and
Financial Contagion
Dissertation
zur Erlangung des akademischen Grades Doktor der Wirtschaftswissenschaften (Dr. rer. pol.)
im Fachbereich Wirtschaftswissenschaften der Universität Konstanz
vorgelegt von: Adrian Alter
Tag der mündlichen Prüfung: 29.08.2013
Referenten:
1. Prof. Dr. Dr. h.c. Günter Franke 2. Prof. Dr. Almuth Scholl
Contents . . . I List of Figures. . . V List of Tables . . . VII Acknowledgements . . . IX Summary . . . XI Zusammenfassung. . . XVII 1 Credit spread interdependencies of European states and banks 1
1.1 Introduction . . . 2
1.2 Related literature . . . 5
1.3 Hypotheses, data, and econometric methodology. . . 7
1.3.1 Hypotheses . . . 7
1.3.2 Bailout specific characteristics . . . 9
1.3.3 Data and sub-sample selection . . . 10
1.3.4 Econometric methodology . . . 12
1.4 Results . . . 14
1.4.1 Cross-country analysis . . . 15
1.4.2 Specific country analysis . . . 20
1.5 Conclusion . . . 31
Appendix 1.A Further issues on methodology . . . 36
1.A.1 VEC-analysis - Selection of sub-stages . . . 36
1.A.2 Pre-analysis of the data, model specification, and estimation 36 1.A.3 Interpretation of long-run relations in a VECM . . . 37
Appendix 1.B Specific country analysis . . . 38
1.B.1 France . . . 38
1.B.2 Germany. . . 39
1.B.3 Ireland . . . 40
1.B.4 Italy . . . 41
1.B.5 The Netherlands . . . 42
1.B.6 Portugal . . . 43
2 A Contagion Index for the Euro Area 47
2.1 Introduction . . . 48
2.2 Related Literature . . . 51
2.3 Econometric Methodology and Data Description . . . 53
2.3.1 Vector autoregressive model with exogenous variables (VARX) 54 2.3.2 Generalized impulse response functions (GIRF) . . . 55
2.3.3 The spillover matrix . . . 56
2.3.4 Contagion indices . . . 59
2.4 Results . . . 60
2.4.1 A static perspective . . . 60
2.4.2 The dynamics of potential spillover effects . . . 66
2.4.3 The euro area Contagion Index . . . 68
2.4.4 The spillover and net spillover matrices . . . 71
2.4.5 The systemic contribution of sovereigns . . . 74
2.4.6 The impact of different economic/policy events on the conta- gion index . . . 74
2.4.7 Critical spillover thresholds for contagion . . . 75
2.5 Robustness and motivation of setup parameters . . . 78
2.5.1 Differences in distributions of residuals . . . 78
2.5.2 Relaxing restrictions imposed on impulse responses . . . 79
2.6 Conclusion and Outlook . . . 81
Bibliography . . . 85
Appendix 2.A Description of variables and events. . . 87
Appendix 2.B The explicit VAR model with exogenous common factors 90 Appendix 2.C Other versions of the contagion indices and systemic con- tribution of sovereigns . . . 91
Appendix 2.D Spillover and Net Spillover Matrices . . . 94
Appendix 2.E Optimal rolling window size . . . 96
3 Centrality-based Capital Allocations and Bailout Funds 99 3.1 Introduction . . . 100
3.2 Data and methodology . . . 106
3.2.1 Methodology . . . 106
3.2.2 Data sources. . . 107
3.3.2 Centrality measures. . . 112
3.4 Credit risk model . . . 116
3.4.1 Modeling the returns of large loans . . . 116
3.4.2 Modeling the returns of small loans . . . 119
3.5 Modeling contagion . . . 120
3.5.1 Losses and bankruptcy costs . . . 121
3.5.2 “Eisenberg and Noe” - interbank contagion algorithm . . . . 122
3.6 Optimization . . . 123
3.6.1 Capital allocations . . . 124
3.6.2 Target function(s) . . . 126
3.6.3 Setting capital allocation(s) procedure . . . 127
3.6.4 Bailout fund mechanism . . . 128
3.7 Results . . . 129
3.7.1 Capital allocations . . . 129
3.7.2 Bailout fund mechanism . . . 135
3.8 Robustness checks. . . 138
3.8.1 Interbank liabilities . . . 138
3.8.2 Network structure. . . 139
3.8.3 Credit risk parameters . . . 140
3.9 Conclusion . . . 140
Appendix 3.A Risk and contagion mechanism . . . 146
Appendix 3.B Centrality measures - technical details . . . 150
3.B.1 Eigenvector centrality . . . 150
3.B.2 Betweenness centrality . . . 150
3.B.3 Closeness centrality . . . 151
3.B.4 (Local) Clustering coefficient. . . 151
Appendix 3.C Modeling returns of small loans . . . 152
3.C.1 Estimating the granularity of small-loans exposures (Herfind- ahl Index) . . . 154
Appendix 3.D Other target functions . . . 155
Appendix 3.E Other results. . . 156
Appendix 3.F Liabilities and network properties . . . 158
Complete Bibliography 163
Erklärung und Eigenabgrenzung 173
1.1 Effects of a Banking Sector Shock on Government Spreads: Before
Government Interventions . . . 18
1.2 Responses on Day 1 after the Shock . . . 19
1.3 Effects of a Sovereign Shock on Bank Spreads: After Government Interventions.. . . 20
1.4 Generalized Impulse Responses for Germany: (Solid) Before, (Dot- ted) During & After Government Interventions . . . 22
1.5 Generalized Impulse Responses for Ireland: (Solid) Before, (Dotted) During & After Government Interventions . . . 26
1.6 Generalized Impulse Responses for Italy: (Solid) Before, (Dotted) During & After Government Interventions . . . 29
1.B.1 France: CDS Level Series . . . 38
1.B.2 Germany: CDS Level Series . . . 39
1.B.3 Ireland: CDS Level Series . . . 40
1.B.4 Italy: CDS Level Series . . . 41
1.B.5 The Netherlands: CDS Level Series . . . 42
1.B.6 Portugal: CDS Level Series . . . 43
1.B.7 Spain: CDS Level Series . . . 44
1.B.8 Cointegration Graph of Germany and Commerzbank (Before Gov- ernment Interventions) . . . 45
1.B.9 Cointegration Graph of Ireland and Allied Irish Banks (During and After Government Interventions) . . . 45
1.B.10 Cointegration Graph of Italy and Intesa Sanpaolo (During and After Government Interventions) . . . 45
2.1 Potential impact of a Spanish sovereign CDS shock on other sovereign CDS spreads . . . 62
2.2 Potential impact on Spanish sovereign CDS from a shock in the other sovereign CDS spreads. . . 62 2.3 Potential impact of a shock in Spanish sovereign CDS on bank CDSs 63 2.4 Potential impact of a shock in bank CDSs on Spanish sovereign CDS 63
2.7 Sovereign CDS series and the EA Contagion Index . . . 69
2.8 EA Contagion Indices . . . 70
2.9 Average potential spillover effects . . . 71
2.10 Systemic contributions of GIIPS countries (left axis) and the Total Net Positive (TNP) Spillover (right axis) . . . 75
2.11 Impact on Contagion Index components at some specific news/policy events . . . 76
2.12 Rejection of the Null hypothesis of the Kolmogorov-Smirnov (KS) test . . . 79
2.13 Moments of the sample distributions of residuals from the VAR model 80 2.14 The Contagion Index with restricted and unrestricted IRs . . . 80
2.C.1 Different versions of the EA Contagion Index of sovereigns . . . 93
2.E.1 Optimal size of the rolling window . . . 97
3.1 Individual bank balance sheet and benchmark capital . . . 108
3.2 A comparison of different capital allocations across network measures 130 3.3 Expected system losses: All defaults, fundamental defaults and con- tagious defaults . . . 132
3.4 Frequency distributions of individual bank PDs . . . 133
3.5 Occurrences of individual bank defaults . . . 134
3.6 Comparison of different capital allocations based on: Total Assets, Opsahl and Weighted Eigenvector . . . 135
3.7 Pdfs of bank defaults AFTER contagion using a bailout fund mech- anism with rules based on: Opsahl (Ops) versus VaR (upper level); Total Asstets (TA) versus VaR (lower level). . . 137
3.8 Bailout efficiency surface: Opsahl centrality . . . 138
3.A.1 Risk model sketch . . . 147
3.B.1 Centrality measures . . . 151
3.E.1 Unconditional distribution of total system losses . . . 156
3.E.2 Conditional distributions of losses for best capital allocation under rule based on: Total Assets (TA), Value-at-Risk (VaR), and Opsahl centrality (Opsahl) . . . 157
3.F.1 Power law vs log-normal diagnostics . . . 158
3.F.2 Comparison between ranked interbank liabilities (by size) . . . 159
1.1 Government Support Measures for Financial Institutions (October
2008 - May 2010) . . . 9
1.2 Results of Granger-Causality Tests for all Countries. . . 15
