). This suggests that bigger banks can more
easily tap wholesale funding markets. It also suggests that smaller banks can meet
their funding needs by resorting to insured customer deposits (negative correlation
between SIZE and DEP; -0,353). For the negative correlation between capital and
deposits it is hard to find an obvious explanation. Because smaller banks tend to
have higher capital ratios and because smaller banks tend to have a higher
proportion of deposits one could expect a positive correlation. It might be that, ceteris
paribus, for a bank that has better access to deposits it is less important to lower
costs for external finance by means of providing a large capital buffer since it is less
dependent on external finance.
With regard to the possible issue of endogeneity of explanatory variables an

interesting observation can be made: One could argue that the growth of the loan
portfolio implies a reduction of the share of non-interest income because the granting
of credit directly generates interest income. This would make the variable NII
endogenous to loans and would probably compromise estimated coefficients. If that
were the case one would be able to observe a relatively high negative correlation
between loans and the share of non-interest income. This is not the case. The
correlation coefficient of -0.014 is very small and certainly much smaller than one
would expect given possible endogeneity concerns of NII. This leads to the
conclusion that endogeneity of NII is not a serious problem in the sample used. To
further reduce possible endogeneity issues all bank characteristics enter the
regressions with one lag.
8.1.2. Baseline analysis
8.1.2.1. Results of the standard specification
Specification
The baseline analysis gives an answer to the majority of the hypotheses stated in
section 7.1. To this end, there is not only one standard specification that is tested.
Rather, the analysis is augmented by further specifications that serve as extensions
of the standard specification and as robustness checks. The tests for those
hypotheses that involve a capital surplus or deposit overhang are described in the
following sections. The reason for this special focus is that their use marks an
innovation compared to the referenced literature on bank lending.
The standard setup begins with one lag of the first difference of the natural logarithm
of loans. Next, to ensure instant comparability between the estimated coefficients
stemming from the crisis and from outside the crisis, the following approach is
chosen: All bank characteristics except for loans (i.e. capitalization, size, the share of
deposits, liquidity, the short-term funding ratio and the share of non-interest income;
all lagged once) enter the equation interacted once with the non-crisis dummy ("nc")
and once with the crisis dummy ("crisis").119 This non-crisis dummy exactly mirrors
the crisis dummy, meaning that it takes on the value of 1 in all normal years, when
the crisis dummy is zero, and vice versa. Comparison of the absolute coefficient
values thus gives an immediate picture regarding the extent to which certain bank
characteristics gain or lose importance during and outside a crisis. As stated in
section 7.3.2, to remove cross-sectional fixed effects, all interaction terms are
transformed using orthogonal deviations within the difference GMM framework. That
is why there is no perfect collinearity between the respective crisis and the non-crisis
interaction terms and, consequently, no dummy variable trap.120
The monetary policy indicator is given by the Euribor. To control for demand factors
in this first specification, GDP enters the equation in the form of its first difference of
the natural logarithm. Furthermore, price effects are controlled for by means of the
inclusion of inflation, represented by the GDP deflator for the euro area. The
inclusion of the crisis dummy variable in the regression allows for possible general
shifts in loan behavior during the crisis that are not accounted for by the bank
characteristics or other control variables.
Validity
The use of lags entails an adjusted sample period ranging from 2001 to 2011. 2,153
cross-sections and 14,061 (unbalanced) observations enter the regression.
119
This is done in order to remedy the problem of endogeneity (in connection with the instrumental
variable procedure within the GMM framework). Recall section 7.5 for details.
120
In fact, if one only uses the crisis dummy and not the dummy variable for the no-crisis period (for
example, only SIZE*CRISIS and SIZE), one will have to add the value of the crisis interaction to
the value of the non-interacted term (the value of SIZE*CRISIS to the value of SIZE) to obtain
exactly the same result. T-statistics are also identical.
The endogenous explanatory loan variable is instrumented by the dynamic

instruments proposed by Arellano and Bond (1991), i.e. in this case the second and
third lags of the natural logarithm of loans in first differences. All other bank
characteristics are instrumented by the second lags of the respective (non-crisis and
crisis) interactions with the bank characteristics. The GDP variable, the monetary
policy indicator and inflation are assumed to be truly exogenous and are
instrumented by themselves. The p-value of the Sargan test is 0.35, which hints at
the validity of the overidentifying assumptions.
Estimation results
The results of the standard specification are summarized in table 8.3 (see figure 8.2
above for an overview of hypotheses tested using the euro area sample) and can be
regarded as the benchmark against which the results of the other estimations are to
be judged. It comprises all basic bank characteristics. As stated above, the impact of
those bank characteristics that are not commonly used in adjacent literature such
as measures of the capital surplus or of a deposit overhang is covered in the
subsequent sections in connection with estimations that are especially geared to the
various hypotheses.
Looking at the growth rate for the lagged value of the natural logarithm of loans, one
can see not only that it is significantly different to zero, but also what is more
interesting that it is significantly below one. If the latter had not been the case, this
would have seriously compromised the validity of the estimations: A value of greater
than one would have implied a self-accelerating, non-stationary process regarding
the growth rate of loans. A value of one or at least close to one would have pointed to
the weakness of the use of lagged variables as instruments.121
The coefficients for capital outside and during a crisis are both statistically significant
at the one per cent level. The crisis coefficient for capital is a little higher than the
non-crisis one. However, the difference is of little economic significance, as the point
estimates differ only slightly. The conjecture that capital might be more relevant to
bank lending during the crisis (hypothesis H3) is only weakly supported on the basis
of this result.
121
See Roodman (2009a) and the literature cited therein for a mathematical perspective on the issue
of non-stationarity.
The estimated parameter for size outside the crisis is a little lower than the crisis
parameter. In addition, the crisis parameter is statistically significant at the one per
cent level, while the non-crisis parameter is not. This can be interpreted as support
for the notion that bigger banks loan portfolios are more robust in turbulent times
(hypothesis H6).
Expected Actual
Variable relationship relationship Coefficient Std.error Prob.
LOG(LOANS(1)) + + *** 0.59 (0.08) 0.000
CAP(1)*NC + + *** 0.46 (0.15) 0.002
CAP(1)*CRISIS + + *** 0.48 (0.17) 0.004
SIZE(1)*NC + + 0.14 (0.09) 0.114
SIZE(1)*CRISIS + + *** 0.23 (0.09) 0.007
DEP(1)*NC + + 0.10 (0.08) 0.223
DEP(1)*CRISIS + + 0.14 (0.09) 0.127
LIQ(1)*NC + + 0.11 (0.36) 0.769
LIQ(1)*CRISIS + + 0.37 (0.53) 0.487
STF(1)*NC + + *** 0.45 (0.12) 0.000
STF(1)*CRISIS <STF*NC(+/) + 0.18 (0.26) 0.498
NII(1)*NC + +* 0.21 (0.12) 0.083
NII(1)*CRISIS <NII*NC(+/) +* 0.08 (0.04) 0.068
LOG(GDP) + + *** 0.21 (0.03) 0.000
EURIBOR ** 0.00 (0.00) 0.041
GDP_DEFLATOR +/ 0.00 (0.00) 0.637
CRISIS 0.00 (0.00) 0.857
Regressionproperties
Sample(adjusted): 20012011 Instrumentrank 28
Periodsincluded: 11 Sargantest(pvalue): 0.35
Crosssectionsincluded: 2,153
Totalpanel(unbalanced)observations: 14,061 Yearfixedeffects no
Theestimatedmodelisgivenbyequation(16).CRISIS isadummyvariablethatis1intheyears
2008and2009and0inallotheryears.NC isthenoncrisisdummy.Ittakesonthevalueofzeroin
2008and2009and1inallotheryearsofthesampleperiod.Themodelallowsforfixedeffects
acrossbanksrealizedbyorthogonaldeviations.ItisestimatedusingdifferenceGMM,2step
estimator,Whiteperiodrobuststandarderrors.***,**and*indicatestatisticalsignificanceon
the1%,5%and10%levelrespectively.
Table 8.3: Results of standard specification
An above-average volume of deposits seems to have a positive impact on lending

both during (hypothesis H8) and outside a crisis, although neither coefficient is
statistically significant. However, the standard error for the crisis coefficients is only
slightly above the ten per cent significance level; and, as the robustness checks
show, it is significant at the ten or even five per cent level most of the time.
The picture regarding holdings of liquid assets is not very conclusive. In standard
literature on the bank lending channel, the impact of liquid asset holdings is usually
motivated by buffer-stock considerations in connection with changes in the monetary
policy rate.122 However, even if one disregards the monetary policy indicator, liquid
assets may be beneficial in a crisis when funding is difficult to obtain: In this case, the
relatively easy reduction of liquid assets reduces the amount of funding needed.
However, this does not seem to be the case. On the basis of this picture, it is hard to
draw any further conclusions.
With respect to the volume of short-term funding, higher-than-average values is

beneficial at normal times. This points to the positive effect of good access to market
funding, such that a bank can take lending opportunities because it can resort to a
broad funding base. In times of crisis, the effect is clearly moderated: As expected,
the coefficient on the crisis term of STF is much smaller than during normal periods.
This supports hypothesis H4, according to which a large share of market funding is
disadvantageous because it is usually accompanied by a more pronounced maturity
mismatch between the asset side and the liabilities side of the balance sheet. In a
crisis, the drying-up of funding markets makes it difficult to roll over the volume of
debt needed to fund loans and other assets. As a consequence, this leads to a
reduced loan growth.
A comparable picture, albeit differently motivated, is obtained from the share of non-
interest-income. At normal times, banks with a higher-than-average share of non-
interest income are usually more profitable, thereby exhibiting a higher chance of
being able to repay debt to investors. However, this positive impact is dampened in
times of crisis. This supports hypothesis H5, which implies that the stronger volatility
of the business from which non-interest income is generated could be perceived as a
sign of heightened risk in turbulent times, thereby negatively affecting both funding
conditions and also lending growth in times of crisis.
The control variables in the baseline regression are inconspicuous insofar as they
show the expected sign: GDP is positive and highly significant, and the coefficient for
Euribor is negative and small in absolute terms but significant at the five per cent
122
The interaction of bank characteristics with the monetary policy indicator is not in the focus of this
study.
level. The crisis dummy is negative, as expected, but like the GDP deflator is not
significant.
8.1.2.2. Robustness checks
The results of the standard specification follow several robustness checks. They
comprise the use of different monetary policy indicators, a different approach to
capture demand effects, an alternative way to model the capitalization of a bank and
the use of time fixed effect.
Expected Actual
LOG(LOANS(1)) + + *** 0.61 (0.13) 0.000
CAP(1)*NC + + *** 0.47 (0.14) 0.001
CAP(1)*CRISIS + + *** 0.50 (0.16) 0.002
SIZE(1)*NC + + 0.28 (0.19) 0.147
SIZE(1)*CRISIS + + *** 0.27 (0.10) 0.007
DEP(1)*NC + + 0.08 (0.05) 0.116
DEP(1)*CRISIS + +* 0.08 (0.05) 0.091
LIQ(1)*NC + + 0.13 (0.36) 0.709
LIQ(1)*CRISIS + + 0.33 (0.52) 0.524
STF(1)*NC + +* 0.24 (0.15) 0.095
STF(1)*CRISIS <STF*NC(+/) + 0.18 (0.16) 0.276
NII(1)*NC + + ** 0.14 (0.06) 0.031
NII(1)*CRISIS <NII*NC(+/) + ** 0.07 (0.03) 0.030
LOG(GDP) + +* 0.20 (0.11) 0.067
EURIBOR * 0.00 (0.00) 0.087
GDP_DEFLATOR +/ *** 0.01 (0.00) 0.000
(NSM) + +* 0.14 (0.07) 0.057
CRISIS 0.00 (0.01) 0.735
Table 8.4: Integration of non-standard monetary policy measures

Monetary policy indicators
A first series of robustness checks involves different ways to model the impact of
monetary policy on loans and to see what influence this has on the results. In the first
robustness check, the NSM (non-standard policy measures) variable is used. The
NSM variable is constructed by taking the size of the ECB's balance sheet over total
euro area GDP. It reflects the effect of the ECB's liquidity injections on the lending
behavior of banks. It enters the estimated equation in first differences.
The results are given in table 8.4. The NSM variable is positive and significant at the
ten per cent level. While one would generally expect this behavior, it is nonetheless,
worthy of comment: It seems as if the ECB's full allotment policy, which has been
primarily responsible for the growth of its assets, has had an impact on bank lending
over above its alleviating effect on monetary interest rates. 123 This points to the
effectiveness of this measure for the functioning of the loan supply.
Compared with the baseline regression, the overall picture of the bank characteristics
does not change much. On a detailed level, this estimation produces significant
results (at the ten per cent level) for the coefficient for deposits during the crisis and
positive but not significant results for the coefficient outside the crisis. This supports
the importance of a solid deposit base as a main element of the funding mix for the
size of the loan portfolio. Overall, hypothesis H8, according to which the relevance of
a large proportion of deposit funding should be more beneficial during the crisis than
at normal times, is supported.
With regard to the impact of the share of non-interest income, significant results at
the five per cent level are obtained for both NII coefficients. (In the baseline
specification they are significant on the ten per cent level.) This is consistent with the
results from the standard specification and with hypothesis H5.
Results are also consistent with the standard specification regarding the coefficients
on STF which lends additional support in favor of hypothesis H4.
Two further robustness checks concern the representation of the monetary policy
indicator. In the first of the two approaches, the Euribor variable is replaced by the
Eonia variable. From a theoretical point of view, the differences should not be large in
123
The alleviating impact of non-standard measures on the monetary policy rates is already reflected
by the monetary policy variable itself.
light of the high degree of co-movement of the two (see figure 7.3). According to the
second approach, the Euribor is replaced by the 3-month Euribor-OIS spread as
motivated in section 7.4.3.2. The results are shown in table 8.5 and table 8.6
respectively.
Expected Actual
LOG(LOANS(1)) + + *** 0.59 (0.12) 0.000
CAP(1)*NC + + *** 0.46 (0.15) 0.002
CAP(1)*CRISIS + + *** 0.48 (0.17) 0.004
SIZE(1)*NC + + 0.24 (0.20) 0.231
SIZE(1)*CRISIS + + ** 0.25 (0.10) 0.012
DEP(1)*NC + +* 0.09 (0.05) 0.062
DEP(1)*CRISIS + + ** 0.10 (0.05) 0.047
LIQ(1)*NC + + 0.11 (0.36) 0.769
LIQ(1)*CRISIS + + 0.37 (0.53) 0.487
STF(1)*NC + + ** 0.32 (0.15) 0.034
STF(1)*CRISIS <STF*NC(+/) + 0.19 (0.14) 0.197
NII(1)*NC + +* 0.11 (0.06) 0.077
NII(1)*CRISIS <NII*NC(+/) + ** 0.06 (0.02) 0.020
LOG(GDP) + + ** 0.22 (0.11) 0.044
EONIA ** 0.00 (0.00) 0.011
GDP_DEFLATOR +/ *** 0.01 (0.00) 0.000
(NSM) + + 0.14 (0.10) 0.165
CRISIS 0.00 (0.01) 0.857
Table 8.5: Eonia as a monetary policy indicator
As can be seen, the general results are relatively robust regarding the way changes
in the monetary policy stance are modeled. There are no major coefficient shifts
among the bank characteristics. In both cases, the coefficients for capitalization,
deposits, and non-interest income are significant, as is the non-crisis coefficient for
short-term funding.
Expected Actual
LOG(LOANS(1)) + + *** 0.59 (0.12) 0.000
CAP(1)*NC + + *** 0.49 (0.15) 0.001
CAP(1)*CRISIS + + *** 0.52 (0.17) 0.002
SIZE(1)*NC + + 0.15 (0.14) 0.254
SIZE(1)*CRISIS + + 0.16 (0.10) 0.111
DEP(1)*NC + +* 0.09 (0.05) 0.063
DEP(1)*CRISIS + + ** 0.10 (0.05) 0.047
LIQ(1)*NC + + 0.15 (0.34) 0.658
LIQ(1)*CRISIS + + 0.44 (0.51) 0.391
STF(1)*NC + + ** 0.27 (0.11) 0.016
STF(1)*CRISIS <STF*NC(+/) + 0.18 (0.17) 0.304
NII(1)*NC + +* 0.15 (0.08) 0.050
NII(1)*CRISIS <NII*NC(+/) +* 0.05 (0.02) 0.054
LOG(GDP) + + 0.08 (0.09) 0.395
EURIBOR_OIS ** 0.01 (0.00) 0.045
GDP_DEFLATOR +/ *** 0.01 (0.00) 0.000
(NSM) + + *** 0.30 (0.09) 0.001
CRISIS 0.01 (0.00) 0.300
Table 8.6: The 3-month Euribor-OIS spread as a monetary policy indicator
One difference concerns the observation that the crisis coefficient for size is a little
lower and insignificant when using the Euribor-OIS spread. It is hard to think of an
explanation why there should be a connection. Since this is the only occurrence, not
too much importance has been attached to this phenomenon.
Alternative approach to capture loan demand factors
The next robustness check is geared to the way in which loan demand is represented.
As outlined in section 7.4.3.1, the results of the euro area bank lending survey are
used as a proxy for loan demand and enter the regression in first differences. The
results are shown in table 8.7.
One important thing to note is that, compared to the standard specification, one loses
one third of all observations. This is due to the reduction in the number of periods
included, as a consequence of the fact that the bank lending survey was first carried
out in 2003. A further observation concerns the validity of the overidentifying
restrictions: The p-value of the Sargan test is only 0.23, which is not completely
satisfying (see Roodman (2009a)).
Expected Actual
LOG(LOANS(1)) + + *** 0.63 (0.16) 0.000
CAP(1)*NC + + ** 0.72 (0.31) 0.020
CAP(1)*CRISIS + + ** 0.70 (0.31) 0.023
SIZE(1)*NC + + 0.28 (0.30) 0.344
SIZE(1)*CRISIS + + ** 0.22 (0.11) 0.042
DEP(1)*NC + + ** 0.19 (0.08) 0.024
DEP(1)*CRISIS + + ** 0.20 (0.08) 0.012
LIQ(1)*NC + + 0.08 (0.36) 0.814
LIQ(1)*CRISIS + + 0.36 (0.50) 0.472
STF(1)*NC + + ** 0.21 (0.09) 0.019
STF(1)*CRISIS <STF*NC(+/) +* 0.18 (0.10) 0.080
NII(1)*NC + + 0.10 (0.07) 0.165
NII(1)*CRISIS <NII*NC(+/) +* 0.03 (0.02) 0.067
EURIBOR 0.00 (0.00) 0.396
GDP_DEFLATOR +/ ** 0.03 (0.01) 0.035
(LEND_SURV) + * 0.00 (0.00) 0.057
(NSM) + + ** 0.39 (0.15) 0.011
CRISIS *** 0.01 (0.00) 0.001
Table 8.7: Loan demand proxied by results of the ECB bank lending survey
Despite the theoretical soundness of the idea of using the data from the bank lending
survey, the practical behavior of the variable hints at certain problems. While there is
little change regarding the point estimates and the standard errors for the bank
characteristics, the estimated parameter for the lending survey variable is negative
contrary to expectations and is significant at the ten per cent level. Moreover, the
absolute value is very small and much smaller than the value of the coefficient for
GDP. This general picture remains even when different specifications are used.124
Why is this the case? One obvious explanation could be due to the loss of
observations. Given that the loss is mainly due to the shorter sample period and not
due to a loss in cross-sections and, moreover, that there is little change in the
estimates of the bank characteristics, it is rather questionable whether a longer time
series on the bank lending survey would have produced different results.
Another possible explanation may be that the 90 banks participating in the survey is
not a large enough number to draw a representative picture of loan demand in the
entire euro area. It could also be the case that the loan officers' answers tend to
exhibit procyclical behavior. In "good" times, for example, slightly favorable
developments may already be overstated as a sign of an upcoming boom. This could
lead to overly optimistic answers with which actual demand is unable to keep up.
Furthermore, the fact that the composition of the euro area has changed over the
years might also induce a bias (of unknown direction), because the countries from
which the surveyed loan officers originate are not exactly the same countries that are
part of the sample.
A further reason might emerge from the following insight: The lending survey is
carried out on a quarterly basis and the answers distinguish between the different
customer groups that demand loans (i.e. enterprise loans, consumer credit and
housing loans). The construction of a single lending survey variable thus requires
some form of data compilation to make it applicable and to match it to the annual
structure of the data. It is possible that the optimal compilation procedure has not yet
been found, although, as stated above, different emphases in the loan demand
groups have been tested, each with unsatisfactory results.
124
Unreported analyses in which both variables (GDP and lending survey) are included do not
produce more convincing results. The same applies to the use of the lending survey variable in
levels (unit root tests lead to a rejection of the hypothesis that a unit root is present in the levels of
this variable) instead of first differences or the use of lags.
Whatever explanation there may be, the bottom line is that, although the use of the
lending survey variable does not seem to improve the results, the coefficients for the
bank characteristics remain fairly stable which is not a bad result at all.125
Another robustness check related to the control for demand factors takes a different
approach, allowing for period fixed effects. Technically speaking, a dummy variable is
introduced for each year in the sample period. In case one is not convinced to
succeed in capturing all relevant factors that affect the demand for loans by
observable proxies such as by GDP and the GDP deflator, using time fixed effects is
a way to address this issue. In doing so, the estimated coefficients should be a
representation of the full cross-sectional impact of the bank characteristics on loans
(see Peydro (2010)).
The estimation with time fixed effects produces the results shown in table 8.8. One
initial observation concerns the macro control variables and the variables that
capture monetary policy actions (Euribor and NSM): None of these are significantly
different from zero (at the 30 per cent level). This is not surprising, since these
variables do not change by cross-sections but over time, and since the variation over
the years is to a large extent reflected in the dummy variables for the different years.
With regard to the instrumentation of the variables in this estimation, it can be noted
125
A further (unreported) robustness check concers the use of the harmonized index of consumer
prices (HICP) as an alternative measure of inflation. The result is that the GDP deflator performs
consistently better than the HICP. From a theoretical perspective, a possible explanation is that
the GDP deflator is the more convincing concept compared to the HICP when trying to capture the
effects of price increases on the volume of loans. While the HICP is explicitly geared to consumer
prices, it does not capture increases in asset prices (asset price inflation). For example, a rise in
real estate prices may lead to increases in loan volumes on banks balance sheets that are not
reflected in the HICP as long as real estate prices do not feed through into consumer prices via the
various possible channels. This feed-through into consumer prices could, for example, work via
rents. In the case of rents for residential real estate, the HICP is directly affected because housing
rents are part of the basket based on which the HICP is calculated. In the case of commercial
properties, rent increases might affect sales prices, which are then passed on until they enter
consumer prices. Either way, given that prices are sticky, e.g. due to menu costs, it is most
realistic to assume that it takes some time before (at least parts of the) asset price changes have
fed-through into consumer prices.
that the p-value of 0.58 in the Sargan test suggests that the instruments behave well
in this case.
Expected Actual
LOG(LOANS(1)) + + *** 0.68 (0.11) 0.000
CAP(1)*NC + + *** 0.38 (0.14) 0.006
CAP(1)*CRISIS + + ** 0.40 (0.16) 0.014
SIZE(1)*NC + + 0.11 (0.07) 0.134
SIZE(1)*CRISIS + +* 0.14 (0.08) 0.080
DEP(1)*NC + + 0.07 (0.05) 0.145
DEP(1)*CRISIS + +* 0.08 (0.05) 0.094
LIQ(1)*NC + 0.00 (0.43) 0.995
LIQ(1)*CRISIS + + 0.22 (0.58) 0.705
STF(1)*NC + + *** 0.32 (0.12) 0.008
STF(1)*CRISIS <STF*NC(+/) + 0.15 (0.15) 0.314
NII(1)*NC + +* 0.12 (0.06) 0.069
NII(1)*CRISIS <NII*NC(+/) + ** 0.06 (0.03) 0.042
LOG(GDP) + + 0.42 (0.42) 0.320
EURIBOR + 0.00 (0.00) 0.445
GDP_DEFLATOR +/ + 0.00 (0.01) 0.974
(NSM) + + 0.15 (0.32) 0.645
Totalpanel(unbalanced)observations: 14,061 Yearfixedeffects yes
Table 8.8: Estimation with time fixed effects
Looking at the point estimates of the bank characteristics, it is apparent that most of
them are slightly lower than the standard specifications or indeed many of the other
specifications. This can be interpreted as supporting the assumption that some of the
cross-sectional variation in the supply of loans might be due to demand factors that
are not completely captured by the variables for GDP and prices. However, since the
difference in the point estimates is, as already stated, relatively small, this
observation does not raise serious concerns about the general validity and accuracy
of the other estimates. Moreover, the estimated signs of all bank characteristics
remain unchanged (with the exception of liquidity) and the estimated standard errors
are also comparable to the standard specification, for example. The latter
observations imply that the significance of the estimates does not change to a large
extent either.
Expected Actual
LOG(LOANS(1)) + + *** 0.59 (0.12) 0.000
CAP_TCE_TCA(1)*NC + + *** 0.46 (0.15) 0.002
CAP_TCE_TCA(1)*CRISIS + + *** 0.48 (0.17) 0.004
SIZE(1)*NC + + 0.16 (0.10) 0.102
SIZE(1)*CRISIS + + *** 0.23 (0.08) 0.006
DEP(1)*NC + +* 0.09 (0.05) 0.065
DEP(1)*CRISIS + + ** 0.09 (0.05) 0.049
LIQ(1)*NC + + 0.10 (0.36) 0.786
LIQ(1)*CRISIS + + 0.36 (0.53) 0.497
STF(1)*NC + + *** 0.37 (0.13) 0.004
STF(1)*CRISIS <STF*NC(+/) + 0.15 (0.24) 0.531
NII(1)*NC + +* 0.18 (0.10) 0.071
NII(1)*CRISIS <NII*NC(+/) + ** 0.08 (0.04) 0.044
LOG(GDP) + + ** 0.22 (0.11) 0.043
EONIA ** 0.00 (0.00) 0.010
GDP_DEFLATOR +/ *** 0.01 (0.00) 0.000
(NSM) + + 0.14 (0.10) 0.165
CRISIS 0.00 (0.01) 0.858
Table 8.9: Capitalization measured in terms of tangible common equity over tangible common assets
Taking all the above into account, the conclusion is that the robustness of the results
shows that controlling for loan demand effects by means of variables for GDP and
prices is successful. Concerns about a (supposedly) insufficient account of the
demand side are unfounded.
Capitalization measures
To check the robustness of the results with respect to the choice of a different
capitalization variable, the ratio of total equity capital to total assets is replaced by the
ratio of tangible common equity to tangible common assets (see table 8.9 for the
results).
As expected on the basis of the strong correlation between the two measures, the
results are very robust with respect to the choice of capitalization variable. The
estimates of the capitalization variables in particular differ only as of the second
decimal place. Regarding the coefficients for some other variables (e.g. the deposit
variables), the equation with tangible common equity shows a higher level of
significance.
Another possible robustness check regarding different measures of capitalization

