1
The toxic legacy of the Greek crisis*
Jordy Daniël Koetsier 5779537
University of Amsterdam
MSc Business Economics, Finance track Master Thesis
Dr. Rafael Matta July 2015
Abstract
This paper contributes to prior research in the area of credit risk interdependence by exploring private-to-public credit risk transfer (i.e. the Irish episode), the public-to-private credit risk transfer (i.e. the Greek episode) and the public-to-public credit risk transfer within the Euro-zone. In other words, this paper evaluates the short-run and long-run relations between CDS series of sovereigns and banks. This paper documents evidence of private-to-public credit risk transfer (i.e. the Irish episode) running from the domestic banking sector to the host sovereign in the long-run – i.e. after and during bank bailouts. In contribution to prior research, this paper documents evidence of short-run Granger causality running from CDS series of sovereign states to CDS series of the banking sector cross-border – i.e. after controlling for common risk factors. No signs of long-run suggestive causality running from sovereign states to the international banking sector are detected.
2
Table of Contents
Part one: introduction ... 3
1.1 Introduction to the European sovereign capital market ... 3
1.2 Introduction to the research area – European capital market interdependence ... 4
1.3 Research question ... 5
1.4 Relevance of research topic ... 5
1.5 Contribution to prior research ... 6
1.6 Outline of thesis ... 6
Part two: literature review ... 7
2.1 Distinctive nature of the sovereign financial markets within the Euro-zone ... 7
2.2 Interdependence of private and public financial markets ... 8
2.3 Empirical evidence on private-to-public and public-to-private credit risk transfer ... 11
2.4 How this paper relates to prior research ... 15
Part three: methodology ... 16
3.1. Type of data and usage ... 16
3.2. Econometric model ... 17
3.3 Hypotheses testing ... 18
Part four: data and descriptive statistics ... 20
4.1 Collected data and sources of withdrawal ... 20
4.2 Sample specification ... 21
4.3 Summary statistics ... 22
Part five: results ... 23
5.1 Pre-analysis and model specification ... 23
5.2 Co-integration and Granger causality analysis ... 24
5.2.1 Interpretation ... 24
5.2.2 Stage 1 ... 25
5.2.3 Stage 2 and 3 ... 26
5.3 Wrap-up ... 28
Part 6: conclusion ... 29
6.1 Key findings in context of financial literature ... 29
3
Part one: introduction
1.1 Introduction to the European sovereign capital market
Athens was the symbol of wealth in the fourth century Before Christ. The ancient prosperity of Athens remained captured in the Greek saying “bringing owls to Athens”, which referred to “doing something useless”. The silver Drachmas were denoted as owls – since a picture of the bird was printed on the back of the coin – and the saying points toward the believe that it was considered pointless to bring any more wealth to Athens at the time. However, expensive warfare bankrupted the entire empire a few decades later. Ten out of thirteen Athens states, funded by the temple of Delos, defaulted on their debt obligations and orchestrated the first sovereign default in history.
As of January first 2001, the Greek Republic was acknowledged as a full-fledged member of the European Monetary Union (EMU). The origination of union initiated a comprehensive change in the structure of the European sovereign capital market. The main driver of sovereign yield spreads within capital markets, exchange rate fluctuation, vanished with the inauguration of the monetary union (Geyer et al., 2004). In addition, a Stability and Growth Pact was established in order to further diminish the dissimilarities in terms of fiscal vulnerability which obligated the member states to meet prescribed economic criteria. The removal of the exchange rate risk along with a converged fiscal programme paved the way for an integrated and competitive capital market (Gómez-Puig, 2008). Prior to the formation of the monetary union, sovereign bond yields were determined by the market perceptions of future exchange rate fluctuations, domestic fiscal policy and sovereign specific credit risk along with dissimilar liquidity levels of segmented international capital markets. As of the origination of the union, “riskless” sovereign debt instruments were believed to be traded as perfect substitutes in a frictionless market against a similar yield curve. However, the persistence of yield spreads within the monetary union today emphasises the fact that the market has assessed dissimilar default rates to individual member states. In other words, European sovereign bond yields have not fully converged (Geyer et al., 2004).
The market perception regarding EMU sovereign creditworthiness deteriorated in the spring of 2010. Sovereign yield spreads and default insurance premiums, determined by state specific fundamental variables such as the debt-to-GDP ratio and the current account balance, rose dramatically and indicated a lack of market confidence regarding a sovereign’s willingness and ability to meet its debt obligations. The Greek Republic has experienced severe funding difficulties, due to large jumps in their debt service costs, until the European Central Bank (ECB)
4 decided to initiate a “bailout” programme in order to ensure economic stability for the entire monetary system in the spring of 2010. The first Greek rescue programme had proven to be insufficient. It did not restore the creditworthiness of the Greek Republic and it did not re-establish access to international capital markets. As a result, Greece has the doubtful honour to become the first advanced economy which underwent a debt restructuring programme since the Second World War.
1.2 Introduction to the research area – European capital market interdependence
The recent sovereign debt crisis has shown that the European Monetary Union (EMU) sovereign capital markets are far more fragile than stand-alone capital markets (De Grauwe, 2012). The disconnect between the credit spread on sovereign debt and the underlying fundamentals, such as the debt-to-GDP ratio, were shown by the development of bond yields of peripheral member states. The increase in spread has been far more significant than the change in underlying fundamentals as a result of negative market sentiments driven by asymmetric information and political reputation (De Grauwe and Ji, 2012). The capital market shifts to turmoil mode and displays high volatility in bond prices and interest rates due to different assessments of asset quality. Investors either lost confidence in the underlying asset or are reluctant to reassess the fundamentals. These negative pricing effects are amplified by “animal spirits” which have led to self-reinforcing bubbles and herding behaviour (Herzog and Müller, 2014).
The distinctive nature of the EMU capital markets is driven by fundamental dissimilarities compared to stand-alone capital markets. First, financial theory suggests that a sovereign bond yield remains free of default risk as long as the issued debt is dominated in the domestic currency. That is, the market presumes that a sovereign is able to repay its debt obligation by simply printing additional funds. EMU member states are unable to control their debt obligations by adjusting monetary policy, due to a lack of monetary power. Second, member states are in principle prohibited to assume liabilities other than their own. Article 125 of the Lisbon Treaty forbids the ECB and the European Union to bailout a distressed government. However, the decision to rescue the Greek Republic contradicts the credibility of the article and, more important, inherits a severe moral hazard problem. That is, the rescue programme points toward a lack of incentive for a distressed government to cut back spending or raise additional income. Economist James Buchanan describes this situation as a “Samaritan’s dilemma”. The dilemma points toward the believe that a father will always support his troubled son even if it will lead to further misbehaviour in the future, due to the fact that they share an unique bond.
5 Members of the monetary union share a unique bond through a common denominator, i.e. the Euro. Hence the fact that the member states, almost unanimous, agreed to “bailout” the Greek Republic driven by the fear that a default would trigger a series of countless banking failures, a dry-up of the interbank market, growing sovereign funding costs and an ultimate destabilisation of the Euro (Mink et al. 2013).
