Cross-border banking: the implementation of the Single Resolution
Mechanism in the European Union
Master thesis for MSc International Economics & Business Faculty of Economics & Business, University of Groningen
4th January 2016 Joran Heeman S1884298 A.J.Heeman@student.rug.nl Supervisor: dr. M.J. Gerritse Co-assessor: dr. D.J. Bezemer
Abstract
The EU recently implemented the Single Resolution Mechanism (SRM) to address the systemic fragility of the increasingly globalized financial system. This paper tests whether the SRM solves the in the literature identified problem of biased intervention incentives of national authorities. A difference-in-difference test results in significant evidence for a relation between the introduction of the SRM and a decrease in default risk for banks with a higher degree of foreignness. This could imply a diminution of the problem. The European Banking Union should further improve regulation to create convergence between banks inside and outside the European Union.
1. INTRODUCTION
National governments have not succeeded in keeping up with the pace of globalization of the financial sector with regard to regulation (Claessens, Herring & Schoenmaker, 2010). This is why on August the 19th in 2014 the Single Resolution Mechanism (SRM) came in to force in the European Union. This supra-national resolution mechanism should help to address the systemic fragility of the European banking system. This could not be fought by following national policies in terms of regulation in the banking sector (European Commision, 2015). Current literature is mainly focused on problems arising from national authorities regulating cross-border banking and possible solutions for these problems. Now the Single Resolution Mechanism has come in to force, it is time to find out if such a supra-national supervisor could solve those problems. Since the introduction of the SRM has been approximately a year ago, these days provide excellent testing ground for the effectiveness of the SRM. Default risk of banks with a higher and lower degree of foreignness is compared before and after the introduction of the SRM by doing a difference-in-difference regression. The results of this regression should help to determine whether the introduction of the SRM could be seen as successful.
Beck, Todorov & Wagner (2013) analysed the effect of distortions created by an increase of cross-border banking on regulatory interventions during the financial crisis of 2007-2009. National authorities were in charge of supervision in the financial sector before the introduction of the European Banking Union in 2014. Resolutions where also conducted by these national authorities. Beck et al. found that during the recent crisis, national authorities were biased in taking action in terms of intervention in financial institutions. They find that nationally operating banks were intervened earlier than internationally operating banks. Financial globalization led to rapid growth in the balance sheets of many banks (Lane, 2013). This made it difficult for national regulators to adequately police risk profiles of banks. Beck et al. propose a solution for the problem of inefficient regulation of financial institutions by national authorities. They suggest a supra-national supervisor. This supervisor could improve welfare because it would take into account the effects that materialize outside the country.
bank bail-outs, which have cost taxpayers hundreds of billions of euros (EC, 2014). As the need for a mechanism like the SRM is largely discussed in the literature, the European Commission recognized this need. The SRM could potentially solve the problem addressed by Beck et al. The bias in intervention incentives should have reduced with the introduction of this SRM. This would mean that more internationally operating banks should see a decrease in default risk.
The severe effects of the recent financial crisis on the world economy show the importance of a properly functioning financial system. Liikanen (2012) discusses the costs of the recent financial crisis. These are (1) a significant loss in jobs, (2) increased government debt that will increase debt-servicing costs in the future, (3) a sharp fall in output and (4) a significant deterioration of financial positions of European households. The financial system is important for economic growth. Levine (1997) argues that there is a preponderance of theoretical reasoning that suggests a positive relationship between financial development and economic growth. This proper functioning of the financial system is especially important in Europe, since financing economic activities of households and companies by banks has traditionally been higher than financing through capital markets (Liikanen, 2012 p. 88). The ECB (2011) shows that in 2005-2009 the relative share of credit financing was 51% in Europe compared to 18% in the United States. The importance of a properly functioning financial system underwrites the need for proper and efficient regulation.
If current regulation would still have shortcomings, this poses serious threats for potential future financial crises. As shown by Beck et al., an inefficient regulation authority can exacerbate default risk of banks, especially in a crisis, with all the consequences that this entails for the global economy. Literature proposes a supervisor like the SRM, but also provides alternative solution. It is thus important to find out if this SRM has had the desired effect. If not, it may not be designed in the right way and the solution might lie in differently designed financial sector supervision.
addition to addressing the problems, a couple of proposed solutions can be found in the literature. These proposed solutions slightly differ from each other, but they all point in the direction of a supra-national supervisor.
It is attempted to find a relation between the introduction of the Single Resolution Mechanism and default risk of banks using Credit Default Swap (CDS) spreads, a widely accepted measure of default risk. A difference-in-difference test is done to examine whether variation in default risk could be assigned to the introduction of the SRM. A comparison is made between the effect of the degree of foreignness on default risk before and after the introduction of the SRM. Significant evidence is found for a relation between the introduction of the SRM and a decrease in default risk for banks with a higher degree of foreignness. Robustness checks confirm this relation, but also bring additional insight in interpreting these results.
The paper is structured as follows. The literature review in section 2 gives insight in research relevant to this paper. The empirical strategy and the data part in section 3 and 4 describe the used model and the data collected for the analysis. Section 5 shows the results of the difference-in-difference regression, which are confirmed and questioned by robustness checks in section 6. The results and its limitations are discussed in section 7 and 8. Section 9 discusses the implications and possibilities for further research and section 10 gives a conclusion on the findings of this paper.
2. THEORETICAL BACKGROUND
2.1 The problem of financial regulation by national authorities
Cross-border banking is introduced in the model as the possibility for a bank to be partially financed by foreign deposits and/or equity and the possibility to have foreign assets. This does not change the efficient intervention threshold, but it can change intervention incentives because the national authority only cares about domestic pay-offs. In deriving the domestic intervention threshold they show the implications of shares of foreign deposits, foreign equity and foreign assets. A high share of foreign deposits reduces external costs of failure for a national authority and makes early intervention less likely. Because the costs of a bank failure are partly borne by debt holders and firms, a higher share of domestic shareholders makes early intervention less likely. A high share of domestic assets increases external costs, so early intervention is more likely.
The theoretical model is also empirically tested in the paper of Beck et al. They examine 55 banks that were intervened between 2007-2009, using CDS spread at the time of intervention as a measure of regulatory lenience or strictness. In this examination they assume that regulators will intervene exactly at the moment that the health of the bank has deteriorated to the degree that the critical point is reached (p. 24). They test the hypothesis that the CDS spread at the time of intervention decreases in the share of foreign equity, increases in the share of foreign assets and increases in the share of foreign deposits (p. 25). They find that banks with a higher share of foreign equity were intervened relatively early compared to banks with a higher share of foreign deposits and assets (p. 35).