concerns the use of the Tier 1 ratio, which takes on a regulatory perspective on
capital. 126 Unfortunately, data availability is poor, which is reflected by the loss of
almost 90 per cent of all observations. In part, this is due to the fact that the concept
of the Tier 1 ratio only recently gained widespread attention. Longer time series are
therefore unavailable. The consequence for the estimation is poor behavior by the
coefficients and results that are unsatisfactory.127
The overall conclusion regarding the use of different capitalization measures is that
results are quite robust to these kind of variations.
Securities that must be marked to market
According to hypothesis H7, situations in which a large share of securities have to be

marked to market (over total assets) should have a dampening impact on a bank's
propensity to lend during a crisis. This is tested by incorporating the corresponding
crisis and non-crisis variable (AFS) in the equation to be estimated. Beyond this, the
126
See Basel Committee on Banking Supervision (1998) for details of the instruments eligible for
inclusion in the Tier 1 capital ratio.
127
Another robustness check is carried out to further strengthen our understanding of capital: In order
to control for probable cyclical patterns in capital endowment, the idea is to interact GDP with
capital. The (unreported) result is that the coefficient for the interaction term of GDP with capital is
negative but clearly insignificant. The coefficients for the capital variables do change marginally.
This picture does not change if one introduces the distinction between the interaction of GDP and
capital during and outside of a crisis.
same bank characteristics enter the regression as in the standard specification. The
monetary policy indicator is the Eonia and the macro controls are GDP (first
difference of logs) and deflation. Instruments are chosen in the same way as in the
standard specification; only the appropriate instruments for AFS are added.
Expected Actual
LOG(LOANS(1)) + + *** 0.79 (0.23) 0.001
CAP(1)*NC + +* 0.85 (0.47) 0.070
CAP(1)*CRISIS + + ** 0.92 (0.41) 0.027
SIZE(1)*NC + + 0.30 (0.28) 0.279
SIZE(1)*CRISIS + + 0.25 (0.31) 0.424
DEP(1)*NC + + 0.18 (0.17) 0.307
DEP(1)*CRISIS + * 0.13 (0.08) 0.099
LIQ(1)*NC + + 1.75 (1.32) 0.186
LIQ(1)*CRISIS + + 2.29 (1.66) 0.167
STF(1)*NC + +* 0.34 (0.20) 0.089
STF(1)*CRISIS <STF*NC(+/) + 0.13 (0.19) 0.486
NII(1)*NC + + 0.06 (0.49) 0.897
NII(1)*CRISIS <NII*NC(+/) + 0.02 (0.07) 0.735
AFS(1)*NC +/ +* 0.22 (0.12) 0.075
AFS(1)*CRISIS <AFS*NC(+/) + 0.19 (0.15) 0.209
LOG(GDP) + + ** 0.57 (0.28) 0.041
EONIA *** 0.01 (0.00) 0.008
GDP_DEFLATOR +/ 0.01 (0.01) 0.238
CRISIS * 0.01 (0.00) 0.092
Crosssectionsincluded: 806
Table 8.10: Estimation including the share of mark-to-market securities
Unfortunately, the main item needed to construct this variable, "available-for-sale

securities", is missing in many periods for many banks. Compared to the standard
specification, 78% per cent of total panel observations are lost, resulting in the
inclusion of only 806 cross-sections. With a p-value of 0.17, the Sargan test is only
just satisfactory.
The main variables of interest in this equation are the two AFS variables. The result
for the non-crisis is that the coefficient is positive and significant at the five per cent
level (see table 8.10).
The crisis coefficient for AFS is positive and lower than for the non-crisis coefficient.
However, it is not significant at conventional significance levels. This results is too
weak to be able to conclude the validity of hypothesis H7. It is likely that the massive
loss of observations leads to flawed estimations. This impression is confirmed by
estimates for the other coefficients: Although they do not differ too greatly from the
previous estimations (standard specification and robustness checks), they appear to
be less reliable, as can be seen by higher standard error values, for example.
As a result, hypothesis H7, which states that a large proportion of securities that have
to be marked to market has a negative impact on bank lending, cannot be confirmed.
Nevertheless this result should be handled with caution, given the possibility that
these results could be caused by a lack of observations and other factors.
8.1.2.3. Summary of main results and relationship to existing literature
What are the main results from the regressions conducted so far, and how do these
results relate to the evidence provided in existing literature?
Starting with the bank characteristics, one important outcome is that "pure" capital is
important both at normal times and during a crisis. The magnitude of both coefficients
is always positive, significant and higher than for the other characteristics,
underscoring its special role. Moreover, it is interesting to note that, in most
specifications, there is a small difference in absolute terms, suggesting that capital is
a little more important during a crisis than it is at normal times, in accordance with
hypothesis H3.
Regarding comparison with the available literature, it can be noted that the positive
impact of capital in a crisis in particular is also supported by Gambacorta and
Marques-Ibanez (2011), while the evidence concerning the non-crisis effect depends
on the specification.
A further robust outcome of the estimations carried out is that the size of a bank has
a positive and significant impact in most cases during times of crisis and a positive
but insignificant impact at normal times. This result is consistent with existing
literature: While the generally positive impact of size in the US has been confirmed
by e.g. Kashyap and Stein (1995) euro area results obtained for the time before the
crisis by, among others, Ehrmann et al. (2003), Ehrmann and Worms (2004) and
Worms (2003) have not found a significant influence. However, for the period of the
recent crisis, Gambacorta and Marques-Ibanez (2011) also detected a positive and
significant effect of size on lending thus size seems to be relevant in a crisis because
it can reduce informational asymmetries. It must be borne in mind, however, that their
sample also includes banks in the US and the UK. Moreover, a connection between
size and lending can be established if one considers that the bigger the institute, the
more diversified its client base and its income sources are likely to be with respect to
industries, geography, client size, etc. Consequently, investors should "trust" bigger
banks to a higher degree regarding their ability to meet their repayment obligations,
especially when this ability is called in question during a crisis. In addition, a bigger
size might also imply a higher degree of professionalism regarding reporting
structures, risk management, etc. Taken together, the evidence provided lends
support to hypothesis H6.
In the case of deposits, one focus is on the effect of a deposit overhang on lending.
This relationship is discussed in detail in section 8.1.4. However, the result regarding
the impact of the pure volume of deposits is also interesting, since a larger share of
deposit funding is, according to hypothesis H8, assumed to influence lending
positively during a crisis. This hypothesis can be confirmed on the basis of the
empirical results. The estimated crisis coefficient is positive and, in most of the
specifications, significant at the ten or five per cent level. This supports the view that
banks with a strong base of insured deposits are less prone to being subject to credit
constraints by having too little funding available when the financial markets are in
turmoil.
In the literature, the effect of a broad (insured) deposit funding base has not yet been
debated intensively.128 However, this picture is shared by Ivashina and Scharfstein
(2010) for the US and Gambacorta and Marques-Ibanez (2011) for a mixed sample
consisting of banks from the euro area, the US and UK.
With respect to the share of short-term funding, the following pattern can be
observed: The crisis coefficient is positive and lower than the non-crisis coefficient
128
One reason for this might be the negative correlation between deposit funding and
alternative/market funding.
but generally not significant, while the non-crisis estimate is positive and significant in
most cases. This can be viewed as supporting hypothesis H4, according to which it
becomes harder for banks to obtain funding during a crisis. This effect seems to be
more pronounced for banks that rely on market funding to a large extent.
The effect of the funding structure and the role of the ease of access to market
funding has attracted particular attention against the background of the recent crisis.
The results in this study are consistent with the findings of Gambacorta and
Marques-Ibanez (2011) as far as the negative impact during the crisis is concerned.
However, the results contrast with those of Brei et al. (2013), who do not observe
clear results regarding the direction of the impact of short-term funding during the
crisis and who observe negative coefficients outside of the crisis. It should be noted
that their approach is not entirely comparable insofar as they distinguish between
banks that have been subject to some sort of rescue measure and those they have
not, and that they employ a different estimation methodology in some areas.129
The results of this study are also consistent with hypothesis H5: A higher-than-
average proportion of non-interest income seems to be positive for the supply of
bank loans at normal times, while the impact is much less positive in a crisis.
According to theory, this relative negative influence is due to the higher volatility of
the business that generates non-interest income, while interest income is a more
stable source. This increases the risk premium on funding and results in a smaller
loan portfolio. The estimated coefficients are significant in many of the specifications.
So far, the impact of the proportion of non-interest income has not been regarded at
all in the relevant literature although this concept is sound from a theoretical point of
view as it fits well in the framework presented in section 4.2.
The only bank characteristic that does not behave as expected is liquidity. The
estimated coefficients are usually positive but associated with a high standard error,
which makes them unreliable. Even in interaction with monetary policy indicators, the
results are not much clearer. However, the positive sign on liquidity is generally
consistent with Ehrmann et al. (2003), Gambacorta (2005) and Gambacorta and
Marques-Ibanez (2011), for example.
129
Alongside the GMM estimator, they also use OLS. Whether the use of OLS is justified in this
context is debatable, as discussed in section 7.5.
The main results are robust to changes in the specification regarding the
implementation of monetary policy measures, the measurement of loan demand
factors, the use of different capitalization measures and other specifications.
Taken together, the hypotheses that have been developed from the respective
appropriate theoretical context are confirmed in large part by the baseline analysis.
The next two sections are devoted to two questions of particular interest to this study:
The first examines the effect of a capital surplus or shortfall relative to a self-imposed
target, and the second concerns the role a deposit overhang plays in the propensity
to lend.
8.1.3. Capital surplus
An innovative approach to the character and role of capital in the context of bank
lending and one of the most interesting questions in the course of this research is
the one regarding the impact of a capital surplus or shortfall relative to a self-chosen
target. While a shortfall implies a (binding) capital constraint, a surplus should give a
bank the opportunity to extend its loan supply (hypothesis H1). To shed light on this
issue, three different specifications are tested. In the first one, the "plain" capital
measure is replaced by the capital surplus variable, which is constructed as
explained in section 7.4.2. In the second one, robustness is checked by the use of
both "pure" capital and the capital surplus variable. In the third specification, time
fixed effects emphasize the true cross-sectional effects.
Specification
Hypothesis H1 that a shortfall of capital relative to a target leads to a reduction in

the loan supply is first tested by means of a standard specification. In this
specification, the capital surplus variable, interacted once with the non-crisis dummy
and once with the crisis dummy, is used. Furthermore, the bank characteristics size,
deposits, liquidity, short-term funding and non-interest income are part of both
regressions. All bank characteristics are lagged once to remedy the problem of
endogeneity (in connection with the instrumental variable procedure). GDP, Euribor,
the GDP deflator, the non-standard monetary policy measures variable (NSM) and
the crisis dummy act as control variables.
Validity
Due to the lagged regressors and the choice of instruments, the sample period is
adjusted and includes the years from 2001 to 2011. The incorporation of 2,150 cross-
sections leads to a total of 14,050 panel observations (unbalanced). To instrument
the endogenous explanatory variable, dynamic instruments along the lines of
Arellano and Bond (1991) are used, i.e. the second and third lags of the first-
differenced natural logarithm of loans. All other bank characteristics, or rather their
interactions with the non-crisis and crisis dummies that are used in the regressions,
are instrumented by the second lags of the relevant (non-crisis and crisis)
interactions with the bank characteristics. The variables measuring GDP, monetary
policy and inflation are assumed to be truly exogenous and are instrumented by
themselves. As a result, the instrument rank is 28. Sargan's J-statistics yield a value
that translates into a probability of almost 0.3 (0.29), which exceeds conventional
significance levels. This is a sign of the validity of the overidentifying restrictions and
hints at a proper specification.
Estimation results
Table 8.11 shows the results for the standard estimation containing capital surplus as
the only capitalization measure. The coefficients for the non-crisis and the crisis
variable are both statistically significant at the 10 per cent level. The point estimate is
0.17 for the non-crisis surplus variable and almost twice as high (0.30) for the crisis
version. It should be noted, however, that the standard error for the crisis coefficient
is also almost twice as large, which calls into question whether the difference
between the two is economically significant. Either way, the results support the view
that a capital surplus or, equivalently, the deviation from an individual capital target
does indeed play a role in explaining the lending response of banks.
What can be said about the other variables included in the regressions under
discussion? The parameters for size are significant during times of crisis at the ten
per cent level, and the point estimates are in a similar range to the specifications in
the baseline analysis. At normal times, however, the coefficient for size is positive but
not statistically significant.
The estimates for deposits are positive but not significantly different to zero.
An observation that is similar to the baseline analysis is the one that the coefficients
for liquidity are far removed from conventional significance levels. The only pattern
that seems to persist across the different specifications is that the non-crisis
estimates of liquidity are positive, while the crisis coefficients are all negative.
However, whether this pattern is real or has to be ascribed to pure chance is not
clear.
Expected Actual
LOG(LOANS(1)) + + *** 0.46 (0.09) 0.000
CAPSUR(1)*NC + +* 0.17 (0.09) 0.061
CAPSUR(1)*CRISIS + +* 0.30 (0.17) 0.077
SIZE(1)*NC + + 0.09 (0.07) 0.153
SIZE(1)*CRISIS + +* 0.22 (0.11) 0.051
DEP(1)*NC + + 0.05 (0.07) 0.509
DEP(1)*CRISIS + + 0.06 (0.08) 0.409
LIQ(1)*NC + + 0.17 (0.46) 0.708
LIQ(1)*CRISIS + 0.43 (0.56) 0.446
STF(1)*NC + + *** 0.31 (0.11) 0.004
STF(1)*CRISIS <STF*NC(+/) + 0.13 (0.08) 0.127
NII(1)*NC + + *** 0.25 (0.08) 0.002
NII(1)*CRISIS <NII*NC(+/) + *** 0.07 (0.03) 0.003
LOG(GDP) + + *** 0.25 (0.07) 0.000
EURIBOR ** 0.00 (0.00) 0.040
GDP_DEFLATOR +/ 0.00 (0.00) 0.316
(NSM) + + *** 0.26 (0.10) 0.009
CRISIS 0.01 (0.01) 0.294
Table 8.11: Results of estimations including capital surplus only
The coefficient for short-term funding is positive, as expected, and highly significant
outside a crisis (p-values <0.01), which lends further support to the view that the
ability to access alternative forms of funding has a positive impact on the volume of
the loan portfolio. In the recent crisis, though, the relationship vanishes, indicated by
estimates that are not significant. The sign is less positive than the one the crisis
parameter, however, being broadly in line with the baseline analysis. Handled with
caution, this could be an additional indication that reliance on short-term funding
causes problems when it becomes harder to roll over debt during a crisis which is the
core consideration underlying hypothesis H4.
One interesting result is obtained regarding the share of non-interest income, which
is positive outside of a crisis and negative in relative terms during a crisis. These
results are statistically significant at the one per cent level and consistent with the
baseline analysis. According to the theoretical foundation of hypothesis H5, in a crisis,
the higher volatility of non-interest income leads investors to suspect higher risks and
demand a higher risk premium for uninsured market funding. This, in turn, increases
banks' difficulties in funding their loan portfolio. Why is this pattern reversed
compared to normal times? The evidence supports the view that, at normal times, a
higher share of non-interest income is associated with higher profitability, since no
equity capital has to be provided for these kinds of business (e.g. investment banking
business, derivatives, transaction banking, securities business, etc.). Under these
circumstances, higher profitability thus reduces the risk of default. This relationship is
reversed in a crisis, however, due to a change in the revenue potential of the
underlying business.
The control variables behave as expected, with a positive sign for GDP, a negative
sign for the monetary policy indicator (both significant) and a positive but insignificant
coefficient for deflation. The magnitude of GDP is in the same range as the
coefficient for the baseline analysis, which hints at the robustness of this variable
relative to different specifications. The same holds for Euribor as the monetary policy
indicator. Moreover, as expected, the estimate for NSM is positive and significant.
Hereafter, the robustness of these results, especially regarding the capital surplus
variable, is tested by means of two additional specifications.
"Plain" capital in addition to capital surplus
According to this specification, in addition to the capital surplus variable, "plain"

capital enters the estimated equation (both non-crisis and crisis interacted). As in the
standard specification above, the bank characteristics size, deposits, liquidity, short-
term funding and non-interest income are part of the regression. Further controls are
GDP, Euribor, the GDP deflator, NSM and the crisis dummy.
Expected Actual
LOG(LOANS(1)) + + *** 0.56 (0.08) 0.000
CAP(1)*NC + + *** 0.72 (0.23) 0.001
CAP(1)*CRISIS + + *** 0.81 (0.24) 0.001
CAPSUR(1)*NC + + 0.14 (0.16) 0.400
CAPSUR(1)*CRISIS + +* 0.16 (0.09) 0.087
SIZE(1)*NC + + 0.18 (0.37) 0.618
SIZE(1)*CRISIS + + *** 0.27 (0.08) 0.001
DEP(1)*NC + + 0.13 (0.09) 0.139
DEP(1)*CRISIS + +* 0.16 (0.10) 0.094
LIQ(1)*NC + + 0.12 (0.61) 0.847
LIQ(1)*CRISIS + 0.35 (0.64) 0.590
STF(1)*NC + + *** 0.39 (0.11) 0.000
STF(1)*CRISIS <STF*NC(+/) +* 0.11 (0.06) 0.090
NII(1)*NC + + *** 0.20 (0.05) 0.000
NII(1)*CRISIS <NII*NC(+/) + *** 0.08 (0.02) 0.000
LOG(GDP) + + *** 0.21 (0.03) 0.000
EURIBOR 0.00 (0.00) 0.114
GDP_DEFLATOR +/ + 0.00 (0.00) 0.881
(NSM) + + ** 0.25 (0.10) 0.011
CRISIS 0.01 (0.01) 0.328
Table 8.12: Results of estimations including capital surplus and capital
The choice of instruments is also analogous to the standard specification (apart from
the additional instruments for the "plain" capital variables), yielding a p-value of 0.31
for the Sargan test statistic. The estimation is based on an equal number of
observations compared to the standard specification.
The main result is that the importance of a capital surplus variable for bank lending is
cautiously encouraged by this specification. (All results are displayed in table 8.12.)
The estimated coefficient of 0.14 is very close to that of the standard specification.
Unfortunately, the standard error is relatively high, leading to an insignificant
coefficient (at conventional levels). However, the crisis coefficient is significantly
different from zero at the ten per cent level (p-value: 0.087). The coefficient of 0.16 is
close to the non-crisis coefficient and also to those of the standard specification,
which includes capital surplus only. It should be remembered that constructing the
capital surplus variable involves the estimation procedure outlined in section 7.4.2. It
is therefore reasonable not to expect the same level of accuracy as for
variables/items that can be deducted more or less directly from items on the balance
sheet or income statement. In light of these results, there is reason to conclude that
the estimation of the capital surplus variable was successful or, at least, that the
results do not provide any arguments to the contrary.
As a further result, the coefficients for "plain capital" are both highly significant (at the
one per cent level) and less than one standard deviation away from the values of the
baseline analysis. Taken together, the overall picture is that the departure from a
capital target helps explain the banks' lending response. Furthermore, there is some
evidence to support the view that whether a target is missed during a crisis or at
normal times is of no relevance to the loan supply.
Regarding the other bank characteristics in the estimation, there are no serious
deviations compared to the standard estimation. One minor exception is that the
crisis coefficient for deposits is significant at the 10 per cent level. Apart from this, the
coefficients are in a similar dimension. Moreover, the control variables behave as
expected.
Estimation with time fixed effects
To analyze the true cross-sectional impact of the bank characteristics, the standard
specification is augmented by dummies for the years in the sample period, thereby
capturing time fixed effects. With the exception of the time dummies, the specification
is identical to standard estimation. Accordingly, the choice of instruments too is
analogous, yielding a p-value of 0.57 for the Sargan test with an instrument rank
of 38.
As a result (see table 8.13), the coefficients for the capital surplus variables are both
positive and significantly different to zero at the ten per cent level. In terms of
absolute values, the coefficients are close to the estimates of the two preceding
specifications. This is additional evidence for the fact that the capital endowment
relative to a self-chosen target helps explain the banks' lending responses. Also
consistent with the results obtained previously with regard to capital surplus is the
observation that, although there is a difference between the crisis and the non-crisis
coefficient, the magnitude of the standard errors is such that one cannot clearly
conclude that the difference is economically significant.
Expected Actual
LOG(LOANS(1)) + + *** 0.60 (0.12) 0.000
CAPSUR(1)*NC + +* 0.19 (0.10) 0.064
CAPSUR(1)*CRISIS + +* 0.20 (0.12) 0.097
SIZE(1)*NC + +* 0.09 (0.05) 0.056
SIZE(1)*CRISIS + + ** 0.09 (0.05) 0.050
DEP(1)*NC + +* 0.09 (0.05) 0.094
DEP(1)*CRISIS + + ** 0.10 (0.05) 0.049
LIQ(1)*NC + + 0.05 (0.43) 0.912
LIQ(1)*CRISIS + + 0.28 (0.58) 0.625
STF(1)*NC + + 0.25 (0.21) 0.232
STF(1)*CRISIS <STF*NC(+/) +* 0.07 (0.04) 0.088
NII(1)*NC + + 0.12 (0.08) 0.106
NII(1)*CRISIS <NII*NC(+/) + 0.06 (0.04) 0.149
LOG(GDP) + + 0.85 (0.65) 0.192
EURIBOR + 0.00 (0.00) 0.489
GDP_DEFLATOR +/ + 0.01 (0.01) 0.635
(NSM) + 0.10 (0.35) 0.777
Table 8.13: Capital surplus estimation with time fixed effects