1.3 Percentage of Significant/Insignificant Responses in the Long Run (after 22 days) . . . 15
1.4 Results of Cointegration Analysis for all Countries. . . 16
1.5 Generalized Impulse Responses . . . 17
1.B.1 France: Bivariate Cointegration Tests . . . 38
1.B.2 Germany: Bivariate Cointegration Tests . . . 39
1.B.3 Ireland: Bivariate Cointegration Tests . . . 40
1.B.4 Italy: Bivariate Cointegration Tests . . . 41
1.B.5 The Netherlands: Bivariate Cointegration Tests . . . 42
1.B.6 Portugal: Bivariate Cointegration Tests . . . 43
1.B.7 Spain: Bivariate Cointegration Tests . . . 44
2.1 The Spillover Matrix . . . 57
2.2 The spillover matrix of EA sovereigns and banks (on 21 June 2012) 64 2.3 Net Spillover matrix of EA sovereigns and banks (on 21 June 2012) 65 2.4 Ranking of NET senders and receivers of spillover effects on the 18 July 2011 . . . 72
2.5 Ranking of NET senders and receivers of spillover effects on the 21 June 2012 . . . 73
2.6 Critical spillover levels . . . 78
2.A.1 Composition and description of bank-specific and exogenous vari- ables . . . 87
2.A.2 Descriptive Statistics . . . 88
2.A.3 Country-specific bank assets and the weight in the country bank index 89 2.A.4 Selected events and the cumulative returns of contagion indices . . 90
2.D.1 The spillover matrix of EA sovereigns and banks (on 18 July 2011) 94 2.D.2 Net spillover matrix (on 18 July 2011) . . . 95
3.A.2 Model parameters . . . 146
3.A.3 Credit risk parameters . . . 148
3.A.4 S&P’s credit ratings transition matrix, in percent . . . 149
3.F.1 Network properties - 2009 Q1 . . . 160
3.F.2 Network properties - 2007 Q1 . . . 161
3.F.3 Network properties - 2005 Q1 . . . 162
This thesis concludes four remarkable years in which I have been a Ph.D. student in Economics and Finance at the University of Konstanz.1 It is time to thank everyone who helped me walk this path and shared with me great moments, interesting ideas, and thoughtful discussions.
First, I would like to express my deepest gratitude to Professor Günter Franke, whose unconditional support and supervision has been invaluable. His thoughts and advice helped me develop most of my knowledge related to systemic risk, macro- prudential regulation, and financial contagion. He also backed my research visits at the European Central Bank and Deutsche Bundesbank. His assistance has been clearly invaluable.
I am very thankful to the members of my thesis evaluation committee: Professors Almuth Scholl and Heinrich Ursprung. Moreover, Professors Ralf Brüggemann and Winfried Pohlmeier have contributed with technical expertise in econometrics and I am very grateful to them. I am also indebted to my colleagues Moritz Heimes, Ferdinand Graf, Steffen Seemann, Matthias Draheim who provided me useful com- ments on my research papers. Elvira Grübel, Michal Marenčák and Angelina Jegel helped me with many administrative issues but the most important achievement is the successful organization of the Final Marie Curie Conference in April 2013.
Second, I would like to thank my former colleagues at the ECB. My supervisor and co-author, Andreas Beyer, has been always committed to our joint work. I am also grateful to him for teaching me the first steps in central banking and financial stability policy. I had amazing moments and I am thankful to my colleagues and friends made during my internship at the ECB: Fiona, Jasmien, Jolyn, Nora, Valerie, Galen, Guillaume, Daniel, David, Marco, Nicollo, Nicola, and Wolf. I am sure the list should be much longer here, hence, I am also thankful to all omitted ones.
During my research secondment at the Deutsche Bundesbank I have collaborated with Peter Raupach and Ben Craig. I am grateful for their research commitment and
1Marie Curie Doctoral Fellowship in Risk Management from European Community’s Sev- enth Framework Programme FP7-PEOPLE-ITN-2008 (grant agreement PITN-GA-2009-237984) is gratefully acknowledged.
Ruprecht, Klaus Düllmann, Thomas Kick, and Barbara Meller.
The Marie Curie network and support has been very useful during my Ph.D.
research. Special thanks to Professor Ser-huang Poon, Claire Faichnie, Fady Bar- soum, Anton Golub, Eberhard Mayerhofer, Kyle Moore, Pengfei Sun, Zhen Guo, Yiran Zhang, Shibashish Mukherjee, Hossein Khatami, Heikki Seppala, Armin Eder, Kebin Ma, and Rachel Lidan.
Third, I want to express my appreciation to Camelia Minoiu, who has been always supportive and contributed with enlightening comments and discussions to my research and career path. Thank you for being part of this success.
I had a great time in Konstanz with two of my best friends Yves Schüler, with whom I wrote my first research paper, and Fabian Krüger. Moreover, I want to express my gratitude to my friends Bianca, Tudor, Denisa, Recca, Ruben, Tim, Anna, Daniel, Anca and Stefano, with whom I spent memorable moments.
Finally, and most importantly, I am thankful to my family. Without the support of my parents, Iancu and Marilena, and my grandparents, things would have been much more difficult. They have always shown their love and encouragement and I am very grateful to them.
Adrian Alter
Konstanz, 19 June 2013
One of the lessons that emerged from the global financial crisis 2007-2009 and the recent European sovereign debt crisis is that the institutional framework for super- vising and regulating financial systems has to be reformed. One of the novel policy direction that gained interest and attention is macro-prudential supervision. The interaction with micro-prudential regulation and monetary policy, among others, adds to the complexity of the global financial system.
This dissertation has at its core the concept of systemic risk. Seen as a feature of financial systems, systemic risk can be related to the likelihood of an institution, an asset class or a group of institutions to harm the financial stability with reper- cussions on the real economy (see e.g. Kaufman (1995)). The aim of this thesis is to tackle this concept from different angles. First, it sheds light on the implications of the nexus between the default risk of sovereigns and the financial sector. This direction is covered by an empirical analysis of the credit default swap (CDS, here- after) market. The dynamics of interconnectedness between financial institutions and countries are at the roots of a novel feature observed during the last four years in the Eurozone: the feedback loop between the stability of the banking system and the macroeconomic health of sovereigns. Moreover, the phenomenon of financial contagion has attracted the attention of academia, policy makers and market par- ticipants. Constâncio (2012), the vice-president of the ECB, refers to contagion as
“one of the mechanisms by which financial instability becomes so widespread that a crisis reaches systemic dimensions.[. . . ] As a consequence, crisis management by all competent authorities should also focus on policy measures that are able to contain and mitigate contagion.”Allen and Gale(2000) explain “contagion” as a consequence of excess spillover effects. For example, a banking crisis in one region may spread to other regions. Contagion in their view is the phenomenon of extreme amplifica- tion of spillover effects. Therefore, spillover effects are a necessary, but not sufficient, condition for contagion. Finally, this thesis is motivated by an acute need for policy- relevant methodologies and frameworks to deal with systemically important financial institutions (SIFIs hereafter), especially for large financial systems. Using measures
of centrality derived from the interbank network, I intend to provide more insight to the too-big-to-fail versus too-interconnected-to-fail discussion. Capital allocations based on both, riskiness and size of individual bank assets combined with metrics of interconnectedness constructed from the entire banking network appear to improve the robustness of financial system.