As a solution for this sub-optimal outcome they propose a supra-national supervisor, which could in principle always improve welfare, because this supervisor would also take into account the effects that materialize outside the country. They also argue that the success of a supra-national supervisor depends on coordination with the local supervisors. After all, the supra-national supervisor may not have perfect knowledge about the probability of success of a bailout. That is why the costs of introducing a supra-national supervisor should be weighed against the benefits of a national supervisor (Beck et al., 2013). One important condition for success of a supra-national supervisor Beck et al. give is that the jurisdiction of the supervisor should cover all activity of the institution that it supervises. When distortions arise in the banking system, the supra-national supervisor is only able to alleviate these distortions when they appear within the jurisdiction of the supervisor, in this case the European Union.
national governments base their decisions on the effect of government intervention on the local economy and financial system. They try to minimize costs that have to be incurred by taxpayers, who in the end fund a resolution of a financial institution. This shows that the focus of national authorities has been to minimize national externalities.
Acharya, Dreschler & Schnabl (2014) confirm that supervision of the financial sector by national authorities has not always led to optimal outcomes. They show that bailouts trigger sovereign credit risk and thereby weaken the financial sector. According to them, financial regulation should be well organized because intervention in the form of bailouts could be very costly, not only for governments, but also for the financial sector as a whole. If supervising authorities fail and bailouts are necessary, a supra-national supervisor could be advantageous. Due to the options of burden sharing on a larger scale, sovereign credit risk would increase much less with resolution on supra-national level.
The existence of the problem of regulation by national authorities in an increasingly integrated financial sector in the European Union is summarized in the financial trilemma by Schoenmaker (2011). The financial trilemma implies the incompatibility of financial stability, financial integration and national financial policies. The trilemma holds that two out of three aspects of the trilemma can be achieved, but not all three at the same time. The trilemma shows the trade-off between financial integration and national financial autonomy. Schoenmaker describes financial stability as being closely related to systemic risk. This is the risk that an event will trigger a loss of economic value or confidence in a substantial portion of the financial system that is serious enough to have significant adverse effects on the real economy (p. 57). The key issue of maintaining financial stability is the question if governments can still produce the public good of
financial stability at the national level (Schoenmaker, 2011).
2.2 Cross-border banking and financial integration
interconnectedness of these countries is especially high. France, Germany, the UK, the USA, Switzerland and the Netherlands accounted for around 47% of global cross-border banking assets in 2009. The USA, the UK, France, Germany, Japan and the Netherlands held 50% of gross global cross-border bank liabilities.
The Financial Stability Board has noticed this group of countries as the ones that house the largest and most important group of financial institutions. They publish a list of 30 global systemically important banks every year. In 2014, 14 banks were located in the European Union. 2 of these banks were located in Switzerland. This means that 53,3% of the global systemically important banks in 2014 were located in Europe. This confirms the importance of a properly functioning financial sector in the European Union, even for the global economy.
Claessens, (2010, p. 7) show that countries have become increasingly financially intertwined, as financial claims have grown much faster than trade and GDP. They distinguish two trends in this increasing financial integration. Firstly, a small number of large “too-big-to-fail” institutions dominate financial markets. Secondly, these institutions that dominate the markets are becoming more and more international in its activities. This increase is caused by financial innovation in advanced countries, but also by the increased degree of openness in emerging markets. Claessens et al. also discuss cross-border externalities related to the impact that Systemically Important Financial Institutions (SIFI’s) have on the global financial system and the global economy. These SIFI’s are apparent all over the world, but especially in Europe. Measured in foreign assets, the European SIFI’s have the largest international presence (67%), when compared to the USA (32%) and Asia (26%).
the international intra-bank funding. They also argue that they see diminishing effectiveness of standard policy tools of regulators, with a need for broader and international coordination.
Due to the unprecedented degree of globalization, foreign banks have become more important in domestic and financial intermediation, especially in the last two decades. Claessens & van Horen (2012) argue that in terms of loans, deposits and profit, current market shares of foreign banks average 20% in OECD countries and close to 50% in emerging markets and developing countries (p. 3). In this increase in global financial integration, bank entry tends to be regionally concentrated. Both in 1995 and 2009, the share of foreign banks coming from countries within each region is always more than 50%. This could be because banks may benefit from operating in a certain region, acquire specific knowledge and re-use that specific knowledge by entering new countries within that region. Claessens & van Horen found that on average foreign banks reduced lending more compared to domestic banks during the global crisis. This contributes to financial instability in host countries. On the other hand, when a foreign bank was relatively dominant in a banking system, it was a more stable source of credit than domestic banks. This means that when foreign banks are confronted with exogenous shocks like a global crisis, their actions do not only influence the home country, but also the host country.
2.3 Contagion and financial stability
Increased financial integration has led to an increase of contagion effects. This increase of (cross-border) contagion is one of the reasons that national authorities have had more and more trouble with regulation and supervision of financial institutions.
Allen & Gale (2000) explain contagion by focusing on one channel of contagion, which is the channel of claims that banks have on one another. They use a framework in which small shocks can lead to large effects by means of contagion. Their results suggest that the degree of completeness of the interbank market combined with the degree of interconnectedness determines the degree of contagion. A liquidity shock in one region will spread by contagion to others in a banking system with high interconnectedness and an incomplete interbank market. When a banking system has high interconnectedness, but a complete interbank market, no contagion will occur after a liquidity shock in one region. In a banking system with low interconnectedness and an incomplete interbank market, low contagion arises.
National interbank markets in local currencies have shifted to an integrated and deep euro interbank market (Schoenmaker & Oosterloo, 2005). The effect of this shift on financial stability is not clear-cut, but because of this increase in border banking activities, the chance of cross-border contagion has increased. Schoenmaker & Oosterloo (2005) analyse the social benefits of a bailout. If these would be sufficiently high in the rest of Europe, then a centralized supervisory mechanism would work. They find a clear and statistically significant upward trend of cross-border penetration within the EU, when looking at a sample of 30 large EU banking groups (from 2000-2003). The number of groups that could have significant cross-border externalities has grown from 6 to 9. They therefore argue that policymakers may need to consider broader European solutions for financial supervision and stability to deal effectively with potential cross-border externalities (p. 24).
From October 2008 to end 2012, European governments have used a total of 1.6 trillion euros of financial aid in the form of guarantees, liquidity support, recapitalisation and asset relief measures (Liikanen, 2012 p. 20). Government support in financial institutions is concentrated in a small number of European countries (90% of the subsidies is transferred to the largest institutions), in a small number of banks. This does not mean that these small numbers of banks were the only ones that profited from this aid. Due to contagion, intra-bank linkages and spill-over effects, other banks indirectly profited from this government aid as well.