The parameters for the other bank characteristics are broadly in line with the
estimates in the preceding specifications and also with those in the baseline analysis.
Inspecting the control variables it is noticeable that they are not significantly different
from zero and, in the case of NSM and Euribor, have the "wrong" sign. This is not
surprising, however, since much of the variation of the controls is captured by the
time dummies.
8.1.3.3. Summary of main results and relationship to existing literature
What are the main results from the preceding regressions? One robust result is that a
surplus of capital relative to a self-chosen target has a positive impact on bank
lending both during a crisis and outside of a crisis. The difference between the
coefficients is relatively small in all cases. The stability of this result across the
different specifications leads to the conclusion that, ceteris paribus, missing a capital
target has a similar or at least comparable effect on a bank's supply of loans,
irrespective of whether the said target is missed during a crisis or at normal times. It
should be noted that, given a causal interpretation, the positive sign on the capital
surplus variable implies that a shortfall in capital relative to a target leads to a
reduction of the supply of bank loans. This can be attributed to the fact that a shortfall
relative to a target even if the target is "only" self-chosen has the effect of a
capital constraint.
Taken together, this outcome is consistent with hypothesis H1. Moreover, the results
also lend support to the hypotheses H8 (the role of deposits in a crisis), H6 (size
being relevant in a crisis), H3 ("plain" capital of greater importance during a crisis,
indicated by higher absolute values than at normal times), H4 (a large proportion of
short-term funding is negative in relative terms in a crisis) and H5 (a large proportion
of NII is negative in relative terms in a crisis).
How do the results, especially the one for the capital surplus, fit in with the existing
literature? One first strand of literature concerned with bank lending and capital
constraints tackles the subject in the context of the credit crunch in the early 1990s in
the US. Despite the fact the focus of authors such as Bernanke and Lown (1991),
Peek and Rosengren (1995b), Peek and Rosengren (1995c) and Hancock and
Wilcox (1994) is clearly on a capital constraint induced by regulatory considerations,
their general conclusion is consistent with the findings regarding a surplus or shortfall
relative to self-chosen targets in this study. Any constraint, however motivated, has a
negative effect on bank lending.
More recently, on the basis of estimated individual capital ratios, Francis and
Osborne (2009) show a modest but positive effect of a capital surplus on bank
lending in the UK and Berrospide and Edge (2010) do so for the US.
To the knowledge of the author, there is as yet no study that is specially geared to
individual capital targets and their impact on lending for banks in the euro area and
against the background of the recent financial crisis.130 The results presented herein
thus constitute an innovation, linking the target capital estimation technique with the
crisis in the context of bank lending.
8.1.4. Deposit overhang
A further innovation in this study concerns the function of a deposit overhang when
the amount of insured deposits is higher than the amount of loans for the supply of
bank loans. According to hypothesis H9, an overhang of deposits should, in the
recent crisis, generally have supported the lending capacity of banks, since those
banks had less need to resort to the dysfunctional wholesale funding markets than
banks without a deposit overhang.
In addition, hypotheses H2a to H2c state that, if a bank does have a deposit
overhang, the impact of a higher-than-average share of market funding, of non-
interest income and of securities that need to be marked to market should be
significant compared to those banks that do not have a deposit overhang. These
relationships are postulated to be valid irrespective of the economic conditions, i.e.
during a crisis or outside of a crisis. The expected result is that a significant
relationship exists for those banks that do not have a deposit overhang. In fact, given
that one result of the baseline analysis is that STF has a positive impact outside of a
crisis and a negative one in relative terms during a crisis, and given that there are
130
There are, however, studies of the cyclicality of capital regulation, of the countercyclicality of
capital buffers in the context of lending and the crisis (see e.g. Drehmann and Gambacorta (2012)
or Repullo and Suarez (2013)) and on other related issues. These studies usually adopt a
regulatory perspective.
only two crisis years, the overall sign for STF is expected to be positive and closer to
the non-crisis impact. The same holds for NII.
The case of the share of securities that need to be marked to market is difficult. As
already stated in the context of the baseline analysis, the loss in the number of
observations is relatively severe. As a consequence, the tests carried out in this
section involving the AFS variable produce neither stable nor meaningful results.
Hence, unfortunately, hypothesis H2c can neither be validated nor rejected.
To test hypotheses H2a and H2b, the deposit overhang-related variables are of
particular interest. A distinction can be drawn between two ways of constructing
these variables:
As outlined in section 7.4.1, the first way involves constructing the variable
OVERHANG. This is built by subtracting one from the result of dividing customer
deposits by loans. Any value above zero means that a bank has a deposit overhang.
The second way uses two dummy variables: DUMMY_OVERHANG assumes the
value of one whenever the amount of insured deposits exceeds the amount of loans
and is otherwise zero. The counterpart is the variable DUMMY_NO_OVERHANG,
which is one if a bank does not have a deposit overhang and is otherwise zero. Both
dummy variables are interacted with STF and NII to test hypotheses H2a and H2b.131
This procedure allows for a direct comparison of the two different groups banks
with and without a deposit overhang with a focus on the crisis period.
Three different specifications are used. In the first one, the overhang variable is used
alongside the "normal" deposit variable. The overhang dummy variables are used in
the second specification. In the third one, this specification is checked for robustness
by including time fixed effects.
131
Again, due to the orthogonal deviations procedure, this does not induce perfect collinearity and
does not lead to the dummy variable trap. See also section 7.3.2 and footnote 120.
Specification
As already mentioned, in the standard specification the focus of the analysis is on the
OVERHANG variable in order to test hypothesis H9. The variable is interacted with
the non-crisis and the crisis dummy. In addition, the bank characteristics
capitalization, size, liquidity, deposits, short-term funding and non-interest income
enter the estimated equation. All bank characteristics are lagged once to further
reduce the problem of endogeneity (together with the instrumental variable
procedure). GDP, Euribor, the GDP deflator, the non-standard monetary policy
measures variable (NSM) and the crisis dummy serve as control variables.
Validity
As a consequence of the lags used in the regression and the choice of instrumental
variables, the sample period is adjusted to the years 2001 to 2011. A total of 2,153
cross-sections are included, leading to 14,061 (unbalanced) observations.
Due to the use of endogenous explanatory variables, dynamic instruments are

chosen in accordance with Arellano and Bond (1991). This means that the second
and third lags of the natural logarithm of the growth rate of loans act as instruments.
Other bank characteristics and their interactions with the non-crisis and crisis
dummies are instrumented by the second lags of the (non-crisis and crisis)
interaction terms. The variables for GDP, monetary policy and inflation are assumed
to be truly exogenous and are instrumented by themselves. This results in a p-value
of 0.33 for the Sargan test at an instrument rank of 30. The p-value is satisfactory as
an indication of the validity of the overidentifying restrictions.
Estimation results
The results of the standard specification are summarized in table 8.14. The main
variables of interest, i.e. the OVERHANG variables, are both positive and significant
at the ten per cent level. With regard to a deposit overhang in normal times it was not
clear a priori which sign to expect. During times of crisis, however, the sign should be
positive in line with the theoretical underpinnings of hypothesis H9, according to
which access to an amount of insured deposits which at least equals the size of the
loan portfolio helps to shield the loan portfolio from the negative consequences of the
disruption of the wholesale funding markets. Hence, although the absolute value of
the coefficient is not excessively high (0.04), the result is consistent with
hypothesis H9.
One interesting observation concerns the coefficients for both "normal" deposit
variables. In the presence of the overhang variables, they turn clearly insignificant
(p-values of 0.76 and 0.96). On the one hand this has to do with the correlation
between the two variables. On the other hand this leads to the conclusion that it is
not necessarily the share of deposits over total assets that matters for the lending
reaction of banks, but that (the degree of) a deposit overhang is what matters most.
Expected Actual
LOG(LOANS(1)) + + *** 0.45 (0.15) 0.003
CAP(1)*NC + + *** 0.46 (0.16) 0.004
CAP(1)*CRISIS + + *** 0.49 (0.18) 0.005
SIZE(1)*NC + + ** 0.08 (0.04) 0.044
SIZE(1)*CRISIS + +* 0.15 (0.08) 0.065
LIQ(1)*NC + + 0.03 (0.34) 0.924
LIQ(1)*CRISIS + + 0.12 (0.53) 0.828
DEP(1)*NC + 0.02 (0.08) 0.763
DEP(1)*CRISIS + 0.00 (0.07) 0.963
OVERHANG(1)*NC +/ +* 0.05 (0.03) 0.061
OVERHANG(1)*CRISIS + +* 0.04 (0.02) 0.083
STF(1)*NC + + 0.14 (0.08) 0.107
STF(1)*CRISIS <STF*NC(+/) + 0.08 (0.06) 0.138
NII(1)*NC + + 0.16 (0.13) 0.230
NII(1)*CRISIS <NII*NC(+/) +* 0.10 (0.06) 0.097
LOG(GDP) + + ** 0.26 (0.11) 0.015
EURIBOR *** 0.00 (0.00) 0.002
GDP_DEFLATOR +/ *** 0.01 (0.00) 0.000
(NSM) + + 0.11 (0.10) 0.271
CRISIS + 0.00 (0.01) 0.611
Theestimatedmodelisgivenbyequation(16).CRISIS isadummyvariablethatis1intheyears2008and
2009and0inallotheryears.NC isthenoncrisisdummy.Ittakesonthevalueofzeroin2008and2009and
1inallotheryearsofthesampleperiod.Themodelallowsforfixedeffectsacrossbanksrealizedby
orthogonaldeviations.ItisestimatedusingdifferenceGMM,2stepestimator,Whiteperiodrobust
standarderrors.***,**and*indicatestatisticalsignificanceonthe1%,5%and10%levelrespectively.
Table 8.14: Results of estimations including the deposit overhang variable
Regarding the other bank characteristics included in the estimation, capitalization is,
as in all other specifications, positive and significant at the one per cent level while
the crisis coefficient is a little higher than the non-crisis one (consistent with
hypothesis H3). Familiar patterns also apply to size: Not only are both size
coefficients positive and significant at the five and ten per cent levels, but the crisis
coefficient is also higher than the non-crisis one in absolute terms. This provides
additional support for hypothesis H6.
As in most of the previous specifications, the estimates for liquidity are positive but
insignificant.
The coefficients for STF are positive for the non-crisis period and positive but lower
for the years of the crisis. Although they marginally miss the ten per cent significance
level, these results are consistent with the ones obtained by the previous
specifications.
A similar observation is shared for NII, which is positive outside of the crisis and
positive but smaller and significant at the ten per cent level during the crisis.
The results for STF and NII can be viewed as lending further support for the
hypotheses H4 and H5.
The results for the control variables are in line with expectations, including a positive
and significant result for the GDP variable and a negative and significant result (albeit
small in absolute terms) for Euribor and inflation.
In the following section, the robustness of the result obtained in this section the
relevance of a deposit overhang is tested by means of two additional specifications.
Deposit overhang dummy variable
According to the first of the alternative specifications that are designed to check the
robustness of the importance of a deposit overhang for the loan supply, an overhang
is captured by a dummy variable (DUMMY_OVERHANG) which assumes the value
of one whenever the amount of insured customer deposits exceeds the size of the
loan portfolio. A second dummy variable (DUMMY_NO_OVERHANG), which is the
exact mirror, assumes the value of zero in the event of a deposit overhang and one
otherwise.132 This setup allows for a direct comparison of the two groups of banks.
According to the hypotheses H2a and H2b, a higher-than-average share of short-

term funding and non-interest income for banks that have no overhang of deposits
should have a significant effect compared to banks that do have an overhang. Hence,
both overhang dummies (DUMMY_OVERHANG) and (DUMMY_NO_OVERHANG)
are interacted with the short-term funding variable and the non-interest income
variables. As a result, the coefficients for STF and NII are expected to be significant
for banks with no overhang in deposits and insignificant for banks with a deposit
overhang.
In addition, the bank characteristics capitalization, size, liquidity, non-interest income

and short-term funding are part of the regression. GDP, Euribor, the GDP deflator,
NSM and the crisis dummy serve as further control variables.
The choice of instruments is analogous to the standard specification, yielding a

p-value of 0.27 for the Sargan test statistic. The estimation is based on 14,030
(unbalanced) observations, including 2,144 cross-sections.
The results of the estimation (see table 8.15) are consistent with hypotheses H2a
and H2b. Indeed, the coefficients for STF and NII for banks that have no deposit
overhang are significantly different from zero at the five and ten per cent level
respectively, while the corresponding coefficients for banks that do have a deposit
overhang are insignificant. As expected, the signs on the no-overhang dummies are
lower than on the overhang dummies for this analysis.
Interestingly, inclusion of the "normal" STF and NII variables (interacted with the
crisis and non-crisis dummy) yields insignificant results throughout. This could be
interpreted to indicate that a deposit overhang (or the absence of it) is a main trigger
for the importance of STF and NII.
The parameters for the other bank characteristics are in line with those of the
previous estimations. Coefficients are positive and significant for capital, size and
deposits, whereas the results for liquidity are of no use.
132
Note that, due to the orthogonal deviations procedure (which removes cross-sectional fixed
effects), this induces neither perfect multicollinearity nor a dummy variable trap.
The results for the control variables are also as expected. In particular, demand for
loans seems to be well captured by the GDP variable, as indicated by highly
significant values. Moreover, the point estimate is in the range of earlier estimations.
Expected Actual
LOG(LOANS(1)) + + *** 0.68 (0.11) 0.000
CAP(1)*NC + + ** 0.42 (0.18) 0.018
CAP(1)*CRISIS + + ** 0.45 (0.18) 0.013
SIZE(1)*NC + + 0.14 (0.09) 0.139
SIZE(1)*CRISIS + +* 0.14 (0.08) 0.085
DEP(1)*NC + + ** 0.10 (0.05) 0.044
DEP(1)*CRISIS + + ** 0.11 (0.05) 0.030
LIQ(1)*NC + + 0.00 (0.39) 0.990
LIQ(1)*CRISIS + + 0.27 (0.56) 0.629
DUMMY_OVERHANG*STF(1) + + 0.13 (0.12) 0.285
DUMMY_NO_OVERHANG*STF(1) +/ + ** 0.05 (0.02) 0.011
DUMMY_OVERHANG*NII(1) + + 0.02 (0.03) 0.488
DUMMY_NO_OVERHANG*NII(1) +/ +* 0.00 (0.00) 0.075
STF(1)*NC + + 0.10 (0.08) 0.198
STF(1)*CRISIS <STF*NC(+/) + 0.09 (0.06) 0.145
NII(1)*NC + 0.06 (0.07) 0.417
NII(1)*CRISIS <NII*NC(+/) 0.06 (0.08) 0.419
LOG(GDP) + + *** 0.56 (0.17) 0.001
EURIBOR 0.00 (0.00) 0.177
GDP_DEFLATOR +/ ** 0.00 (0.00) 0.015
(NSM) + 0.12 (0.09) 0.177
CRISIS ** 0.02 (0.01) 0.040
Table 8.15: Estimation including a deposit overhang dummy variable
Deposit overhang dummy variable and time fixed effects
An additional robustness check repeats the preceding specification, with the

difference that it is estimated with a time fixed effect to underscore the cross-
sectional effects of the bank characteristics. Time dummies are thus included for
each year of the sample period.133
The instruments are chosen analogous to the preceding estimation, yielding a

satisfying p-value of 0.53 for the Sargan test. 2,144 cross-sections enter the
estimation with 14,030 (unbalanced) observations.
Expected Actual
LOG(LOANS(1)) + + *** 0.71 (0.11) 0.000
CAP(1)*NC + +* 0.45 (0.24) 0.062
CAP(1)*CRISIS + + ** 0.49 (0.25) 0.050
SIZE(1)*NC + + 0.14 (0.09) 0.117
SIZE(1)*CRISIS + + 0.14 (0.09) 0.104
DEP(1)*NC + + 0.08 (0.05) 0.141
DEP(1)*CRISIS + +* 0.09 (0.05) 0.082
LIQ(1)*NC + 0.02 (0.43) 0.970
LIQ(1)*CRISIS + + 0.14 (0.58) 0.806
DUMMY_OVERHANG*STF(1) + + 0.13 (0.13) 0.306
DUMMY_NO_OVERHANG*STF(1) +/ +* 0.06 (0.03) 0.074
DUMMY_OVERHANG*NII(1) + + 0.02 (0.03) 0.401
DUMMY_NO_OVERHANG*NII(1) +/ +* 0.00 (0.00) 0.069
STF(1)*NC + + 0.08 (0.07) 0.225
STF(1)*CRISIS <STF*NC(+/) + 0.07 (0.07) 0.304
NII(1)*NC + 0.06 (0.07) 0.381
NII(1)*CRISIS <NII*NC(+/) 0.06 (0.07) 0.427
LOG(GDP) + + 0.61 (0.59) 0.299
EURIBOR 0.01 (0.01) 0.256
GDP_DEFLATOR +/ + 0.00 (0.00) 0.108
(NSM) + + 0.39 (0.27) 0.144
Table 8.16: Deposit overhang dummy variable and time fixed effects
The results (table 8.16) confirm the conclusions reached earlier: Again, the
coefficients for STF and NII for banks with no deposit overhang are significant at the
133
The only difference to the preceding specification is that the crisis dummy is omitted here, since
the shift in the offset for the crisis years is already captured by relevant the time dummies.
ten per cent level and lower as expected, while the results for banks with an
overhang are not significant. Absolute values and standard error change only
marginally.
The robustness of the results for the other bank characteristics is relatively high. All
coefficients are in the same dimension and differ only slightly, pointing to the overall
validity of the results obtained.
8.1.4.3. Summary of main results and relationship to existing literature.
What can be concluded concerning the main results of the estimations for a deposit
overhang? The first result is that a deposit overhang has a positive impact on bank
lending during a crisis. In the presence of a deposit overhang variable, the "normal"
deposit variable becomes insignificant. This points to the conclusion that the fact that
a bank has an overhang of deposits marks a threshold from which its dependence on
wholesale funding is considerably lower. This "jump" seems to dominate the impact
of normal deposits to a certain extent.
A second result is that, if a bank has a deposit overhang, a higher-than-average

share of short-term funding and non-interest income places a burden on the loan
portfolio. Hypotheses H2a and H2b suggest that this is due to the lower relevance of
the asymmetric information issue, since a deposit overhang is a sign of a reduced
need of wholesale funding.
Taken together, the results highlight the fact that a deposit overhang helps to explain
the loan supply behavior of banks, thereby lending support to hypotheses H2a, H2b
and H9. Moreover, the results for the other bank characteristics are consistent with
hypotheses H3 (impact of capital during the crisis) and H6 (size being relevant in a
crisis).
How do the results compare to the related literature? Up to now, the effect of a
deposit overhang has scarcely been debated. The few available studies on the
subject focus on the role of deposits rather than on the impact of an overhang. As
mentioned in the context of the baseline analysis, there are two studies that refer to
deposits in a crisis context: One is Ivashina and Scharfstein (2010), who find that
banks with low deposits experienced the sharpest decline in their loan portfolios in
the US. They conclude that this can be attributed to the less acute need to roll over
debt and raise alternative market funding. This notion is broadly supported by
Gambacorta and Marques-Ibanez (2011) for a sample of banks in the euro area, the
US and the UK during the recent crisis. To the knowledge of the author, no further
studies are available that deal explicitly with deposits in a crisis context or with
deposit overhangs.
8.2. Results for major euro area countries
To ascertain whether the main results from the preceding section are also valid for
individual euro area countries or whether they only hold for the euro area as a whole,
the four most important euro area countries Germany, France, Italy and Spain
are analyzed in more detail. Another question regards the differences between the
individual countries which is also addressed.
Tested using
Hypotheses country
samples
H1
hypo-

H4
is lower (in terms of the value of the coefficient) compared to
normal times
crisis H6
context
H8
impact on lending

H9
lending
Hypothesis tested Hypothesis not tested (refer to text for further remarks)
Figure 8.3: Tested hypotheses based on individual country samples

8.2 Results for major euro area countries 171
The country analyses follow a standardized procedure according to which three types
of regressions are run for each of the countries under scrutiny: The first is the
baseline model, the second one focuses on the impact of a capital surplus, and the
third one deals with the effect of a deposit overhang on bank lending. Each of these
three types of regressions is accompanied by several robustness checks, such that a
total of seven specifications are tested for each of the countries. A summary of tested
hypotheses which are based on individual country samples is given in figure 8.3
above.
The subsequent sections are organized as follows: First, the composition of the
country samples and the exact specifications to be tested are described. The results
of the country analyses are then presented with a focus on the hypotheses-related
results. 134 The subsequent section is devoted to the differences between the
individual countries and their underlying causes.
8.2.1. Composition of the country samples and tested specifications
The composition of the country samples is straightforward: The basis for the data is
given by the euro area sample after applying the purging steps described in section
7.4.4 (and figure 7.5), from which the data on the banks in each of the four countries
is extracted.
This procedure ensures the generation of true subsamples. This is desirable in the
case in point in order to compare the results for the individual countries with the
overall euro area outcomes. If all purging steps had been applied to the country
samples individually, this would have implied that a number of the banks represented
in the euro area sample would not be included in the individual country samples
which could have an impact on the results.
Accordingly, all variables based on bank characteristics are constructed in exactly

the same way as for the euro area, as outlined in section 7.4. The only difference is
that normalization of the bank characteristics regarding their respective averages
134
Because the analysis is basically carried out in the same way as for the euro area as a whole,
large parts of the more technical considerations (number of cross-sections and observations,
variables used etc.) are not repeated in this section. This includes the summary statistics and the
correlation tables can be found in the appendix.
(across all banks in a country)135 is applied to the country samples individually. This
yields indicators that add up to zero across all observations, meaning that the
estimated parameters can be directly interpreted as the loan response of an average
bank.
In case of the macroeconomic control variables GDP and the GDP deflator, the
respective country variables are used (German GDP and GDP deflator for Germany,
French GDP for France, etc.). Since individual country data from the euro area bank
lending survey is not available, the data from the lending survey is not included in the
country samples.
The monetary policy indicators are euro area-wide figures, because monetary policy
indicators for individual countries are neither available nor would they be appropriate.
For all countries, the Euribor and the Euribor-OIS spread are used. Moreover, the
variable that captures non-standard monetary policy measures (NSM) is incorporated
in some of the specifications.
As stated above, three types of regressions are run, leading to seven different
specifications that are tested for each country. The specifications are briefly
explained below and can also be inferred from the information shown in table 8.17 to
table 8.24.
The first group of estimations is the one for the baseline specifications. In the first
model (I), based on equation (16), the first difference of the natural logarithm of loans
is regressed against its first lag, against the bank characteristics capital, size,
deposits, liquidity, short-term funding and non-interest income, and against the
control variables GDP, the GDP deflator, Euribor and the crisis dummy.136 Liquidity is
used only in interaction with Euribor as the monetary policy indicator. On the one
hand, this takes account of the buffer stock motive for liquidity holdings in the case of
adverse monetary policy changes, as stated by bank lending channel theory. On the
other hand, it is also in accordance with the findings from the euro area that liquidity
holdings themselves do not add much explanatory power regarding the lending
response of banks at conventional significance levels.
135
The only bank characteristics that are not normalized with respect to their averages are loans and
the variables related to a capital surplus and a deposit overhang.
136
The exact construction of the variables is explained in section 7.4 and table 7.2.
In the second model (II), the robustness of the results is checked with respect to
changes in the way monetary policy measures are integrated. The Euribor is
replaced by the Euribor-OIS spread and the non-standard monetary policy measures
variable is used.
The third specification (III) is the baseline model estimated using time fixed effects
and including NSM.
The second group of estimations focuses on the effect of a capital surplus or shortfall
relative to an individual capital target on the supply of bank loans. To this end, the
capital surplus variables enter the model (IV) in the interacted form (interacted with
the crisis dummy and the non-crisis dummy) in addition to the "pure" capital variables.
In a robustness check (V), the same specification is estimated with time fixed effects.
The third group is geared to testing the impact of a deposit overhang. Accordingly,
the two (interacted) deposit overhang variables enter the equation (VI). Again,
robustness is checked by estimating the same specification with time fixed effects
(VII).
These seven specifications for each country allow all the hypotheses stated in
section 7.1 to be tested, with the exception of hypotheses H2 and H7. Hypotheses
H2a to H2c are the only ones that are not related to comparing the crisis to normal
periods and are therefore less in focus for the country samples. 137 Moreover,
Hypotheses H2c and H7 both concern the impact of securities that have to be
marked to market (AFS). As discussed in connection with the hypothesis tests for the
euro area as a whole, the number of available data points is relatively low. Dividing
the euro area sample into country subsamples thus further aggravates the issue of
the poor availability of AFS-related data. As a result, no meaningful hypothesis tests
are possible, such that the effect of securities that have to be marked to market on
lending on an individual country basis is not analyzed.
Having explained the various specifications, the more technical details of the
estimation should not be overlooked. The estimation method is exactly the same as
the one used for the euro area. This includes the GMM methodology, the removal of
individual fixed effects by orthogonal deviations and the choice of instruments. As in
137
Hypotheses H1 in its formulation is also not specifically related the crisis period but, first, this
distinction is accounted for in the empirical design of the respective estimations and, second, the
impact of a capital surplus on lending is of distinct interest to this study.
all other regressions, loans serving as the endogenous explanatory variable are
instrumented by the dynamic instruments proposed by Arellano and Bond (1991), i.e.
the second and third lags of the natural logarithm of loans in first differences. All
other bank characteristics are instrumented by the second lags of (non-crisis and
crisis) interactions with the bank characteristics. GDP, the GDP deflator variable, the
monetary policy indicator and inflation are assumed to be truly exogenous and are
instrumented by themselves.
While this approach to the estimation is an important element to deal with the
possible problem of endogenous or predetermined regressors, a second step to
address this issue is the inclusion of bank characteristics that are all lagged once.138
The descriptive statistics and a correlation matrix for each of the four countries can
be found in the appendix.
The following sections are devoted to the main results from the four countries under
scrutiny, with a focus on those aspects that are directly linked to the hypotheses.
8.2.2. Main results for Germany
Before coming to the results, it is worth noting that, with more than 1,200 cross-
sections and almost 10,000 observations included in the regressions, Germany is the
country that provides the richest amount of data. The regression outcomes can
therefore be viewed as reliable and resilient in the sense that there is no reason to
believe they suffer from an insufficiently broad set of data.
What are the main insights from the baseline regressions (I-III) with regard to the
hypotheses to be tested? (See figure 8.3 for an overview of hypotheses tested using
the individual country sample for Germany and figure 8.6 in section 8.3 for a
summary of the results.)
The results (see table 8.17) are consistent with hypothesis H3: In all specifications,
the crisis coefficient for capitalization is higher than the non-crisis coefficient.
The estimation outcome also supports hypothesis H4, according to which a larger139
share of short-term funding has a negative impact in relative terms on lending, i.e.
138
This issue of possible endogeneity is discussed in section 7.5. See also footnote 114 on page 120.
compared to the non-crisis coefficient, during a crisis. The result is statistically

significant at the ten per cent level in most of the cases. In addition, a persistent
pattern is that the effect of short-term funding is consistently more positive outside of
a crisis.
GERMANY (I) Baseline model (II) Baseline with Euribor_OIS (III) Baseline model with time-
and non-standard measures fixed effects
Dependent variable:
LOG(LOANS) Coeff. Std. Error Coeff. Std. Error Coeff. Std. Error
LOG(LOANS(-1)) 0.56 *** (0.10) 0.52 *** (0.10) 0.52 *** (0.10)
CAP(-1)*NC 0.45 ** (0.21) 0.57 *** (0.20) 0.45 ** (0.19)
CAP(-1)*CRISIS 0.46 *** (0.17) 0.60 *** (0.16) 0.51 *** (0.18)
CAPSUR(-1)*NC
CAPSUR(-1)*CRISIS
SIZE(-1)*NC 0.06 (0.04) 0.06 (0.04) 0.05 * (0.03)
SIZE(-1)*CRISIS 0.06 ** (0.03) 0.08 ** (0.04) 0.07 ** (0.03)
DEP(-1)*NC 0.00 (0.04) 0.02 (0.04) 0.00 (0.05)
DEP(-1)*CRISIS 0.06 * (0.03) 0.07 * (0.04) 0.05 (0.03)
OVERHANG(-1)*NC
OVERHANG(-1)*CRISIS
LIQ(-1)*EURIBOR*NC 0.31 (0.21) 0.17 * (0.10) 0.05 (0.11)
LIQ(-1)*EURIBOR*CRISIS 0.29 (0.28) 0.17 (0.15) 0.03 (0.17)
ST F(-1)*NC 0.19 * (0.10) 0.18 * (0.10) 0.16 (0.13)
ST F(-1)*CRISIS 0.06 * (0.03) 0.06 * (0.03) 0.05 (0.03)
NII(-1)*NC 0.03 (0.10) 0.08 (1.76) 0.08 (0.09)
NII(-1)*CRISIS 0.01 (0.09) 0.03 (0.05) 0.03 (0.10)
LOG(GDP) 0.24 * (0.13) 0.13 * (0.07) 0.45 (0.30)
EURIBOR 0.00 *** (0.00) 0.00 (0.00)
EURIBOR_OIS -0.03 *** (0.01)
GDP_DEFLAT OR -0.01 *** (0.00) 0.00 (0.00) 0.00 (0.00)
(NSM) 0.36 ** (0.17) -0.11 (0.39)
CRISIS -0.01 (0.01) -0.01 ** (0.00)
Time dummies: NO NO YES