The deterioration of a sovereign’s creditworthiness affects the bank’s balance sheet. Financial institutions are the prime holders of sovereign debt. A rise in sovereign credit risk effectively impairs the bank’s balance sheet and diminishes the bank’s access to market funding as the collateral value on sovereign debt decreases. Uncertainty regarding the actual exposure of the particular bank on a sovereign in distress might even result in capital outflow by deposit holders. Moreover, a weakened sovereign’s creditworthiness diminishes the implicit government guarantee when a bank faces distress as the government’s ability and willingness to support its financial sector declines. On the other hand, the implied government guarantee to support financial institutions of systemically importance are contingent liabilities and their impact on future deficits depends on their size and likelihood of being called (Ejsing and Lemke, 2011). That is, a systemic banking crisis weakens the entire economy and effectively transfers risk from the private-to-public sector.
1.3 Research question
This paper aims to evaluate the interdependence of credit risk between sovereigns and the banking sector within the monetary union since the start of the financial crisis up to February 2012 – i.e. the second bailout of the Greek Republic. This paper aims to evaluate the short-run and long-run relationship between CDS markets of sovereign states and the banking sector. That is, this paper aims to evaluate whether a surge in sovereign risk compounds the credit risk within the monetary union. Therefore, the following research question is defined.
“To what extent are EMU financial markets interdependent in terms of credit risk?”
1.4 Relevance of research topic
The potential advantages of economic, monetary and market integration are well known. Prior literature has identified that the construction of the monetary union has not only eliminated exchange rate risk, it has also shield monetary policy from domestic political authorities and
6 simultaneously shared risks among members of the union. On the contrary, the aftermath of the recent financial turmoil has emphasised the fact that the levels of contagion have intensified dramatically, due to more advanced market integration (Baele, 2005). The dynamic linkage within capital markets is an important consideration in the light of the portfolio diversification theory, the asset pricing theory and allocation theory regarding marketable securities (Skintzi et al., 2006). Accurate knowledge regarding dynamic suggestive causal effects is essential in order to develop effective strategies for hedging and insurance purposes. In addition, monetary and regulatory authorities require proper knowledge regarding volatility transmission within international capital markets in order to effectively asses capital adequacy principles and to acquire perfect understanding regarding the suggestive causal effect of proposed policy alterations. Gray (2009) argue that policy makers have not anticipated on the level of interdependence of bank and sovereign CDS markets in the light of exposure to sovereign credit risk and potential causal effects to other sectors of the economy.
1.5 Contribution to prior research
This paper contributes to prior research on interdependence of credit risk by exploring private-to-public credit risk transfer (i.e. the Irish episode), the public-to-private credit risk transfer (i.e. the Greek episode) and the public-to-public credit risk transfer within the Euro-zone. The latter episode refers to the potential transfer of credit risk associated with the supply of aid packages to Greek Republic by the collective of Euro-zone sovereigns. The transfer of public-to-public credit risk might reflect in CDS series. As suggested by Alter and Schüler (2012) and Dieckmann and Plank (2011), documenting the contrast between private-to-public and public-to-public interdependencies may further gain important insights in the credit risk transfer mechanism within the Euro-zone. In addition, this paper aims to capture the cross-border interdependencies through the bank’s holdings in foreign public debt and their exposure through interbank markets. Cross-border interdependencies have had little attention in financial literature as prior empirical evidence is based on the impression that financial institutions hold a biased preference towards public debt attributable to their home country.
1.6 Outline of thesis
This paper is constructed as follows. First, the related literature and empirical evidence covering this research topic are discussed (“part two”). Followed with a brief discussion on the methodology in place (“part three”) and an elaboration of the data required to run the
7 methodology (“part four”). Finally, the results are discussed (“part five”) with associated findings presented in the appendices along with a wrap-up of this paper (“part six”).
Part two: literature review
This paper relates to two important strands in recent financial literature covering (i) the distinctive nature of the Euro-zone sovereign capital markets in terms of interdependence and (ii) the dynamic linkage between private (i.e. banking sector) and public (i.e. sovereign) financial markets in terms of credit risk (i.e. CDS spreads serve as a proxy for credit risk). This section starts with a brief discussion regarding the distinctive nature of the European capital market. Following with an outline of the main theory and predictions concerning the linkage between public and private financial markets. Rounding up with empirical evidence on that matter along with the contribution of this paper to prior research.
2.1 Distinctive nature of the sovereign financial markets within the Euro-zone
Investors seem to regard EMU sovereign capital markets as one single capital market since of the formation of the Euro-zone (Pagano et al., 2004). Investors shared the wide held believe that the sovereign capital markets of individual member states shall converge to one integrated capital market – “convergence-trading hypothesis” (Codogno et al., 2003, Geyer et al., 2004 and Pagano et al., 2004). Sovereign debt instruments issued by member states were believed to be traded as perfect substitutes in a frictionless market against a similar yield curve. The main effect of the formation of the monetary union was therefore limited to an enhanced level of substitution among debt instruments along with an improved level of competition among issuers to attract funding. In other words, the founding of the Euro-zone enhanced the level of capital market integration (Gómez-Puig 2009, and Abad et al., 2010). The persistence of yield differentials today illustrate that the level of market integration within the Euro-zone is incomplete (Gómez-Puig 2009 and Abad et al., 2010).
Geyer et al. (2004) defines the persistence of yield differentials within the Euro-zone by means of several hypotheses. First, sovereign-specific default risk may occur due to a remained dissimilarity in fiscal policy or due to different levels of fiscal vulnerability to external shocks. Second, yield differentials may occur due to a persistent segmentation of capital markets in terms of liquidity and taxation. Finally, there is a probability of overall failure of the monetary union which will lead to the reintroduction of exchange rate risk at some date prior to the redemption
8 of a sovereign debt instrument. Empirical evidence suggests that sovereign capital markets of the Euro-zone incorporate substantial systemic risk, which in turn limits the benefits of portfolio diversification and support the presence of some positive probability of overall failure which should be incorporated in market prices of sovereign debt instruments. Intuitively, the persistence of yield spreads may even reflect dissimilar market assessment regarding the potential influence of overall failure of the monetary union for an individual member state (Geyer et al., 2004).
Financial markets seemed to pay less attention to sovereign’s indebtedness as opposed to their behaviour prior to the formation of the monetary (Bernoth et al., 2012). Sovereign debt instruments of peripheral states traded as near-substitutes of equivalent German ‘Bunds’ – up to the financial crisis – illustrating an implied credit risk of effectively zero as German sovereign debt is assumed to be free of default risk. However, financial markets’ reaction to fiscal loosening changed extensively after the collapse of Lehman on September 15th of 2008 (Bernoth
et al., 2012). The failure of Lehman Brothers emphasised the importance of the solvency of a particular issuer and the liquidity of the instrument which in turn reinforced market discipline (Gómez-Puig, 2008). As a result, the dramatic rise in yield spread – disconnected from underlying increase in debt-to-GDP ratio and primarily driven by negative market sentiment – have indicated that the sovereign capital markets within the Euro-zone are far more fragile than stand-alone sovereign capital markets and vulnerable for self-fulfilling liquidity crises (De Grouwe and Ji, 2012). Therefore, interdependence of private (i.e. banking sector) and public (i.e. sovereign) capital markets in terms of credit risk (i.e. CDS spreads) is particular relevant for the Euro-zone given the prominent role of bank credit in financing the economy (Moody’s, 2014).