Lack of coordination and failure of working together by national authorities has been present during the financial crisis of 2007-2009. Other problems were failure to recognize insolvency, slow reactions and randomly assigned resolutions (Lehman Brothers failing as opposed to AIG receiving a massive bail-out). Claessens et al. (2010) argue that in 2008, 16 failures took place. These involved bankruptcy, conservatorship, government takeover or assisted mergers. This amount of 16 failures is more than all the failures in the 20 years preceding 2008 combined (p. 21). Bailout costs have appeared to be higher than just the costs of intervening a bank in danger of default. Acharya et al. (2014) analysed bank bailouts through government support, focusing on the ex post fiscal costs of bailouts. By examining CDS spreads of both governments and banks, they explored the link between bank bailouts and sovereign credit risk. They find a so-called “two-way feedback loop” where issuance of bonds helps to fund bank bailouts but dilutes existing bondholders and raises sovereign credit risk. A two-way feedback loop between sovereign and financial sector credit risk occurs due to financial intermediaries being exposed to government debt directly through the holding of government bonds or indirectly through the value of explicit government guarantees. Their findings indicate that bank bailouts triggered sovereign credit risk in the Eurozone and thereby weakened the financial sector.
undermines the Single Market principle in the EU. Their model predicted possible cost savings of some 15% due to economies of scale when centralizing financial supervision in the European Union.
2.5 The design of a supra-national supervisor
There seems to be consensus in the literature that, due to increased financial integration, an increase in cross-border banking and more severe contagion effects, financial regulation on national level does not seem maintainable. A supra-national supervisor could solve this problem. Some different opinions and insights have been presented in the literature on how a supra-national supervisor should be designed and act in regulation of cross-border banking.
Goodhart & Schoenmaker (2009) developed two mechanisms for ex ante fiscal burden sharing of a fiscal crisis in Europe. The first is a general fund, financed by the participating countries, to alleviate the burden sharing. The second is a mechanism of specific burden sharing where the burden of recapitalisation in case if a bailout is assigned to countries where the financial institution operates. Both mechanisms are subject to free-rider problems, as the financial stability created by these mechanisms is general European economic good.
Brunnermeier et al. (2009) argue that the country or entity that bears the burden in case of a bailout of a financial institution should be in charge of regulating that financial institution (p. 29). This means that foreign branches of international banks should be seen as separate entities that should hold a separate “pot” of capital. The country where the branch operates should regulate the institution, not the home country. According to Brunnermeier et al., this would reduce synergies of cross-border banking. The more European approach they propose is one with regulation transferred to a European institution. This would only work if burden sharing could be agreed upon within the Euro area. The separate-entity principle of Brunnermeier et al. would work if financial intermediaries could be seen as separate institutions.
that these actions maximize value for the financial institution. The territoriality approach is an approach where the subsidiaries of a financial institution are seen as separate, where they have to be financially and functionally independent. Each of these subsidiaries are separately supervised and resolved by local laws. The modified universal approach holds that the home country is in charge of the resolution but cooperation between different jurisdictions is possible to impose such resolution. The first approach can be seen as giving up national sovereignty. When this would be combined with burden sharing between countries, it would incentivize countries for better coordination in supervision and resolution (p. 89). In the second approach there is no need for central burden sharing or central coordination of supervisory and regulatory activity. The third approach addresses financial stability, financial integration and national financial autonomy in an intermediate way, and makes sure that not one of the three has got to give (p. 90).
In trying to find a solution for the problem addressed in the financial trilemma, Schoenmaker (2011) proposes to reverse the level of financial integration, or to create a supra-national supervisor. The foundation of a supra-supra-national supervisor would mean that regulation, supervision and stability would be taken care of at central European level.
Schoenmaker & Siegmann (2014) also evaluate possible solutions, by looking at the body of literature on international policy coordination. The first solution they analyse is to create a supra-national approach to financial stability. This is similar to the supra-supra-national approach to monetary stability with the establishment of the ECB. The second solution is a binding rule among national governments to share the burden of failing banks in order to maintain financial stability (p. 335).
2.6 The European Banking Union
financial crisis with negative effects on depositors and tax-payers (Single Resolution Mechanism). Thirdly, the EU-wide coverage of deposit insurance (not implemented yet). All these measures together comprise the single rulebook, which according to the European Commission, is necessary to achieve stability and integrity in the EU’s internal financial markets.
The Single Resolution Mechanism (SRM), which came into force in August 2014, applies to all countries under the Single Supervisory Mechanism (SSM). The SRM is implemented after realisation of the SSM because the supervision mechanism first had to be functional (Breuss, 2013). It consists of a Single Resolution Board (SRB), which manages a resolution of a failing bank. The resolution is financed by the Single Resolution Fund (SRF), which is funded by the banking sector of all countries under the SRM (EC, 2015).
The task of the ECB is to detect risks that may threaten the stability and viability of banks. To fulfil this task, the ECB is endowed with investigatory and supervisory powers. When the ECB signals a bank in danger of failing or facing severe difficulties, the SRM steps in. The SRB prepares a resolution, in which authorities are closely involved. The SRB then recommends the European Commission whether or not to resolve the failing bank and the European Commission makes the final decision. If the concerning bank has been decided to be resolved, execution of the resolution is in hands of national authorities. The SRB monitors this executed resolution (Breuss, 2013).
The focus of this paper is on the Single Resolution Mechanism, not the Single Supervisory Mechanism or the centralized deposit insurance mechanism. The SSM takes care of prudential regulation, which is not expected to solve biased incentives in intervening in financial institutions. The SRM complements the SSM to a full supra-national supervisory mechanism. This could possibly solve the problem of biased intervention incentives of national authorities.
different countries is central. In this approach, they however do not suggest a supra-national mechanism that takes over the power of national authorities, as is the case for the SRM.
3. EMPIRICAL STRATEGY
The main goal of this paper is to identify whether or not the introduction of the SRM has led to a reduction or disappearance of the bias that was apparent when banks were regulated by national authorities. The empirical analysis of Beck et al. (2013) suggests that national regulators had biased incentives when dealing with cross-border banks in the recent financial crisis of 2007-2009. Banks with a higher share of foreign equity were intervened relatively early while banks with a higher share of foreign deposits and assets were intervened relatively late. In their conclusion, they suggest the establishment of a national supervisor, which could decrease this bias. As this supra-national supervisor is present since the 19th of August 2014, the effects of this introduction could be visible at this moment. It should be possible to determine whether the introduction of the SRM has already had the desired effect with regard to the biased intervention incentives.
3.1 Why the SRM could solve the problem of biased incentives
They write this in their paper as:
λR - (1 - λ)c = 1 (1.1)
Solving for λ gives
𝜆 ∗=
%&'(&' (1.2)
Where λ is the probability that an investment of a bank succeeds, R is the return from this investment and c represents the external costs of a failure. The initial investment can be recovered when a government intervenes, as shown by the right hand side of equation 1.1. λ* represents the intervention threshold, which equates the expected returns from continuation with the return from immediate liquidation (p. 14). In other words, when λ* is higher, the return on intervention is higher than the return on continuation of the bank, so the bank will be intervened.
As already argued, when cross-border banking is introduced in the model, this does not change the most efficient intervention threshold but it does change the pay-offs to domestic agents. Taking foreign assets as an example. When a bank has a high share of foreign assets, the external costs of a failure for domestic agents is lower than when the bank has a high share of domestic assets. This means that when a bank has a high share of foreign assets, external costs of failure are low, because a national authority only cares about external costs of the failure that accrue to domestic agents. The intervention point of the national authority thus will be later (intervention threshold increases) than the efficient intervention point.