Sample period (adjusted): 2001-2011 2001-2011 2001-2011
Cross-sections included: 1,241 1,241 1,241
Total (unbalanced) obs.: 9,974 9,974 9,974
Instrument rank: 27 28 38
Sargan test (p-value): 0.59 0.68 0.72
Theestimatedmodelisgivenbyequation(16).CRISISisadummyvariablethatis1intheyears2008and2009and0inallother
years.NCisthenoncrisisdummy.Ittakesonthevalueof0in2008and2009and1inallotheryearsofthesampleperiod.The
modelallowsforfixedeffectsacrossbanksrealizedbyorthogonaldeviations.ItisestimatedusingdifferenceGMM,2step
estimator,Whiteperiodrobuststandarderrors.***,**and*indicatestatisticalsignificanceonthe1%,5%and10%level
respectively.
Table 8.17: Baseline results for Germany
139
Because all bank characteristics (except for loans, capital surplus and deposit overhang) are
normalized with respect to their respective averages, "larger" means a larger share than that of an
average (German) bank.
A similar observation can be made with respect to the impact of non-interest income
and hypothesis H5: It is positive at normal times but, relative to that, negative in times
of crisis. The difference is, however, that the estimated coefficients miss the ten per
cent significance level in most cases.
GERMANY (IV) Capital surplus (V) Capital surplus and (VI) Deposit overhang (VII) Deposit overhang
time-fixed effects and time-fixed effects
Dependent variable:
LOG(LOANS) Coeff. Std. Error Coeff. Std. Error Coeff. Std. Error Coeff. Std. Error
LOG(LOANS(-1)) 0.54 *** (0.10) 0.52 *** (0.10) 0.48 *** (0.12) 0.44 *** (0.12)
CAP(-1)*NC 0.53 *** (0.16) 0.48 *** (0.13) 0.45 ** (0.19) 0.47 *** (0.18)
CAP(-1)*CRISIS 0.76 ** (0.32) 0.74 *** (0.26) 0.46 *** (0.17) 0.54 *** (0.19)
CAPSUR(-1)*NC 0.06 *** (0.02) 0.06 *** (0.02)
CAPSUR(-1)*CRISIS 0.05 * (0.03) 0.06 ** (0.03)
SIZE(-1)*NC 0.05 * (0.03) 0.05 (0.03) 0.06 (0.05) 0.08 * (0.04)
SIZE(-1)*CRISIS 0.07 ** (0.03) 0.08 ** (0.04) 0.08 ** (0.04) 0.10 *** (0.03)
DEP(-1)*NC -0.03 (0.05) -0.03 (0.05) -0.04 (0.05) -0.04 ** (0.02)
DEP(-1)*CRISIS 0.03 (0.04) 0.01 (0.05) 0.03 (0.04) 0.03 (0.02)
OVERHANG(-1)*NC 0.06 *** (0.01) 0.06 *** (0.01)
OVERHANG(-1)*CRISIS 0.08 *** (0.02) 0.05 *** (0.01)
LIQ(-1)*EURIBOR*NC 0.42 ** (0.21) 0.20 (0.21) 0.26 * (0.15) 0.06 (0.10)
LIQ(-1)*EURIBOR*CRISIS 0.39 * (0.21) 0.18 (0.26) 0.28 (0.22) 0.05 (0.16)
STF(-1)*NC 0.23 ** (0.10) 0.21 * (0.12) 0.17 (0.12) 0.14 (0.19)
STF(-1)*CRISIS 0.10 * (0.05) 0.10 * (0.06) 0.04 * (0.02) 0.03 (0.02)
NII(-1)*NC 0.05 (0.45) 0.11 (0.22) 0.06 (1.09) 0.14 (0.13)
NII(-1)*CRISIS 0.03 (0.08) 0.05 * (0.03) 0.01 (0.09) 0.04 (0.05)
LOG(GDP) 0.33 ** (0.13) 0.06 (0.42) 0.38 *** (0.13) 0.17 (0.40)
EURIBOR 0.00 *** (0.00) 0.00 (0.00) 0.00 *** (0.00) 0.00 (0.00)
EURIBOR_OIS
GDP_DEFLATOR -0.01 *** (0.00) -0.01 (0.01) -0.01 *** (0.00) 0.00 (0.00)
(NSM) -0.06 (0.10) -0.08 (0.42) -0.10 (0.09) 0.29 (0.37)
CRISIS 0.00 (0.01) -0.01 (0.02)
Time dummies: NO YES NO YES

Sample period (adjusted): 2001-2011 2001-2011 2001-2011 2001-2011
Cross-sections included: 1,238 1,238 1,241 1,241
Total (unbalanced) obs.: 9,939 9,939 9,974 9,974
Instrument rank: 30 40 30 40
Sargan test (p-value): 0.60 0.72 0.54 0.65
respectively.
Table 8.18: Capital surplus and deposit overhang results for Germany
Regarding hypothesis H6, the role of the size of a bank is positive and significant at
the five per cent level in all specifications during the crisis, which is consistent with
what the hypothesis claims. Nevertheless, it should be noted that, although the crisis
coefficient is consistently higher, the difference between the crisis and the non-crisis
estimate is relatively small.
The parameters for deposits are consistent with hypothesis H8 in the sense that a
higher-than-average share of deposits has a positive impact during the crisis.
Turning to the regressions that center around the capital surplus variable, the fact
that a capital surplus helps to explain the lending response of banks can be recorded
as a key result (table 8.18). The crisis and non-crisis parameters are significant at the
ten and one per cent levels respectively in both of the tested specifications (IV and V).
It is worth noting that they are positive even in the presence of the "pure" capital
variables, whose coefficients also show the expected positive sign and are
statistically significant at the five and one per cent levels. All in all, the capital surplus
variables seem to work well for Germany and the results very strongly support
hypothesis H1.
The last two specifications (VI and VII) are devoted to the role that a deposit
overhang plays in the supply of bank loans. The result for the German sample is, first,
that the overhang parameters are positive and significant, particularly in times of
crisis (but also during normal times) in both specifications. In this sense, the
coefficients are as expected and consistent with hypothesis H9. A second result is
that, in presence of the deposit overhang variables, the coefficients on normal
deposits turn negative for the non-crisis parameter while they stay positive for the
crisis version. Although not statistically significant at conventional levels, this can be
seen as further supporting the positive impact of normal deposits on the loan supply
during a crisis.
To sum up the results for Germany: The results are consistent with the formulated
hypotheses to a large extent. Support for hypothesis H5 (non-interest income) alone
seems slightly vague, which can be attributed to a lack of statistically significant
results and not to unexpected deflections for the parameters. In particular, the
hypotheses regarding the impact of a capital surplus or a deposit overhang are
supported by the data.
8.2.3. Main results for Italy
ITALY (I) Baseline model (II) Baseline with Euribor_OIS (III) Baseline model with time-
Dependent variable:
LOG(LOANS(-1)) 0.07 ** (0.04) 0.05 (0.05) 0.05 (0.05)
CAP(-1)*NC 1.19 (0.74) 1.20 ** (0.59) 1.12 (0.82)
CAP(-1)*CRISIS 1.16 * (0.61) 1.09 * (0.65) 1.11 (0.89)
CAPSUR(-1)*NC
CAPSUR(-1)*CRISIS
SIZE(-1)*NC 0.13 *** (0.04) 0.07 * (0.04) 0.24 *** (0.06)
SIZE(-1)*CRISIS 0.35 ** (0.14) 0.06 (0.04) 0.23 *** (0.06)
DEP(-1)*NC 0.45 ** (0.21) 0.58 *** (0.17) 0.44 ** (0.20)
DEP(-1)*CRISIS 0.48 ** (0.21) 0.64 *** (0.18) 0.47 ** (0.21)
OVERHANG(-1)*NC
OVERHANG(-1)*CRISIS
LIQ(-1)*EURIBOR*NC 0.18 (0.30) 0.08 (0.24) 0.05 (0.27)
ST F(-1)*NC 0.21 (0.19) 0.07 (0.30) 0.07 (0.27)
ST F(-1)*CRISIS 0.11 (0.19) 0.04 (0.19) 0.04 (0.27)
NII(-1)*NC 0.22 (1.26) 0.13 (0.52) 0.16 ** (0.07)
NII(-1)*CRISIS 0.02 * (0.01) 0.03 * (0.02) 0.08 (0.07)
LOG(GDP) 0.47 ** (0.20) 0.30 (0.40) 0.16 (0.18)
EURIBOR -0.01 *** (0.00) -0.01 (0.02)
EURIBOR_OIS -0.01 (0.02)
GDP_DEFLAT OR 0.02 *** (0.00) 0.00 (0.01) -0.06 (0.05)
(NSM) -0.06 (0.26) -0.18 (1.28)
CRISIS -0.03 *** (0.01) -0.02 (0.02)

Cross-sections included: 494 494 494
Total (unbalanced) obs.: 1,594 1,594 1,594
respectively.
Table 8.19: Baseline results for Italy
The second biggest euro area country in terms of the number of banks in the sample
is Italy. With less than 500 banks and about 1,600 observations, however, the Italian
sample includes only 40 per cent of the banks and only 15 per cent of the
observations compared to Germany. Especially the comparison of the number of
observations leads to the conclusion that, in the Italian sample, on average the
number of observations per bank is available to a lesser extent than for Germany.
Nevertheless, the number of observations should still be large enough to produce

meaningful results.
ITALY (IV) Capital surplus (V) Capital surplus and (VI) Deposit overhang (VII) Deposit overhang
Dependent variable:
LOG(LOANS(-1)) 0.09 ** (0.03) 0.06 (0.04) 0.34 (0.24) 0.04 (0.03)
CAP(-1)*NC 0.26 (1.27) 0.67 * (0.38) 0.92 * (0.51) 0.89 (0.65)
CAP(-1)*CRISIS 0.39 (1.38) 0.92 * (0.50) 1.05 * (0.56) 0.84 (0.69)
CAPSUR(-1)*NC 0.18 ** (0.07) 0.10 (0.06)
CAPSUR(-1)*CRISIS 0.14 * (0.08) 0.12 * (0.07)
SIZE(-1)*NC 0.27 *** (0.05) 0.27 *** (0.05) 0.31 *** (0.06) 0.32 *** (0.05)
SIZE(-1)*CRISIS 0.25 *** (0.05) 0.25 *** (0.05) 0.30 *** (0.06) 0.31 *** (0.04)
DEP(-1)*NC 0.59 *** (0.14) 0.35 ** (0.16) 0.52 * (0.29) 0.24 (0.15)
DEP(-1)*CRISIS 0.57 *** (0.15) 0.34 ** (0.17) 0.58 * (0.32) 0.26 * (0.16)
OVERHANG(-1)*NC 0.22 * (0.12) 0.12 *** (0.02)
OVERHANG(-1)*CRISIS 0.20 * (0.12) 0.12 *** (0.03)
LIQ(-1)*EURIBOR*NC 0.25 (0.30) 0.13 (0.28) 0.14 (0.34) 0.11 (0.28)
LIQ(-1)*EURIBOR*CRISIS 0.15 (0.24) 0.10 (0.22) 0.10 (0.30) 0.06 (0.20)
STF(-1)*NC 0.24 (0.22) 0.16 (0.18) 0.42 (0.33) 0.17 (0.51)
STF(-1)*CRISIS 0.06 (0.04) 0.04 (0.18) 0.19 (0.15) 0.08 (0.21)
NII(-1)*NC 0.18 (0.59) 0.20 (0.19) 0.74 * (0.40) 0.24 * (0.13)
NII(-1)*CRISIS 0.04 (0.02) 0.07 (0.07) 0.47 ** (0.20) 0.16 (0.20)
LOG(GDP) 1.03 *** (0.31) 1.29 (2.29) 1.97 *** (0.62) 0.09 (1.40)
EURIBOR -0.02 *** (0.00) -0.06 * (0.03) -0.03 *** (0.01) 0.00 (0.01)
EURIBOR_OIS
GDP_DEFLATOR 0.02 *** (0.01) 0.10 ** (0.04) 0.04 *** (0.01) -0.04 (0.03)
(NSM) -0.20 (0.19) -0.16 (0.14) -0.68 * (0.35) -0.02 (0.96)
CRISIS -0.01 (0.02) -0.03 (0.02)

Cross-sections included: 494 494 494 494
Total (unbalanced) obs.: 1,594 1,594 1,594 1,594
respectively.
Table 8.20: Capital surplus and deposit overhang results for Italy
Starting with the results from the baseline specifications (I-III; see table 8.19), it can
be observed that the role of the share of non-interest income is supported in
accordance with hypothesis H5. (See figure 8.3 on page 170 for an overview of
hypotheses tested using the individual country sample for Italy and figure 8.6 in
section 8.3 for a summary of the results.) The estimated coefficients for the crisis
version have are lower, as expected, and are significant in the majority of cases.
Regarding the impact of short-term funding (hypothesis H4), it must be noted that the
estimates produce the expected lower coefficients for the crisis (and positive but
higher ones for normal times), but the results are not statistically significant at the ten
per cent level. Despite this, the results can still be interpreted as supporting
hypothesis H4 at least weakly.
The impact of size, particularly in the crisis, is positive and significant at the five per
cent level. One exception is the second specification, in which the estimate of size is
not significant and is considerably smaller in absolute terms. Hypothesis H6 is not
supported, however, since the coefficient on size in a crisis is not higher than outside
of a crisis.
Hypothesis H3 is also not supported by the data. The estimates for the crisis version
of the capital variable are not consistently higher than for the non-crisis one (but only
in roughly half of the cases). Still, the effect of capital on lending is generally positive.
For Italy, the notion that a larger-than-average amount of deposits is beneficial during
a crisis regarding the size of the loan portfolio (hypothesis H8) is well confirmed,
although the difference between the crisis and the non-crisis coefficients is not very
large.
Turning to the evidence provided by tests of specifications IV and V on the capital

surplus (table 8.20), the estimated parameters are significant at the ten per cent level
for interaction with the crisis dummy. For non-crisis interaction, the sign is also
positive but the significance is not robust for the specification using time fixed effects
(p-value: 0.12). Nevertheless, taking all factors into account, the results are
consistent with hypothesis H1.
The last two regressions focus on the role of a deposit overhang. According to theory
and to hypothesis H9, an overhang in deposits in crisis years should largely shield
banks from turmoil on the wholesale funding markets. On the basis of the data, this
assumption can be confirmed with significantly positive results throughout all
specifications.
To sum up: Hypotheses H3 and H6 ("pure capital" and size) are not supported by the
data. The evidence in favor of hypothesis H5 (non-interest income) is relatively weak.
However, this hypothesis is not contradicted by the data as measured by the
expected difference between the crisis and the non-crisis coefficients. The other
hypotheses, H1, H4, H8 and H9, are well supported, especially the one regarding the
impact of a deposit overhang.
8.2.4. Main results for France
The third biggest country in the sample is France. 120 cross-sections (banks) are
included in the sample, producing almost 900 observations. On average, coverage in
terms of the number of observations for each bank is clearly better than in the Italian
sample. The total number of observations is still satisfactory.
This better coverage of French banks is reflected in the consistency and robustness
of the results across different specifications as far as the signs and absolute values
are concerned (I-III; see table 8.21). (See figure 8.3 on page 170 for an overview of
hypotheses tested using the individual country sample for France and figure 8.6 in
section 8.3 for a summary of the results.)
This does not mean, however, that all hypotheses are supported by the data. : An
interesting observation that can be made given the baseline analysis in specifications
I to III is that the parameter for capital is consistently higher at normal times than
during the crisis. This contrasts with the results for the euro area as a whole, the
other countries individually (apart from Italy) and the prediction of hypothesis H3.
Regarding hypotheses H4 and H5 the results for the share of short-term funding and
the share of non-interest income line up with expectations regarding the direction and
difference of the impacts, especially during the crisis. It must yet be noted that they
are not statistically significant. Accordingly, the evidence in favor of hypotheses H4
and H5 is relatively weak.
As a further result, the coefficients for size are positive both during and outside of a
crisis and are statistically significant at least at the ten per cent level in most cases. In
particular, the crisis coefficients are higher in absolute terms than the non-crisis ones,
although the difference in absolute terms is small. Altogether, this is still consistent
with hypothesis H6 which is weakly supported.
Regarding the role of "pure" deposits during the crisis (hypothesis H8), the evidence
is somewhat unconvincing. Though the sign is positive and the absolute value for the
crisis coefficient is consistently higher than for the non-crisis one, the difference
between the non-crisis and crisis estimates is very small and the coefficients are not
statistically significant in specifications I to III.
FRANCE (I) Baseline model (II) Baseline with Euribor_OIS (III) Baseline model with time-
Dependent variable:
LOG(LOANS(-1)) 0.26 *** (0.06) 0.18 *** (0.06) 0.15 *** (0.05)
CAP(-1)*NC 0.82 *** (0.23) 0.46 *** (0.14) 0.45 *** (0.13)
CAP(-1)*CRISIS 0.55 ** (0.23) 0.17 * (0.10) 0.16 * (0.10)
CAPSUR(-1)*NC
CAPSUR(-1)*CRISIS
SIZE(-1)*NC 0.02 * (0.01) 0.07 ** (0.03) 0.12 *** (0.03)
SIZE(-1)*CRISIS 0.03 ** (0.01) 0.08 ** (0.03) 0.13 *** (0.03)
DEP(-1)*NC 0.01 (0.09) 0.02 (0.06) 0.12 (0.12)
DEP(-1)*CRISIS 0.11 (0.12) 0.06 (0.08) 0.16 (0.11)
OVERHANG(-1)*NC
OVERHANG(-1)*CRISIS
LIQ(-1)*EURIBOR*NC 0.54 ** (0.22) 0.22 (0.15) 0.50 (0.32)
LIQ(-1)*EURIBOR*CRISIS 0.35 ** (0.15) 0.22 * (0.12) 0.29 (0.18)
ST F(-1)*NC 0.10 (0.11) 0.09 (0.61) 0.26 (0.52)
ST F(-1)*CRISIS 0.01 (0.08) 0.07 (0.08) 0.11 (0.10)
NII(-1)*NC 0.02 (0.03) 0.03 (0.12) 0.02 (0.03)
NII(-1)*CRISIS 0.00 (0.00) 0.03 (0.02) 0.01 (0.02)
LOG(GDP) 0.64 *** (0.19) 0.32 ** (0.14) 0.56 (1.07)
EURIBOR 0.00 (0.00) 0.00 (0.02)
EURIBOR_OIS -0.05 * (0.03)
GDP_DEFLAT OR -0.01 * (0.00) -0.05 * (0.03) -0.01 (0.02)
(NSM) 0.99 *** (0.20) 0.35 (0.36)
CRISIS -0.01 (0.01) -0.01 (0.01)

Total (unbalanced) obs.: 896 896 896
respectively.
Table 8.21: Baseline results for France
Interestingly, this picture changes in specifications VI and VII in the presence of the
deposit overhang variables. In these specifications, the estimated parameters for
deposits are both negative and significant, whereas those for the overhang variables
are positive and significant. This peculiarity of the French data supports hypothesis
H9 well but is only very weakly consistent with hypothesis H8 based on specifications
VI and VII.
The results for the capital surplus variables are consistent with hypothesis H1. Both
during and outside of the crisis, a capital surplus has a positive impact on bank
lending. It is worth noting that the impact of "pure" capital is stronger in absolute
terms compared to the other specifications without the surplus variables.
FRANCE (IV) Capital surplus (V) Capital surplus and (VI) Deposit overhang (VII) Deposit overhang
Dependent variable:
LOG(LOANS(-1)) 0.13 * (0.07) 0.13 * (0.08) 0.41 *** (0.04) 0.04 *** (0.01)
CAP(-1)*NC 1.56 ** (0.61) 1.81 *** (0.57) 0.57 *** (0.22) 0.47 *** (0.16)
CAP(-1)*CRISIS 1.69 ** (0.71) 1.93 *** (0.46) 0.29 (0.22) 0.15 (0.17)
CAPSUR(-1)*NC 0.05 * (0.03) 0.08 *** (0.02)
CAPSUR(-1)*CRISIS 0.07 ** (0.03) 0.08 *** (0.03)
SIZE(-1)*NC 0.06 *** (0.02) 0.10 *** (0.01) 0.03 * (0.02) 0.10 *** (0.03)
SIZE(-1)*CRISIS 0.06 *** (0.02) 0.11 *** (0.01) 0.04 ** (0.02) 0.10 *** (0.03)
DEP(-1)*NC 0.13 (0.09) 0.01 (0.06) -0.12 * (0.06) -0.22 * (0.13)
DEP(-1)*CRISIS 0.16 * (0.08) 0.07 (0.05) -0.11 * (0.07) -0.18 * (0.10)
OVERHANG(-1)*NC 0.05 *** (0.02) 0.07 * (0.04)
OVERHANG(-1)*CRISIS 0.07 *** (0.01) 0.07 ** (0.03)
LIQ(-1)*EURIBOR*NC 0.66 *** (0.23) 0.50 *** (0.18) 0.26 (0.16) 0.57 * (0.34)
LIQ(-1)*EURIBOR*CRISIS 0.32 ** (0.15) 0.35 *** (0.13) 0.31 ** (0.12) 0.36 * (0.19)
STF(-1)*NC 0.26 ** (0.12) 0.12 (0.35) 0.11 (0.74) 0.27 (0.19)
STF(-1)*CRISIS 0.11 (0.09) 0.10 * (0.05) 0.06 (0.05) 0.13 (0.13)
NII(-1)*NC 0.02 (0.02) 0.03 (0.03) 0.10 (0.07) 0.04 *** (0.01)
NII(-1)*CRISIS 0.01 (0.02) 0.01 (0.02) 0.05 *** (0.02) 0.02 (0.01)
LOG(GDP) 0.43 * (0.24) 1.48 ** (0.62) 0.24 (0.19) 0.68 (1.00)
EURIBOR -0.03 *** (0.00) -0.03 *** (0.01) -0.01 *** (0.00) 0.00 (0.01)
EURIBOR_OIS
GDP_DEFLATOR -0.04 *** (0.01) -0.03 ** (0.01) -0.03 *** (0.00) -0.03 (0.02)
(NSM) 0.44 *** (0.14) 1.16 ** (0.47) 1.18 *** (0.17) 0.76 (0.73)
CRISIS -0.01 (0.01) -0.01 (0.01)

Total (unbalanced) obs.: 896 896 896 896
respectively.
Table 8.22: Capital surplus and deposit overhang results for France
To sum up: The estimates based on the sample for France produce the strongest
evidence in favor of hypotheses H1 and H9 (capital surplus and deposit overhang).
Weak support is provided for hypotheses H4 H5 and H6 (STF, NII and SIZE), while
hypotheses H3 and H8 (capital and deposits) are not consistently backed up by the
data.140
8.2.5. Main results for Spain
Of the four individual countries examined, Spain is the one with the least number of
cross-sections and observations in its country sample. Each of the specifications
I to VII contains 74 cross-sections and about 220 observations. This is still marginally
satisfactory.
What are the main results for the Spanish sample (see table 8.23 and table 8.24)?
(See figure 8.3 on page 170 for an overview of hypotheses tested using the individual
country sample for Spain and figure 8.6 in section 8.3 for a summary of the results.)
Regarding the impact of capital during the crisis (hypothesis H3), the coefficient is
consistently higher than the one outside of a crisis. Although the crisis coefficient is
not statistically significant in all cases, it nevertheless speaks well for the validity of
hypothesis H3.
Regarding the share of short-term funding and non-interest income, the pattern
observable in France is repeated here: For both variables, the crisis and non-crisis
coefficients show the anticipated difference but are not statistically significant (at the
ten per cent level) in most cases. Accordingly, hypotheses H4 and H5 once again
receive only weak support from the data.
The impact of size on lending (hypothesis H6) is not clear for Spain. Although the
impact is positive in most cases, the results are not significant. Moreover, the crisis
coefficient is not higher than the non-crisis one. Altogether, the results do not
substantiate the hypothesis.
The estimated coefficients for deposits during the crisis (hypothesis H8) are positive
and statistically significant. In addition, the absolute values are consistently higher
than those for the non-crisis coefficients, which suggests that deposits are more
important during the crisis.
140
The case of pure deposits (H8) may be considered as borderline, since specifications IV to VII
support hypothesis H8 at least weakly. Nevertheless, in an overall assessment and applying strict
standards, results are interpreted as not supporting H8.
SPAIN (I) Baseline model (II) Baseline with Euribor_OIS (III) Baseline model with time-
Dependent variable:
LOG(LOANS(-1)) 0.12 * (0.06) 0.21 ** (0.09) 0.10 (0.12)
CAP(-1)*NC 1.10 *** (0.29) 0.77 (0.51) 0.31 (0.75)
CAP(-1)*CRISIS 1.33 ** (0.53) 1.08 (0.66) 1.09 (0.75)
CAPSUR(-1)*NC
CAPSUR(-1)*CRISIS
SIZE(-1)*NC 0.03 (0.06) 0.00 (0.07) 0.09 (0.24)
SIZE(-1)*CRISIS 0.02 (0.06) -0.01 (0.08) 0.09 (0.24)
DEP(-1)*NC 0.39 *** (0.11) 0.08 (0.10) 0.33 * (0.19)
DEP(-1)*CRISIS 0.75 *** (0.20) 0.41 * (0.22) 0.60 ** (0.28)
OVERHANG(-1)*NC
OVERHANG(-1)*CRISIS
LIQ(-1)*EURIBOR*NC 0.38 *** (0.06) 0.26 *** (0.07) 0.25 ** (0.12)
ST F(-1)*NC 0.43 *** (0.09) 0.39 (0.26) 0.35 (0.22)
ST F(-1)*CRISIS 0.22 (0.30) 0.17 (0.13) 0.01 (0.14)
NII(-1)*NC 0.27 (0.17) 0.35 * (0.20) 0.33 (0.57)
NII(-1)*CRISIS 0.07 (0.10) 0.08 (0.17) 0.16 (0.13)
LOG(GDP) 0.67 *** (0.19) 1.51 *** (0.42) 0.59 (0.73)
EURIBOR -0.02 *** (0.00) 0.00 (0.02)
EURIBOR_OIS -0.03 * (0.02)
GDP_DEFLAT OR -0.02 *** (0.00) 0.00 (0.00) -0.01 (0.02)
(NSM) -0.73 ** (0.32) -0.23 (1.67)
CRISIS -0.03 *** (0.01) -0.05 ** (0.02)