Let me give an overview of this dissertation. Chapter 1 presents an analysis of the interdependencies between the credit risk of Eurozone sovereigns and financial sector during the financial crisis. Generalizing Chapter 1, in Chapter 2 we pro- pose a framework of tracking and monitoring financial contagion in the Eurozone.
Furthermore, this chapter intends to provide a toolbox that allows policy makers to determine the impact of political/policy events on stemming or fueling financial contagion. Finally, Chapter 3 presents a tractable framework to deal with SIFIs. By proposing two policy directions, we show that network measures derived from the interbank market and the size of financial institutions could help to improve the re- silience of large financial systems. In the rest of this summary I provide an overview of the main ideas and results of each chapter and highlight the contributions to the literature.
Chapter 1 investigates empirically the relationship between the credit risk of several Eurozone countries (France, Germany, Italy, Ireland, the Netherlands, Por- tugal, and Spain) and their domestic banking sectors during the period 2007 - 2010, using daily CDS spreads.2 Bank bailouts changed the composition of both banks’
and sovereign balance sheets and, moreover, affected the linkage between the default risk of governments and their local banks. Our main findings suggest that in the pe- riod before bank bailouts the contagion disperses from bank credit spreads into the sovereign CDS market. After bailouts, a financial sector shock affects sovereign CDS spreads more strongly in the short run. However, the impact becomes insignificant in the long term. Furthermore, government CDS spreads become an important deter- minant of banks’ CDS spreads. Yet, there exist clear-cut differences between strong and weak member states. The relationship between government and bank credit risk is heterogeneous across countries, but homogeneous within the same country. These findings help to better understand the interaction between bank and sovereign risks and shed light on the private-to-public risk transfer.
Moreover, focusing on the effects of bank bailouts on the linkage between CDS spreads of governments and their local banks, we contribute to the literature in the following ways: i) relying on previous studies that emphasize the importance
2Chapter 1 is a reprint of the published paper “Credit spread interdependencies of European states and banks during the financial crisis”, Journal of Banking and Finance, Volume 36, Issue 12, December 2012, pp. 3444 - 3468, joint work with Yves S. Schüler (University of Konstanz).
of the domestic financial sector as a determinant of sovereign CDS spreads, we provide detailed empirical evidence for its influence during the financial crisis; ii) in contrast to other studies, we research on the credit risk interdependence of banks and governments during the last financial turmoil. Using this approach we highlight stark alterations of the latter linkage after bank bailouts and we contrast differences in the private-to-public risk transfer both within a country but also across the Eurozone.
In Chapter 2, we develop an analytical and empirical framework for measuring spillover effects by extending the econometric setup proposed by Diebold and Yilmaz (2009;2011). 3 We illustrate our method by an empirical application to the interlink- ages between sovereigns and banks in the Eurozone. By analyzing daily CDS data, we quantify those spillover effects based on an 80-days rolling window. We combine sovereign and bank CDS spreads in a vector autoregressive framework, augmented by several control variables (i.e. in order to deal with common risk factors, omitted variables). In our model we focus on 11 sovereigns and nine country-specific banking groups from the euro area, over the period October 2009 - July 2012. Furthermore, we rely on a generalised impulse response approach to assess the systemic effect of an unexpected shock to the creditworthiness of a particular sovereign or country- specific bank index. We aggregate this information into a Contagion Index. This index has four main components. Average potential spillover: i) among sovereigns, ii) among banks, iii) from sovereigns to banks and iv) vice-versa. Our measure can be used to highlight the potential contagion at a certain point in time or the time-dependent feature of the contagion index. This toolbox allows us to identify systemically relevant entities (i.e. country specific banking sectors and sovereigns) from the proposed set of sovereigns and banks in our system. Based on empirical distributions of CDS changes, we propose a simple method to compute thresholds for “excessive” spillovers. Excessive spillovers are a characteristic of dysfunctional markets and can be regarded as a source of contagion and systemic risk. When financial variables (e.g. markets, participants, intermediaries) are characterized by extreme dependence, financial systems become unstable.
Furthermore, we show the dynamics of the nexus between banks and sovereigns, that represents a potential source of systemic risk. Euro area sovereign creditworthi- ness carries a growing weight in the overall financial market picture, with a subset of sovereigns that can potentially produce negative externalities to the financial sys- tem. We find that several previous policy interventions had a mitigating impact on spillover risks. A potential unexpected shock to Spanish sovereign CDS spread
3Chapter 2 is a partial rewrite of the CFS Working Paper No. 2012/13 and ECB Working Paper No. 2013/1558 “The dynamics of spillover effects during the European sovereign debt turmoil”, joint work with Andreas Beyer (ECB).
reveals an elevated impact on both spreads of euro area sovereigns and banks during the first half of 2012, compared to 2011. Moreover, spillover effects from a shock to Spanish sovereign CDS to Eurozone core countries and to non-core countries become more similar in magnitude during 2012. We also highlight the stark amplification of the nexus between sovereigns and banks until the end of June 2012, the announce- ment of the Banking Union. Focusing on policy relevant interventions, we observe that the systemic contributions of Greece, Portugal and Ireland decrease remarkably after the implementation of IMF/EU programs. Moreover, the setup of the Euro- pean Financial Stability Facility (EFSF) and the decision of the two Long-Term Refinancing Operations (LTRO) in December 2012 have a mitigating impact on all four contagion index components. This chapter suggests a macro-prudential toolbox for measuring the potential contagion in the euro area using market data.4 It can be adapted to the needs of policy makers by integrating other banks or sovereigns or extending it to real economy variables. Furthermore, we attempt to show its use- fulness in quantifying the potential effects of different policy measures on containing spillovers across the system.
Applying network theory and analysis, we determine inChapter 3 different capi- tal rules with the scope of minimizing total system losses.5 Using the German Credit Register database, several measures of centrality are constructed from the network topology of the interbank market that help us designing systemic capital buffers.
Capital is reallocated among banks based on network centrality measures in order to minimize the expected losses in the banking system. The so-called Opsahl cen- trality turns out to dominate any other centrality measure tested, apart from total assets. Furthermore, we use centrality measures to implement a bailout fund mecha- nism. The bailout fund offers insurance against the default of certain banks, however with priorities depending on banks’ size and centrality. Finally, we compare the to- tal system loss across different types of capital allocation and sizes of the bailout fund. We attempt to draw policy conclusions related to too-interconnected-to-fail versus too-big-to-fail discussion. We show that there are certain capital allocations that are able to improve financial stability. Focusing on the system as a whole and assigning capital allocations based on network metrics yields better results than the benchmark capital allocation that is based solely on the individual bank balance sheet. The improvement comes from getting the big picture of the entire system where interconnectedness and centrality play a major role in triggering and ampli- fying contagious defaults. What is interesting is that capital allocations based on
4This toolbox has been applied also in ECB2012, pg. 73, Box 5.
5Chapter 3 has been partially written together with Peter Raupach and Ben Craig, both from the Deutsche Bundesbank.
total assets dominate any other centrality measure tested. These results strengthen previous findings that systemic capital requirements should depend mainly on total assets as proposed by Tarashev et al. (2010). One could improve even further the system’s stability, by combining total assets and network metrics on top of individual bank asset riskiness (measured for example by the individual bank Value-at-Risk).