When the supra-national supervisor is introduced into the model, the biased incentives to intervene would reduce or even disappear. In the simplest way, it could be said that the intervention model of a supra-national supervisor is the same as the model for the national supervisor where no foreign assets/deposits/equity exist. This means that the intervention model with a supra-national supervisor could be written as shown in formula 1.1 and 1.2.
For the sake of simplicity, it is assumed that all cross-border banking happens within the jurisdiction of the supra-national supervisor. This means that if banks have foreign assets/deposits/loans within this jurisdiction then all external costs of a failure accrue to agents within this jurisdiction. This would then in turn mean that possible intervention will take place at the most efficient intervention threshold, so there are no biased incentives to intervene in an earlier or later phase.
Banks is around 47% (in 2007) but this also includes foreign activity outside the jurisdiction (EU) of the SRM. Thus in practice, the intervention threshold of a supra-national supervisor is still influenced by the share of foreign assets/deposits/equity of the bank. The intervention point of the supra-national supervisor will however be closer to the most efficient intervention threshold. This can be explained by looking at the external costs of a failure. The supra-national supervisor cares about all external costs of a failure that accrue to the agents within its jurisdiction. National interbank markets in local currencies have shifted to an integrated and deep euro interbank market (Schoenmaker & Oosterloo, 2005). This means that it can be assumed that the largest part of external costs related to a failure accrue to agents inside the jurisdiction of the SRM. In theory, the introduction of the SRM would mean that intervention incentives will be less biased and more in line with the most efficient intervention threshold.
3.2 Analysing default risk before and after the introduction of the SRM
As Beck et al. (2013) have done, the strategy here is to look at the default risk of banks. Before the introduction of the SRM, banks that would be intervened relatively late, or after the efficient intervention point, would see higher default risk than banks that would be intervened relatively early or before the efficient intervention point. Banks that at first would be intervened relatively late would now be intervened earlier, because of the reduction of the bias due to the introduction of the SRM.
Due to the introduction of the SRM, it could be expected that the incentives for intervention will be more in line with the most efficient intervention threshold. Thus, banks with a higher share of foreign assets/deposits, or degree of foreignness, would see lower default risk. This would be the case because of earlier intervention when necessary, due to diminished bias in intervention incentives. When formulating similar to Beck et al. (2013) this would mean the following. Banks with a higher share of foreign equity and/or lower share of foreign assets and deposits would see an increase in default risk, because now they would be intervened relatively late with respect to the period before the introduction of the SRM. Banks with a higher share of foreign deposits and assets would see a decrease in default risk because with a supra-national supervisor, they would be intervened earlier because of the diminished bias.
before the introduction of the SRM; the second period is after the introduction of the SRM. Using the difference-in-difference test, the following model is used:
Default risk = β1 Foreignness + β2 SRM + β3 Foreignness x SRM + ε (1.3)
By using this model (1.3), it is attempted to find out if variation in default risk of banks (= the difference between the log of CDS spread of the bank and the log of CDS index) can be assigned to the introduction of the SRM. The coefficient of interest β3 is an interaction between Foreignness and SRM. Foreignness is a continuous variable that represents the degree of foreignness of a bank. This is measured as the share of foreign assets or foreign loans as a percentage of total assets or loans. SRM is a dummy variable that is 1 after the introduction of the SRM on August the 19th 2014. The coefficient of interest β3 shows to what extent default risk of a bank would decrease or increase more with an increase of the degree of foreignness after the introduction of the SRM compared to the period before the introduction of the SRM. In this way, default risk of banks dependent on the degree of foreignness can be compared in the two periods. Based on theoretical reasoning by means of the model of Beck et al. (2013) and literature on cross-border banking and regulation the following is expected. After the introduction of the SRM, an increase in the degree of foreignness of a bank should be related to a decrease in default risk. This would be shown by a negative coefficient β3.
4. DATA
4.1 Credit Default Swap spreads
ways. In a cash settlement, the protection buyer receives an amount of cash to compensate for the loss caused by the credit event. In a physical settlement, the buyer delivers the reference obligation to the seller, and the buyer receives the notional amount in return (Norden & Weber, 2004). When no credit event occurs, the protection buyer still pays the premium and the derivative contract terminates at maturity.
The use of CDS spreads is often seen in the literature. Longstaf & Mital (2005) use credit default swap premia to provide direct measures of the size of the default and non-default component in corporate yield spreads (p. 2214). They use CDS spreads as a measure of the default component, to measure the default component in corporate yield spreads.
Advantages of using credit default swap spreads instead of corporate bond yield spreads are discussed by Ericsson, Jacobs & Oviedo (2009). The first advantage comprises the fact that credit default swap spreads are already spreads. This means no need to find a benchmark risk-free yield curve, and it thus no suffering from added noise to the model from this risk-free yield curve. The second advantage is discussed by Blanco, Brennan & Marsh (2005). CDS spreads may reflect changes in credit risk more accurate than bond yield spreads. These might be related to the more important part of non-default component in bond yield spreads, that may blur the default component in these bond yields. Another advantage is the fact that CDS are traded more frequently than corporate bonds. This means that the amount and frequency of the data is more useful because CDS data is available on daily basis. Corporate bond yields data is mainly available on monthly basis.
4.2 Degree of foreignness
Table 1. Descriptive statistics of the size of banks per measure of degree of foreignness.
Variable Observations Mean Std. Dev. Min Max log Size (assets) 4410 19.33889 2.066204 10.29612 21.46736 log Size (loans) 4263 19.31636 1.271979 15.7338 21.56841 log Size (assets+loans) 8673 19.32782 1.722157 10.29612 21.56841
The approach to create a measure of the degree of foreignness is somewhat different from Beck et al. They use the share of foreign loans or deposits when data on foreign assets are missing and the available data on foreign assets as loans is used as a complement for missing shares of foreign deposits (p. 27). As can be seen in table 1, the data on foreign assets seems to cover the whole spectrum, from the smallest bank to the largest bank. Data on foreign loans is mainly found in annual reports of the larger banks in the sample.
When looking at balance sheets of banks, foreign assets as a share of total assets could be said to cover the widest range of foreign activity of a bank, where foreign loans are only part of this activity. Foreign deposits are not used as a measure of the degree of foreignness in this paper, because banks rarely publish data on foreign deposits. Either way, foreign assets and foreign loans are not interchangeably used as a measure of the degree of foreignness.
4.3 Statistical details
The dataset comprises a total of 46 banks with home countries in the European Union. These are all under the Single Supervisory Mechanism. The data on the degree of foreignness and CDS spreads contains observations of these banks from January 2013 until October 2015. The data collected on CDS spreads is on daily basis, but averaged on weekly basis. The variable that measures the degree of foreignness is on yearly basis.
Table 2. Descriptive statistics of the sample of banks that are under the SRM.