Total (unbalanced) obs.: 222 222 222
respectively.
Table 8.23: Baseline results for Spain
This picture is confirmed by specifications VII and VIII. When the deposit overhang
variables are integrated in the equations, "plain" deposits remain significant during
the crisis and the parameters are higher than the non-crisis ones. The coefficients for
the overhang variables are also positive and significant at least at the five per cent
level and are higher during the crisis than at normal times. This further underlines the
importance not only of deposits, but also of a deposit overhang for Spain during the
crisis (hypotheses H8 and H9).
SPAIN (IV) Capital surplus (V) Capital surplus and (VI) Deposit overhang (VII) Deposit overhang
Dependent variable:
LOG(LOANS(-1)) 0.08 (0.08) 0.03 (0.10) 0.24 *** (0.07) 0.39 ** (0.17)
CAP(-1)*NC 0.24 (0.69) 0.78 (1.72) 1.06 * (0.55) 1.00 (0.65)
CAP(-1)*CRISIS 0.69 (0.93) 1.06 (2.07) 1.65 ** (0.69) 1.65 ** (0.64)
CAPSUR(-1)*NC 0.12 ** (0.06) 0.02 (0.05)
CAPSUR(-1)*CRISIS 0.18 ** (0.08) 0.08 (0.08)
SIZE(-1)*NC 0.03 (0.06) 0.07 (0.08) 0.10 (0.07) 0.25 *** (0.09)
SIZE(-1)*CRISIS 0.01 (0.07) 0.10 (0.08) 0.10 (0.07) 0.26 *** (0.09)
DEP(-1)*NC 0.34 ** (0.17) 0.26 (0.25) 0.07 (0.23) 0.26 (0.33)
DEP(-1)*CRISIS 0.81 ** (0.37) 0.61 ** (0.30) 0.60 * (0.33) 0.60 ** (0.26)
OVERHANG(-1)*NC 0.17 ** (0.08) 0.50 *** (0.13)
OVERHANG(-1)*CRISIS 0.22 *** (0.08) 0.55 *** (0.13)
LIQ(-1)*EURIBOR*NC 0.40 ** (0.18) 0.22 ** (0.10) 0.45 * (0.25) 0.42 *** (0.14)
LIQ(-1)*EURIBOR*CRISIS 0.63 * (0.37) 0.45 (0.38) 0.60 * (0.33) 0.56 ** (0.27)
STF(-1)*NC 0.42 * (0.23) 0.31 (0.35) 0.39 *** (0.13) 0.30 (0.21)
STF(-1)*CRISIS 0.19 (0.15) 0.18 * (0.10) 0.16 (0.15) 0.13 (0.10)
NII(-1)*NC 0.32 ** (0.15) 0.45 ** (0.20) 0.20 (0.18) 0.29 (0.18)
NII(-1)*CRISIS 0.10 ** (0.05) 0.09 (0.15) 0.02 (0.12) 0.01 (0.15)
LOG(GDP) 1.83 *** (0.50) 1.73 ** (0.78) 0.31 (0.54) 1.90 ** (0.76)
EURIBOR -0.01 ** (0.01) -0.02 (0.03) 0.00 (0.00) 0.00 (0.01)
EURIBOR_OIS
GDP_DEFLATOR -0.01 ** (0.00) 0.02 (0.02) 0.05 *** (0.01) 0.00 (0.01)
(NSM) -1.11 *** (0.40) 0.63 (1.76) -0.38 * (0.22) 0.18 (0.29)
CRISIS -0.05 ** (0.02) 0.02 (0.02)

Total (unbalanced) obs.: 222 222 222 222
respectively.
Table 8.24: Capital surplus and deposit overhang results for Spain
The estimates for a capital surplus produce somewhat mixed results. All parameters
are positive as expected. In the presence of time fixed effects, however, they lose
their significance and are a little smaller in absolute terms. On the basis of this
observation, hypothesis H1 still receives support.
To sum up the results for Spain: It must be noted that the deposit-related and deposit
overhang-related hypotheses are most strongly reflected in the data. Strong support
is given to the hypothesis that capital is more important during a crisis than at normal
times. The anticipated impacts of a capital surplus relative to a bank-specific target,
the share of short-term funding and non-interest income are only weakly backed up
by the sample. Size does not seem to be important to Spanish banks during the crisis.
8.2.6. Discussion of inter-country differences and differences between countries and

the euro area
Having examined the results for the individual countries, it makes sense to put them
in a broader context. The differences between the individual countries themselves
and between the countries and the euro area as a whole are therefore discussed in
this section.
Generally speaking, most of the patterns seen for the euro area sample are also
present in the individual country samples. However, the few notable exceptions are
worth highlighting: They are not merely identified in this section, but are also
attributed to differences in the countries' economic situations and the institutional
setups of the various financial systems. In this respect, this section has an
explorative character.
To facilitate comparison, the results of the three specification with time fixed effects
(specifications III, V and VII) are reproduced in the appendix, juxtaposing the four
countries in tabular form. Another summary of the results is given in figure 8.6 in the
following section.
One general reason for the differences between the results for the euro area as a
whole and the individual countries and for the different behavior of important
variables might be that it is harder to properly filter out demand effects at the country
level. Significant elements of the demand companies face in euro area countries are
driven not only by domestic factors. Rather, they also receive demand from many
other European countries (and other countries around the world). For this reason,
focusing only on domestic demand-related variables may be a too narrow approach
to capture all the factors that drive loan demand in a given country.141 This possibility
cannot be ruled out entirely.
141
One could be tempted to propose testing the inclusion of the euro area GDP variable in addition or
as a substitute to the domestic GDP variables in order to capture broader demand factors.
Unreported results, however, suggest that this does not improve the estimates.
A further general reason could be the size of the (sub)samples. Especially in the
case of Spain, the number of results that are not statistically significant may be due
to the relatively low number of cross-sections and observations included in the
sample. In connection with lower average coverage in terms of observations per
bank included in the sample, this may not allow for sufficient variation in the data. To
some extent, this also applies to the Italian sample, for which coverage per bank is
also relatively low. The difference to Spain is that the overall number of observations
is much higher.
Governmental debt in percentage of national GDP

140
120
100 Italy
France
Spain
80 Germany
Percent
60
40
20
0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Year
Source: ECB
Figure 8.4: Government debt as a percentage of national GDP
One interesting observation is that in Italy and Spain deposits seem to be more
relevant to banks' lending response than in other countries. Another is that the
absolute values on deposits are consistently higher in the two countries. The cause
of this peculiarity can probably be found in the fact that the level of government debt
in Italy has been consistently higher during the entire sample period than in the other
countries in the crisis years, as shown in figure 8.4. In Spain, it was the public deficit
that grew more rapidly compared to the other countries (figure 8.5). The imbalances
in public accounts and the dwindling creditworthiness of these states may also have
negatively impacted domestic banks' ability to raise uninsured funding on the
wholesale funding markets, since domestic banks are the most important providers of
public debt. Consequently, the availability of deposits gained importance to the extent
that access to uninsured sources of funding became more difficult.
Interestingly, the same effect is also observable for the deposit overhang. For both
countries, Italy and Spain, the absolute values are higher than for the other countries
and for the euro area sample; the results are also statistically significant. This can be
taken as further evidence of the increased economic significance of deposits in the
two countries, especially during a crisis.
Governmental deficit or surplus in percentage of national GDP

4
-2
Germany
Italy
Percent -4
France
Spain
-6
-8
-10
-12
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Year
Source: ECB
Figure 8.5: Government deficit or surplus as a percentage of national GDP
The same type of consideration may be responsible for the observation that "pure"
capital, measured in terms of the magnitude of absolute values, is more important in
Spain and Italy. These values are higher than in other countries and in the euro area,
not only but especially so in the crisis. As a result, providers of uninsured funding
perceive that substantial holdings of public debt add risk to a bank's portfolio a
factor that banks try to outweigh by higher equity capital ratios.
In the case of Spain, the situation was exacerbated by the overvaluation of house
prices and the crash of the Spanish real estate market during the crisis (see Barrel
and Davis (2008), for example). An explanation consistent with the results of the
estimations is that these economically harmful developments may to some extent

have been anticipated by informed market participants. In this was the case, they
may have provided funds only on condition of an ample capital endowment in order
to cover possible futures losses.
Taken together, these considerations lead to the conjecture that this Spanish
singularity among the analyzed countries led to a structural shift in the importance of
capital during the crisis, an assertion reflected in the higher absolute values for the
coefficients of the capital variable.
Furthermore, this factor could also have overshadowed the impact of size on lending
in Spain. Compared to the euro area sample and to the other countries, the size of a
bank seems to be generally less important (apart from Italy).
A further observation concerns the impact of the share of short-term funding during
the crisis. As pointed out above, a higher-than-average share of short-term funding is
usually associated with a relatively pronounced maturity mismatch. During a crisis,
this can be particularly disadvantageous to banks when it becomes harder to raise
the necessary amounts of uninsured funding on dysfunctional funding markets in
order to maintain the loan portfolio. For Italy, however, the estimated coefficients are
lower (in absolute terms) than for the euro area as a whole, as well as being lower
than in most of the other countries.142 Furthermore, the difference between the crisis
impact and the non-crisis impact is smaller compared to other countries. How can
this be? The solution is probably given by a characteristic of the Italian financial
system: Traditionally, as reported by Ehrmann et al. (2003) (and originally by Borio
(1996)), the proportion of short-term loans is high in Italy.143 This means that even
when it is not easy to borrow on the wholesale funding markets, Italian banks are not
impacted as severely as banks in other countries because, due to shorter maturities,
it is easier to adjust their lending portfolios downward until they match the amount of
funding available at acceptable rates. Hence, the low economic significance of short-
term funding in the crisis for Italian banks may not be overly surprising against the
background of this country's maturity structure for bank loans.
142
The only exception is Germany for which the estimated coefficients are similarly low.
143
According to Ehrmann et al. (2003) this is different for Germany, France and Spain.
8.3 Conclusion Main research hypotheses confirmed 191
8.3. Conclusion Main research hypotheses confirmed
Before discussing all the implications of the results in the next chapter, this section
provides a concise summary of the results. It briefly outlines whether the hypotheses
geared to the euro area as a whole and to the four major euro area countries are
verified on the basis of the empirical analysis.
Verified for Verified for country samples

Hypotheses euro area
sample Germany Italy France Spain
H1
For banks that do not face a deposit overhang, lending depends on
> H2a: The share of uninsured short-term funding
H2 n.a.
> H2c: The share of securities that must be marked to market n.a.

H4 is lower (in terms of the value of the coefficient) compared to
normal times
H5 income on lending is lower (in terms of the value of the coefficient)
compared to normal times
During a crisis, the size of a bank has a more positive impact on
H6
H7 market securities on lending is lower (in terms of the value of the n.a.
H8
impact on lending
H9
lending
Hypothesis very strongly supported Hypothesis not supported

Hypothesis strongly supported n.a. Hypothesis not tested / no indication possible
Hypothesis weakly supported
Figure 8.6: Summary of the results of the hypothesis test
Figure 8.6 provides a graphical overview of the results. Each hypothesis is assessed
for the euro area and for each of the four individual countries on basis of the sign of
the coefficients measured against the expected sign and the statistical significance.
In case of each hypothesis, the strength of validation is indicated by the number of
plus signs, where three plus signs stand for very strong evidence, two plus signs
stand for strong evidence and one plus sign stands for weak evidence. Whenever a
hypothesis is not supported by the data, this indicated by a minus sign.
The assessment of the degree of support that the hypotheses receive (and the
assignment of plus signs or minus signs in figure 8.6 respectively) is not limited to the
results of the standard specifications or specifications used within the context of the
baseline analyses. In fact, the results of all estimated specifications are taken into
account. In case of the euro area sample results of 14 estimated specifications are
included in the assessment. In case of the individual country samples results of
seven estimated specifications are accounted for.144
For the euro area, only hypotheses H2c and H7, both of which involve the share of
securities that have to be marked to market (AFS), are either not verified or not
tested due to data issues. All other hypotheses are supported or strongly supported
by the results of the estimates. This includes, in particular, the role of a surplus or
shortfall relative to a bank-specific, self-imposed target a concept that has not been
studied in the context of bank lending in the euro area so far. The surplus or shortfall
has an impact on the loan supply, irrespective of the economic circumstances
(normality or crisis).
The hypotheses that test whether certain characteristics have a different impact on
lending in the crisis compared to normal periods, hypotheses H3 to H6 and H8 to H9
can be validated on basis of the estimations. The claim that there is a difference
under crisis conditions is thus generally supported for the recent crisis. Among these
characteristics are the capitalization of a bank, the share of short-term funding, the
share of non-interest income, the size of a bank, the share of deposit funding and the
fact of whether or not a bank has an overhang in deposits (over the amount of loans).
Compared with the euro area, the overall picture in the individual countries is quite
similar. As figure 8.6 shows, all hypotheses are (at least weakly) supported for the
four countries analyzed individually, except from H3 and H6. The collective evidence
for H5 for the German sample is a borderline case. 145 The support is particularly
strong for hypotheses H1 and H9, capital surplus and deposit overhang. For most
hypotheses there are only subtle differences across countries. Where bigger
144
See figure 8.1 for an overview of all tested specifications for the euro area sample as well for the
individual country samples.
145
Despite missing significance of the coefficients on defined levels in most cases the crisis
coefficient is consistently lower. Based on this, the assessment "weakly supported" seems to be
just warrantable.
8.3 Conclusion Main research hypotheses confirmed 193
differences exist between the countries, this has been discussed in the previous
section.
This summary concludes the empirical part of the study and opens the way for the
final discussion.
194 9 Final discussion and implications
9. Final discussion and implications
9.1. Overall summary of results
The results of this study imply that the economic significance of certain bank
characteristics in determining the loan supply changed during the recent crisis. Some
characteristics that are less important at normal times gained economic significance
during the crisis, while the impact of other characteristics on lending was amplified or
even reversed.
Figure 4.4 and figure 6.2 provide a framework within which the relevance of certain
bank characteristics can be identified. It was developed by referring to literature on
the bank lending channel and to the theoretical foundations of the loss spiral and the
liquidity spiral. It has been adapted to the bank-centric view of this study. Rigorous
reference to a theoretical framework provides the basis on which the exact effect of
the analyzed bank characteristics on lending has been determined empirically.
Consideration of the crisis period in this context means that this study is, to date, one
of the few that explicitly deals with the determinants of lending during the crisis in the
euro area, thereby contributing to the understanding of the impact and its implications.
The detailed findings of this study for the euro area and for most of the four individual
countries are as follows: 146 As discussed in detail above, almost all financial
institutions target a certain level of equity capital which they judge to be adequate for
a number of reasons. One finding of this study is that a downward deviation from
these self-chosen capital targets has a negative effect on the supply of loans
(cf. hypothesis H1). Its quantitative importance does not depend on whether a target
is missed at normal times or during a crisis. The explanation for this behavior is that
falling short of the self-chosen target implies a capital constraint, in which case a
bank is reluctant to expand its lending and will instead constrain the granting of credit.
The concept of a bank-specific capital target and its use in the context of bank
lending in the euro area is a novelty in the relevant literature landscape.
146
The results for the four countries (Germany, Italy, Spain and France) analyzed individually differ
only slightly from those for the euro area as a whole, as shown in figure 8.6.

9.1 Overall summary of results 195
Banks that need to avail themselves of (uninsured) wholesale funding a need

indicated by the fact that they have no overhang in deposits generally (i.e.
irrespective of the economic situation) find that their lending is impacted significantly
and negatively by their volume of short-term funding (cf. hypothesis H2a). The likely
reason is that a large proportion of short-term funding is perceived by providers of
uninsured market funding as a sign of a greater risk, because a larger share of short-
term funding is accompanied by a greater maturity mismatch. A maturity mismatch
becomes a pressing problem when wholesale funding markets dry out and when the
yield curve is inverted.
A generally heightened perceived riskiness from providers' of uninsured funding point

of view is also likely to be responsible for the significant and negative impact on
credit granting of a higher-than-average share of non-interest income for those banks
that do not have an overhang in deposits (cf. hypothesis H2b). The business that
generates non-interest income is more vulnerable to variations relating to the
business cycle. In an economic downturn, this leads to a steeper decline in income
compared to banks with a more stable business model. By consequence, investors
demand a higher risk premium, which banks pass on in the form of higher lending
rates. This upward shift in the supply curve for bank loans in turn leads to a reduced
growth of the loan portfolio.
One main objective of this study is a better understanding of the recent crisis and
how and in what ways it affected bank lending. Only a very small number of studies
on this issue have been published to date. Hence, the results of empirical analyses
that focus on the crisis period are all the more valuable and can be summarized as
follows: A higher-than-average capital ratio has a more positive impact on lending
during a crisis than at normal times (cf. hypothesis H3). According to the theoretical
foundation for the loss spiral and the margin spiral, the adverse consequences of
these two mechanisms are aggravated by high leverage. A higher capital ratio thus
has the ability to cushion this effect at least to some extent. It is clear that more
capital does not only provide buffer against losses stemming from these two effects
but also against losses resulting from other events. Accordingly, higher capital ratios
have the quality to generally increase the confidence in an institution which is all the
more important during a crisis ("flight to quality").
The issue of the share of short-term funding, which is an indication for a maturity
mismatch, is exacerbated during a crisis when the associated risk materializes
(cf. hypothesis H4): When investors withdraw their funds and funding markets begin
to dry out, rolling over short-term debt becomes much more difficult. This increases
the risk that a bank may need to liquidate long-term assets at short notice and on
unfavorable terms. In such a situation, it is very difficult to expand lending.
A related mechanism is likely to be responsible for the negative impact in relative

terms of a higher-than-average share of non-interest income during the crisis as
opposed to at normal times (cf. hypothesis H5): The greater vulnerability of non-
interest income generating business is no longer a "theoretical" risk, but
consequences in the form of income losses become real and apparent. The
banks hit by the resultant income losses are forced to pay higher risk premiums on
wholesale funding markets, which they then pass on to their borrowers. This results
in a slower growth of the loan portfolio compared to banks with a higher share of
more stable interest income.
During the crisis, larger banks, measured in terms of total assets, reduced their loan
portfolio to a lesser extent (cf. hypothesis H6). In accordance with the theoretical
framework, the likely cause is that bigger banks generally have more advanced
controlling and reporting procedures in place. For providers of uninsured funding, this
results in a greater degree of transparency. In addition, a bank's size is likely to be
correlated with the degree of diversification of income sources and professionalism
(with regard to management capacity and process quality, for example). This should
contribute to a lower risk and a higher probability that the bank can repay borrowed
funds. Both factors are especially important during a crisis, when investment
decisions are all the more challenging. Accordingly, the lower risk premium
demanded by investors allows a bank to borrow at lower interest rates. Logically, the
bank can then lend money on more attractive terms to their clients and will also have
a bigger loan portfolio.
A higher-than-average share of deposits has a positive impact on the loan portfolio

(cf. hypothesis H8). The explanation for this is simply that good access to insured
deposits as a source of funding makes it less necessary to resort to borrowing on
wholesale funding markets which lowers overall funding costs. As a consequence,
9.2 Theoretical contributions 197
the overall funding costs of a bank are lower, so the bank can lend money at lower
rates. This in turn leads to a bigger loan portfolio.
In addition, a very much similar logic applies to banks that have an overhang in
deposits over the amount of loans (cf. hypothesis H9). The deposit overhang acts as
a proxy for the point from which independence of wholesale funding markets is
theoretically possible. 147 It is therefore a more accentuated measure than "pure"
deposits. The results suggest that the overhang dominates the effect of "pure"
deposits to some extent which hints at the notion that independence of uninsured
funding is better captured by the deposit overhang.
The question of what overriding theoretical contribution and practical implications

these results imply is addressed in the following sections.
9.2. Theoretical contributions
The contribution that these results make to literature, especially with regard to the
role of the bank characteristics at normal times and in times of crisis, have already
been discussed in the context of the results of the empirical analysis. This section
therefore summarizes the main contributions and distinguishes between those that
concern the role of bank characteristics and those that relate to literature on the bank
lending channel.
9.2.1. Contributions to research regarding the determinants of bank lending
One important finding concerns the setting of bank-specific capital targets and the
consequences of missing or overshooting these self-imposed thresholds. The idea of
bank-specific capital targets is rooted in a connection between two different research
streams.
The first stream deals with the capital structures of non-financial firms and builds on
the work of Flannery and Rangan (2006), who finds that non-financial firms target
capital ratios to which they partially adjust if they are missed, and the study by
Lemmon et al. (2008), who add that, besides the explanatory power of certain firm
147
Even if an overhang does not mean that a bank needs no wholesale funding at all apart from
loans, there are other assets that also have to be funded the necessary amount should,
nevertheless be significantly reduced.
characteristics, the bulk of the variation in non-financial firms' leverage ratios can be
explained by time-invariant fixed effects.
The second stream addresses target capital ratios for banks, albeit in the form of
minimum capital ratios with reference to a regulatory capital requirement. Hancock
and Wilcox (1994) and Hancock and Wilcox (1998) use capital shortfall or surplus
variables measured relative to regulatory not self-chosen standards.
In connecting these two streams, Gropp and Heider (2010) find that, like non-
financial corporations, banks too target individual capital ratios. These can be
explained by bank-specific fixed effects and by the variation in certain bank
characteristics. Along the same lines, Berrospide and Edge (2010) and Francis and
Osborne (2009) model bank target capital ratios for the US and the UK respectively.
For the euro area and context of bank lending, however, this has not yet been done.
In this respect, using the concept of a capital shortfall relative to a bank-specific
target and finding that this has an impact on a bank's credit growth allows this study
to contribute to the related body of literature. In addition, the result can be interpreted
as general confirmation of the notion that banks target individual capital ratios that
are usually well above regulatory minimum rates.
Another important strand of findings in this study is related to the impact of certain
bank characteristics especially during the recent banking crisis as opposed to normal
times. To date, only the work of Gambacorta and Marques-Ibanez (2011) is remotely
comparable although it is not exclusively geared to the euro area but to banks from a
broader set of European countries, the UK and the US. 148 Without repeating the
detailed overall summary of results given in the previous section, many of the effects
found by Gambacorta and Marques-Ibanez are confirmed by this study, while others
such as the impact of a deposit overhang on lending are put forward for the first
time in the present study.
In addition, this study contributes to literature by separately analyzing the four biggest
euro area countries, namely Germany, Italy, France and Spain. As demonstrated
above, subtle differences exist with respect to the effect of some bank characteristics
148
Other euro area studies that deal with bank lending in the crisis context either have a narrowed
geographical scope (e.g. Albertazzi and Marchetti (2010) or Bonaccorsi di Patti and Sette (2012)
analyze the situation in Italy only) or a different topical focus (e.g. Brei et al. (2013) concentrate on
the effect of rescue packages on bank lending.
on lending during the crisis. These differences can be attributed to the specific
economic situations in these countries during the crisis, or to the respective
institutional contexts.
A further contribution with respect to possible ways in which a bank's capitalization

can be measured is that the difference between the tested options is only marginal.
Given the strong correlation, it is not surprising that the results are very similar
irrespective of whether total equity or tangible common equity is used. Subject to
certain limitations, this also applies to the Tier 1 capital ratio: Since the available time
series on the Tier 1 ratio is limited because it is a relatively new concept, it was not
possible to test this matter exhaustively in the empirical analyses. However, strong
correlation to the two other capital measures suggests that the difference in the
overall results is, again, not substantial.
9.2.2. Contributions to bank lending channel-related research
9.2.2.1. General contributions
In addition to the analysis of the impact of certain bank characteristics on bank

lending, which is the focus of this study, it also contributes to literature relating to the
bank lending channel, from which important aspects of the theoretical background
have been derived.
In chapter 4, the traditional theoretical justification for a bank lending channel is