In this last chapter, we propose a tractable framework to assess the impact of different capital allocations on financial stability. We integrate a sound credit risk engine (i.e. CreditMetrics) to generate correlated shocks to credit exposures of the entire German banking system, where about 1750 banks are active in the interbank market. This engine gives us the opportunity to focus on correlated tail events, endogenously determined by the composition of bank balance sheets. Moreover, we model interbank contagion based on a clearing mechanism, as described firstly by Eisenberg and Noe(2001), and extend it to include bankruptcy costs as proposed by Elsinger et al. (2006). This feature allows us to measure expected contagion losses and to observe the propagation process in the interbank market. To empirically exemplify our framework, we use several sources of information: German central credit register (for large exposures), aggregated credit exposures (for small loans), balance sheet data (i.e. total assets, total sector exposures), and market data (to compute correlations between real economy sectors or credit spreads). We focus on capital reallocationsand try to minimize a target function with the scope of improv- ing financial stability. We intend to use severaltarget functions: total system losses, approximated by total bankruptcy costs, second-round contagion effects (i.e. conta- gious defaults) or losses from fundamental defaults (i.e. banks which default due to
“real-economy” portfolio losses). Furthermore, we determine capital allocations that improve the resilience of the financial system based on interconnectedness measures constructed from the interbank network or associated with the size and riskiness of bank balance sheets (e.g. total assets, total interbank liabilities, eigenvector centralities, Opsahl centrality, closeness or the clustering coefficient). Our method which deals with interconnected financial systems is an alternative to market-based systemic measures (e.g Acharya et al. (2010); Adrian and Brunnermeier (2011);
Gauthier et al.(2012)) when tackling the too-interconnected-to-fail externality. The main advantage of our framework is that policy makers can deal with large banking systems where market data is not available for most of the institutions.
To sum up, this dissertation intends to familiarize the reader with several impor- tant concepts related to macro-prudential supervision and financial stability. More- over, it provides empirical applications to show case these concepts and ignite further research on these topics.
Keywords: interdependencies, systemic risk, financial contagion, interconnect- edness, private-to-public risk transfer, sovereign - financial sector feedback loop, SIFIs, large financial system, credit default swap (CDS), systemically important market, contagion index, systemic contribution, liquidity risk, credit risk, macro- prudential supervision, regulation, mutual exposures, interbank network, common shocks, collateral, bailout fund, resolution mechanism, lender-of-last-resort, trans- parency, moral hazard, uncertainty, stress tests, cointegration, externalities, too-big- to-fail, degree, betweenness, closeness, eigenvector centrality.
Eine der Lektionen aus der globalen Finanzkrise 2007-2009 und der Europäischen Staatsschuldenkrise, die Anfang des Jahres 2010 begann, ist, dass eine Reformierung des institutionellen Rahmens für die Überwachung und Regulierung des Finanzsys- tems notwendig ist. Eine neue politische Strategie, die bereits auf grosses Interesse stiess, ist die makroprudentielle Aufsicht. Die Interaktion zwischen der Geldpolitik und der unter anderem mikroprudentiellen Regulierung verstärkt die Komplexität des globalen Finanzsystems.
Der Kern dieser Dissertation ist das Konzept des systemischen Risikos. Das systemische Risiko, das als Eigenschaft des Finanzsystems gesehen wird, stellt die Wahrscheinlichkeit einer Institution, einer Anlagenklasse oder einer Gruppe von In- stitutionen dar, die Finanzmarktstabilität zu gefährden, was wiederum Auswirkun- gen auf die Realwirtschaft hat (z.B.Kaufman(1995)). Das Ziel der Dissertation ist es das systemische Risiko von mehreren Seiten zu betrachten.
Zuerst werden die Auswirkungen der Verbindung des Ausfallrisikos staatlicher Organe mit dem Finanzsektor analysiert. Diese Analyse umfasst eine empirische Auswertung des Credit-Default-Swap Marktes. Die Dynamiken der Vernetzung der Finanzinstitutionen mit den Ländern sind die Ursache für eine neuentdeckte Eigenschaft: eine Rückkopplungsschleife zwischen den Problemen des Bankensys- tems und der makroökonomischen Stabilität der staatlichen Organe. Diese Eigen- schaft wurde während der letzten vier Jahre in der Eurozone beobachtet. Zudem hat der finanzielle Contagion die Aufmerksamkeit von Akademikern, Politikern und Marktteilnehmern während der jüngsten Europäischen Staatsschuldenkrise auf sich gezogen. Constâncio (2012), der Vize-Präsident der EZB, beschreibt Contagion als
“einen der Mechanismen, durch die die finanzielle Instabilität so weit verbreitet wird, dass eine Krise systemische Dimensionen erreicht [. . . ] Folglich sollte das Krisenman- agement durch alle kompetenten Behörden auch politische Massnahmen beachten, die in der Lage sind, den Contagion- Effekt einzugrenzen und abzuschwächen.” Nach Allen and Gale (2000) ist Contagion eine Konsequenz von übermässigen Spillover - Effekten. Zum Beispiel könnte die Bankenkrise einer Region auf andere Regio-
nen übergehen. Ihrer Ansicht nach ist Contagion daher die extreme Verstärkung der Spillover - Effekte. Deshalb sind Spillover - Effekte notwendige - jedoch keine hinreichenden - Bedingungen für den Contagion- Effekt. Letztlich wurde die Thesis motiviert durch einen akuten Bedarf an politisch relevanten Methoden und Rah- menbedingungen, um mit systemisch wichtigen Finanzinstitutionen - insbesondere mit den grossen Finanzsystemen - umzugehen. Ich beabsichtige, unter der Verwen- dung der Network Theorie und der Analyse mehr Einsicht für Diskussionen über die Auswirkungen der Theorien Too-Big-to-Fail und Too-Interconnected-to-Fail zu geben. Die Kapital-Allokation, die von der Grösse und dem Risikogehalt einer An- lage einer individuellen Bank abhängt, verknüpft mit Massen für die finanzielle Ver- netzung, die von dem gesamten Banken-Netzwerk konstruiert wurde, scheint die Widerstandsfähigkeit des Finanzsystems zu verbessern.
Im Folgenden gebe ich Ihnen einen Überblick über meine Dissertation. Im ersten Kapitel präsentieren wir eine Analyse der Interdependenzen des Kreditrisikos eines europäischen Staates mit dem Finanzsektor während der Finanzkrise 2007 − 2010.
Den ersten Teil verallgemeinernd, stellt das zweite Kapitel einen Rahmen für die Nachverfolgung und Beobachtung des finanziellen Contagion- Effekts in der Euro- zone auf. Weiterhin stellt es eine Toolbox bereit, die es den Politikern ermöglicht, die Auswirkungen von politischen Ereignissen auf das Verringern oder Verstärken des finanziellen Contagion- Effekts zu ermitteln. Schlussendlich präsentiert das dritte Kapitel lenkbare Rahmenbedingungen, um mit SIFIs umzugehen. Wir zeigen, dass Network - Masse des Interbankenmarktes und die Grösse der Finanzinstitution dabei helfen könnten, die Widerstandsfähigkeit grosser Finanzsysteme zu verbessern, in- dem wir zwei Politikrichtungen vorschlagen. Die wichtigsten Ideen und Ergebnisse jedes Kapitels werden im Folgenden kurz zusammengefasst.