The weekly CDS spreads are positively skewed, with a skewness of 4.58. The data on CDS spreads has some extreme outliers. The variable in normal state has a mean of 255.65, with a standard deviation of 318.35 and a maximum of 6357.93. Since it can be expected that those extreme outliers can bias the results, these have to be controlled for. The main part of the extreme values of CDS spreads is situated in the highest percentiles, with a value 1602.58 at the 99th percentile and the maximum of 6357.93. To make sure that these extreme values do not bias the results, the extreme values are so-called “Winsorised”. This means that the 99th percentile replaces the values higher than the 99th percentile. The CDS spread variable now has a maximum of 1602.58. CDS spreads are now less skewed, with a skewness of 2.94. To control for this skewness, the log of the Winsorised CDS data is taken.
In the regression, CDS spreads are used as a measure of default risk of banks. To make sure that the explained variation in CDS spreads is not caused by other economic or political factors such as interest rates, GDP growth, sentiment in the industry or other laws, regulations or political sentiment, the dependent variable is the difference between the log of the CDS spreads per bank and the log of the CDS index per region (which are Europe and the UK, where Asia and North America are added in the robustness section).
CDS spread data is on daily basis. Five year CDS spreads are used because this five-year maturity derivative is the most widely traded under the CDS spread credit derivatives. To control for the sharp daily movements and spikes due to irregular trading, the CDS spreads are smoothened by using the weekly average of the CDS spread, as Eichengreen et al. (2012) have done. In figure 1 the mean of CDS spreads is plotted per quartile of degree of foreignness. Table 3 shows the mean degree of foreignness, mean weekly CDS spreads and mean weekly CDS spreads before and after the introduction of the SRM.
Table 3. Mean degree of foreignness and mean weekly CDS per quartile of foreignness, in total, before and after the introduction of the SRM
Figure 1. Mean weekly CDS spreads per quartile of degree of foreignness from a year before until a year after the introduction of the SRM.
The graph in figure 1 shows the mean of weekly CDS spreads per quartile of degree of foreignness. In calculating the quartiles, both foreign assets and foreign loans as percentage of total assets/loans are included. The vertical line marks the introduction of the Single Resolution Mechanism on August 19th 2014 (week 34). It can be seen that around the introduction of the SRM, mean CDS spreads converged. Shortly after the introduction, there seems to be a joint decrease in CDS spreads. This is in line with the expectations that a supra-national supervisor would lead to a decrease in default risk of internationally operating banks. The mean CDS of Q1 and Q3, or the mean CDS of banks with a lower and higher degree of foreignness respectively, have been converging around the introduction of the SRM. Before this convergence, they have been jointly moving, apart from of January 2014 until July 2014. Banks with a higher degree of foreignness experienced lower default risk than banks with a lower degree of foreignness.
A couple of months after the introduction of the SRM, the mean CDS of Q1 and Q3 are diverging. This could mean that the default risk of banks with a low degree of foreignness increases when compared to banks with a high degree of foreignness. Banks with a median degree of
foreignness and a high degree of foreignness, or Q2 and Q4, seem to have profited the most from the introduction of the SRM, assuming that variation in CDS spreads can be assigned to the introduction of the SRM. These groups show the lowest mean CDS spreads. The mean CDS spreads of Q4 also moves in the expected manner, however it does not show a very significant change around the introduction of the SRM.
4.3 Timing
On the 19th of August 2014 the SRM came in to force. This day is used as the starting point of the treatment period of the difference-in-difference regression. In analysing the effect of the introduction of the SRM on default risk, CDS spreads of a year before the introduction of the SRM until a year after the introduction of the SRM are analysed.
5. EMPIRICAL RESULTS
5.1. Difference-in-difference regression
Table 4. Main result: difference-in-difference regression
(1) RE (foreign assets) VARIABLES CDS weekly Foreignness 1.199*** (0.465) SRM 0.643 (0.498) 1.SRM x Foreignness -1.743* (1.035) Constant -0.628*** (0.232) Observations 1,330 Number of panels R-squared 28 0.0631
Note: Robust standard errors in parentheses (clustered by SRM x bank)
*** p<0.01, ** p<0.05, * p<0.1. Model 1.3 is tested in a random effects regression with time fixed effects, in the period of a year before until a year after the
The (1) random effects regression in table 4 shows a negative coefficient that is significantly different from zero at the 90% level. The degree of foreignness is measured as the percentage of foreign assets as a share of total assets. This coefficient can be interpreted as follows. After the introduction of the SRM, a 1 per cent point increase in the degree of foreignness is related to a decrease in default risk (= the difference between the log of the CDS spread and log CDS index) by 1.743% more than it would have decreased before the introduction of the SRM, ceteris paribus. This is the main result and others will follow.
The interpretation of the results can also be approached in another way. Since the dependent variable is a log transformed variable. This means that 𝑒*%.,-.∗%∗/./% represents the increase of the
degree of foreignness by 0.01 (or 1%). The outcome is 0.9827. With an increase of 0.01 of the degree of foreignness, the dependent variable is multiplied by 0.9827. This is a decrease of 1.727%.
As already mentioned, CDS spreads are averaged on weekly basis to rule out spikes and other volatility. Since the degree of foreignness is measured on yearly basis, a regression is also done with CDS spreads averaged on quarterly basis to bring both measures more in line with regard to variation in time. The results are shown in the table in appendix 1. The coefficient of interest here is a slightly less negative (-1.11) and not significant. The results are more or less in line with the results from the main regression in table 4.
Before running the regression, a Wooldridge test is done look for serial correlation. With an F-statistic of 640.728 and a p-value of 0.000, the test shows enough evidence to reject the null hypothesis of no serial correlation in panel data. A test for the presence of heteroscedasticity shows enough evidence to reject the null hypothesis of homoscedastic standard errors. To control for this serial correlation and heteroscedasticity, clustered standard errors are used. Since default risk of banks under the SRM is expected to be different in the period before and after the introduction of the SRM, standard errors are not clustered by bank. It would not be logical to assume that standard errors do not differ before and after the introduction of the SRM. That is why they are clustered per bank before and per bank after the introduction of the SRM.
Serial correlation thus is fought by clustering standard errors by bank before and after the introduction of the SRM. A Dicky Fuller test is conducted to find out in which panels the serial correlation is found. 15 panels do not show evidence for serial correlation. The main regression is also conducted with only these panels, however this did not result in significant results for the coefficient of interest. These results are shown in the table in appendix 2.
Foreign assets is chosen as a measure of the degree of foreignness, because this measure covers the broadest range of banks in terms of size. Foreign loans and both foreign loans and foreign assets together will be used as a measure in the robustness section to find out if this shows any different results. Time fixed effects on quarterly basis are used. To confirm the need for using time fixed-effects in the regression, a test is done to find out whether all time fixed-effects coefficients are jointly equal to zero. Enough evidence is found to reject the null hypothesis that these are jointly zero, which means that time fixed-effects are needed. The use of time fixed effects on quarterly basis is justified by looking at the frequency that banks publish their results.
6. ROBUSTNESS CHECK
6.1 Banks not under SRM
To do a robustness check, an extra group of countries is added to the sample. Adding this group of countries, a difference-in-difference-in-difference model can be tested to compare default risk of banks under the SRM with banks not under the SRM, before and after the introduction of the SRM.