reconsidered in light of developments regarding the integration of financial markets
and the importance of financial innovation. According to the traditional view, the
central bank exerts control over the volume and terms of available reserves by
changing key interest rates. In so doing, it affects the creation of deposits and, at the
same time, banks' ability to grant loans. This view has been criticized for not
adequately reflecting modern central banks' operational framework and for omitting
the importance of wholesale funding markets as the marginal funding source today:
The ECB, for example, supplies any amount of reserves demanded by the banking
system, not just some amount that the ECB itself determines in advance. It follows
that the direction of causality between reserves and bank lending has not been
established correctly.
Instead, according to the new view on the bank lending channel, the ECB changes
key interest rates in order to affect the terms and availability of uninsured forms of
funding, which have evolved into an economically significant pillar of banks' funding
strategies. In this way, the ECB impacts the overall funding costs/terms for banks,
which are then reflected in banks' own lending rates and the growth of their credit
portfolio.
Having already been put forward by Disyatat (2011), these considerations are
consolidated in this study and brought into a single framework that, to date, has
never been explicitly expressed. This framework, captured in figure 4.4 and figure 6.2,
may not be complete and will probably be extended in future. However, it provides a
benchmark framework into which bank characteristics that have not yet been
considered can be integrated. It connects the volume of required funding and its cost
to the other conditions that must be met in order for the bank lending channel to exist.
As this study has proven, the underlying framework is valid and applicable even in
the context of a crisis.
9.2.2.2. Monetary policy indicator
A further contribution this study makes with respect to literature on the bank lending
channel concerns the way in which the monetary policy indicator is modeled. The
general aim is to find a representation for the impact that monetary policy measures
have on bank lending (monetary policy stance). The most widespread and best-
established approach is to use the Eonia or the 3-month Euribor rate. The results
confirm the adequacy of these two measures.
Mindful of the turbulence on interbank markets during the recent crisis, which raised
questions about the significance of the Eonia or Euribor for this period, a novel
approach has been adopted: Its main element is the use of the Euribor-OIS spread.
As explained in section 7.4.1, the approach focuses on reflecting the risk that is
involved when banks lend to other banks. It should therefore be a very important
measure for banks' credit supply without being blurred by the turbulence on financial
markets. A second element of the novel approach is the inclusion of a measure that
captures the non-standard monetary policy measures adopted by the ECB during the
crisis, as suggested by Gambacorta and Marques-Ibanez (2011). The non-standard
measures have also been of relevance to the supply of loans, because they helped
overcome the shortage in central bank reserves that was caused by the dysfunction
of interbank markets. The non-standard measures are represented by the ECB's
assets as a share of nominal GDP.
The results show that both measures are relevant to the explanation of banks' loan
supply. Accordingly, this study contributes by proposing the Euribor-OIS spread as a
novel and theoretically sound concept that has proven to work well in practice.
Moreover, the significance of the use of non-standard measures and its positive
impact on lending is confirmed.
9.2.2.3. Disentanglement of loan supply and loan demand
One important issue in the context of bank lending and a major empirical challenge
when trying to identify the impact of bank characteristics on lending is the
disentanglement of supply and demand factors. Theoretically, an observed change in
the volume of loans recorded on banks' balance sheets could be attributed to shifts in
either the loan supply or demand for loans. This problem occurs whenever changes
in demand factors coincide with changes in the bank characteristics that are
postulated to affect the supply of loans. As outlined in section 7.4.3.1, the correlation
between loan demand and the factors that have an impact on the supply of loans
works through two main patterns that are linked to business cycle fluctuations.
In this study, two main approaches are made to address this issue: Following the
majority of related studies, the first approach consists of a combination of GDP and
the GDP deflator as the variable that captures inflation. Furthermore, all major
specifications are also tested using time fixed effects in addition. The use of time
fixed effects involves the use of time dummies for each year of the sample period
and helps to capture demand effects that might not have been accounted for with the
GDP and inflation variables. However, as the results show, the difference between
the estimates with and without time fixed effects is relatively small. This holds for the
euro area sample as well as for the country subsamples. Taken together, this is a
strong indication that loan demand effects are already well captured by the
combination of GDP and inflation.
The second approach involves using the results of the euro area bank lending survey,
in the course of which senior loan officers at selected euro area banks are, for
example, asked about past and expected future developments in loan demand. In
theory, this is a very sound concept, because it gathers information directly at the
point where demand for loans should be most apparent: at the banks themselves.
Unfortunately, however, in practice the use of a variable based on the answers to the
lending survey does not work well: The sign for the lending survey variable is not as
expected, and the results are generally unsatisfactory.
As discussed above, a combination of reasons may be responsible for this finding.

First, the ECB carried out the bank lending survey for the first time in 2003. This
leads to a loss of observations that might in some way negatively impact the results,
although an obvious rationale is not easy to establish. Second, the set of 90
participating banks might not be large enough to draw a representative picture of
loan demand for the entire euro area. Third, the fact that the composition of the euro
area has changed might induce a small bias, because the countries from which the
surveyed loan officers originate are not exactly the same countries that are part of
the sample. Fourth, it could be the case that the loan officers' answers tend to exhibit
procyclical behavior. In "good" times, for example, slightly favorable developments
may already be overstated as a sign of an upcoming boom. This could lead to overly
optimistic answers compared to actual loan demand. Fifth, the lending survey is
carried out on a quarterly basis and the answers distinguish between the different
customer groups that demand loans (i.e. enterprise loans, consumer credit and
housing loans). The construction of a single lending survey variable thus requires
some form of data compilation to make it applicable and to match it to the annual
structure of the data. It is possible that there is a superior compilation procedure,
although different approaches have been made and tested, each with unsatisfactory
results.
Taking everything into account, the conclusion, in confirmation of the approach

adopted by the majority of related studies, is that using a combination of GDP and
inflation together with time fixed effects works well in capturing demand for loans.
Moreover, although theoretically sound, the use of a variable based on the euro area
bank lending survey does not work well in practice.
9.3. Implications for bank management
The findings of this dissertation have several implications for the management of
banks. Since the lending business is banks' strongest pillar overall and the most
9.3 Implications for bank management 203
reliable source of income for banks, having an understanding of the factors that
determine it is an important piece of knowledge. In particular with regard to the recent
crisis, some aspects that have not been accounted for in the past are worth
considering.
The first aspect is the share of uninsured short-term funding in the overall funding
mix. Broad access to wholesale funding markets may look attractive at first sight. In
fact, however, banks with a large proportion of short-term funding run a higher
business risk, because this practice is associated with maturity mismatches. The
greater the mismatch between maturities on the asset and liabilities sides, the
greater the risk premium that investors demand especially in a crisis. This is
because investors anticipate the higher risk that a bank may be unable to roll over its
debt when turbulence hits the funding markets, causing them to tend to dry out. In
the worst case, a bank might not only be forced to constrain its lending but also to
sell assets under fire-sale conditions at prices considerably below their book value.
This could lead to a situation in which a bank finds itself plunged into the loss and the
margin spiral. The bottom line is that a bank is well advised not to "play the yield
curve" too hard, and to be aware of the risks that such a business model entails.149
This leads to another funding-related implication of the results: Good access to

insured retail deposits turned out to be most valuable during the crisis. Banks that
have a large share of deposit funding might look "boring", like many savings and
cooperative banks. Nevertheless, this kind of bank managed to master the crisis
better than other banks for which higher lending growth was a clear sign. On this
basis, bank managers should consider strengthening the acquisition of retail deposits,
because this is cheap on the one hand and helps a bank to avoid being hit too hard
by adverse developments on the wholesale funding markets in a crisis on the other
hand.
In addition, a large share of non-interest income has turned out to negatively impact
the credit supply of a bank. While it seems attractive to tap non-interest related
sources of income because no capital is required for this kind of business, which
149
German bank Hypo Real Estate provides a striking example in a negative sense of the risks
inherent in a business model based on the transformation of maturities (see Dettmer and Weiland
(2009) and Siemers (2009)): After it was bailed out in October 2008, the bank was completely
taken over by the German Financial Markets Stabilisation Fund (SoFFin) in April 2009.
strengthens the profitability of an institution under normal business circumstances,

this practice had a negative effect during the crisis. Because these income sources
are more prone to business cycle fluctuations and because non-interest income
declined significantly, investors that lent funds to these banks demanded a higher
risk premium. This results in higher interest rates for the banks' borrowers and,
consequently, less credit growth. What is more, it usually causes high expenses for
personnel and materials (especially IT expenses) to serve the business that
generates non-interest income. When this business declines, capacity is left unused,
leaving banks with a hefty cost block that is hard to reduce quickly. Bank managers
should therefore carefully rethink the balance between interest and non-interest
income, even if some consultants never tire of advising them to go for a larger share
of the latter.
The observation that size has a positive impact on the loan portfolio does not
necessarily equate with a call for internal or external growth. A more general
interpretation, according to which size is a proxy for transparency and business risk,
is needed: The more transparent a bank's controlling and reporting procedures are
the more "trust" investors will place in a bank and the lower the risk premium they will
demand by way of compensation. The implication for bank managers is therefore that
investors should explicitly be targeted with company communication to give them as
transparent a picture as possible. A similar logic applies to the size of a bank as a
proxy for the degree of income sources and professionalism (e.g. regarding process
quality handling of risk management etc.). The higher the degree of diversification,
the better the quality of processes and the better staff is trained the more positive
should this impact the risk perception of a bank from an investor's point of view. Bank
managers should therefore evaluate room for improvement particularly in these areas.
This is especially important against the background of a crisis.
Last but not least, one probably obvious piece of advice is for a bank to strengthen its
equity capital ratio in order to support its lending business. More capital has a
positive effect both outside of a crisis and during a crisis, with a slightly stronger
effect in the latter case. Due to the design of the empirical analysis, this is clearly the
impact that capital has on lending, and not the effect that more lending requires more
9.4 Implications for monetary policymakers 205
capital.150 Higher capital ratios send a signal to investors that the problem of moral
hazard and asymmetric information is reduced. They also serve as a buffer against
losses. This is positively reflected in the risk premium that investors demand.
9.4. Implications for monetary policymakers
This study does not have implications only for the management of banks, but also for
monetary policymakers. Although references to monetary policy are not the focus of
the empirical analyses, two main points are worth mentioning.
The first regards the non-standard measures taken by the ECB during the recent
financial crisis. In light of the risk that some banks could lose their access to the
interbank markets which would have resulted in substantially higher rates for these
banks and generally higher money market rates the ECB responded in particular by
providing an unlimited supply of central bank money at a fixed rate (full-allotment
policy), and by expanding the scope of securities that are accepted as collateral. If
the risks had materialized during a protracted period and the ECB had not done
anything about it, this would not only have impaired the efficiency of monetary policy
(by muting the bank lending channel), but would also have had serious negative
consequences for the ability of banks to grant credit, not to mention the effects for the
real economy.
As the results of the empirical analysis show, the ECB's non-standard measures had
a positive impact on bank lending and were effective in countering the threats
stemming in particular from distortions in the interbank market. The implication for
monetary policymakers is thus that, in a comparable situation in future, it seems likely
that these measures would again be successful.
The second implication concerns the importance of non-bank financial intermediaries

and their possible consequences for the effectiveness of monetary policy. In
particular, it has been argued that the importance of the bank lending channel may
have been eroded by the emergence and substantial growth of non-bank financial
150
See 7.5 for the respective details on the estimation method. The endogeneity issue not only of
capital but also of other variables is addressed by using lagged values of the explanatory variables
in all estimations and by the employment of the GMM methodology that involves an instrumental
variable procedure.
intermediaries. These thoughts have, at least, been suggested by Kashyap and Stein
(1994) and Ashcraft (2006). Such a view relies very much on the "traditional view" of
the bank lending channel, according to which the central bank influences lending by
exercising control over the terms and availability of reserves. Non-bank financial
intermediaries do not resort to insured retail deposits at all and are not part of reserve
requirements schemes. Apparently, however, they have gained importance in terms
of their capacity to provide non-bank credit and in terms of total assets. This has led
some scholars to conclude that the influence of monetary policy on the real economy
via banks has declined.
Although this contention is not tested empirically in the current study, the theoretical
framework provided in the context of the "new view" of the bank lending channel
strongly indicates that this conclusion is premature. The new view focuses on
wholesale funding, which has taken the place of insured retail deposits as a main
funding source. There is hardly any reason to believe that the mechanism according
to which monetary policy can affect the terms on which wholesale funding is available
should not also apply to non-bank financial intermediaries. Non-banks too have to
raise funds on the funding markets, and the cost depends on their overall financial
condition, their riskiness as perceived by investors, and so on. These costs will then
be passed on to borrowers. Ultimately, there is good reason to assume that the
theoretical framework outlined in section 4.2 is helpful in explaining how funding
costs are determined not only for banks, but also for non-bank financial
intermediaries.
For monetary policymakers, the implication is that the bank lending channel has not
lost its significance. On the contrary, the likely applicability of the theoretical
framework to also non-banks may actually have increased its significance.
9.5. Implications for the discussion of banking supervision
Although the present study is not geared toward contributing to the ongoing debate
surrounding banking regulation and macroprudential supervision, there is one result
in particular that has implications for this field. The aim of macroprudential
supervision is to strengthen the robustness of the financial and banking system as a
whole in its role as a provider of credit by going beyond the perspective of the
9.6 Limitations and outlook 207
individual bank and instead considering the economy-wide effects of regulatory and
supervisory measures.151
In this context, many instruments and tools have been proposed (see Hannoun
(2010), for example). Two of the more prominent ones are limiting leverage and
achieving higher capital ratios, depending to some extent on the business cycle
("counter-cyclical capital buffers)".
One important result of this study is the indication that banks expand their lending
only if their capital endowment is above their self-chosen capital target. Consequently,
when higher capital requirements are introduced too abruptly this might lead to the
situation that banks are instantly constrained by their own capital targets. This calls
for a very careful introduction of new capitalization rules because banks will only
continue to grant credit as long as banks' capital endowments are sufficiently strong.
9.6. Limitations and outlook
This last section provides a summary of issues that could not be considered in this
study, as well as pointing out suggestions and possible directions for future research.
One first issue regards the Bankscope database maintained by the Bureau van Dijk
in cooperation with Fitch Solutions as the data source. Bankscope covers a smaller
number of banks than the number claimed by the ECB.152 However, as already noted,
the studies that compare the results obtained on the basis of a sample from
Bankscope with results based on a sample from national banking authorities or
Eurosystem datasets report neither systematic nor substantial differences. There is
therefore no reason to conclude that the choice of Bankscope data leads to any bias.
A second issue has to do with the question whether capitalization (understood as an

economic concept) is adequately proxied. There are different ways to measure a
bank's capital (adopting a regulatory focus, for example, or using a broader approach
that also considers generally liable equity capital). Due to the fact that regulation-
151
For a literature review on macroprudential supervision, see Galati and Moessner (2012). Bianchi
and Mendoza (2011) provide a study that shows how positive externalities can emerge by
adopting a macroprudential perspective.
152
A constantly updated record of the number of monetary financial institutions is provided by the
ECB online at http://sdw.ecb.europa.eu.
inspired measures such as the Tier 1 capital ratio are relatively new and that, hence,
the data availability is not yet sufficient, these measures could not be compared and
contrasted with the total equity or common tangible equity in the course of this study.
Accordingly, the question of whether using the Tier 1 capital ratio would have made a
greater difference cannot be answered.
Another measurement issue regards construction of the capital surplus variable. The
use of a variable that is the result of an estimation is associated with a higher level of
uncertainty concerning the accuracy of the variable. Despite the careful selection of
variables that have explanatory power for the capital target, one could think of other
variables that might improve this accuracy. Although using the surplus variable does
not involve any conspicuously negative results, some residual doubt remains whether
all banks' capital targets are estimated as accurately as possible.
A further limitation of this study is the fact that, due to the poor availability of data
about securities that must be marked to market, their influence on lending could not
be clarified empirically. It is hoped that more light will be shed on this issue in future
as soon as enough data becomes available.
A related issue is the impact on lending of a bank's affiliation to a wider bank network.
A number of studies report a difference between banks that are thus affiliated and
those that are not. Unfortunately, due to the structure of the data, this information
was not available for the present study.
Worthy of additional consideration could be the fact that the non-standard monetary
policy measures adopted by the ECB during the crisis might not be fully captured by
the ratio of the ECB's assets to total euro area GDP. There is no question that other,
qualitative measures exist too, one example being the acceptance of a broader range
of assets that serve as collateral. Although this places a limitation on the
interpretation of the corresponding coefficient, it must be noted that the qualitative
measures coincided to a large extent with the more quantitative measures. Hence,
one could conclude that the non-standard measures as a whole might not be too
badly proxied by the NSM variable.
A number of suggestions for future research emerge from this study. What has not
been addressed is the impact of bank characteristics on different loan classes (e.g.
mortgage loans, consumer credit and enterprise loans), as this distinction is only
available when drawing on data collected via national banking authorities. Especially
9.6 Limitations and outlook 209
for the US, a number of studies differentiate between the loan groups, e.g. Kashyap
and Stein (2000), Kishan and Opiela (2000) and Kishan and Opiela (2006). However,
this makes the task of disentangling loan supply and loan demand even more difficult.
For example, levels of monetary interest rates that are perceived as relatively low
compared to historical standards lead to a temporary boom in, say, real estate
investments and the corresponding loan classes. When trying to filter this demand-
side effect for certain asset classes using only broad measures of aggregate demand,
the results must necessarily be weak. In addition, when the trigger for this demand-
side effect (low interest rates in this example) coincides with changes in the loan
supply (e.g. due to the better availability of uninsured funding), the identification
problem becomes even more challenging. Altogether, analyzing the different loan
classes is an interesting field for further research. The identification of a bank lending
channel, though, is even more challenging and has yet to be completed satisfactorily.
A different possible avenue for further research could focus even more closely on the
differences between individual countries. Probably due to the integration of Europe's
financial markets, the fact that differences in the characteristics of individual euro-
area countries might well cause the same bank characteristics to have a dissimilar
impact on lending seems to have fallen into oblivion. Some studies, such as that of
Gambacorta and Marques-Ibanez (2011), control for these factors by using country
dummies. However, the present study suggests that country-specific differences can
be attributed to economic particularities or to institutional circumstances in the
individual countries. A more thorough understanding of the exact causes of the
differences could enable better forecasts of the impact of monetary policy in the
individual countries, or indeed in the euro area as a whole.
A very interesting research enterprise would be to empirically validate the notion that
the significance of the bank lending channel has not been eroded by financial
innovation, the integration of financial markets and the rapid growth of non-bank
financial intermediaries, as has been argued above on theoretical grounds.
Unfortunately, poor data availability regarding this largely unregulated aspect of the
financial system still imposes limits on this kind of project.
Finally, the question whether the findings in this study regarding the recent crisis also
apply to other historical banking crises would be worth answering in future. This
might contribute to more universal recommendations for bank managers, monetary
policymakers and banking regulators/supervisors, helping them to equip themselves

better for and avoid the most damaging consequences of the next crisis.
References 211
References
Adrian, Tobias; Shin, Hyun Song (2008a): Financial intermediaries, financial stability,
and monetary policy, Staff Reports 346, Federal Reserve Bank of New York.
Adrian, Tobias; Shin, Hyun Song (2008b): Liquidity, monetary policy, and financial
cycles, Current Issues in Economics and Finance, 14 (1), 17.
Adrian, Tobias; Shin, Hyun Song (2010a): Liquidity and leverage, Journal of
Financial Intermediation, 19 (3), 418437.
Adrian, Tobias; Shin, Hyun Song (2010b): The Changing Nature of Financial
Intermediation and the Financial Crisis of 20072009, Annual Review of
Economics, 2 (1), 603618.
Albertazzi, Ugo; Marchetti, Domenico J. (2010): Credit supply, flight to quality and
evergreening: an analysis of bank-firm relationships after Lehman, Temi di
discussione (Working papers) 756, Banca d'Italia.
Alper, Koray; Hlag, Timur; Keles, Grsu (2012): An Empirical Study on Liquidity
and Bank Lending, Working Paper 12/04, Central Bank of the Republic of Turkey.
Altunbas, Yener; De Bondt, Gabe; Marques-Ibanez, David (2004): Bank capital,
bank lending and monetary policy in the euro area, Kredit und Kapital, 37 (4),
443464.
Altunbas, Yener; Fazylov, Otabek; Molyneux, Philip (2002): Evidence on the bank
lending channel in Europe, Journal of Banking & Finance, 26 (11), 20932110.
Altunbas, Yener; Gambacorta, Leonardo; Marques-Ibanez, David (2009):
Securitisation and the bank lending channel, European Economic Review, 53 (8),
9961009.
Altunbas, Yener; Gambacorta, Leonardo; Marques-Ibanez, David (2010): Bank risk
and monetary policy, Journal of Financial Stability, 6 (3), 121129.
Angeloni, Ignazio; Kashyap, Anil K.; Mojon, Benoit; Terlizzese, Daniele (2003):
Monetary Transmission in the Euro Area: Does the Interest Rate Channel Explain
it All?, NBER Working Papers 9984, National Bureau of Economic Research.
Arellano, Manuel; Bond, Stephen (1991): Some Tests of Specification for Panel
Data, Review of Economic Studies, 58 (2), 277297.
Arellano, Manuel; Bover, Olympia (1995): Another look at the instrumental variable
estimation of error-components models, Journal of Econometrics, 68 (1), 2951.
Ashcraft, Adam B. (2006): New Evidence on the Lending Channel, Journal of
Money, Credit and Banking, 38 (3), 751775.
Barrel, Ray; Davis, E. Philip (2008): The Evolution of the Financial Crisis of 2007-8,
National Institute Economic Review 206, National Institute of Economic and Social
Research (UK).
Basel Committee on Banking Supervision (October 27, 1998): Instruments eligible for
inclusion in Tier 1 capital (press release). Basel.
Basel Committee on Banking Supervision (September 12, 2012): Group of
Governors and Heads of Supervision announces higher global minimum capital
standards (press release). Basel.

DOI 10.1007/978-3-658-07175-2, Springer Fachmedien Wiesbaden 2015
212 References
Behr, Andreas (2003): A comparison of dynamic panel data estimators: Monte Carlo
evidence and an application to the investment function, Discussion paper 05/03,
Deutsche Bundesbank.
Beirne, John; Dalitz, Lars; Ejsing, Jacob; Grothe, Magdalena; Manganelli, Simone;
Monar, Fernando; Sahel, Benjamin; Suec, Matja; Tapking, Jens; Vong, Tana
(2011): The impact of the Eurosystem's covered bond purchase programme on
the primary and secondary markets, Occasional Paper Series 122, European
Central Bank.
Berger, Allen N.; DeYoung, Robert; Flannery, Mark J.; Lee, David; ztekin, zde
(2008): How Do Large Banking Organizations Manage Their Capital Ratios?,
Journal of Financial Services Research, 34 (2), 123149.
Berger, Allen N.; Udell, Gregory F. (1994): Do Risk-Based Capital Allocate Bank
Credit and Cause a "Credit Crunch"' in the United States?, Journal of Money,
Credit and Banking, 26 (3), 585628.
Berg, Japser; van Rixtel, Adrian; Ferrando, Annalisa; De Bondt, Gabe; Scopel, Silvia
(2005): The bank lending survey for the euro area, Occasional Paper Series 23,
European Central Bank.
Bernanke, Ben S. (1983): Nonmonetary Effects of the Financial Crisis in
Propagation of the Great Depression, American Economic Review, 73 (3), 257
276.
Bernanke, Ben S. (2007): The financial accelerator and the credit channel, Speech
given at the 'The Credit Channel of Monetary Policy in the Twenty-first Century'
conference, Federal Reserve Bank of Atlanta, Atlanta, Georgia, June 15, 2007.
Bernanke, Ben S.; Blinder, Alan S. (1988): Credit, Money, and Aggregate Demand,
American Economic Review, 78 (2), 435439.
Bernanke, Ben S.; Blinder, Alan S. (1992): The Federal Funds Rate and the
Channels of Monetary Transmission, American Economic Review, 82 (4), 901
921.
Bernanke, Ben S.; Gertler, Mark (1989): Agency Costs, Net Worth, and Business
Fluctuations, American Economic Review, 79 (1), 1431.
Bernanke, Ben S.; Gertler, Mark (1995): Inside the Black Box, Journal of Economic
Perspectives, 9 (4), 2748.
Bernanke, Ben S.; Gertler, Mark; Gilchrist, Simon (1996): The Financial Accelerator
and the Flight to Quality, The Review of Economics and Statistics, 78 (1), 115.
Bernanke, Ben S.; Gertler, Mark; Gilchrist, Simon (1999): The financial accelerator in
a quantitative business cycle framework, in: Handbook of Macroeconomics, ed.
by Taylor, John B.; Woodford, Michael. Amsterdam: Elsevier, 13411393.
Bernanke, Ben S.; Lown, Cara S. (1991): The Credit Crunch, Brookings Papers on
Economic Activity, 22 (2), 205248.
Berrospide, Jose M.; Edge, Rochelle M. (2010): The Effects of Bank Capital on
Lending: What Do We Know, and What Does It Mean?, International Journal of
Central Banking, 6 (34), 554.
Bhattacharya, Sudipto; Nyborg, Kjell G. (2010): Bank Bailout Menus, CEPR
Discussion Papers 7906, Centre for Economic Policy Research.
References 213
Bianchi, Javier; Mendoza, Enrique G. (2011): Overborrowing, Financial Crises and