Im ersten Kapitel untersuchen wir die Wechselbeziehung zwischen dem Ausfall- risiko einiger europäischen Länder (Frankreich, Deutschland, Italien, Irland, Nieder- lande, Portugal und Spanien) und deren inländischen Bankensektoren im Zeitraum 2007−2010. Dabei nutzten wir die täglichen CDS-Spreads. Banken Bailouts verän- derten die Zusammensetzung der Bilanzen sowohl von Banken als auch von Staat- sorganen. Zudem wirkten sich die Banken-Bailout-Programme auf die Koppelung des Ausfallsrisikos von Staaten und deren lokalen Banken aus. Unsere wichtig- sten Ergebnisse deuten darauf hin, dass der Contagion- Effekt in der Zeit vor den Banken- Bailouts auf den CDS- Markt überging. Nach den Bailouts beeinträchtigt ein Schock in den Finanzsektoren die staatlichen CDS-Spreads kurzzeitig sogar noch stärker. Langzeitig hatte das jedoch keine signifikanten Auswirkungen. Darüber hinaus wurden staatliche CDS-Spreads ein wesentlicher Bestandteil in den CDS-
Reihen der Banken. Noch immer existieren deutliche Unterschiede zwischen den starken und schwachen Mitgliedsstaaten. Zwischen verschiedenen Ländern sind die Interdependenzen der Kreditrisiken von Regierungen und Banken heterogen, aber homogen innerhalb eines Landes. Diese Ergebnisse führen zu einem besseren Ver- ständnis der Interaktion zwischen Banken und staatlichen Risiken und geben Auf- schluss über den Private-to-Public Risikotransfer. Wir konzentrieren uns auf die Effekte der Banken-Bailout- Programme auf die Verbindungen zwischen staatlichen CDS-Spreads und ihren lokalen Banken. Dabei verwenden wir den Beitrag der Lit- eratur folgendermassen: i) Indem wir uns auf frühere Studien, die die Bedeutung von heimischen Finanzsektoren als einen Bestandteil der staatlichen CDS-Spreads hervorheben, stützen, liefern wir detaillierte empirische Nachweise für ihren Ein- fluss während der Finanzkrise. ii) Im Gegensatz zu anderen Studien erforschen wir die Interdependenz des Kreditrisikos von Banken und Staaten während der letzten Finanzkrise. Unter Verwendung dieses Ansatzes heben wir reine Änderungen der let- zteren Verbindung nach Banken- Bailouts hervor. iii) Wir stellen die Unterschiede in Private-to-Public-Risikotransfers innerhalb eines Landes und innerhalb der Euro- zone gegenüber.
Durch die Erweiterung des ökonometrischen Aufbaus von Diebold and Yilmaz (2009; 2011) stellen wir im zweiten Kapitel analytische und empirische Rahmenbe- dingungen für die Messung von Spillover- Effekten vor. Durch eine empirische An- wendung von Verknüpfungen zwischen Staatsorganen und Banken innerhalb der Eurozone veranschaulichen wir unsere Methode. Bei der Analyse täglicher CDS- Daten quantifizierten wir diese Spillover- Effekte basierend auf einem 80 Tagen langen Zeitfenster. Wir kombinierten die CDS- Spreads von Staatsorganen und Banken in einem autoregressiven Vektorsystem. Dieser Rahmen wurde durch einige Kontrollvariablen erweitert (z.B. um gewöhnliche Risikofaktoren oder weggelassene Variablen zu handhaben). Der Fokus unseres Modells lag auf elf Staatsorganen und neun länderspezifischen Bankengruppen der Eurozone im Zeitraum Oktober 2009
− Juli 2012. Um den systemischen Effekt von unerwarteten Schockzuständen auf die Kreditwürdigkeit eines bestimmten Staatsorganes oder einen länderspezifischen Bankenindex zu bewerten, berufen wir uns auf den Impulse-Response Ansatz. Die Informationen sammelten wir in einem Contagion-Index. Dieser Index beinhaltet vier Komponenten. Er kann in durchschnittliche potenzielle Überschüsse aufges- plittet werden: i) unter Staatsorganen, ii) unter Banken, iii) von Staatsorganen zu Banken und iv) umgekehrt. Unser Mass kann genutzt werden, um den Zus- tand des potenziellen Contagion zu einem bestimmten Zeitpunkt oder die zeitab- hängige Eigenschaft des Contagion-Index hervorzuheben. Diese Toolbox ermöglicht
es uns, systemrelevante Einrichtungen (z.B. länderspezifische Bankensektoren und Staatsorgane) der in unserem System befindlichen Staatsorganen und Banken zu identifizieren. Wir stellen eine einfache Methode bereit, um die Schwelle der exzes- siven Überschüsse zu berechnen. Diese Methode basiert auf empirischen Verteilun- gen der CDS-Veränderungen. Exzessive Überschüsse sind ein Charakteristikum von dysfunktionalen Märkten und können als eine Quelle des Contagion und des sys- temischen Risikos betrachtet werden. Finanzsysteme neigen zur Instabilität, wenn Finanz-Variablen (z.B. Märkte, Marktteilnehmer, Finanzvermittler) durch eine ex- treme Abhängigkeit ausgezeichnet sind. Eine solche extreme Abhängigkeit kann während ruhiger Zeiten nicht beobachtet werden. Ausserdem zeigen wir die Dy- namiken der Verknüpfung zwischen Banken und Staatsorganen, die eine potentielle Quelle des systemischen Risikos darstellen. Die staatliche Kreditwürdigkeit in der Eurozone gibt dem Gesamtbild des Finanzmarktes zusätzliches Gewicht, mit einem Teil der Staatsorgane, die möglicherweise negative externe Effekte auf das Finanzsys- tem haben. Wir fanden heraus, dass einige der früheren politischen Interventionen einen mildernden Einfluss auf das Spillover- Risiko hatten. In unserer Anwendung bemerkten wir, dass ein Schock in den staatlichen CDS Spaniens während des er- sten Halbjahres 2012 einen erhöhten Einfluss auf Staaten und Banken der Eurozone hatte, verglichen mit dem Jahr 2011. 2012 waren zudem die Auswirkungen der Spillover, die aus einem Schock in den staatlichen CDS Spaniens resultierte, auf die Kernländer der Eurozone ähnlich zu den Auswirkungen auf die Nicht-Kernländer der Eurozone. Wir fanden auch einen erheblichen Hinweis dafür, dass die Verknüpfung zwischen Staatsorganen und Banken bis Ende Juni 2012 verstärkt wurden. Jedoch verringerte sich der systemische Beitrag von Griechenland, Portugal und Irland nach der Einführung des IMF/EU Programms beachtlich. Das Einführen der EFSF und die Beschlüsse der beiden LTROs im Dezember 2012 haben einen mildernden Ein- fluss auf alle vier Contagion-Index-Komponenten. Dieses Kapitel präsentiert eine makroprudentielle Toolbox, um den potentiellen Contagion- Effekt in der Eurozone unter Verwendung von Marktdaten zu messen. Dies kann den Bedürfnissen der Politiker angepasst werden, indem man andere Banken und Staatsorgane integri- ert oder es zu realwirtschaftlichen Variablen erweitert. Ausserdem versuchen wir, ihre Nützlichkeit darzustellen, indem wir die potenziellen Effekte von verschiedenen politischen Massnahmen, die Spillovers im System mit sich bringen, quantifizieren.
Im dritten Kapitel bestimmen wir verschiedene Kapitalregeln mit der Absicht, den Totalverlust des Systems zu minimieren. Dabei verwendeten wir die Network- Theorie und -Analyse. Aus der Network- Struktur des Interbankenmarktes wurden einige Zentralitätsmasse konstruiert, die uns dabei helfen, systemische Kapitalre-
serven zu kreieren. Als Quelle nutzten wir die Deutsche Kreditregister- Daten- basis. Unter Verwendung der Netzwerk-Zentralitätsmasse ist das Kapital unter den Banken neu verteilt worden, um die erwarteten Verluste im Bankensystem zu minimieren. Abgesehen von der Bilanzsumme dominiert die sogenannte Opsahl- Zentralität jedes andere getestete Zentralitätsmass. Ausserdem nutzen wir Zentral- itätsmasse, um Bailout-Fond-Mechanismen einzuführen. Der Rettungsfond bietet Versicherungsschutz gegen den Ausfall bestimmter Finanzinstitutionen, jedoch wer- den die Finanzinstitute abhängig von ihren Ratings bezüglich der Bankengrösse und Zentralität bevorzugt. Schliesslich vergleichen wir den Gesamtsystemverlust in Bezug auf die verschiedenen Arten der Kapitalverteilung und die verschiedenen Grössen des Bailout-Fonds. Wir versuchen politische Schlussfolgerungen bezüglich der externen Effekte,Too-Interconnected-to-FailundToo-Big-to-Fail, zu ziehen. Wir zeigen, dass es gewisse Kapitalallokationen gibt, die die Finanzstabilität verbessern.