The table in appendix 3 shows the descriptive statistics of the group of banks added to the sample. This added group of banks has to be utmost similar to the banks in the sample that are under the SRM. The banks added to the sample are banks of which the home countries are appointed by the IMF as advanced economy, as is the case for the home countries of banks that are under the SRM.
When the introduction of the SRM would have had the desired effect on default risk of banks, CDS spreads would be converging towards CDS spreads of banks that have not been affected by the biased incentives of national authorities as happened in the EU.
Figure 2. Mean CDS spreads of banks under and not under the SRM from 01/2013 until 09/2015.
In figure 2, the mean CDS spread of banks under the SRM and banks not under the SRM is plotted. The vertical line represents the introduction of the SRM (19th August 2014). Analysing the path of both means of CDS spreads from the beginning of 2013 until July 2014, CDS spreads have been converging. After the introduction of the SRM, the mean CDS spreads are diverging again. This is not in line with the expected effect of the introduction of the SRM. Theory suggests that the introduction of a supra-national supervisor would lead to a decrease in default risk of banks with a higher degree of foreignness.
50 10 0 15 0 20 0 25 0 30 0 Me an C D S sp re a d 01/'13 01/'14 01/'15 01/'16 Time
6.3 Difference-in-difference-in-difference model
To find out whether there is difference in default risk between banks not under the SRM and banks under the SRM, before and after the introduction of the SRM, a difference-in-difference-in-difference model is used:
Default risk = β1 Foreignness + β2 SRM + β3 EBA + β4 Foreignness x SRM + β5 Foreignness x EBA + β6 Foreignness x SRM x EBA + ε
(1.4)
The added variable is EBA, which is a dummy that is 1 when the bank is under the SRM and 0 when it is not. The dependent variable is default risk, expressed as the difference between the log CDS spread of a bank and the log CDS index of the world region it operates in. Four region CDS indices are available, for Europe, the United Kingdom, North America and Asia. There are four banks with Australia as home country. For these countries the difference between the log of CDS spread of the bank and log of CDS index of Asia is taken as a dependent variable. This is the region that is most nearby for Australia and the Asian CDS index is expected to capture variation in CDS spreads that can be assigned to economic and political events. Standard errors are again clustered by bank and before introduction or after introduction of the SRM.
Table 5 shows the results of the regression, where foreign assets as a percentage of total assets is used as a measure of the degree of foreignness. The coefficient of interest here is β6. This coefficient indicates to what extent a bank that is under the SRM experiences higher or lower default risk after the introduction of the SRM when increasing the degree of foreignness, relative to banks that are not under the SRM.
Unfortunately running the regression does not result in any significant results. The direction of the coefficient however is noteworthy. It shows that banks under the SRM experience a decrease of 0.973% more in default risk with a one-percentage point increase in the degree of foreignness, after the introduction of the SRM than banks that are not under the SRM.
Table 5. Difference-in-difference-in-difference regression
RE diff-in-diff-in-diff
VARIABLES (Foreign assets)
Foreignness 0.0868 (1.024) SRM -0.0450 (0.352) EBA 0.213 (0.504) SRM x Foreignness -0.183 (1.063) EBA x Foreignness 0.528 (1.295) 0. SRM x 1.EBA -0.492 (0.664) SRM x EBA x Foreignness -0.973 (1.621) Constant -0.123 (0.321) Observations 4,398 R-squared 0.0294 Number of panels 60
Note: Robust standard errors in parentheses (clustered by SRM x bank) *** p<0.01, ** p<0.05, * p<0.1. Model 1.4 is tested in a random effects regression with time fixed effects, in the period of a year before until a year after the introduction of the SRM. Foreign assets are used as measure of the degree of foreignness.
6.2 Control variables
Next, a group of control variables is added to the regression. Adding these control variables to the regression could increase the explanatory power of the model. Part of the variation in default risk could now be assigned to these variables for which is controlled. The remaining part could be explained by the coefficient of interest. This means that the variation in default risk that could be assigned to the coefficient of interest is not influenced by other factors. These factors are now covered by the control variables in the regression. All these control variables are found to be explaining variation in CDS spreads. Literature in the field of CDS spreads identifies variables that are found to be explaining some of the variation in CDS spreads.
unreserved impaired loans/equity (%), capital (tier 1 ratio (%), leverage; equity/total assets (%)), operations (ROA (%), ROE(%)) and liquidity (net loans/deposits & short term funding(%), liquid assets/deposits & short term funding (%)). Following a fixed effects panel data regression, they find that nearly 58% of the variation in CDS spreads could be explained by these ratios. All these variables are included as control variables. The table in appendix 4 describes these variables; the table in appendix 5 shows descriptive statistics.
Beck et al. (2013) use several control variables in their model. They include bank size, defined as the log of bank total assets to control for the fact that banks might be too-big-to-fail and thus intervened relatively late. This control variable is also included, because the size of banks might also be a factor that explains variation in CDS spreads. Larger banks in terms of assets might have a lower perceived default risk, because they are perceived as too-big-to fail. This is confirmed by looking at correlation between default risk and size, which shows a negative correlation of -0.157 (as can be seen in the table in appendix 6).
The Financial Stability Board (FSB) publishes a list every year with 30 financial institutions that are considered Global Systemically Important Banks. These banks are considered systemically important for the global financial system. A designation to be systemically important by the FSB, a renowned institution, could affect the default risk of banks. To control for variation caused by being a Global Systemically Important Bank, a dummy variable is added which is 1 if a bank is declared to be Global Systemically Important in 2014.
Table 6. Difference-in-difference regression with control variables.
RE with control vars. VARIABLES (Foreign assets)
Foreignness 0.902 (1.025) SRM 0.571 (0.401) avliqas -0.00476 (0.00735) LLR 0.00166 (0.0311) lTotAs 0.167 (0.315) Tier1 -0.00765 (0.0123) ROAA 0.0624 (0.0852) ROAE -0.00321 (0.00307) GSIBdum -0.0489 (0.539) 1.SRM x Foreignness -1.618 (1.071) Constant -3.676 (5.947) Observations R-squared 5,878 0.066 Number of panels 80
Note: Robust standard errors in parentheses (clustered by SRM x bank) *** p<0.01, ** p<0.05, *p<0.1. Model 1.3 is tested in a random effects regression with time fixed effects, in the period of a year before until a year after the introduction of the SRM. Foreign assets are used as measure of the degree of foreignness.
Table 7. Difference-in-difference regression with all events regarding the SRM.