2.2 Interdependence of private and public financial markets
The recent European financial turmoil emphasised the appearance of contagion risk within the financial system. Enhanced credit risk on sovereign debt of peripheral states resulted in contagion risk in terms of direct capital loss incurred by the banking sector, effectively impairing the collateral value of these instrument as sovereign debt serve as deposit to raise wholesale funding or ECB liquidity. Contagion risk has been observed in credit events prior to the current European crisis. Holders of sovereign debt in the 1980s, mostly held by the banking sector, suffered significant losses due to a default of Latin-American nations on their outstanding sovereign liabilities. In the 1990s, these investors incurred another substantial loss due to a default of Russia, Mexico and Argentina on their debt obligations. However, in the recent
9 financial market the contagion risk has fundamentally changed. Sovereign bonds are traded continuously in secondary markets. Therefore non-bank investors are also holding government debt, resulting in a far more fast-tracking growth of contagion risk. That is, when the creditworthiness of a sovereign issuer starts being questioned, problems of a particular sovereign could spread among the entire financial system and may trigger a reluctance to hold debt instruments of highly indebted sovereigns.
The European financial system witnessed private and public credit risk linkage through the “Irish episode” – where sovereign credit risk rises through the extent of government support to distressed banks – and the “Greek episode” – where a distress sovereign triggers fragility of the banking sector. The Irish banking crisis and the subsequent sovereign default starting at the autumn of 2008 have illustrated the potential private-to-public credit risk transfer (Acharya et al., 2014). The Republic of Ireland guaranteed to back all deposits of the six banks of systemic importance. The effect of this announcement was demonstrated overnight on the CDS market. Protection fee on Irish banks moved downward whilst CDS rates on Irish sovereign debt rose sharply – effectively transferring credit risk from the banking sector to the sovereign (i.e. private-to-public). The credit risk within the banking sector and the credit risk on sovereign debt instruments are linked through the implied government guarantee to bailout a distressed bank of systemic importance (Acharya et al., 2014, Alter and Schüler, 2012). Leveraged financial institutions may experience insolvency problems when the market switches to turmoil mode and therefore offer limited contribution to the overall economy as the debt overhang problem restricts their power to invest. The elimination of this so-called “under-investment problem” drives a sovereign to initiate a bailout program which subsequently benefits the entire economy. Acharya et al. (2014) argue that the choice of funding may have an inefficient fallout, in terms of a negative welfare effect to the entire economy. First, a higher tax burden – attributable to the non-financial sector to fund the bailout – impairs the incentive to invest and thus reduces economic growth in the mid-term. Second, a sovereign may seek funding with the use of the issuance of new debt instruments which in turn dilutes current sovereign debt holders and simultaneously deteriorates sovereign’s creditworthiness, effectively feeding capital loss back to the banking sector. The latter approach weakens the implied value of future government guarantee – exemplifying a potential two-way credit risk transfer. The Irish-episode may have pushed the capital market on edge as multiple Western-European bailouts followed with similar private-to-public credit risk transfers.
Prior to the financial turmoil, no signs of sovereign credit risk occurred in developed economies – i.e. shown in Graph 1 attached to this paper. However, the Greek episode showed
10 otherwise and unambiguously captured the fragility of the banking sector to signs of credit risk on outstanding sovereign debt. Impending sovereign distress primarily attributable to peripheral states negatively affected several European banks – heavily invested in Greek and other peripheral states’ public debt such as Dexia, Crédit Agricole, Société Générale and several German Landesbanken – due to rising concern regarding the sustainability of these instruments at the start of 2009. Gennaioli et al. (2014) argue that the potential capital loss incurred by the banking sector is subject to the willingness of a sovereign to repay its debt obligation rather than to preserve access to international capital markets or to avoid international sanctions as textbook financial knowledge may suggests. Gennaioli et al. (2014) claims that sovereign defaults are witnessed infrequently due to the fact that a sovereign chose to repay its debt to avoid large cost to the domestic economy rather than to avoid exclusion from international capital markets or international sanctions as these interventions are usually short-lived. Financial literature demonstrates that banks choose to hold public debt to store liquidity for pending investments (Holmström and Tirole, 1993). Sovereign default will therefore dry up bank’s liquidity and effectively reduces private credit or other investments going forward – i.e. the “under-investment problem”. Sovereigns have to discriminate between domestic and foreign bondholders in order to avoid any default cost attributable to their domestic economy. However, sovereigns are unable to shield the domestic economy by selectively default on foreign public bondholders as perfect discrimination in today’s financial market is close to impossible due the fact that sovereign debt instruments are traded continuously in secondary markets. In addition, Gennaioli et al. (2014) denotes that the incentive to repay public debt is higher when the domestic financial sector is relatively more developed as the ultimate cost of default to the domestic economy is higher. In other words, the cost of default to the domestic economy is higher when developed financial institutions boost leverage – to allow for higher real investments – and therefore hold a large share of domestic public debt as collateral to attract foreign capital. The surge for leverage simultaneously amplifies the vulnerability of the bank’s balance sheet to adverse shocks. The foreign capital inflow offers easy access to boost leverage. Gennaioli et al. (2014) argue that the latter effect reduces the share of public debt held by foreign investors. Imminent sovereign distress will ensue disruption of the domestic economy with the impairment on the collateral value of public debt and the direct capital loss incurred by the banking sector.
The deterioration of sovereign’s creditworthiness weakens the implied value of government guarantee. This raises the question whether financial institutions of systemic importance are rather too-big-to-save than too-big-to-fail? Demirgüç-Kunt and Huizinga (2013) argue that banks with outstanding liabilities exceeding a certain threshold level measured with
11 respect to their attributable country’s GDP may become too-big-to-save if a sovereign becomes fiscally strapped. Over the years leading up to the financial crisis, multiple European banks reached to enormous size – primarily attributable to foreign capital inflow – and even outgrown the national GDP level. Demirgüç-Kunt and Huizinga (2013) have identified 30 listed banks – of which c. 50% headquartered in the Euro-zone – with outstanding liabilities exceeding at least half of the GDP level attributable to their home country. For example, listed financial institutions in the Netherlands, France, Belgium, Ireland and Denmark had outstanding liabilities up to 2 times national GDP, by the end of 2008. The financial sector of Switzerland and the UK grew even larger with respectively 6.3 and 5.5 times GDP. The prime example, the overall outstanding liabilities of the banking sector of Iceland reached 9 times GDP in 2007. Demirgüç-Kunt and Huizinga (2013) argue that the protection fee on bank’s debt should fall if a bank reach the too-big-to-fail status, which implies lower funding cost under normal business conditions – i.e. effectively feeding credit risk to the public sector through contingent liabilities of the sovereign. However, the deterioration of public creditworthiness raises doubt regarding a sovereign’s ability to provide a sufficient safety net for its banking sector – i.e. effectively downgrading the too-big-to-fail status to a too-big-to-save banner and illustrating a public-to-private credit risk transfer.