Diese gewissen Kapitalallokationen sind in dieser Thesis definiert. Mit Fokus auf das System als ein ganzes und zuordnendes System produzieren Kapitalallokatio- nen auf Basis der Network -Masse hervorragende Ergebnisse, verglichen mit der Bezugs-Kapitalallokation, die nur auf den individuellen Bankenbilanzen beruht. Die Ähnlichkeiten der Portfolios machen das Finanzsystem gegenüber normalen Makro- Schocks anfällig. Bekommt man einen grossen Überblick des gesamten Systems, in dem Vernetzung und Zentralität eine wesentliche Rolle des Contagion-Effekts übernehmen, so kann eine Verbesserung erzielt werden. Das Interessante daran ist, dass Kapitalallokationen, die auf dem Gesamtvermögen basieren, jedes an- dere getestete Zentralitätsmass dominieren. Diese Ergebnisse verstärken die Un- tersuchungsergebnisse von Tarashev et al. (2010), dass die systemischen Kapitaler- fordernisse hauptsächlich von dem Gesamtvermögen abhängen sollten. Indem man das Gesamtvermögen zusätzlich zum individuellen Risiko des Bankenvermögens mit denNetwork-Massen kombiniert, könnte die Systemstabilität sogar weiter verbessert werden (das z.B. durch das VaR individueller Banken gemessen wird).
Des Weiteren schlagen wir lenkbare Rahmenstrukturen vor, um den Einfluss von den verschiedenen Kapitalallokationen auf die Finanzmarktstabilität zu bew- erten. Wir integrieren ein gut fundiertes Kreditrisikomodell (z.B. CreditMetrics), um korrelierende Schocks von Kreditrisiken auf das gesamte deutsche Bankensys- tem zu generieren (1764 Sind Monetäre Finanzinstitute (MFIs) im Interbanken- markt (IB) aktiv). Diese Modelle ermöglichen, dass wir uns auf korrelierende Nach- effekte fokussieren können (endogen bestimmt durch gewöhnliche Risiken in der Realwirtschaft). Zudem modellieren wir auf Basis von Eisenberg and Noe (2001) den Interbank-Contagion und erweitern ihn, indem wir die Bankrottkosten von
Elsinger et al. (2006) einbeziehen. Dieses Element erlaubt uns, die erwarteten Con- tagion Verluste zu messen und den Verbreitungsprozess zu beobachten. Um unsere Rahmenbedingungen empirisch zu belegen, nutzen wir einige Informationsquellen:
Die Gross- und Millionenkreditstatistik der Deutschen Bundesbank (Grosse Kred- ite), die Kreditnehmerstatistik der Deutschen Bundesbank (Kleine Kredite), Bilanz- daten (Kapital, Gesamtvermögen) und Marktdaten (z.B. um die Korrelation zwis- chen realwirtschaftlichen Sektoren, Ratingtabellen oder Credit Spreads zu berech- nen). Wir fokussieren uns auf die Kapital-Neuverteilungen und versuchen, die ver- schiedenen Zielfunktionen in der Absicht die Finanzmarktstabilität zu verbessern, zu minimieren. Wir beabsichtigen einige Zielfunktionen zu nutzen: Gesamtsys- temverlust, Contagion Effekte der zweite Runde (z.B. Ausfällen durch das Conta- gion) oder Verluste von fundamentalen Ausfällen (z.B. Banken, die aufgrund von realwirtschaftlichen Portfolioverlusten ausfallen). Zum Beispiel können Gesamtsys- temverluste als Gesamtkosten des Bankrotts ausgefallener Banken definiert wer- den. Ausserdem bestimmen wir Kapitalallokationen, die die Belastbarkeit des Fi- nanzsystems verbessern (wie durch unsere Zielfunktionen definiert ist) auf Basis der Verknüpfungsmassnahmen des IB Network. Um neue Kapitalregeln zu bestimmen und diese mit unseren Rahmenbedingungen zu testen, verwenden wir einige Risiko- undNetwork- basierende Konnektivitätsmasse: Gesamtvermögen, die gesamte Inter- bank Aktiva und Passiva, den Grad, den Eigenvektor und die gewichteten Eigenvektor- Zentralitäten, das gewichtete Betweenness -Mass, die Opsahl- Zentralität, der Nähe- und der Clusterkoeffizient. Diese Rahmenbedingungen können auf jedes Land oder auf jede Gruppe von Ländern, in denen solche Informationen verfügbar sind, angewen- det werden. Diese Methode ist eine Alternative zu marktbasierenden systemischen Massen, um die Verknüpung von Finanzsystemen zu handhaben (z.B.Acharya et al.
(2010); Adrian and Brunnermeier (2011); Gauthier et al. (2012)). Der hauptsäch- liche Vorteil unserer Rahmenbedingungen ist, dass Politiker mit grossen Bankensys- temen, in denen Marktdaten für den grössten Teil der Institutionen - auch einigen sehr grossen Einrichtungen - nicht verfügbar sind, umgehen können.
Bibliography
Acharya, V., L. H. Pedersen, T. Philippon, and M. Richardson (2010). Measuring systemic risk. Mimeo.
Adrian, T. and M. K. Brunnermeier (2011). Covar. Federal Reserve Bank of New York Staff Reports No. 348.
Allen, F. and D. Gale (2000). Financial contagion.Journal of Political Economy 108, 1–33.
Constâncio, V. (2012). Contagion and the european debt crisis. Banque de France, Financial Stability Review No. 16, 109 – 121.
Diebold, F. X. and K. Yilmaz (2009). Measuring financial asset return and volatility spillovers, with application to global equity markets. Economic Journal 119, 158 – 171.
Diebold, F. X. and K. Yilmaz (2011). On the network topology of variance decom- positions: Measuring the connectedness of financial firms. PIER Working Paper, 11 – 031.
ECB (2012). Financial stability review. December.
Eisenberg, L. and T. Noe (2001). Systemic risk in financial systems. Management Science 47(2), 236–249.
Elsinger, H., A. Lehar, and M. Summer (2006). Risk assessment for banking systems.
Management Science 52, 1301–1314.
Gauthier, C., A. Lehar, and M. Souissi (2012). Macroprudential capital requirements and systemic risk. Journal of Financial Intermediation 21, 594–618.
Kaufman, G. (1995). Research in Financial Services: Banking, Financial Markets, and Systemic Risk, Chapter "Comment on Systemic Risk", pp. 47–52. JAI Press.
Tarashev, N., C. Borio, and K. Tsatsaronis (2010). Attributing systemic risk to individual institutions. BIS Working Papers No. 308.
Credit spread interdependencies of
European states and banks
“The scope and magnitude of the bank rescue packages also meant that signifi- cant risks had been transferred onto government balance sheets. This was particu- larly apparent in the market for CDS referencing sovereigns involved either in large individual bank rescues or in broad-based support packages for the financial sector.”
(BIS,2008, p. 20)
1.1 Introduction
During the recent financial crisis extraordinary measures were taken by central banks and governments to prevent a potential collapse of the financial sector that threat- ened the entire economy.1 However, it was widely unknown what the effects would be on the interdependence of the financial and the sovereign sector. Gray(2009, p. 128) argues that “regulators, governments, and central banks have not focused enough on the interconnectedness between financial sector risk exposures and sovereign risk exposures and their potential interactions and spillovers to other sectors in the econ- omy or internationally”. The lack of theoretical macroeconomic models that are able to incorporate contagion mechanisms between government and financial sectors have amplified the uncertainty related to the implications of government interventions.
Nevertheless, regulators and policy makers need to understand the complex dy- namics of risk transmission in order to be able to formulate effective policies and be aware of the risk that may be transferred from the financial sector to the government.
This chapter proposes a framework for investigating in detail the interdependence of banks’ and sovereign credit risk in the Eurozone. Our setup highlights the important changes that have occurred due to the bank bailouts.