'Macro-prudential' Policy?, IMF Working Papers 11/24, International Monetary
Fund.
Blinder, Alan S. (1984): Notes on the Comparative Statics of a Stiglitz-Weiss Bank,
Mimeo.
Blinder, Alan S. (1987): Credit Rationing and Effective Supply Failures, The
Economic Journal, 97 (386), 32752.
Blundell, Richard; Bond, Stephen (1998): Initial conditions and moment restrictions
in dynamic panel data models, Journal of Econometrics, 87 (1), 115143.
Bofinger, Peter (2001): Monetary Policy: Goals, Institutions, Strategies, and
Instruments. New York: Oxford University Press.
Bonaccorsi di Patti, Emilia; Sette, Enrico (2012): Bank balance sheets and the
transmission of financial shocks to borrowers: evidence from the 2007-2008 crisis,
Temi di discussione (Working papers) 848, Banca d'Italia.
Bond, Stephen (2002): Dynamic panel data models: a guide to micro data methods
and practice, Portuguese Economic Journal, 1 (1), 141162.
Borio, Claudio (1996): Credit Characteristics and the Monetary Policy Transmission
Mechanism in Fourteen Industrial Countries: Facts, Conjectures and Some
Econometric Evidence, in: Monetary Policy in a Converging Europe, ed. by Alders,
Koos J.; Koedijk, Kees K.; Kool, Clemens C.; Winder, Carlo C. Dordrecht,
Netherlands: Kluwer Academic Publishers, 77-115.
Borio, Claudio; Disyatat, Piti (2010): Unconventional Monetary Policies: An
Appraisal, The Manchester School, 78 (s1), 5389.
Borio, Claudio; Zhu, Haibin (2012): Capital regulation, risk-taking and monetary
policy: a missing link in the transmission mechanism?, Journal of Financial
Stability, 8 (4), 236251.
Brner, Christoph J.; Grichnik, Dietmar; Reize, Frank (2010):
Finanzierungsentscheidungen mittelstndischer Unternehmen - Einflussfaktoren
auf die Kapitalstruktur deutscher KMU, Zeitschrift fr betriebswirtschaftliche
Forschung, 62 (3), 227250.
Brei, Michael; Gambacorta, Leonardo; Peter, Goetz von (2013): Rescue packages
and bank lending, Journal of Banking & Finance, 37 (2), 490505.
Brunnermeier, Markus K. (2009a): Deciphering the Liquidity and Credit Crunch
2007-2008, Journal of Economic Perspectives, 23 (1), 77100.
Brunnermeier, Markus K. (2009b): Financial Crises: Mechanisms, Prevention and
Management, in: Macroeconomic stability and financial regulation: Key issues for
the G20, ed. by Dewatripont, Mathias; Freixas, Xavier; Portes, Richard. London:
Centre for Economic Policy Research, 91103.
Brunnermeier, Markus K.; Pedersen, Lasse Heje (2009): Market Liquidity and
Funding Liquidity, Review of Financial Studies, 22 (6), 22012238.
Bun, Maurice J.G.; Windmeijer, Frank (2010): The weak instrument problem of the
system GMM estimator in dynamic panel data models, Econometrics Journal, 13
(1), 95126.
214 References
Campello, Murillo (2002): Internal Capital Markets in Financial Conglomerates,

Journal of Finance, 57 (6), 27732805.
Carlson, Mark; Shan, Hui; Warusawitharana, Missaka (2013): Capital ratios and
bank lending: a matched bank approach, Journal of Financial Intermediation,
22 (4), 663687.
Carlstrom, Charles T.; Fuerst, Timothy S. (2001): Monetary policy in a world without
perfect capital markets, Working Paper 01-15, Federal Reserve Bank of
Cleveland.
Cecchetti, Stephen G. (1995): Distinguishing theories of the monetary transmission
mechanism, Federal Reserve Bank of St. Louis Review (May/June), 8397.
Ciccarelli, Matteo; Maddaloni, Angela; Peydro, Jose-Luis (2010): Trusting the
bankers: a new look at the credit channel of monetary policy, Working Paper 1228,
Congleton, Roger D. (2009): On the political economy of the financial crisis and
bailout of 20082009, Public Choice, 140 (3), 287317.
De Bondt, Gabe; Maddaloni, Angela; Peydro, Jose-Luis; Scopel, Silvia (2010): The
euro area Bank Lending Survey matters - empirical evidence for credit and output
growth, Working Paper 1160, European Central Bank.
Del Giovane, Paolo; Eramo, Ginette; Nobili, Andrea (2011): Disentangling demand
and supply in credit developments: A survey-based analysis for Italy, Journal of
Banking & Finance, 35 (10), 27192732.
Dell'Ariccia, Giovanni; Igan, Deniz; Laeven, Luc (2012): Credit Booms and Lending
Standards: Evidence from the Subprime Mortgage Market, Journal of Money,
Dettmer, Markus and Weiland, Severin (2009): BaFin-Prfbericht: Wie die HRE in
die Katastrophe schlitterte, Spiegel Online,
http://www.spiegel.de/politik/deutschland/bafin-pruefbericht-wie-die-hre-in-die-
katastrophe-schlitterte-a-627127.html. Accessed September 30, 2013.
Diamond, Douglas W.; Dybvig, Philip H. (1983): Bank Runs, Deposit Insurance, and
Liquidity, Journal of Political Economy, 91 (3), 401419.
Disyatat, Piti (2004): Currency crises and the real economy: The role of banks,
European Economic Review, 48 (1), 7590.
Disyatat, Piti (2011): The Bank Lending Channel Revisited, Journal of Money,
Douglas, Diamond W.; Rajan, Raghuram G. (2001): Liquidity Risk, Liquidity Creation,
and Financial Fragility: A Theory of Banking, Journal of Political Economy, 109 (2),
287327.
Drehmann, Mathias; Gambacorta, Leonardo (2012): The effects of countercyclical
capital buffers on bank lending, Applied Economics Letters, 19 (7), 603608.
Ehrmann, Michael; Gambacorta, Leonardo; Martinez-Pages, Jorge; Sevestre, Patrick;
Worms, Andreas (2003): Financial systems and the role of banks in monetary
policy transmission in the Euro area, in: Monetary policy transmission in the Euro
area: A study by the Eurosystem Monetary Transmission Network, ed. by Angeloni,
References 215
Ignazio; Kashyap, Anil K.; Mojon, Benoit. Cambridge: Cambridge University Press,
235269.
Ehrmann, Michael; Worms, Andreas (2001): Interbank lending and monetary policy
transmission - evidence for Germany, Working Paper 73, European Central Bank.
Ehrmann, Michael; Worms, Andreas (2004): Bank Networks and Monetary Policy
Transmission, Journal of the European Economic Association, 2 (6), 11481171.
European Central Bank (21 November, 2002): Bank lending survey for the euro area
(press release). Frankfurt am Main.
European Central Bank (2009): Recent Developments in the Balance Sheets of the
Eurosystem, the Federal Reserve System and the Bank of Japan, ECB Monthly
Bulletin, October (-), 8194.
European Central Bank (2010a): Financial Stability Review December 2010.
Frankfurt am Main: European Central Bank.
European Central Bank (2010b): The ECB's monetary policy stance during the
financial crisis, ECB Monthly Bulletin, January (-), 6371.
European Central Bank (2011a): The ECB's non-standard measures - impact and
phasing-out, ECB Monthly Bulletin, July (-), 5569.
European Central Bank (2011b): The monetary policy of the ECB. Frankfurt am Main:
Favero, Carlo A.; Giavazzi, Francesco; Flabbi, Luca (1999): The Transmission
Mechanism of Monetary Policy in Europe: Evidence from Banks' Balance Sheets,
NBER Working Papers 7231, National Bureau of Economic Research.
Fiore, Fiorella de; Uhlig, Harald (2005): Bank finance versus bond finance - what
explains the differences between US and Europe?, Working Paper 547, European
Central Bank.
Fisher, Irving (1928): Money illusion. New York: Adelphi.
Fisher, Irving (1933): The Debt-Deflation Theory of Great Depressions,
Econometrica, 1 (4), 337357.
Flannery, Mark J.; Rangan, Kasturi P. (2006): Partial adjustment toward target
capital structures, Journal of Financial Economics, 79 (3), 469506.
Francis, William; Osborne, Matthew (2009): Bank regulation, capital and credit
supply: Measuring the Impact of Prudential Standards, Occasional Paper Series
36, UK Financial Services Authority.
Fratianni, Michele; Marchionne, Francesco (2010): The Banking Bailout of the
Subprime Crisis: Size and Effects, PSL Quarterly Review, 63 (254), 187233.
Friedman, Milton (1960): A Program for Monetary Stability. New York: Fordham
University Press.
Friedman, Milton (1968): The Role of Monetary Policy, American Economic Review,
58 (-), 117.
Friedman, Milton; Schwartz, Anna J. (1963): A Monetary History of the United States,
1867-1960. Princeton: Princeton University Press.
216 References
Furfine, Craig (2001): Bank Portfolio Allocation: The Impact of Capital Requirements,
Regulatory Monitoring, and Economic Conditions, Journal of Financial Services
Research, 20 (1), 3356.
Galati, Gabriele; Moessner, Richhild (2012): Macroprudential Policy A Literature
Review, Journal of Economic Surveys (-).
Gambacorta, Leonardo (2005): Inside the bank lending channel, European
Economic Review, 49 (7), 17371759.
Gambacorta, Leonardo; Marques-Ibanez, David (2011): The bank lending channel:
lessons from the crisis, Economic Policy, 26 (66), 135182.
Gambacorta, Leonardo; Mistrulli, Paolo E. (2004): Does bank capital affect lending
behavior?, Journal of Financial Intermediation, 13 (4), 436457.
Gambacorta, Leonardo; Rossi, Carlotta (2010): Modelling bank lending in the euro
area: a nonlinear approach, Applied Financial Economics, 20 (14), 10991112.
Geanakoplos, John (2010): The Leverage Cylce, in: NBER Macroeconomics
Annual 2009, ed. by Acemoglu, Daron; Rogoff, Kenneth S.; Woodford, Michael.
Chicago: National Bureau of Economic Research, 165.
Gertler, Mark; Gilchrist, Simon (1993): The Role of Credit Market Imperfections in
the Monetary Transmission Mechanism: Arguments and Evidence, Scandinavian
Journal of Economics, 95 (1), 4364.
Gonzlez-Pramo, Jos Manuel (2011): The ECBs monetary policy during the
crisis, Closing speech at the Tenth Economic Policy Conference, Mlaga,
October 21, 2011.
Greenlaw, David; Hatzius, Jan; Kashyap, Anil K.; Shin, Hyun Song (2008):
Leveraged Losses: Lessons from the Mortgage Market Meltdown, U.S. Monetary
Policy Forum Report No. 2, Rosenberg Institute, Brandeis International Business
School and Initiative on Global Markets University of Chicago Graduate School of
Business.
Griliches, Zvi; Hausman, Jerry A. (1986): Errors in variables in panel data, Journal
of Econometrics, 31 (1), 93118.
Gropp, Reint; Heider, Florian (2010): The Determinants of Bank Capital Structure,
Review of Finance, 14 (4), 587622.
Hall, Brian J. (1993): How Has the Basle Accord Affected Bank Portfolios?, Journal
of the Japanese and International Economies, 7 (4), 408440.
Hancock, Diana; Wilcox, James A. (1993): Has There Been a Capital Crunch in
Banking? The Effects on Bank Lending of Real Estate Market Conditions and
Bank Capital Shortfalls, Journal of Housing Economics, 3 (1), 3150.
Hancock, Diana; Wilcox, James A. (1994): Bank Capital and the Credit Crunch: The
Roles of Risk-Weighted and Unweighted Capital Regulations, Real Estate
Economics, 22 (1), 5994.
Hancock, Diana; Wilcox, James A. (1998): The "credit crunch" and the availability of
credit to small business, Journal of Banking & Finance, 22 (6-8), 9831014.
Hannoun, Herve (2010): Towards a global financial stability framework, Speech
given at the 45th SEACEN Governors' Conference, Siem Reap province,
Cambodia, February 26-27, 2010.
References 217
Hansen, Lars P. (1982): Large Sample Properties of Generalized Method of

Moments Estimators, Econometrica, 50 (4), 102954.
Hanson, Samuel G.; Kashyap, Anil K.; Stein, Jeremy C. (2011): A Macroprudential
Approach to Financial Regulation, Journal of Economic Perspectives, 25 (1), 3
28.
Harris, David; Matyas, Laszlo (1999): Introduction to the Generalized Method of
Moments Estimation, in: Generalized Method of Moments Estimation, ed. by
Matyas, Laszlo. Cambridge: Cambridge University Press, 330.
Healy, Paul M.; Palepu, Krishna G. (2001): Information asymmetry, corporate
disclosure, and the capital markets: A review of the empirical disclosure literature,
Journal of Accounting and Economics, 31 (1-3), 405440.
Holod, Dmytro; Peek, Joe (2007): Asymmetric information and liquidity constraints,
Journal of Banking & Finance, 31 (8), 24252451.
Hubbard, Glenn R. (1995): Is there a "credit channel" for monetary policy?, Federal
Reserve Bank of St. Louis Review, 77 (May/June), 6377.
Iacoviello, Matteo (2005): House Prices, Borrowing Constraints, and Monetary Policy
in the Business Cycle, American Economic Review, 95 (3), 739764.
International Accounting Standards Board (October 2008): IASB amendments permit
reclassification of financial instruments (press release). London.
Ivashina, Victoria; Scharfstein, David S. (2010): Bank lending during the financial
crisis of 2008, Journal of Financial Economics, 97 (3), 319338.
Jaffee, Dwight M.; Russell, Thomas (1976): Imperfect Information, Uncertainty, and
Credit Rationing, The Quarterly Journal of Economics, 90 (4), 65166.
Jarchow, Hans-Joachim (2003): Theorie und Politik des Geldes. Gttingen:
Vandenhoeck und Ruprecht.
Jimenez, Gabriel; Ongena, Steven; Peydro, Jose-Luis; Saurina, Jesus (2008):
Hazardous times for monetary policy: What do twenty-three million bank loans
say about the effects of monetary policy on credit risk-taking?, Documentos de
Trabajo (Working Paper) 0833, Banco de Espana.
Kashyap, Anil K.; Rajan, Raghuram G.; Stein, Jeremy C. (2002): Banks as Liquidity
Providers: An Explanation for the Coexistence of Lending and Deposit-Taking,
Kashyap, Anil K.; Stein, Jeremy C. (1994): Monetary Policy and Bank Lending, in:
Monetary Policy, ed. by Mankiw, Gregory N. Chicago: The University of Chicago
Press for NBER, 221261.
Kashyap, Anil K.; Stein, Jeremy C. (1995): The impact of monetary policy on bank
balance sheets, Carnegie-Rochester Conference Series on Public Policy, 42 (1),
151195.
Kashyap, Anil K.; Stein, Jeremy C. (2000): What Do a Million Observations on Banks
Say about the Transmission of Monetary Policy?, American Economic Review, 90
(3), 407428.
Kashyap, Anil K.; Stein, Jeremy C.; Wilcox, David W. (1993): Monetary Policy and
Credit Conditions: Evidence from the Composition of External Finance, American
Economic Review, 83 (1), 7898.
218 References
Keynes, John Maynard (1936): The General Theory of Employment, Interest and
Money. London: Macmillan.
Keys, Benjamin J.; Mukherjee, Tanmoy; Seru, Amit; Vig, Vikrant (2010): Did
securitization lead to lax screening? Evidence from subprime loans, The Quarterly
Khan, Sajawal; Qayyum, Abdul (2007): Measures of Monetary Policy Stance,
Macroeconomics Working Papers 22201, East Asian Bureau of Economic
Research.
King, Stephen R. (1986): Monetary Transmission: Through Bank Loans or Bank
Liabilities?, Journal of Money, Credit and Banking, 18 (3), 290303.
Kishan, Ruby P.; Opiela, Timothy P. (2000): Bank Size, Bank Capital, and the Bank
Lending Channel, Journal of Money, Credit and Banking, 32 (1), 121141.
Kishan, Ruby P.; Opiela, Timothy P. (2006): Bank capital and loan asymmetry in the
transmission of monetary policy, Journal of Banking & Finance, 30 (1), 259285.
Kiyotaki, Nobuhiro; Moore, John (1997): Credit Cycles, Journal of Political Economy,
105 (2), 21148.
Lemmon, Michael L.; Roberts, Michael R.; Zender, Jaime F. (2008): Back to the
Beginning: Persistence and the Cross-Section of Corporate Capital Structure,
Lenza, Michele; Huw, Pill; Lucrezia, Reichlin (2010): Monetary policy in exceptional
times, Economic Policy, 25 (62), 295339.
Loutskina, Elena; Strahan, Philip E. (2009): Securitization and the Declining Impact
of Bank Finance on Loan Supply: Evidence from Mortgage Originations, Journal
of Finance, 64 (2), 861889.
Ludvigson, Sydney (1998): The Channel of Monetary Transmission to Demand:
Evidence from the Market for Automobile Credit, Journal of Money, Credit and
Banking, 30 (3), 365383.
Maddaloni, Angela; Peydro, Jose-Luis (2011): Bank Risk-taking, Securitization,
Supervision, and Low Interest Rates: Evidence from the Euro-area and the U.S.
Lending Standards, Review of Financial Studies, 24 (6), 21212165.
Marques-Ibanez, David (2009): Banks, credit and the transmission mechanism of
monetary policy, Research Bulletin 8, European Central Bank.
Marques-Ibanez, David; Scheicher, Martin (2010): Securitisation: Instruments and
Implications, in: The Oxford Handbook of Banking, ed. by Berger, Allen N.;
Molyneux, Philip; Wilson, John. Oxford: Oxford University Press, 530555.
Mayer, Colin (1990): Financial Systems, Corporate Finance, and Economic
Development, in: Asymmetric Information, Corporate Finance, and Investment, ed.
by Hubbard, Glenn R. Chicago: University of Chicago Press, 307332.
Mercier, Paul (2009): The Liquidity Management of the Eurosystem during the
Financial Turmoil (August 2007 to February 2009), Revue bancaire et financire,
73 (2/3), 96104.
Mishkin, Frederic S. (1995): Symposium on the Monetary Transmission Mechanism,
Journal of Economic Perspectives, 9 (4), 310.
References 219
Mishkin, Frederic S. (2010): The economics of money, banking and financial markets.
Boston, MA: Pearson Education.
Modigliani, Franco (1971): Monetary Policy and Consumption, in: Consumer
Spending and Monetary Policy - The Linkages, ed. by Federal Reserve Bank of
Boston: Federal Reserve Bank of Boston, 984.
Modigliani, Franco; Miller, Merton H. (1958): The cost of capital, corporation finance
and the theory of investment, American Economic Review, 48 (3), 261297.
Mojon, Benoit (2000): Financial structure and the interest rate channel of ECB
monetary policy, Working Paper 40, European Central Bank.
Myers, Stewart C. (1977): Determinants of corporate borrowing, Journal of Financial
Economics, 5 (2), 147175.
Myers, Stewart C. (1984): The Capital Structure Puzzle, Journal of Finance, 39 (3),
57592.
Myers, Stewart C.; Majluf, Nicholas (1984): Corporate financing and investment
decisions when firms have information that investors do not have, Journal of
Financial Economics, 13 (2), 187221.
Nakamura, Leonard (1984): Bankruptcy and the Informational Problems of
Commercial Bank Lending, Financial Research Center Memorandum 54,
Princeton University.
Newey, Whitney K.; McFadden, Daniel (1994): Large Sample Estimation and
Hypothesis Testing, in: Handbook of Econometrics, ed. by Engle, Robert F.;
McFadden, Daniel. Amsterdam: Elsevier, 21112245.
Nickell, Stephen J. (1981): Biases in Dynamic Models with Fixed Effects,
Econometrica, 49 (6), 14171426.
Oliner, Stephen D.; Rudebusch, Glenn D. (1996a): Is there a broad credit channel
for monetary policy?, Economic Review (), 313.
Oliner, Stephen D.; Rudebusch, Glenn D. (1996b): Monetary Policy and Credit
Conditions: Evidence from the Composition of External Finance: Comment,
American Economic Review, 86 (1), 300309.
Peek, Joe; Rosengren, Eric S. (1995a): Bank lending and the transmission of
monetary policy, in: Is Bank Lending Important for the Transmission of Monetary
Policy?, ed. by Peek, Joe; Rosengren, Eric S.: Federal Reserve Bank of Boston
Conference Proceedings No. 39, 4768.
Peek, Joe; Rosengren, Eric S. (1995b): Bank regulation and the credit crunch,
Journal of Banking & Finance, 19 (3-4), 679692.
Peek, Joe; Rosengren, Eric S. (1995c): The Capital Crunch: Neither a Borrower nor
a Lender Be, Journal of Money, Credit and Banking, 27 (3), 62538.
Peek, Joe; Rosengren, Eric S. (1997): The International Transmission of Financial
Shocks: The Case of Japan, American Economic Review, 87 (4), 495505.
Peek, Joe; Rosengren, Eric S. (2000): Collateral Damage: Effects of the Japanese
Bank Crisis on Real Activity in the United States, American Economic Review, 90
(1), 3045.
220 References
Peek, Joe; Rosengren, Eric S. (2010): The Role of Banks in the Transmission of
Monetary Policy, in: The Oxford Handbook of Banking, ed. by Berger, Allen N.;
Molyneux, Philip; Wilson, John. Oxford: Oxford University Press, 257278.
Peydro, Jose-Luis (2010): Discussion of "The Effects of Bank Capital on Lending:
What Do We Know, and What Does It Mean?", International Journal of Central
Banking, 6 (34), 5569.
Phelps, Edmond S. (1967): Phillips Curves, Expectations of Inflation and Optimal
Unemployment over Time, Economica, 34 (-), 254281.
Phillips, A. W. (1958): The Relation Between Unemployment and the Rate of
Change of Money Wage Rates in the United Kingdom, 1861-1957, Economica,
25 (100), 283299.
Rajan, Raghuram G. (1992): Insiders and Outsiders: The Choice between Informed
and Arm's-Length Debt, Journal of Finance, 47 (4), 13671400.
Repullo, Rafael; Suarez, Javier (2013): The Procyclical Effects of Bank Capital
Regulation, Review of Financial Studies, 26 (2), 452490.
Romer, Christina D.; Romer, David H. (1990): New Evidence on the Monetary
Transmission Mechanism, Brookings Papers on Economic Activity, 21 (1), 149
214.
Roodman, David (2009a): A Note on the Theme of Too Many Instruments, Oxford
Bulletin of Economics and Statistics, 71 (1), 135158.
Roodman, David (2009b): How to do xtabond2: An introduction to difference and
system GMM in Stata, Stata Journal, 9 (1), 86136.
Samuelson, P.A; Solow, R. M. (1960): Problem of Achieving and Maintaining a
Stable Price Level: Analytical Aspects of Anti-Inflation Policy, American Economic
Review, 50 (2), 177194.
Sharpe, Steven A. (1990): Asymmetric Information, Bank Lending, and Implicit
Contracts: A Stylized Model of Customer Relationships, Journal of Finance, 45 (4),
106987.
Shleifer, Andrei; Vishny, Robert W. (1992): Liquidation Values and Debt Capacity: A
Market Equilibrium Approach, Journal of Finance, 47 (4), 134366.
Siemers, Ansgar (2009): Hypo Real Estate: Die groe Peer-Steinbrck-Show,
Focus Online, http://www.focus.de/finanzen/banken/hypo-real-estate-die-grosse-
peer-steinbrueck-show_aid_428079.html. Accessed September 30, 2013.
Solon, Gary (1985): Comment on Benefits and Limitations of Panel Data by C.
Hsiao, Econometric Reviews, 4 (-), 183186.
Standard & Poor's (2013): S&P/Case-Shiller Home Price Indices Methodology, S&P
Dow Jones Indices LLC.
Stein, Jeremy C. (1998): An Adverse-Selection Model of Bank Asset and Liability
Management with Implications for the Transmission of Monetary Policy, RAND
Stein, Jeremy C. (2012): Monetary Policy as Financial Stability Regulation, The
Quarterly Journal of Economics, 127 (1), 5795.
References 221
Stiglitz, Joseph E.; Weiss, Andrew (1981): Credit Rationing in Markets with Imperfect
Information, American Economic Review, 71 (3), 393410.
Thornton, Daniel L. (2009): What the Libor-OIS spread says, Economic Synopses,
24 (-), 12.
Tobin, James (1969): A General Equilibrium Approach to Monetary Theory, Journal
of Money, Credit and Banking, 1 (1), 1529.
Tobin, James (1982): The Commercial Banking Firm: A Simple Model,
Scandinavian Journal of Economics, 84 (4), 495530.
Trichet, Jean-Claude (2010): State of the Union: The Financial Crisis and the ECB's
Response between 2007 and 2009, Journal of Common Market Studies, 48 (s1),
719.
van Rixtel, Adrian; Gasperini, Gabriele (2013): Financial crises and bank funding:
recent experience in the euro area, BIS Working Papers 406, Bank for
International Settlements.
Weber, Axel A.; Gerke, Rafael; Worms, Andreas (2009): Has the monetary
transmission process in the euro area changed? Evidence vased on VAR
estimates, BIS Working Papers 276.
Wooldridge, Jeffrey M. (2002): Econometric Analysis of Cross Section and Panel
Data. Cambridge, MA: MIT Press.
Worms, Andreas (2001): Monetary Policy Effects on Bank Loans in Germany: A
Panel-econometric Analysis, Discussion paper 17/01, Deutsche Bundesbank.
Worms, Andreas (2003): The reaction of bank lending to monetary policy measures
in Germany, in: Monetary policy transmission in the Euro area: A study by the
Eurosystem Monetary Transmission Network, ed. by Angeloni, Ignazio; Kashyap,
Anil K.; Mojon, Benoit. Cambridge: Cambridge University Press, 270283.
222 Appendix
Appendix
DescriptivestatisticsGERMANY
observations
Dependentvariable
LOG(LOANS) 13,655 0.029 0.019 0.593 0.683 0.076
Bankcharacteristics
CAP 15,486 0.000 0.005 0.061 0.939 0.037
CAPSUR 15,486 0.020 0.020 0.950 0.120 0.041
SIZE 15,168 0.000 0.063 4.341 7.612 1.387
DEP 15,114 0.000 0.015 0.715 0.281 0.118
OVERHANG 15,110 0.296 0.177 1.000 11.025 0.589
LIQ 15,150 0.000 0.000 0.022 0.372 0.010
STF 15,114 0.000 0.015 0.168 0.712 0.096
NII 15,132 0.000 0.008 1.221 1.726 0.098
EURIBOR 19,065 3.022 3.078 0.814 4.644 1.245
EURIBOR_OIS 16,523 0.263 0.245 0.121 0.774 0.242
(NSM) 16,523 0.013 0.005 0.011 0.072 0.024
LOG(GDP) 17,794 0.022 0.023 0.041 0.050 0.021
GDP_DEFLATOR 19,065 0.753 0.800 0.700 1.600 0.554
DescriptivestatisticsITALY
observations
Dependentvariable
LOG(LOANS) 2,729 0.103 0.097 0.517 0.693 0.088
Bankcharacteristics
CAP 3,580 0.000 0.012 0.099 0.870 0.071
CAPSUR 3,580 0.053 0.040 0.920 0.060 0.071
SIZE 3,266 0.000 0.195 4.532 7.384 1.539
DEP 3,238 0.000 0.008 0.517 0.441 0.145
OVERHANG 3,238 0.106 0.269 1.000 15.207 0.759
LIQ 3,261 0.000 0.002 0.010 0.099 0.008
STF 3,238 0.000 0.046 0.059 0.864 0.125
NII 3,266 0.000 0.006 1.064 1.114 0.132
EURIBOR 7,815 3.022 3.078 0.814 4.644 1.245
EURIBOR_OIS 6,773 0.263 0.245 0.121 0.774 0.242
(NSM) 6,773 0.013 0.005 0.011 0.072 0.024
LOG(GDP) 7,294 0.029 0.035 0.036 0.055 0.021
GDP_DEFLATOR 7,815 2.187 2.400 0.400 3.200 0.712