The perceived transfer of credit risk between financial markets is subject to changes in common risk factors. Ejsing and Lemke (2011) argue that the seeming transfer of credit risk between banks and sovereigns reflect some impact of deteriorating macroeconomic outlook in terms of lower profit expectation and higher default risk in the banking sector along with downward pressure on tax benefits and subsequent higher fiscal deficit in the public sector. That is, the widening of both bank and sovereign CDS spreads simultaneously – e.g. co-movement of financial markets – may partially reflect changes in the state of the (local) economy. In addition, movements in overall risk aversion of investors influence both the private and public CDS markets. In order to assess the level and direction of the perceived credit risk transfer between financial markets, one should control for time-varying control parameters that document changes in common risk factors.
2.3 Empirical evidence on private-to-public and public-to-private credit risk transfer
Acharya et al. (2014) evaluates a series of bank bailouts in the Euro-zone, Nordics, Switzerland and the United Kingdom and find empirical evidence supporting the private-to-public credit risk transfer hypothesis. Their model distinguish between distinctive periods in time to assess the
12 transfer of credit risk. The first period refers to the start of financial crisis as the turmoil evolves in the banking sector and the spreads on bank CDS widened whilst sovereign CDS rates remain constant. The second period starting late September 2008 with the Irish’ announcement to bailout its financial sector. Spreads of bank’s CDS decreases whilst the CDS spreads across sovereigns widened. The third “post-bailout” period is denoted for a feedback of sovereign credit risk back to the bank’s balance sheet. Their model is based on the outcome of the European stress test which suggests that banks hold on average approximately one-sixth of their risk weighted assets in sovereign bonds, with a bias towards the home country in which the participating bank is headquartered. The stress test serves as their key assumption to limit the assessment of bank’s exposure to sovereign credit risk associated with their home country. The empirical evidence of Acharya et al. (2014) points toward an intimate link between sovereign credit risk and bailouts of the banking sector. They document a rise in sovereign credit risk driven by distress of the banking sector. They do not capture a relationship between banking and sovereign CDS series prior to government interventions. In contrast, they show that bailouts trigger a rise in sovereign CDS spreads. In addition, they find that an increase in sovereign credit risk is associated with a rise in bank credit risk post-bailout. That is, bailouts effectively transfer credit risk from the bank’s balance sheet to that of the sovereign. Acharya et al. (2014) argue that sovereign debt instruments are more sensitive to macroeconomic shocks post-bailout. Their evidence illustrate a strong positive relationship between public debt-to-GDP and sovereign CDS series as opposed to no signs prior to government interventions. Their result points toward a strengthened private-to-public credit risk transfer for sovereigns with relative higher debt-to-GDP ratio’s. In addition, their result illustrates that movements in sovereign CDS spreads explain movements in bank CDS spreads post-bailout, which indicates that a rise in sovereign credit risk weakens the domestic banking sector after a government intervention.
In order to determine a “pure” domestic assessment of credit risk transfer, Acharya et al. (2014) control for variations in macroeconomic conditions and dissimilarities in terms of exposure to these variations – i.e. time fixed effects and bank fixed effects. In addition to time and bank fixed effects a third control mechanism is introduced to control for holdings in foreign sovereign debt. A “foreign exposure” parameter is incorporated to control for the impact of changes in sovereign credit risk of these holdings. Acharya et al. (2014) argue that changes in macroeconomic fundamentals generate potential correlation between bank and sovereign credit risk, which is documented by Ejsing and Lemke (2011) in their evaluation of the Euro-zone bank bailouts. They find empirical evidence for co-movement in weekly CDS spreads of banks and sovereigns with the iTraxx index consisting of non-financial CDS spreads which serves as a
13 proxy for a common risk factor. Their empirical work documents high level of fit for all sovereign CDS series and for some of the corporate issuers incorporated in their sample. They argue that the iTraxx index serves as a sufficient parameter to capture changes in risk aversion and the impact of deteriorating macroeconomic outlook. In addition, they suggests the use of regressions allowing for structural breaks in time series and time-varying parameters in order to assess the dynamic nature of credit risk transfer. Their empirical evidence documents a decrease in slope of the common factor for bank CDS spreads post-bailout along with an opposite effect documented in the sovereign CDS market. Their empirical evidence points toward an increase in sensitivity of the sovereign’ CDS spreads to further aggravation of the macroeconomic outlook and investor’s risk aversion post. The documented co-movements reflect the sensitivity of CDS spreads to the aggravation of macroeconomic outlook and changes in risk aversion. These results contribute to the pre-crisis research of Geyer et al. (2004) that documents a common long-term risk factor affecting EMU sovereign bond yields. Favero et al. (2010) and Bernoth et al. (2004) find an aggregate level of risk driving spreads on Euro-zone sovereign bonds and illustrate a shift in risk aversion regarding credit risk within the EMU capital markets. Codogno et al. (2003) illustrate that the spread on sovereign bonds driven by global risk factors reflect a change in investor’s view on the underlying credit risk.
In connection with the empirical evidence of Acharya et al. (2014), Demirgüç-Kunt and Huizinga (2013) document signs of dynamic linkage of global private and public financial markets. They conduct a global research on the impact of bank size and sovereign debt on bank equity prices and sovereign CDS rates and show that banks are rather big-to-save than too-big-to-fail. They examine the effect on systemically large banks and smaller banks – in terms of outstanding liabilities with respect to GDP – with the main body of their work concentrated on the top-30 largest listed banks in 2008. They use sovereign CDS spreads to evaluate the impact of expected losses to bank’s balance sheet through their sovereign debt holdings. In other words, they measure the impact of sovereign credit risk to the market value of bank equity and use instrumental variables to control for business cycle and the level of international trade. Their work points toward a negative relationship between the market-to-book value of bank equity and the size of bank’s outstanding liabilities-to-GDP as well as to absolute size measured as the natural logarithm of assets under management. Their work suggests that banks are considered too-big-to-save if the size of their outstanding liabilities – with respect to the GDP level of their home country – imposes a negative effect on the market-to-book ratio of bank equity. In other words, the implied value of the financial safety net is no longer considered sufficient to harvest an adequate aid package for distressed banks at some point in the future. Demirgüç-Kunt and
14 Huizinga (2013) document a positive effect of leverage on market-to-book ratio’s whilst negatively affecting bank’s CDS spreads. Simultaneously, the interaction term – used to incorporate the bank’s liabilities-to-GDP ratio with sovereign fiscal balance – indicates a negative effect on market-to-book ratio whilst positive affecting CDS spreads. This result exemplifies that some of the credit risk is transferred to the financial safety net as bank’s valuation increases among higher leverage. Prior research documents several examples of positive wealth effects associated with the too-big-to-fail status for shareholders. However, Penas and Unal (2004) on the other hand demonstrate that the benefit of the too-big-to-fail effect is not limited to shareholders as their research on US bank mergers in the 1990s illustrates a similar wealth effect for bondholders. The too-big-to-fail status affects a bank’s risk profile which should prevail within CDS markets. The empirical evidence of Demirgüç-Kunt and Huizinga (2013) suggest that banks considered too-big-to-save should be split up or sufficiently downsized in order to unwind pressure on the market-to-book value or to restore the favourable too-big-to-fail status. Their empirical evidence demonstrates that sovereign creditworthiness imposes constrains on the implied value of government guarantee which reflect in bank valuation. Their findings points toward an enhanced downward pressure on equity valuation of systemically large banks if a sovereign faces financial distress – illustrating a public-to-private risk transfer and a too-big-to-save effect.