As pointed out by Gray et al. (2008), using arguments from contingent claims analysis (CCA)2, there are several channels linking the banking and sovereign sec- tors, which are affected by implicit as well as explicit guarantees. A systemic bank- ing crisis can induce a contraction of the entire economy, which will weaken public finances and transfer the distress to the government. This contagion effect is am- plified when state guarantees exist for the financial sector. As a feedback effect, risk is further transmitted to holders of sovereign debt. An increase in the cost of sovereign debt will lead to a devaluation of government debt, which will impair the
1This chapter is a reprint of the published paper “Credit spread interdependencies of European states and banks during the financial crisis”, Journal of Banking and Finance, Volume 36, Issue 12, December 2012, pp. 3444 - 3468, joint work with Yves S. Schüler.
2This approach is based on Merton’s and Black-Scholes’ (1973) option pricing work. It can also be employed for measuring sovereign-bank interaction, taking into account the implicit and explicit contingent liability for the financial system.
balance sheets of banks holding these assets. Acharya et al. (2011) have recently used the term “two-way feedback” to describe these interdependencies. The authors construct a novel theoretical framework to model the link between bank bailouts and sovereign credit risk. In this chapter, we empirically study this feedback effect and show how the linkage between the sovereign and financial sectors was affected during the recent period of turmoil.
The interconnectedness through balance sheets of governments and banks has been described in the context of the financial crisis in other recent empirical studies.
For instance, Gerlach et al. (2010) find that, as a consequence of macroeconomic imbalances, especially in peripheral European countries (e.g. Greece, Ireland), a jump in sovereign bond and credit default swap (CDS) spreads may be transmit- ted from the banking sector. The authors claim that systemic and sovereign risk became more interwoven after governments began to issue guarantees for banks’
liabilities. This result is supported by Ejsing and Lemke (2011), who argue that the sensitivity of sovereign CDS spreads to the intensifying financial crisis increased after the bailout of the financial sector. Dieckmann and Plank (2011) also present evidence of a private-to-public risk transfer in the countries where governments sta- bilized the financial system after the Lehman Brothers’ event. Banks’ and sovereign CDS became closely linked, with financial institutions holding significant amounts of government debt and states bearing vital contingent liabilities from the finan- cial system. Furthermore, Acharya et al. (2011) provide empirical evidence of the interconnection of financial and sovereign sector credit risk as a result of bailout programs. Our study contributes to the literature in three ways: First, relying on previous studies that emphasize the importance of the domestic financial sector as a determinant of sovereign CDS spreads, we provide detailed empirical evidence of the influence of the domestic financial sector during the financial crisis. Second, in contrast to other studies, we research the credit risk interdependence of banks and governments during the recent turmoil. Using this approach we highlight stark changes that occurred in that interdependence after bank bailouts. Third, we study differences in the private-to-public risk transfer both within countries and across the Eurozone.
In more detail, we study the lead-lag relation between governments’ and banks’
default risk, with a focus on the effect of the bank bailouts in the midst of the recent financial crisis. First, we investigate whether, prior to the government interventions, an increase in the default risk of banks and states originates mainly from the finan- cial sector. Second, we assess whether public contingent liabilities for the financial sector affected governments’ default risk. In tandem, this study examines whether
the default risk of the banking sector is influenced by the sovereign default risk.
Finally, we investigate the following two questions: i) Does the perceived degree of a bank’s participation in a national rescue scheme influence its dependency on the development of the sovereign spread? ii) Are country-specific bailout characteristics reflected in the impact of government bailout programs?
Methodologically, we consider the relationship between government and banks’
CDS spreads, as they provide a proxy for default risk.3 We conduct this analysis by applying the theory of cointegration, Granger-causality, and impulse responses to daily CDS series, which are able to capture changes in the dynamic relation between government and bank credit risk. We consider sovereign CDS from seven EU mem- ber states (France, Germany, Italy, Ireland, the Netherlands, Portugal, and Spain) together with a selection of bank CDS from these states. We divide the analyzed period, i.e. June 2007 until May 2010, into the time before and after bank bailout programs were implemented.
Our main findings suggest that in the periodpreceding government intervention, the contagion from bank credit spreads disperses into the sovereign CDS market. This finding can be interpreted as evidence of the systemic feature of the recent financial crisis. The default risk spills over from the financial system to the entire econ- omy and calls into question the government’s capacity to repay its liabilities. After government interventions, due to changes in the composition of both banks’ and sovereign balance sheets, we find that the government CDS spreads have increased importance in the price discovery mechanism of the banks’ CDS series. Further- more, a financial sector shock affects the sovereign CDS spreads more strongly in the short run. However, the impact becomes insignificant in the long term. Based on a bank’s dependency on future government aid, we are able to capture differences and similarities in the outcomes of bank bailouts within the same country. Finally, our cross-country analysis reveals noticeable differences in the outcomes of state in- terventions.
From a policy perspective, our results imply an elevated financing cost for coun- tries with contingent liabilities from the financial sector and a higher volatility in sovereign yield spreads. In assessing the total cost of bank bailouts, governments need to include increased interest payments due to augmented spreads. Further- more, the banking system is sensitive to the economic health of the host country
3The objective of this chapter is not to investigate the accuracy of this proxy. Our research design takes this link as given, even though there might have been distortions in this proxy during the recent turmoil.
and the credibility of the support measures.
This chapter is organized as follows. In Section 2 we discuss studies related to our research. Section 3 presents our hypotheses, the data, our sub-sample selection pro- cedure, and the methodology. In Section 4 we present our results and Section 5 concludes.
1.2 Related literature
Our study contributes to, at least, two strands of literature: On the one hand, it is linked to the literature that investigates the determinants of bond and CDS spreads and their returns, especially in the midst of financial crisis. On the other, it is related to the analysis of the effects of bank bailouts on the credit risk of governments and banks.
Tied to the first strand and relying on a structural model, Schweikhard and Tsesmelidakis (2009) conclude that credit and equity markets were decoupled dur- ing the financial turmoil. They find support for the “too-big-to-fail” hypothesis, as some companies’ debt holders benefited from government interventions, and a shift of wealth took place from taxpayers to creditors after the bailout programs. During the crisis, some other factors might have influenced CDS prices (e.g. counterparty or liquidity risk). Collin-Dufresne et al. (2001) find that changes in credit spreads are mostly driven by a systematic factor; however, they are not able to identify it.
Berndt and Obreja (2010) study determinants of European corporate CDS returns and identify the common factor, which explains around 50% of the variation, as the super-senior tranche of the iTraxx Europe index, referred to as “the economic catastrophe risk”. Similar to our study, Dieckmann and Plank (2011) find evidence of a private-to-public risk transfer for countries whose governments have intervened in the financial system. By employing panel regressions, the authors analyze the determinants of changes in sovereign CDS spreads, and find that both domestic and international financial systems play an important role in explaining the dynamics of CDS spreads. They also argue that countries in the European Monetary Union (EMU) are more sensitive to the health of the financial system than non-EMU coun- tries. Fontana and Scheicher(2010) identify the main determinants of bond and CDS spreads. They include in their set of explanatory factors proxies for market liquidity and global risk appetite, and these are found to be significant. Furthermore, they employ a lead-lag analysis for bond and CDS markets and find that for France, Germany, the Netherlands, Austria, and Belgium the cash market dominates, while
for Greece, Italy, Ireland, Spain, and Portugal the CDS market is more important in terms of price discovery. Hull et al.(2004) andNorden and Weber(2004) analyze the impact of unique events on CDS markets, such as credit rating announcements.
Both studies find that markets anticipate both news and reviews of downgrades, and that credit rating announcements contain important information and have a significant effect, especially on the CDS market.
Furthermore, there are studies that solely investigate the sovereign bond market. Us- ing a GARCH-in-mean model, Dötz and Fischer(2010) analyze the EMU sovereign bond spreads during the financial crisis and find that the implied probability of de- fault reached unprecedented values and the increased expected loss component made some sovereign bonds lose their status as a “safe haven” investment. Gerlach et al.