RE with all events VARIABLES (For. assets & loans)
Foreignness 0.619 (0.504) SRM1 0.0302 (0.0364) SRM2 -0.0292 (0.0923) SRM3 -0.0357 (0.0954) SRM4 -0.135 (0.0859) SRM5 -0.0749 (0.0478) SRM 0.455* (0.262) 1.SRM x Foreignness -1.207* (0.631) 1.SRM1 x Foreignness 0.116 (0.0775) 1.SRM2 x Foreignness 0.264 (0.201) 1.SRM3 x Foreignness 0.271 (0.223) 1.SRM4 x Foreignness 0.397 (0.254) 1.SRM5 x Foreignness 0.189 (0.139) Constant -0.386* (0.208) Observations 4,022 Number of panels 85
Table 6 shows the results of the regression with all control variables added. Comparing regression 1 in table 6 with regression 1 in table 4, it can be said that the coefficient becomes slightly less negative, but loses its significance. The addition of all these control variables does not help by explaining variance in default risk in terms of significance of the results. What it does show is that when these control variables are included, the coefficient of interest does not radically change. This could mean that adding control variables shows that the probability is higher that the coefficient of interest actually explains the variation in default risk that could be assigned to the introduction of the SRM. This could however not be said with confidence.
6.3 Starting point of treatment period
It could be expected that when it becomes clear that certain policy measures that affect the entity are apparent, the effect of those measures will be reflected in the CDS spreads. As Fama (1970) argues, in the light of theory of efficient markets, there is a point in time where prices fully reflect all available information. A regression is done with all events included in one regression, to find out at what moment in time a significant change in default risk occurs. This is the moment that CDS spreads reflect all available information about the effects of introduction of the SRM.
Table 8 shows a timeline of events with regard to the introduction of the SRM. The first relevant event is the proposition of the Single Resolution Mechanism. Since this was the moment that the mechanism was proposed and not yet agreed upon, this event is not taken into account. After this first event, the following events could be seen as moments where it became clear that the probability increased that resolution mechanism would come into force.
The regression includes dummies that indicate periods between events. SRM1 indicates the period from the 19th of December 2013 until the 20th of March, and so on. Both foreign assets and loans are used as a measure of the degree of foreignness to see at what point in time there is a significant relation between events regarding the introduction in time and default risk of banks.
occurred. The 19th of August thus seems to be a well-chosen starting point of the treatment period in the difference-in-difference regression in the main result.
Table 8. Timeline of different events regarding the Single Resolution Mechanism.
Date Entity Action taken Variable
10/07/2013 European Commission Proposition of the Single Resolution Mechanism - 19/12/2013 European Council
Agreed general approach on Single Resolution
Mechanism SRM1
20/03/2014
European Parliament & European Council
Reached a provisional agreement on proposed
Single Resolution Mechanism SRM2 15/04/2014
European Parliament & European Council
Adoption of the Single Resolution Mechanism
as proposed by the Commission in July SRM3
21/05/2014 26 EU member states
Signed intergovernmental agreement on transfer and mutualisation of contributions to the Single
Resolution Fund SRM4
30/07/2014 European Union
Regulation no 806/2014 establishing Single Resolution Mechanism published into Official
Journal of EU SRM5
19/08/2014
European Banking Union
Single Resolution Mechanism entering into
force SRM
6.4 Measures of degree of foreignness
Until now, the degree of foreignness is measured mainly by using foreign assets as a percentage of total assets. Foreign loans as a percentage of total loans as a measure of the degree of foreignness does not measure all foreign activity of a bank. That is why foreign assets as a percentage of total assets seems the most legitimate measure of the degree of foreignness of a bank. This measure comprises most international activity of a bank. As mentioned, larger banks appear to be publishing foreign loans more often than foreign assets. Taking foreign assets as a measure of the degree of foreignness seems to cover de broadest spectrum in terms of size of banks.
Unfortunately not all banks publish information about the foreignness of their assets, so in some cases the data collected on the degree of foreignness is foreign loans. Using this information on foreign loans increases the sample of banks and their degree of foreignness.
measures are included in the (3) regression, the coefficient of interest (SRM x Foreignness) is still negative and significant, even at the 95% level.
Assuming that both foreign loans and foreign assets are legitimate measures of the degree of foreignness; it could be said that a 1% increase in the degree of foreignness of a bank is related to a decrease in default risk of between 1.358% and 1.743% more after the introduction of the SRM than before the introduction of the SRM.
Table 10. Difference-in-difference regressions with different measures as the degree of foreignness.
(1) RE diff-in-diff (2) RE diff-in-diff (3) RE diff-in-diff VARIABLES Foreign assets Foreign loans Foreign assets & loans
Foreignness 1.199*** 0.0738 0.774 (0.465) (1.114) (0.501) SRM 0.643 0.272 0.508** (0.498) (0.422) (0.254) 1.SRM x Foreignness -1.743* -0.677 -1.358** (1.035) (1.171) (0.629) Constant -0.628*** -0.233 -0.473** (0.232) (0.395) (0.203) Observations R-squared 1,330 0.0631 2,692 0.1243 4,022 0.0968 Number of panels 28 57 85
Note: Robust standard errors in parentheses (clustered by SRM x bank) *** p<0.01, ** p<0.05, * p<0.1. Model 1.3 is tested in a random effects regression with time fixed effects, in the period of a year before until a year after the introduction of the SRM. Foreign assets, foreign loans and a combination of foreign assets and foreign loans is used as measure of the degree of foreignness.
7. DISCUSSION
significance. A difference-in-difference-in-difference regression with adding banks outside the scope of the SRM neither shows significant results. Plotting mean CDS spreads shows that banks under the SRM have not been converging to banks not under the SRM.
8. LIMITATIONS
This paper has some limitations. First of all, the difference-in-difference-in-difference regression does not show significant results. It could be expected that such a regression adds to explaining variation in default risk because it separates two groups; one that is under supervision of the European Banking Authority and one that is not. Comparing these two groups could reveal variation that could be assigned the introduction of the SRM. Based on the assumption that all banks in the sample experience the same forces that influence credit default swap spreads, this diff-in-diff-in-diff could show the difference in variation between the groups which could be assigned to the introduction of the SRM. It would be expected that a strong relation between the introduction of the SRM and a decrease in default risk would exist when significant evidence is found in both difference-in-difference and difference-in-difference-in-difference. This is not the case.
Second, the addition of control variables in the difference-in-difference regression results in a loss of significance of the coefficient of interest. When adding the control variables, the coefficient shows the same direction, approximately the same value but no significance. Chiarmonte & Casu (2010) find that the used control variables all explain some variation in CDS spreads. It could be expected that the addition of these control variables facilitate addressing the remaining variation in CDS spreads. This variation not explained by control variables could then be assigned to the introduction of the SRM. The results of the regression with control variables however do not show any significant evidence that this remaining variation could be assigned to the introduction of the SRM. It thus could be the case that other unknown variables cause variation in CDS spreads and not the introduction of the SRM.
and scope advantages, which is confirmed by positive correlation between bank size and foreignness. It might be the case that the smaller banks that have a degree of foreignness that is nil or zero are not included in the sample. This means that comparison over the whole spectrum of banks in terms of size with a high and low degree of foreignness is not made in this paper.