In addition to the evidence on public-to-private risk transfer, Gennaioli et al. (2014) conduct a global research on the vulnerability of leveraged financial institutions to sovereign credit risk. They use a panel of developed and emerging markets with a timespan ranging from 1980s up to 2005 in order to evaluate whether sovereign default should lead to contraction in private credit. On that notion, they aim to evaluate if the level of contraction is higher for developed markets where financial institutions hold relatively more sovereign debt. Keeping in mind that ultimate default of developed sovereigns should be less likely as these events will harm the domestic economy to a greater extent as opposed to emerging markets – i.e. to rationalise the low level of default witnessed in developed financial markets in the past. Their empirical evidence points toward contraction of private credit, post-default. In addition, the documented contraction is more advanced for a developed banking sector. They find lower probabilities of default for developed financial markets with banks holding a relative larger share of public debt which rely to some extent on foreign capital inflow. Their result documents higher “willingness” of developed sovereigns to repay debt obligations as a more developed financial sector provide higher consequences for sovereign default and in turn incentivise sovereigns to repay their outstanding obligation which enables borrowing in the first place. They capture a strengthened
15 effect for open economies, where financial institutions benefit from higher foreign capital inflow. Gennaioli et al. (2014) argue that developed financial markets with strong banking sectors facilitate disciplined sovereign borrowing.
Empirical evidence attributable to the research of other authors covering the transfer of risk between private and public financial markets support the key findings discussed above. For example, Alter and Schüler (2012) find evidence of domestic private-to-public credit risk transfer within the Euro-zone. Their empirical evidence documents dispersion of bank’s credit risk into sovereign CDS markets prior to government interventions – similar to the too-big-to-save effect captured by Demirgüç-Kunt and Huizinga (2013). In addition, the research of Alter and Schüler (2012) illustrates an enhanced importance of sovereigns CDS spreads as explanatory variable for the CDS series of domestic banks (after government intervention) – demonstrating evidence of credit risk feedback to the banking sector and supports the findings of Acharya et al. (2014).
2.4 How this paper relates to prior research
This paper relates to prior research covering the dynamic linkage of public and private financial markets in terms of credit risk transfer. The interdependence of financial markets is in particular relevant for the Euro-zone as banks have a more prominent role within the economy in terms of sources of funding as opposed to other markets. In addition, the distinctive nature of the Euro-zone suggests that capital markets are far more fragile than stand-alone sovereign capital markets. The interdependence of credit risk is measured with the use of financial theory on contagion, which refers to contagion as the explanatory variable to rationalise the transmission of a fundamental shock. Contagion – historically a medical term – refers to the transmission of a decease through direct or indirect contact. In the context of finance literature, contagion risk points toward the transmission of a shock from one link throughout an entire system or a spill-over effect toward another system. One could distinguish between fundamental based contagion – the transmission of credit risk from an infected country through trade and financial linkage – and pure contagion – change in market sentiment without any linkage in terms of economic fundamentals.
This paper contributes to prior research on interdependence of credit risk by exploring private-to-public credit risk transfer (i.e. the Irish episode), the public-to-private credit risk transfer (i.e. the Greek episode) and the public-to-public credit risk transfer within the Euro-zone. The latter episode refers to the potential transfer of credit risk associated with the supply of aid packages to Greek Republic by the collective of Euro-zone sovereigns. The transfer of
16 public-to-public credit risk might reflect in CDS series. As suggested by Alter and Schüler (2012) and Dieckmann and Plank (2011), documenting the contrast between private-to-public and public-to-public interdependencies may further gain important insights in the credit risk transfer mechanism within the Euro-zone. In addition, this paper aims to capture the cross-border interdependencies through the bank’s holdings in foreign public debt and their exposure through interbank markets. Cross-border interdependencies have had little attention in financial literature as prior empirical evidence is based on the impression that financial institutions hold a biased preference towards public debt attributable to their home country.
Part three: methodology
This section starts with a brief discussion on the type of data required to conduct this research followed with a discussion on the selected econometric framework to test the presumed relationships. This section further elaborates on the statistical relationship and suggestive causal effects between 𝑥 and 𝑦 in terms of hypotheses testing.
3.1. Type of data and usage
This research is based on CDS series of the European sovereign states and banks. The CDS data collected is limited to high-volume contracts that incorporate sufficient liquidity to ensure that these instruments serve as a suitable proxy for credit risk. The spreads on these CDS instruments serve as the market’s perception of credit risk on the outstanding liabilities attributable to a particular sovereign or bank as the seller is obligated to repay par value in case of default. Longstaff et al. (2005) argue that a CDS spread is a more sufficient proxy of credit risk as opposed to bond yields which incorporate a liquidity component. Norden and Weber (2004) claim that CDS spreads incorporate a forward looking component as these instruments anticipate on credit events such as downgrades by rating agencies. Acharaya and Johnson (2007) argue that the spread on CDS contracts reflect some inside information as price-driven events are usually revealed in the CDS markets. The data collected is limited to CDS spreads on 5-year contracts as Arakeyyan et al. (2012) points out that dealers update quotes on these maturities more often. The preparation of the data along with more in-depth knowledge of these CDS series is covered in part four. Graph 1 and 2, attached as an appendix to this paper, illustrate the development of sovereign and bank CDS series over time.
17 3.2. Econometric model
The methodology discussed below is based on the approach suggested by Alter and Schüler (2012) as they argue that the methodology is suitable to capture credit risk interdependence between sovereigns and banks. In order to determine short-run and long-run relationships, the bivariate vector error correction model (VECM) and the vector autoregressive model with exogenous variables (VARX) are used. In this paper, three distinctive time periods are formulated to evaluate short and long-run relations. The long-run and co-integrated relations are examined with the use of the VECM framework. The bivariate set-up of a 𝑝 − 1 lag model is constructed as formulated below.
(∆𝑐𝑑𝑠𝑆𝑜𝑣,𝑡 ∆𝑐𝑑𝑠𝐵𝑘,𝑡) = ( 𝛼𝑆𝑜𝑣 𝛼𝐵𝑘) (𝛽𝑆𝑜𝑣𝑐𝑑𝑠𝑆𝑜𝑣,𝑡−1+ 𝛽𝐵𝑘𝑐𝑑𝑠𝐵𝑘,𝑡−1+ 𝛽0) + ∑ [𝛾𝛾𝑆𝑜𝑣𝑆𝑜𝑣,𝑖 𝛾𝑆𝑜𝑣𝐵𝑘,𝑖 𝐵𝑘𝑆𝑜𝑣,𝑖 𝛾𝐵𝑘𝐵𝑘,𝑖] 𝑝−1 𝑖=1 (∆𝑐𝑑𝑠𝑆𝑜𝑣,𝑡−1 ∆𝑐𝑑𝑠𝐵𝑘,𝑡−1) + 𝑢𝑡
Where 𝑐𝑑𝑠𝑗,𝑡 with 𝑗 ∈ {𝑆𝑜𝑣, 𝐵𝑘} refers to the natural logarithm of 𝑐𝑑𝑠𝑗,𝑡 – i.e. for both
sovereign states and the banking sector. 𝛽0 is the restricted constant and 𝑢𝑡 is assumed to be
𝑤𝑛(0, ∑ 𝑢). The “white noise” error term refers to a stochastic process of serial uncorrelated random variables. 𝛾 and 𝛽 indicate short and long-run relationships respectively, where 𝛽𝑆𝑜𝑣 is
normalised with 𝛽𝑆𝑜𝑣 = 1 and 𝛽𝐵𝑘 is estimated. The coefficients α measure the speed of
adjustment in which the CDS series alters to the long-run relationship.