(2010) analyze the determinants of Eurozone sovereign bond spreads. They show that the size of the banking sector has an important explanatory value for changes in bond spreads, suggesting that markets perceive countries with a large stake in this sector at higher risk of stepping up and rescuing the banks. Employing a dynamic panel, Attinasi et al. (2009) highlight the main factors that explain the widened sovereign bond spreads in some Eurozone countries for the period that covers the core part of the financial crisis in Europe.
Within the second strand of literature, Ejsing and Lemke (2011) investigate the co-movement of CDS spreads of Eurozone countries and banks with a common risk factor, i.e. the iTraxx CDS index of non-financial corporations. The authors find that the government bailout and guarantee programs for the financial sector induced a drop in the credit spreads for banks but a jump in governments’ CDS spreads.
Furthermore, the sovereign CDS series became more sensitive to the common risk factor, while the banks’ CDS spreads became less so. Besides providing a model for the interrelation of bank and government credit risk, Acharya et al. (2011) outline the same mechanism empirically, showing a widening of the sovereign and a nar- rowing of the bank CDS spreads. Focusing on the financial crisis, Demirgüç-Kunt and Huizinga (2010) find that bank CDS spreads are significantly affected by the deterioration of public finance conditions. A high sovereign debt burden impairs the ability to provide support to the financial sector and too-big-to-fail banks might thus become too-big-to-be-saved.
1.3 Hypotheses, data, and econometric methodology
1.3.1 Hypotheses
In this subsection we develop the hypotheses to be tested in our study. Firstly we describe the main transmission channels that emerge when either a (systemic) banking crisis develops or sovereign distress appears. Based onAcharya et al.(2011), Gray(2009) andIMF(2010), we present both directions of the contagion mechanism.
If a financial institution faces funding and/or liquidity issues, this can trigger a sharp rise in its default risk and may have specific contagion effects: (I) the bank cannot pay its obligations to another financial counterparty which in turn can set off funding/liquidity difficulties for the latter and increases its perceived default risk; (II) the state might intervene in order to prevent bank bankruptcies. This private-to-public risk transfer augments the probability of default for the state and lowers the default risk of the financial institution. If (I) occurs, difficulties within the entire financial system (e.g. systemic banking crisis) might arise and translate into a contraction of the economy, which would also weaken public finances (e.g. a decrease in the present value of taxes) and, again, the sovereign default risk would increase.
In the case of a country’s distress, in the first wave, the contagion to other entities can be triggered via three direct channels (Chapter 1, IMF (2010)): (i) from the affected state to other countries that are highly interconnected through bilateral trade or share similar problems (e.g. public deficit, funding needs, etc.); (ii) from the distressed country to domestic banks as the market value of government bonds held by these banks decreases, and government support loses credibility; (iii) from the impaired state to foreign banks that hold government (or bank) bonds (or other assets) from the affected country.
Before the recent government interventions, we argue that financial sector issues had a systemic component, leading to contagion mechanism (I). Thus, the rising default risk of banks had an indirect effect on governments’ credit risk. Additionally, state interventions in response to financial sector problems were possibly expected by market participants. Thus, the perceived sovereign default risk increased but was considered of limited importance in terms of having any visible impact on banks’
default risk.
Hypothesis 1. Prior to state interventions, changes in the default risk of banks affect the default risk of European governments, but not vice- versa.
After government interventions, states not only bear an asset exposure to the bank- ing sector but their balance sheets contain contingent liabilities (e.g. government guarantees) as well. Thus, the sensitivity of government default risk to the banking sector risk is expected to increase. Furthermore, through the credibility of govern- ment contingent liabilities, changes in government default risk have a direct impact on the perceived risk of financial institutions.
Hypothesis 2 (a). In the period after a government intervention, changes in the de- fault risk of banks affect the sovereign default risk more strongly than before.
Hypothesis 2(b). After bailout programs have been implemented, an in- crease/decrease in sovereign default risk causes a change in the default risk of the domestic banks in the same direction.
Some banks received direct capital injections from their governments. If the capital injections were sufficient, we would expect the dependency on future bailouts to be the same as for the rest of the financial sector. On the other hand, in case of a partial recapitalization or any other insufficient intervention, the bank in question should be highly sensitive to the health and credibility of the host government. The following hypothesis links the sensitivity of banks’ default risk to the probability of future government support.
Hypothesis 3. The bank’s sensitivity to the sovereign default risk increases with the bank’s reliance on future government aid.
Our last hypothesis compares the outcomes of bailout programs in different coun- tries. The magnitude of different support measures provided by each country was heterogeneous among the analyzed Eurozone countries. This was induced by, at least, three factors: (i)the economic health of the country, (ii) the size of its finan- cial sector relative to the total economy and (iii) the exposure of the banking sector to the systemic crisis.
Hypothesis 4. Heterogeneity of bailout programs across European countries translates into asymmetric interdependence between sovereign and banks’ default risk.
The model introduced by Acharya et al.(2011) describes in detail this feedback mechanism, i.e. how financial sector and sovereign default risk are linked. The
authors present a three-period model, in which a financial and corporate sector jointly produce aggregate output. There exists a potential underinvestment problem.
Bank bailouts are used to help resolve this problem in the financial sector. The framework predicts that bank bailouts increase sovereign credit risk. The latter affects the financial sector as the value of guarantees and bond holdings decreases.
This linkage implies a post-bailout increase in the co-movement of government and financial sector default risk.
1.3.2 Bailout specific characteristics
In order to compare the selected countries, we relate our analysis to the specific bailout schemes provided in each country. Hence, we look at the magnitude of the different support measures utilized by each country, while additionally considering the particular aid offered to each bank. Following Stolz and Wedow (2010), we categorize the general set of measures, emphasizing the differences and similarities across countries. Even though there are differences in the number and types of institutions involved in banking crisis management, there is less variation across the countries in terms of the types of support measures that were applied. The financial aid programs can be classified into four broad categories: capital injections, guarantees for bank liabilities, asset support programs, and deposit insurance (see Table 1.1).
Table 1.1: Government Support Measures for Financial Institutions (October 2008 - May 2010)
Country Capital injection Liability guarantees Asset support Total commitment Deposit insurance Guaranteed Other
Within Outside issuance of guarantees, Within Outside as % of
Schemes Schemes bonds loans Schemes Schemes 2008 GDP in EUR
France 8.3 (21) 3 134.2 (320) 0 - (-) - 18% 70,000
Germany 29.4 (40) 24.8 110.8 (400) 75 17 (40) 39.3 25% Unlimited
Ireland 12.3 (10) 7 72.5 (485) 0 8 (90) - 319% Unlimited
Italy 4.1 (12) - - (-) 0 - (50) - 4% 103,291
Netherlands 10.2 (20) 16.8 54.2 (200) 50 - (-) 21.4 52% 100,000
Portugal - (4) - 5.4 (16) 0 - (-) - 12% 100,000
Spain 11 (99) 1.3 56.4 (100) 9 19.3 (50) 2.5 24% 100,000
Note: All amounts are in billions ofe, except for the last two columns. Figures in brackets denote total committed funds and figures outside brackets are the utilized amounts up to May 2010. “Within schemes” refers to a collective bailout program that can be accessed by any bank that fulfills the requirements for that particular aid scheme.
“Outside schemes” refers to individually tailored aid measures (ad hoc schemes).Source:Stolz and Wedow(2010)
Based on the ratio of total commitment to GDP, the selected countries can be ranked (from high to low): Ireland, the Netherlands, Germany, Spain, France, Portugal, and Italy. Furthermore, the set of countries can be clustered into three groups: Ireland (high commitment - above 75% of GDP); the Netherlands, Germany, Spain, and France (medium commitment - 20% - 75% of GDP); Portugal and Italy (low commitment - below 20% of GDP).