The fourth limitation regards the degree of foreignness of banks. Not for all banks data could be collected on their degree of foreignness. This might strengthen the bias towards large banks. It may be that only the largest banks concern about their degree of foreignness, thus publishing information on foreign assets or loans.
The last limitation of this research concerns the limited possibility to assign the variation in default risk to the introduction of the SRM. Other events that influence default risk of banks could also cause variation in default risk. Examples are the sovereign debt crisis in Europe or effects of the Euro crisis. These causes, together with the introduction of the SRM are hard to uncover separately with a difference-in-difference or difference-in-difference-in-difference model. This is due to the fact that the members of the group of interest are all affected by these events, so even a comparison with a group outside the European Union could not uncover the separate effects.
9. IMPLICATIONS & FURTHER RESEARCH
The results have policy implications for the European Banking Union. This paper shows that there is at least a presumption that the introduction of the Single Resolution Mechanism is related to a decrease in default risk for banks with a higher degree of foreignness. This could imply that the introduction of the SRM has been successful.
Comparing plotted mean CDS spreads of banks under the SRM and banks not under the SRM shows that a significant gap exists. Somehow banks under the SRM experience higher default risk than banks not under the SRM. The EBU should improve regulation and resolution to create convergence in default risk between banks under the SRM and banks in advanced economies not under the SRM. An important thought on increasing the success of a supra-national supervisor is presented by Beck et al. (2013). They argue that the success of a supra-national supervisor depends on coordination with local supervisors. What has to be found out is to what extent the jurisdiction of the EBU covers financial linkages of banks that are under the SRM. If the EBU can cover all activity of the institution that it supervises, this increases the supervisors’ capability to alleviate distortions in its jurisdiction (Beck et al. 2013).
Further research is necessary to find out what the exact effect is of the introduction of the SRM. The sample has to be increased by adding all banks that operate under the SRM. It is however not possible to use CDS spreads as a measure of default risk, because of its limited availability. Default risk could be determined by estimating default risk premia for corporate debt of banks, as Berndt, Douglas, Duffie, Ferguson & Schranz (2004) do. Since not all banks publish information on the degree of foreignness in annual reports, it might be necessary to directly contact the banks and ask them for internal information that reveals their degree of foreignness. Another suggestion for further research is doing an event study. When doing this event study, it is important to find out when the information on the effect of the introduction of the SRM is priced in spreads or premia.
10. CONCLUSION
solved the problem of biased incentives of national authorities in charge of regulation and supervision of the financial sector. Robustness checks confirm the validity of this finding, but also raise doubts whether this decrease in default risk could actually be assigned to the introduction of the SRM and not any other event that might have occurred at the same time. Further research is needed to analyse the effects over the complete banking sector in Europe and increase the possibility to assign a variation in default risk to the introduction of the SRM. The European Banking Union has to further improve regulation to create convergence between banks under the SRM and banks not under the SRM. This could involve coordination with national supervisors or examining cross-border linkages to scrutinize its jurisdiction and increase cross-border coordination outside the EU.
REFERENCES
Acharya, V., Dreschler, I. & Schnabl, P. (2014). A Pyrrhic Victory? Bank Bailouts and Sovereign Credit Risk. The Journal of Finance, 69(6): 2689-2739.
Allen, F. & Gale, D. (2000). Financial Contagion. Journal of Political Economy, 108(1): 1-33. Beck, T., Todorov, R. & Wagner, W. (2013). Supervising cross-border banks: theory, evidence
and policy. Economic Policy, 5-44.
Berndt, A., Douglas, R., Duffie, D., Ferguson, M., & Schranz, D. (2005). Measuring default risk premia from default swap rates and EDFs.
Blanco, R., Brennan, S., & Marsh, I. W. (2005). An empirical analysis of the dynamic relation between investment‐grade bonds and credit default swaps.The Journal of Finance, 60(5), 2255-2281.
Breuss, F. (2013). European Banking Union. WIFO Working papers, no 454.
Brunnermeier, M., Crocket, A., Goodhart, C., Hellwig, M., Persaud, A. & Shin, H. (2009). The Fundamental Principles of Financial Regulation. Geneva Report on the World Economy, 11.
Chiaramonte, L. & Casu, B. (2010). Are CDS spreads a good proxy of bank risk? Evidence from the financial crisis. Working paper 05/10, Centre for Banking research, City University London
Cetorelli, N. & Goldberg, L. (2012). Banking Globalization and Monetary Transmission. The Journal of Finance, 67(5): 1811-1843.
Claessens, S., Herring, R. & Schoenmaker, D. (2010). A Safer World Financial System: Improving the Resolution of Systemic Institutions. Geneva Reports on the World Economy, 12.
De Haan, J., Oosterloo, S. & Schoenmaker, D. (2012). Financial Markets and Institutions, a European Perspective. Second Edition. Cambridge, United Kingdom: Cambridge University Press.
ECB. (2011) – Financial integration in Europe (May) ECB, Frankfurt am Main. European Commision. (27/02/2015). Understanding… Banking Union. Banking and
Finance newsletter. [updated: 27/02/2015 accessed: 10/10/2015] Available at:
<http://ec.europa.eu/information_society/newsroom/cf/fisma/item-detail.cfm?item_id=20758&newsletter_id=166&lang=en>
European Commision. (15/04/2014). Banking Union: restoring financial stability in the Eurozone. Press Release Database. [updated: 24/11/2015, accessed: 10/12/2015] Avaliable at:
<http://europa.eu/rapid/press-release_MEMO-15-6164_en.htm?locale=en>
European Commission. (2012). COMMUNICATION FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL: A Roadmap towards a Banking Union. Available at:
<http://eur-ex.europa.eu/legalcontent/EN/TXT/?uri=CELEX:52012DC0510>
Eichengreen, B., Mody, A., Nedeljkovic, M. & Sarno, L. (2012). How the Subprime Crisis went global: Evidence from bank credit default swap spreads. Journal of International Money and Finance, 31: 1299-1318.
Ericsson, J., Jacobs, K. & Oviedo, R. (2009). The Determinants of Credit Default Swap Premia. Journal of Financial and Quantitative Analysis 44(1): 109-132.
Financial Stability Board. (2014). 2014 update of list of global systemically important banks (G-SIBs). Available at:
<http://www.fsb.org/2014/11/2014-update-of-list-of-global-systemically-important-banks/>
Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work*. The journal of Finance, 25(2), 383-417.
Goodhart, C. & Schoenmaker, D. (2009). Fiscal Burden Sharing in Cross-Border Banking Crises. International Journal of Central Banking, 5: 1-19.
Houweling, P. & Vorst, T. (2005). Pricing default swaps: Empirical evidence. Journal of International Money and Finance, 24: 1200-1225.
Lane, P. (2013). Financial Globalisation and the Crisis. Open Economic Review, 24: 555-580. Levine, R. (1997). Financial Development and Economic Growth: Views and Agenda.
Journal of Economic Literature, 35(2): 688-726.
Longstaf, F. A., Mithal, S. & Neis, E. (2005). Corporate Yield Spreads: Default Risk or