In this research, the exogenous control variable 𝑖𝑇𝑟𝑎𝑥𝑥 is incorporated in the VARX framework to control for the common risk factor in the model – i.e. general market risk is separated from idiosyncratic risk within the CDS series. The VARX framework formulated below is used to test for Granger causality – i.e. to capture short-run suggestive causality. This paper follows the lag augmented approach suggested by Dolado and Lütkepohl (1996). The lag augmented approach considers a fitted VAR process of order 𝑝 + 1 when the true order is 𝑝. Dolado and Lütkepohl (1996) suggest this approach as it assurances the validity of the asymptotic distribution of test statistic regardless of stationary variables and uncertainty regarding co-integration properties (Alter and Schüler, 2012). In addition, the VAR framework is modelled with log-levels as opposed to first differences as dynamic interactions are therefore not cancelled out – i.e. following Alter and Schüler (2012).
18 (𝑐𝑑𝑠𝑆𝑜𝑣,𝑡 𝑐𝑑𝑠𝐵𝑘,𝑡) = 𝑣 + ∑ [ 𝑎𝑆𝑜𝑣𝑆𝑜𝑣,𝑖 𝑎𝑆𝑜𝑣𝐵𝑘,𝑖 𝑎𝐵𝑘𝑆𝑜𝑣,𝑖 𝑎𝐵𝑘𝐵𝑘,𝑖] 𝑝 𝑖=1 (𝑐𝑑𝑠𝑆𝑜𝑣,𝑡−1 𝑐𝑑𝑠𝐵𝑘,𝑡−1) + 𝛽𝑖𝑇𝑟𝑎𝑥𝑥,𝑡−1+ 𝑢𝑡 3.3 Hypotheses testing
The hypotheses formulated below are based on the “transmissions channels” identified in the literature review. As noted, this paper contributes to prior research in the area of credit risk interdependence by exploring private-to-public credit risk transfer (i.e. the Irish episode), the public-to-private credit risk transfer (i.e. the Greek episode) and the public-to-public credit risk transfer within the Euro-zone. In other words, this paper evaluates the short-run and long-run relations between CDS series of sovereigns and banks. As noted, the CDS series serve as a proxy for credit risk. The “transmission” or “transfer” of credit risk is defined in terms of short-run and long-run suggestive causal effects. Three types of relationships are identified in the literature review:
(i) between banks and sovereigns (i.e. private-to-public credit risk transfer) prior to government interventions (i.e. bailouts of banks)
(ii) between sovereigns and banks (i.e. public-to-private risk transfer) during and after bank bailouts
(iii) among sovereigns states within the Euro-zone (i.e. public-to-public credit risk transfer)
As discussed in the literature review, financial distress in the banking sector may trigger contagion effects in terms of credit risk. For example, a distressed bank may not be able to pay its obligation to its counterpart in the banking sector. Subsequently, this counterpart may choose to cut off the distress bank from further liquidity. The CDS price for the distressed bank will rise due to an increase of default risk. In this case, the host sovereign may choose to interfere with the use of a bailout package to (i) prevent a bank run, (ii) avoid contraction of the domestic economy (i.e. “under-investment problem”) and (iii) avoid bankruptcy of the distressed bank. The bailout of the bank effectively transfers the credit risk from the bank’s balance sheet to the balance sheet of the sovereign. The private-to-public credit risk transfer enhances the probability of default of the sovereign – which is translated in an augmented CDS spread. On the other hand, sovereign distress may trigger contagion risk (i) through interwoven bilateral trade with
19 other countries, (ii) from the distressed sovereign to the domestic banking sector holding the debt instruments and (iii) from the distressed sovereign to foreign banking sector holding debt instruments of the distressed sovereign or bank headquartered in the distressed country. The first hypothesis is based on the premise that the credit risk build-up in the banking sector, primarily triggered by the losses on US mortgages, leads to the first contagion mechanism – i.e. government intervention (“bailout”).
Hypothesis 1. Credit risk transfers from the banking sector to the sovereign – prior to bank bailouts
After intervention, sovereigns are exposed to the distressed assets of the banks that have received a rescue package. In addition, the implicit government guarantee serves as a contingent liability on the sovereign’s balance sheet. Meaning that the sovereign has become more sensitive to movements of credit risk in the banking sector. These contingent liabilities may even put pressure on the creditworthiness of the sovereign. The following two hypotheses are therefore tested.
Hypothesis 2a. Credit risk transfers from the banking sector to the sovereign – after/during bank bailouts Hypothesis 2b. Credit risk transfers from the sovereign to the banking sector – after/during bank bailouts
The rise of sovereign credit risk through the extent of government guarantee to distressed banks may have pushed the capital market on edge as multiple Western-European bailouts followed with similar private-to-public credit risk transfers (i.e. the Irish-episode). The series of bank bailouts followed with uncertainty regarding the adequacy of these measures along with doubt concerning the capability of the sovereigns to carry the cost incurred – e.g. distressed banks may considered too-big-to-save or even too-many-to-save.
Hypothesis 3. Credit risk transfers from sovereign to sovereign – after/during bank bailouts
The distinctive time periods within the sample are discussed in more detail in the next section. These time periods are based on a set of empirical evidence covering similar research areas. In
20 this paper, three distinctive periods are chosen to test the hypotheses described above. These hypotheses refer to private-to-public credit risk transfer (“hypothesis 1 and 2a”), public-to-private credit risk transfer (“hypothesis 2b”) and public-to-public risk transfer (“hypothesis 3”). The development of the CDS series along with the distinctive time periods are illustrated in graph 1 and 2 attached as an appendix to this paper.
Part four: data and descriptive statistics
This section elaborates on the collected data, sources of withdrawal and the sample specification with associated selection of relevant time periods. This section further covers the summery statistics of both dependent, independent and control variables over the complete sample period.