DOI 10.1007/978-3-658-07175-2, Springer Fachmedien Wiesbaden 2015
Appendix 223
DescriptivestatisticsFRANCE
observations
Dependentvariable
LOG(LOANS) 1,315 0.070 0.072 0.547 0.675 0.245
Bankcharacteristics
CAP 1,508 0.000 0.011 0.080 0.802 1.649
CAPSUR 1,508 0.039 0.030 0.820 0.050 4.193
SIZE 1,465 0.000 0.117 5.130 6.294 0.068
DEP 1,437 0.000 0.050 0.410 0.498 0.057
OVERHANG 1,437 0.242 0.357 1.000 12.056 0.011
LIQ 1,411 0.000 0.002 0.012 0.077 0.261
STF 1,437 0.000 0.027 0.403 0.512 0.836
NII 1,453 0.000 0.126 159.273 4.615 0.024
EURIBOR 2,130 3.022 3.078 0.814 4.644 0.095
EURIBOR_OIS 1,846 0.263 0.245 0.121 0.774 0.018
(NSM) 1,846 0.013 0.005 0.011 0.072 0.679
LOG(GDP) 1,988 0.033 0.038 0.025 0.052 1.245
GDP_DEFLATOR 2,130 1.580 1.700 0.200 2.600 0.242
DescriptivestatisticsSPAIN
observations
Dependentvariable
LOG(LOANS) 458 0.087 0.090 0.542 0.594 0.126
Bankcharacteristics
CAP 639 0.000 0.011 0.080 0.461 0.058
CAPSUR 639 0.024 0.010 0.490 0.210 0.057
SIZE 585 0.000 0.126 6.588 6.514 2.206
DEP 565 0.000 0.091 0.711 0.217 0.224
OVERHANG 563 0.271 0.075 1.000 33.077 1.854
LIQ 583 0.000 0.007 0.014 0.279 0.024
STF 565 0.000 0.074 0.131 0.729 0.193
NII 577 0.000 0.031 0.502 0.757 0.152
EURIBOR 1,590 3.022 3.078 0.814 4.644 1.245
EURIBOR_OIS 1,378 0.263 0.245 0.121 0.774 0.242
(NSM) 1,378 0.013 0.005 0.011 0.072 0.024
LOG(GDP) 1,484 0.053 0.070 0.037 0.083 0.035
GDP_DEFLATOR 1,590 2.887 3.300 0.100 4.400 1.393
CorrelationmatrixGERMANY
224
LOG(LOANS) CAP CAPSUR SIZE DEP OVERHANG LIQ STF NII EURIBOR EURIBOR_OIS (NSM) LOG(GDP) GDP_DEFLATOR CRISIS
LOG(LOANS) 1.000

CAP 0.013 1.000
0.157
CAPSUR 0.064 0.005 1.000
0.000 0.558
SIZE 0.031 0.192 0.097 1.000
0.001 0.000 0.000
DEP 0.029 0.044 0.047 0.376 1.000
0.001 0.000 0.000 0.000
OVERHANG 0.087 0.043 0.013 0.107 0.557 1.000
0.000 0.000 0.147 0.000 0.000
LIQ 0.033 0.078 0.024 0.072 0.160 0.101 1.000
0.000 0.000 0.006 0.000 0.000 0.000
STF 0.051 0.100 0.004 0.265 0.837 0.498 0.173 1.000
0.000 0.000 0.683 0.000 0.000 0.000 0.000
NII 0.070 0.074 0.002 0.087 0.025 0.040 0.140 0.058 1.000
0.000 0.000 0.848 0.000 0.004 0.000 0.000 0.000
EURIBOR 0.021 0.094 0.031 0.023 0.051 0.049 0.039 0.038 0.070 1.000
0.018 0.000 0.000 0.009 0.000 0.000 0.000 0.000 0.000
EURIBOR_OIS 0.007 0.137 0.020 0.020 0.040 0.082 0.043 0.058 0.154 0.034 1.000
0.417 0.000 0.014 0.023 0.000 0.000 0.000 0.000 0.000 0.000
(NSM) 0.032 0.129 0.046 0.023 0.023 0.043 0.020 0.058 0.152 0.052 0.677 1.000
0.000 0.000 0.000 0.009 0.007 0.000 0.022 0.000 0.000 0.000 0.000
LOG(GDP) 0.031 0.034 0.010 0.015 0.004 0.014 0.002 0.020 0.086 0.213 0.019 0.373 1.000
0.000 0.000 0.250 0.082 0.613 0.115 0.821 0.023 0.000 0.000 0.012 0.000
GDP_DEFLATOR 0.127 0.050 0.009 0.027 0.044 0.055 0.040 0.061 0.041 0.188 0.061 0.040 0.127 1.000
0.000 0.000 0.286 0.002 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
CRISIS 0.028 0.048 0.006 0.002 0.028 0.070 0.005 0.032 0.059 0.029 0.670 0.160 0.639 0.148 1.000
0.001 0.000 0.459 0.841 0.001 0.000 0.576 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Appendix
CorrelationmatrixITALY
LOG(LOANS) 1.000
Appendix

CAP 0.055 1.000
0.005
CAPSUR 0.053 0.012 1.000
0.007 0.464
SIZE 0.053 0.310 0.158 1.000
0.007 0.000 0.000
DEP 0.107 0.006 0.034 0.270 1.000
0.000 0.731 0.057 0.000
OVERHANG 0.086 0.101 0.010 0.168 0.511 1.000
0.000 0.000 0.568 0.000 0.000
LIQ 0.030 0.109 0.004 0.024 0.281 0.054 1.000
0.121 0.000 0.824 0.183 0.000 0.003
STF 0.013 0.188 0.037 0.373 0.550 0.180 0.193 1.000
0.500 0.000 0.042 0.000 0.000 0.000 0.000
NII 0.007 0.100 0.042 0.176 0.064 0.090 0.024 0.094 1.000
0.707 0.000 0.019 0.000 0.000 0.000 0.189 0.000
EURIBOR 0.175 0.081 0.019 0.052 0.012 0.032 0.073 0.007 0.152 1.000
0.000 0.000 0.252 0.004 0.523 0.073 0.000 0.689 0.000
EURIBOR_OIS 0.113 0.066 0.044 0.061 0.023 0.066 0.049 0.102 0.217 0.034 1.000
0.000 0.000 0.010 0.001 0.200 0.000 0.007 0.000 0.000 0.005
(NSM) 0.033 0.057 0.040 0.019 0.005 0.004 0.102 0.075 0.201 0.052 0.677 1.000
0.089 0.001 0.019 0.287 0.773 0.834 0.000 0.000 0.000 0.000 0.000
LOG(GDP) 0.216 0.060 0.016 0.060 0.022 0.054 0.025 0.066 0.026 0.541 0.550 0.110 1.000
0.000 0.000 0.337 0.001 0.215 0.003 0.171 0.000 0.153 0.000 0.000 0.000
GDP_DEFLATOR 0.086 0.084 0.019 0.042 0.032 0.040 0.014 0.043 0.065 0.539 0.412 0.261 0.170 1.000
0.000 0.000 0.261 0.021 0.074 0.027 0.430 0.018 0.000 0.000 0.000 0.000 0.000
CRISIS 0.135 0.017 0.003 0.043 0.040 0.049 0.001 0.063 0.155 0.029 0.670 0.160 0.777 0.107 1.000
0.000 0.327 0.854 0.017 0.027 0.007 0.961 0.000 0.000 0.016 0.000 0.000 0.000 0.000
225
CorrelationmatrixFRANCE
226
LOG(LOANS) 1.000

CAP 0.168 1.000
0.000
CAPSUR 0.114 0.017 1.000
0.000 0.515
SIZE 0.070 0.399 0.229 1.000
0.014 0.000 0.000
DEP 0.150 0.275 0.023 0.114 1.000
0.000 0.000 0.415 0.000
OVERHANG 0.021 0.116 0.016 0.008 0.563 1.000
0.455 0.000 0.574 0.763 0.000
LIQ 0.020 0.055 0.025 0.231 0.451 0.116 1.000
0.498 0.052 0.372 0.000 0.000 0.000
STF 0.049 0.041 0.059 0.158 0.813 0.513 0.375 1.000
0.087 0.148 0.035 0.000 0.000 0.000 0.000
NII 0.002 0.023 0.003 0.017 0.044 0.034 0.025 0.045 1.000
0.934 0.399 0.910 0.534 0.112 0.219 0.371 0.102
EURIBOR 0.108 0.082 0.063 0.057 0.006 0.039 0.011 0.012 0.010 1.000
0.000 0.002 0.017 0.038 0.826 0.164 0.688 0.678 0.706
EURIBOR_OIS 0.028 0.084 0.019 0.087 0.072 0.001 0.038 0.036 0.037 0.034 1.000
0.327 0.001 0.480 0.002 0.010 0.984 0.175 0.191 0.177 0.141
(NSM) 0.079 0.067 0.056 0.056 0.046 0.019 0.008 0.037 0.024 0.052 0.677 1.000
0.005 0.011 0.034 0.042 0.094 0.496 0.782 0.181 0.393 0.026 0.000
LOG(GDP) 0.174 0.056 0.021 0.068 0.037 0.022 0.001 0.000 0.001 0.501 0.367 0.113 1.000
0.000 0.034 0.416 0.013 0.180 0.428 0.974 0.989 0.962 0.000 0.000 0.000
GDP_DEFLATOR 0.072 0.001 0.054 0.005 0.009 0.013 0.025 0.008 0.010 0.592 0.083 0.176 0.380 1.000
0.011 0.976 0.040 0.849 0.745 0.632 0.368 0.780 0.712 0.000 0.000 0.000 0.000
CRISIS 0.120 0.035 0.030 0.067 0.053 0.003 0.006 0.016 0.019 0.029 0.670 0.160 0.754 0.048 1.000
0.000 0.188 0.262 0.015 0.055 0.913 0.817 0.559 0.483 0.208 0.000 0.000 0.000 0.038
Appendix
CorrelationmatrixSPAIN
LOG(LOANS) 1.000
Appendix

CAP 0.105 1.000
0.028
CAPSUR 0.148 0.044 1.000
0.002 0.269
SIZE 0.144 0.382 0.325 1.000
0.002 0.000 0.000
DEP 0.026 0.167 0.143 0.318 1.000
0.586 0.000 0.001 0.000
OVERHANG 0.122 0.093 0.026 0.108 0.161 1.000
0.011 0.032 0.553 0.012 0.000
LIQ 0.053 0.401 0.008 0.070 0.075 0.062 1.000
0.265 0.000 0.856 0.098 0.081 0.150
STF 0.031 0.040 0.038 0.211 0.908 0.177 0.044 1.000
0.514 0.358 0.378 0.000 0.000 0.000 0.309
NII 0.078 0.120 0.149 0.332 0.136 0.265 0.408 0.028 1.000
0.100 0.005 0.000 0.000 0.002 0.000 0.000 0.510
EURIBOR 0.394 0.010 0.026 0.005 0.002 0.067 0.039 0.011 0.145 1.000
0.000 0.795 0.510 0.897 0.971 0.121 0.359 0.790 0.001
EURIBOR_OIS 0.259 0.143 0.062 0.067 0.166 0.006 0.152 0.131 0.096 0.034 1.000
0.000 0.000 0.119 0.110 0.000 0.883 0.000 0.002 0.023 0.203
(NSM) 0.106 0.098 0.076 0.021 0.124 0.053 0.100 0.122 0.108 0.052 0.677 1.000
0.024 0.014 0.056 0.616 0.004 0.222 0.018 0.004 0.010 0.055 0.000
LOG(GDP) 0.560 0.061 0.058 0.021 0.062 0.049 0.085 0.040 0.016 0.591 0.641 0.204 1.000
0.000 0.124 0.149 0.623 0.151 0.250 0.045 0.354 0.706 0.000 0.000 0.000
GDP_DEFLATOR 0.554 0.074 0.069 0.009 0.067 0.038 0.069 0.050 0.031 0.551 0.707 0.298 0.920 1.000
0.000 0.062 0.082 0.835 0.116 0.383 0.101 0.242 0.471 0.000 0.000 0.000 0.000
CRISIS 0.282 0.036 0.006 0.054 0.076 0.038 0.079 0.051 0.085 0.029 0.670 0.160 0.642 0.489 1.000
0.000 0.373 0.888 0.199 0.078 0.372 0.062 0.232 0.046 0.276 0.000 0.000 0.000 0.000
227
228 Appendix
Juxtaposition of countries - specification III

(III) Baseline model with Germany Italy France Spain
time-fixed effects
Dependent variable:
LOG(LOANS(-1)) 0.52 *** (0.10) 0.05 (0.05) 0.15 *** (0.05) 0.10 (0.12)
CAP(-1)*NC 0.45 ** (0.19) 1.12 (0.82) 0.45 *** (0.13) 0.31 (0.75)
CAP(-1)*CRISIS 0.51 *** (0.18) 1.11 (0.89) 0.16 * (0.10) 1.09 (0.75)
SIZE(-1)*NC 0.05 * (0.03) 0.24 *** (0.06) 0.12 *** (0.03) 0.09 (0.24)
SIZE(-1)*CRISIS 0.07 ** (0.03) 0.23 *** (0.06) 0.13 *** (0.03) 0.09 (0.24)
DEP(-1)*NC 0.00 (0.05) 0.44 ** (0.20) 0.12 (0.12) 0.33 * (0.19)
DEP(-1)*CRISIS 0.05 (0.03) 0.47 ** (0.21) 0.16 (0.11) 0.60 ** (0.28)
LIQ(-1)*EURIBOR*NC 0.05 (0.11) 0.05 (0.27) 0.50 (0.32) 0.25 ** (0.12)
LIQ(-1)*EURIBOR*CRISIS 0.03 (0.17) 0.07 (0.22) 0.29 (0.18) 0.05 (0.33)
STF(-1)*NC 0.16 (0.13) 0.07 (0.27) 0.26 (0.52) 0.35 (0.22)
STF(-1)*CRISIS 0.05 (0.03) 0.04 (0.27) 0.11 (0.10) 0.01 (0.14)
NII(-1)*NC 0.08 (0.09) 0.16 ** (0.07) 0.02 (0.03) 0.33 (0.57)
NII(-1)*CRISIS 0.03 (0.10) 0.08 (0.07) 0.01 (0.02) 0.16 (0.13)
LOG(GDP) 0.45 (0.30) 0.16 (0.18) 0.56 (1.07) 0.59 (0.73)
EURIBOR 0.00 (0.00) -0.01 (0.02) 0.00 (0.02) 0.00 (0.02)
EURIBOR_OIS
GDP_DEFLATOR 0.00 (0.00) -0.06 (0.05) -0.01 (0.02) -0.01 (0.02)
D(NSM) -0.11 (0.39) -0.18 (1.28) 0.35 (0.36) -0.23 (1.67)
Time dummies: YES YES YES YES

Cross-sections included: 1,241 494 120 74
Total (unbalanced) obs.: 9,974 1,594 896 222
respectively.
Appendix 229
Juxtaposition of countries - specification V

(V) Capital surplus and Germany Italy France Spain
time-fixed effects
Dependent variable:
LOG(LOANS(-1)) 0.52 *** (0.10) 0.06 (0.04) 0.13 * (0.08) 0.03 (0.10)
CAP(-1)*NC 0.48 *** (0.13) 0.67 * (0.38) 1.81 *** (0.57) 0.78 (1.72)
CAP(-1)*CRISIS 0.74 *** (0.26) 0.92 * (0.50) 1.93 *** (0.46) 1.06 (2.07)
CAPSUR(-1)*NC 0.06 *** (0.02) 0.10 (0.06) 0.08 *** (0.02) 0.02 (0.05)
CAPSUR(-1)*CRISIS 0.06 ** (0.03) 0.12 * (0.07) 0.08 *** (0.03) 0.08 (0.08)
SIZE(-1)*NC 0.05 (0.03) 0.27 *** (0.05) 0.10 *** (0.01) 0.07 (0.08)
SIZE(-1)*CRISIS 0.08 ** (0.04) 0.25 *** (0.05) 0.11 *** (0.01) 0.10 (0.08)
DEP(-1)*NC -0.03 (0.05) 0.35 ** (0.16) 0.01 (0.06) 0.26 (0.25)
DEP(-1)*CRISIS 0.01 (0.05) 0.34 ** (0.17) 0.07 (0.05) 0.61 ** (0.30)
LIQ(-1)*EURIBOR*NC 0.20 (0.21) 0.13 (0.28) 0.50 *** (0.18) 0.22 ** (0.10)
LIQ(-1)*EURIBOR*CRISIS 0.18 (0.26) 0.10 (0.22) 0.35 *** (0.13) 0.45 (0.38)
STF(-1)*NC 0.21 * (0.12) 0.16 (0.18) 0.12 (0.35) 0.31 (0.35)
STF(-1)*CRISIS 0.10 * (0.06) 0.04 (0.18) 0.10 * (0.05) 0.18 * (0.10)
NII(-1)*NC 0.11 (0.22) 0.20 (0.19) 0.03 (0.03) 0.45 ** (0.20)
NII(-1)*CRISIS 0.05 * (0.03) 0.07 (0.07) 0.01 (0.02) 0.09 (0.15)
LOG(GDP) 0.06 (0.42) 1.29 (2.29) 1.48 ** (0.62) 1.73 ** (0.78)
EURIBOR 0.00 (0.00) -0.06 * (0.03) -0.03 *** (0.01) -0.02 (0.03)
GDP_DEFLATOR -0.01 (0.01) 0.10 ** (0.04) -0.03 ** (0.01) 0.02 (0.02)
D(NSM) -0.08 (0.42) -0.16 (0.14) 1.16 ** (0.47) 0.63 (1.76)

respectively.
230 Appendix
Juxtaposition of countries - specification VII

(VII) Deposit overhang Germany Italy France Spain
and time-fixed effects
Dependent variable:
LOG(LOANS(-1)) 0.44 *** (0.12) 0.04 (0.03) 0.04 *** (0.01) 0.39 ** (0.17)
CAP(-1)*NC 0.47 *** (0.18) 0.89 (0.65) 0.47 *** (0.16) 1.00 (0.65)
CAP(-1)*CRISIS 0.54 *** (0.19) 0.84 (0.69) 0.15 (0.17) 1.65 ** (0.64)
SIZE(-1)*NC 0.08 * (0.04) 0.32 *** (0.05) 0.10 *** (0.03) 0.25 *** (0.09)
SIZE(-1)*CRISIS 0.10 *** (0.03) 0.31 *** (0.04) 0.10 *** (0.03) 0.26 *** (0.09)
DEP(-1)*NC -0.04 ** (0.02) 0.24 (0.15) -0.22 * (0.13) 0.26 (0.33)
DEP(-1)*CRISIS 0.03 (0.02) 0.26 * (0.16) -0.18 * (0.10) 0.60 ** (0.26)
OVERHANG(-1)*NC 0.06 *** (0.01) 0.12 *** (0.02) 0.07 * (0.04) 0.50 *** (0.13)
OVERHANG(-1)*CRISIS 0.05 *** (0.01) 0.12 *** (0.03) 0.07 ** (0.03) 0.55 *** (0.13)
LIQ(-1)*EURIBOR*NC 0.06 (0.10) 0.11 (0.28) 0.57 * (0.34) 0.42 *** (0.14)
LIQ(-1)*EURIBOR*CRISIS 0.05 (0.16) 0.06 (0.20) 0.36 * (0.19) 0.56 ** (0.27)
STF(-1)*NC 0.14 (0.19) 0.17 (0.51) 0.27 (0.19) 0.30 (0.21)
STF(-1)*CRISIS 0.03 (0.02) 0.08 (0.21) 0.13 (0.13) 0.13 (0.10)
NII(-1)*NC 0.14 (0.13) 0.24 * (0.13) 0.04 *** (0.01) 0.29 (0.18)
NII(-1)*CRISIS 0.04 (0.05) 0.16 (0.20) 0.02 (0.01) 0.01 (0.15)
LOG(GDP) 0.17 (0.40) 0.09 (1.40) 0.68 (1.00) 1.90 ** (0.76)
EURIBOR 0.00 (0.00) 0.00 (0.01) 0.00 (0.01) 0.00 (0.01)
GDP_DEFLATOR 0.00 (0.00) -0.04 (0.03) -0.03 (0.02) 0.00 (0.01)
D(NSM) 0.29 (0.37) -0.02 (0.96) 0.76 (0.73) 0.18 (0.29)

respectively.

Hartmut Brinkmeyer Auth. Drivers of Bank Lending New Evidence From The Crisis

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Hartmut Brinkmeyer Auth. Drivers of Bank Lending New Evidence From The Crisis

Hochgeladen von

Copyright:

Verfügbare Formate

Schriften zum europischen

Prof. Dr. Rolf Caspers () Prof. Dr. Dr. Karl-Ulrich Rudolph

Prof. Dr. Alfred Ktzle

Drivers of Bank Lending

Dissertation Heinrich-Heine-Universitt Dsseldorf D61 / 2014

ISBN 978-3-658-07174-5 ISBN 978-3-658-07175-2 (eBook)

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;

Library of Congress Control Number: 2014949154

Printed on acid-free paper

Springer Gabler is a brand of Springer DE.

Hartmut Brinkmeyer's dissertation contributes to this field of research not only on a

Dsseldorf, April 2014

An undertaking such as writing a dissertation is a great challenge. Thinking of some

First and foremost, I would like to express my profound gratitude to my thesis

My employer, Roland Berger Strategy Consultants, gave me the opportunity to take

Dsseldorf, June 2014 Hartmut Brinkmeyer

LIST OF FIGURES .................................................................................................. XIII

LIST OF TABLES .................................................................................................... XV

LIST OF NOTATIONS AND ABBREVIATIONS .................................................... XVII

1. Introduction ....................................................................................................... 1

1.1. Motivation ..................................................................................................... 1

2. Transmission channels of monetary policy .................................................... 7

2.1. The money view ........................................................................................... 7

2.2.1. The balance sheet channel ................................................................... 11

3. The bank lending channel in detail ................................................................ 18

3.1. Structure and elements of the bank lending channel.................................. 18

3.1.1.1. Subcondition 1: No complete adjustment to adverse monetary

3.1.2. Condition 2: Publicly issued debt and non-bank intermediated loans

3.2. Conclusion ................................................................................................. 34

4.1. The bank lending channel revisited ............................................................ 35

5.1. The loss spiral ............................................................................................ 45

6.2.1. US evidence based on aggregate data ................................................. 58

6.3. Evidence from the euro area ...................................................................... 68

6.4. Implications of theoretical framework for interpretation of empirical

7. Empirical analysis approach ....................................................................... 82

7.1. Research hypotheses................................................................................. 82

7.1.1. General hypotheses .............................................................................. 83

7.2. Overall empirical strategy and approach .................................................... 89

7.3.1. Derivation of a model of bank behavior ................................................. 91

7.4. Data............................................................................................................ 98

7.4.1. Data sources ......................................................................................... 98

7.4.3. Special challenges .............................................................................. 113

7.4.3.1. Disentangling loan supply and loan demand ................................ 113

7.4.4. Purging the data.................................................................................. 119

7.5. Estimation method.................................................................................... 121

8. Empirical analysis results ......................................................................... 129

8.1. Results for the euro area .......................................................................... 129

8.1.1. Descriptive statistics and correlations ................................................. 131

8.1.2.1. Results of the standard specification ............................................ 134

8.1.3. Capital surplus .................................................................................... 153

8.1.3.1. Results of the standard specification ............................................ 153

8.1.4. Deposit overhang ................................................................................ 161

8.1.4.1. Results of the standard specification ............................................ 163

8.2. Results for major euro area countries ...................................................... 170

8.3. Conclusion Main research hypotheses confirmed ................................. 191

9. Final discussion and implications ............................................................... 194

9.1. Overall summary of results ....................................................................... 194

9.2.1. Contributions to research regarding the determinants of bank lending 197

9.2.2.1. General contributions ................................................................... 199

9.3. Implications for bank management........................................................... 202

References ............................................................................................................ 211

Appendix ............................................................................................................... 222

Figure 1.1: Organization of the research .................................................................... 5