4.1 Collected data and sources of withdrawal
CDS data on both banks and sovereigns are subtracted from Thomson Reuters DataStream. All available weekly CDS data covering banks headquartered in the Euro-zone (dominated in Euro’s) are withdrawn from the database. The collected data on banks are grouped as per country data and labelled as either “principal” or “peripheral” given the member state in which they are headquartered. Unfortunately due to constrains on the availability of CDS series, banks headquartered in principal member states are limited to German and France financial institutions. Banks headquartered in peripheral member states reflect Portuguese, Irish, Italian, Greek and Spanish financial institutions and are referred to as “PIIGS” within this paper going further. In order to document the potential two-way feedback of credit risk on aggregate country level – i.e. private-to-public and public-to-private credit risk transfer – a weighted average CDS spread of banks is constructed with the use available data on assets under management (“AUM”) following the suggested approach of Acharya et al. (2014) with weights determined by bank assets. Information covering movement in AUM is subtracted from quarterly reports of the attributable financial institutions. The selected CDS series of the Euro-zone banking sector are summarised in table 2. Thomson Reuters DataStream comprehends CDS spreads on European bank and sovereigns dominated in both Euro’s and US-dollar ranging back to December 2007. The sovereign CDS data gathered has similar limitations as the CDS spreads on financial institutions. The starting periods of the various CDS spreads gathered differ among the selected issuers. The dissimilar starting periods poses a minor constrain on the perceived analysis as the majority of selected members states starts at the end of 2007 or at the start of 2008. The available data on Irish CDS spreads may limit the assessment of private-to-public credit risk transfer as
21 DataStream starts documenting CDS spreads on Irish sovereign debt by late October 2008, just after the bailout announcement. As noted earlier, common risk factors reflecting macroeconomic outlook and investors risk aversion may generate correlation between private and public financial markets (Acharya et al., 2014 and Ejsing and Lemke, 2011). In order to determine a “pure” assessment of credit risk transfer, this research controls for variations in macroeconomic conditions. This paper follows the argumentation of Ejsing and Lemke (2011) which refers to the iTraxx index as a sufficient parameter to capture changes in risk aversion and the impact of deteriorating macroeconomic outlook.
4.2 Sample specification
In order to assess the short-run and long-run relationship between banks and sovereigns CDS series, a weighted average sample of the banking sector’s CDS series is specified – i.e. which serves as a proxy for credit risk of the aggregate banking sector of a particular country. The quarterly reported AUM of these banks serve as weights for the construction of an aggregate country-level specification of credit risk. Banks that have received an aid package during the sample period are exempt bailout. The CDS series of these rescued banks remain flat post-bailout and therefore do not resemble any valuable information in the light of credit risk transfer. As noted, the weighted average sample specification serves as a proxy and does not reflect a complete set of financial institutions of the attributable countries as the complete set of CDS series of the Euro-zone banks is unavailable.
The full time period ranging from March 2008 until February 2012 is subdivided into several sub-periods to evaluate the nature of credit risk transfer between financial markets (Ejsing and Lemke, 2011). The time period is divided into distinctive periods to evaluate credit risk transfer between banks and sovereigns – i.e. the period leading up to the series of bank bailouts followed with period after government intervention – and the potential transfer of public-to-public credit risk – i.e. the period leading up to the second bailout of the Greek Republic. Prior literature covering this research area documents three distinctive time periods that may incorporate co-movements of sovereign and bank CDS series. The first period refers to the start of financial crisis when the turmoil evolves in the financial sector – i.e. the spreads on bank CDS widened whilst sovereign CDS spreads remain constant. Due to limitation on data availability, the first stage starts at March 2008 and ends with the Irish’ announcement on 30 September 2008. This period is denoted as “Stage 1” for the remainder of this paper and includes 31 observations per CDS series (excluding Irish sovereign series). The second period
22 starts late September 2008 with the Irish’ announcement to bailout its financial sector. This period starts on 30 September 2008 and ends one month later. The CDS spreads of banks diminished whilst the CDS spread across sovereigns widened. During this stage, sovereign and bank CDS spreads underwent a structural market adjustments. As Alter and Schüler (2012) evoke, the CDS processes underwent a structural break due to market adjustments in which the interdependence between the private and public market shifts. This period is therefore exempt from this research. The third “post-bailout” period is denoted for a potential feedback of impending sovereign credit risk back to the balance sheet of the banking sector. This period – denoted as “Stage 2” – starts on 30 October 2008 and ends on 4 November 2009 – i.e. incorporates 54 observations per CDS series. This stage ends one day prior to the announcement of the Greek Republic that the actual fiscal deficit was twice as much as previous disclosed. The fourth period (referred to as “Stage 3”) starts on 5 November 2009 and leads up to the second and decisive bailout of the Greek Republic on 21 February 2012 – i.e. incorporates 120 observations. This period is symbolised by the extreme uncertainty regarding the survival of the Greek Republic as member the monetary union – illustrated by the steep rise in CDS spread during the period leading up to the bailout. Over the course of this period, a rescue package was negotiated in which banks (holding Greek sovereign debt) were obligated to participate to some extent by taking a haircut on sovereign holdings. On the day of rescue and the start of the restructuring of debt, the Greek Republic lost their autonomy in terms of excess to international capital markets. Members of the Eurozone – i.e. prime funders of the bailout package – became the main lender to the Greek Republic. This period may incorporate signs of public-to-public credit risk transfer. The fear that the Greek crisis could spill-over to Portugal, Italy or Spain was symbolised in May 2010. The collective of European sovereigns started a rescue fund to support troubled Euro-zone sovereigns. The distinctive sub-periods with intermediate events over the course of these periods described above are illustrated in graphs attached as an appendix to this paper.
4.3 Summary statistics
Table 3 summarises the data for the selected bank and sovereign CDS series in basis points (bps). The top seven rows document the CDS series of the selected weighted average banks. The average spread of the “principle” banking sector of the Euro-zone (i.e. the German and French weighted average bank CDS series) resembles low credit risk – i.e. 139 and 136 basis points. On average, moderate credit risk is observed in the Italian and Spanish banking sector. Not
23 surprisingly, the Greek and Irish banking sector reveal extreme credit risk with maximum CDS spreads of 2,584 bps and 2,498 bps respectively. The CDS series of Portuguese banks illustrate an average spread of 396 bps, just above Italy and Spain. However, the maximum value observed is far north of the average spread (1,271 bps). Graph 2, attached to this paper, illustrates the development of the weighted average bank CDS series over the sample period in basis points. The bottom seven rows of table 3 illustrate the summary statistics of the sovereign CDS series. The sovereigns CDS series resemble a similar image as the banking sector. Low credit risk is observed in Germany and France – i.e. 34 and 56 basis points on average. Moderate credit risk is observed on average for Italy (147 bps) and Spain (149 bps). The table demonstrates relative high CDS spreads for in Ireland (369 bps) and Portugal (330 bps). In addition, extreme credit risk is displayed by the Greek sovereign CDS series with a maximum observed spread of 14,912 bps – i.e. on the day of second bailout of the Greek Republic. Graph 1, attached to this paper, illustrates the development of sovereign CDS series over the sample period in basis points.
Part five: results
This section covers the key findings of this paper based on the hypotheses set out in the methodology section. The complete overview of the findings are presented in several tables, attached as an appendix to this paper. This section starts with the pre-analysis of the data and the model specification. Furthermore, this section elaborates on the key findings, their economic meaning and their implications in the light of the general research area. This section covers the discussion of the results on private-to-public, public-to-private and public-to-public credit risk transfer in the light of Ganger causality and the co-integration framework to capture evidence on financial markets’ interdependence. The key findings are summarised in table 1a and b, attached as an appendix to this paper. The complete output of the research is attached to this paper for an in-depth overview of the results – i.e. these appendices are not referred to in the discussion below.
5.1 Pre-analysis and model specification
In order to capture interdependences that could trigger financial contagion within the Euro-zone, several diagnostic test are completed to ensure robust results. First, the Augmented Dickey Fuller (ADF) test and the Johansen co-integration test are performed for the various subsamples to determine whether the 𝐴𝑅(𝑝) process is stationary and to define whether the two processes
24 share a common stochastic trend. In this case, the possibility of a zero mean or a trend stationary process are neglected. Unit root testing is therefore limited to an ADF model with an intercept as it is considered irrational to assume that a CDS time series will rise continuously (Alter and Schüler, 2012). Based on that notion, the level of co-integration is determined with the use of a restricted constant. In addition, stationary CDS series are exempt from the VECM analysis as the model cannot capture a common stochastic trend if one of the two processes is stationary – i.e. avoiding any spurious regression testing. Finally, the selection order criteria is used to determine the optimal lag-order by minimising the common information criteria. The bivariate pre-analysis test results are illustrated in appendices attached.
The Johansen co-integration test statistics presented in the tables attached confirms the existence of multiple co-integration relations within the three distinctive time periods. These results support the premise of domestic and cross border co-integrations of sovereign and bank CDS series. As opposed to empirical evidence regarding the pre-bailout “Stage 1” – denoted for a potential credit risk transfer from the domestic banking sector to the host sovereign – no co-integration relations between the host sovereign and bank CDS series are observed. The lack of significant results during this period may be caused by data limitations as DataStream started covering Euro-zone CDS markets late 2007. Surprisingly, various co-integrations of sovereign CDS series are detected over the course of “Stage 1”. In addition, the subsequent periods display multiple co-integration relations. These time periods illustrate co-integrations between the banking sector and sovereigns – i.e. both domestic and cross-border.
5.2 Co-integration and Granger causality analysis 5.2.1 Interpretation
The complete overview of the results of the Granger causality test and the co-integration analysis are illustrated in the tables attached to this paper. These tables report the output of the Granger causality test – with the use of the VARX framework as discussed in the methodology section – for the selected CDS series. The p-values are reported with associated significant levels “***”, “**” and “*” which refers to “1%”, “5%” and “10%” significance respectively. The results indicate whether the independent variable – shown at the far left side of the table – Granger causes the dependent variable. The subsequent tables report the complete output of the co-integration analysis with the use of the VECM framework, as discussed in the methodology section. The coefficients α measure the speed of adjustment in which the CDS series alters to the long-run relationship – found with the Johansen co-integration test. These adjustments take
25 place whenever the series are out of equilibrium. The series revert back to the long-run relationship if αSov has an opposite sign compared to the corresponding βSov (which is equal to
1 in the model setup) and the coefficient of adjustment is significant. The same interpretation applies for the bank CDS series – denoted with 𝐵𝑘 . In addition, one defines “real” co-integration of two series if both adjustment coefficients have opposite signs and are declared significant. The variable is not part of the error correction mechanism if the coefficient α does not have an opposite sign with respect to 𝛽 – i.e. this setup cannot define a long-run relation. The tables attached incorporate diagnostic tests to asses co-integration stability, serial autocorrelation and normally distribution of error terms – i.e. for both the Ganger causality test and the co-integration analysis. Co-integrations that do not incorporate serial correlation, have normally distributed error terms and pass the stability test are considered valid and are discussed below – i.e. denoted with “No”, “Yes” and “Yes” respectively as shown in the tables.
5.2.2 Stage 1
Table 1a does not report significant co-integration relations for the first period. This paper does not document long-run equilibrium relationships between CDS series of sovereign states and the banking sector – both domestic and cross-border. Hypothesis 1 can therefore be rejected in the long-run. However, the results illustrate signs of public-to-public credit risk transfer over the course of this period. The results and their implication are discussed below.
The Granger causality test (table 1b) indicates no evidence of short-run suggestive causality running from the domestic banking sector to the host sovereign. Hypothesis 1 can therefore be rejected in the short-run. The table uncovers short-run suggestive causality running from the sovereign states to the banking sector – i.e. both domestic as cross-border. The empirical evidence documents short-run causality running from peripheral sovereigns to the German banking sector after controlling for common risk factors. The results illustrate that the credit risk of German banks are Granger caused by movements in Portuguese, Italian, Greek and Spanish sovereign credit risk. Similar results are found for the Italian banking sector. Movements in credit risk of Italian banks are Granger caused by the host sovereign as well as the Spanish and Portuguese sovereign credit risk.
The co-integration analysis regarding public-to-public credit risk transfer merely applies to long-run suggestive causality between PIIGS. The evidence documents long-run suggestive causality running from Portugal to Italy. One can interpret coefficient 𝛽 as an elasticity since the
26 variables are expressed in logs. Implying that a 1 % increase in Portuguese sovereign credit risk leads to a 0.98% increase in Italian sovereign CDS spreads. In addition, the results document significant one way suggestive causality running from Spain and Greece to Italy. Implying that a 1% increase in Spanish and Greek CDS spreads leads to a 1.02% and 1.09% increase in Italian CDS spreads, respectively. The documented long-run effects vaporise over the course of the subsequent two selected time periods. Furthermore, the Granger causality test illustrates two-way short-run Granger causality running between Spain and Portugal along with Italy and Portugal.
5.2.3 Stage 2 and 3
Table 1a reports significant long-run relations over the course of the second period of the sample. These findings indicate merely domestic and public-to-public suggestive causal relations. The latter co-integration contribute to the third hypothesis, as stated in the methodology part. First, the co-integration analysis is discussed followed with a brief discussion on the Granger causality tests.
The results (table 1a) indicate that the German domestic sovereign and banking sector CDS series are tied together in the long-run – i.e. after government intervention. The co-integration relation implies that a 1% increase in the weighted average German banking CDS spread leads to increase of 3.34% in the German sovereign CDS spread. The sovereign CDS spread is not part of the error correction mechanism as the coefficient 𝛼𝑆𝑜𝑣 does not have an
opposite sign with respect to 𝛽𝑆𝑜𝑣. Normalising coefficient 𝛽𝐵𝑘 – which measures the potential
feedback of credit risk – does not produce significant results. Furthermore, the co-integration analysis uncovers a second and third domestic long-run equilibrium within the second stage – i.e. Portugal and Ireland. These co-integration relations imply that a 1% increase in the weighted average Portuguese and Irish banking CDS spread leads an increase of 2.59% and 0.70% of the associated sovereign series, respectively. No credit risk feedback from sovereign states to the domestic banking sector is detected over the course of the second stage. The results covering the cross-border co-integration relations display no signs of long-run causality running from sovereign states to the international banking sector. Although the some parameters (𝛼𝑆𝑜𝑣) are
significant, these sovereign CDS spreads are not part of the error correction mechanism and therefore provide no suggestive explanatory power towards the movements in credit risk of European banking sector in the long-run. Stage 2 documents empirical evidence on private-to-public credit risk transfer in the long run. However, the results only involve domestic effects – i.e. form the banking sector to the host sovereign. No cross-border private-to-public credit risk
