The Single Resolution Board Promoting Market
Discipline in the European Banking Sector
Author
Kirsten van de Grift
10371915
Supervisor
Prof. Dr. A.C.J.F. Houben
University of Amsterdam
Msc Economics
Monetary Policy and Banking
July 2018
Keywords
Abstract
In light of the debate on ‘too big to fail’ in the European banking sector, this thesis explores the effect of the ‘too big to fail’ concept on market discipline in the bond market. Analysis of bonds of nineteen banks shows weakened market discipline when a bank is considered systemically important or ‘too big to fail’. The Single Resolution Board has been established to strengthen market discipline, but results show no significant success in strengthening market discipline. First, the legal framework and the credibility of the resolution process have to be improved.
Statement of originality
This document is written by Kirsten van de Grift who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents
1. Introduction 4
2. Literature Review 5
2.1 Market Discipline in the Banking Sector 5
2.2 Too Big to Fail 7
2.3 The Single Resolution Board 10
2.4 Resolution Cases 12
2.5 Quantification of Implicit Subsidies 13
3. Methodology 16
3.1 Model 1: Too Big to Fail 16
3.2 Model 2: Too Big to Fail and the Bank-‐Sovereign Nexus 17
3.3 Sample 18
3.4 Preparation of the Dataset 18
4. Results 19
4.1 Descriptive Statistics 19
4.2 Too Big to Fail and Market Discipline 19
4.3 The Single Resolution Board and Market Discipline 21
4.4 The Bank-‐Sovereign Nexus and Market Discipline 22
5. Conclusion 25 References 27 Appendix 29
1. Introduction
“The public will not, and should not, accept more bailouts. Addressing the problem of ‘too big to fail’ is therefore the next central step in the reform programme,” declared Mario Draghi in September 2010, at that time chairman of the Financial Stability Board. Two years after the onset of the Global Financial Crisis, many large European banks failed and were saved by governments. Financial markets anticipated on the
government guarantees and came to believe that major financial institutions, labelled as systemically important, would be bailed out in case of default. The ‘too big to fail’ belief subsequently reduced the urge of markets to monitor and discipline banks by charging risk in their credit prices. In other words, banks received an implicit subsidy and market discipline in the banking sector weakened. A key question is whether large banks still receive a funding advantage, not because they are safer, but because they are bigger. In this thesis, the ‘too big to fail’ problem is examined by testing the following hypothesis:
Market discipline in the European banking sector is lower for systemically important banks.
Market discipline is a key objective to the Basel Committee on Banking Supervision; Pillar 3 of the Basel requirements explicitly aims at strengthening market discipline. In response to the many bailouts of banks in the Global Financial Crisis, the European Commission established the Single Resolution Board (SRB). The aim of the SRB is to reduce the number of bank bailouts in the European Union and thereby strengthen market discipline in the European banking sector. To see if the establishment of the SRB helped to reach the objective of more market discipline, the following hypothesis is also tested:
Market discipline in the European banking sector has increased since the establishment of the Single Resolution Board.
In this thesis, the hypotheses are empirically tested supported by an overview of the related theoretical and empirical literature. Two models are constructed to empirically test for the effect of systemic importance of banks on their bond spreads. The first model includes bond, bank and macroeconomic control variables, the second also accounts for the bank-‐sovereign nexus. For both models, the effect of systemic
importance on bond spreads before and after July 2014 is compared to see if the establishment of the SRB improved market discipline.
Results suggest market discipline in the European banking sector is indeed
lower for systemically important banks as they enjoy lower bond spreads compared to other banks. The establishment of the Single Resolution Board did not convincingly improve market discipline. Only for very systemically important banks, market
discipline strengthened. For less systemically important banks, no significant difference is found. To further strengthen market discipline, the legal framework and credibility of the resolution process have to be improved.
In this thesis, the existing literature on market discipline will first be discussed.
This includes literature on market discipline in the banking sector, the ‘too big to fail’ concept, the Single Resolution Board, resolution cases and the quantification of implicit subsidies. Second, in the methodology the two models used to test the hypotheses, the sample features and the preparation of the dataset will be described. The third
paragraph will cover descriptive statistics, the results and the analysis. The last paragraph will present the conclusion including a summary of the results, policy implications and suggestions for further research.
2. Literature Review
2.1 Market Discipline in the Banking Sector
Market discipline in the banking sector is described as a situation in which agents face costs that increase with banks’ exposure to risks (Berger, 1991). Bank risk can arise in several ways, alongside each other or reinforcing each other, being (1) credit risk, when a bank’s creditors do not meet their obligations; (2) operational risk, the risk of losses resulting from failed processes, people and systems, either internal or from external events; (3) market risk, the risk of losses due to movement in market prices; (4) liquidity risk, the risk of a bank’s inability to meet its cash flow obligations; (5)
reputational risk, the risk of a loss in a bank’s reputation, which may be triggered by a bank’s activities or by rumours and (6) systemic risk, which will be explained in paragraph 2.2. Also, management may take on more risk because the bank lacks good internal risk controls or because they know someone else will have to face the burden, commonly known as moral hazard. When costs increase due to an increase in risk, agents take action. Agents can be equity holders, depositors and debt holders. For example, an uninsured depositor may withdraw her deposits in response to excessive
risk taking. This can lead to a loss in customers and business, forcing the bank to change its strategy.
According to the Basel Committee on Banking Supervision “market discipline
imposes strong incentives on banks to conduct business in a safe, sound and efficient manner”. The Committee underlines several benefits to effective market discipline in the banking sector (Pillar 3 (Market Discipline), 2001). First, it can reduce moral hazard as a bank is punished when taking on too much risk, which creates an incentive to limit risk taking. Second, it can make the banking sector more efficient when agents force an inefficient bank to become more efficient or to leave the market (Soledad, Peria & Schmukler, 2001).
For market discipline to be effective, agents should be able to monitor and
influence the market (Bliss & Flannery, 2002). To be able to monitor, market prices should be a reflection of the bank’s exposures to risks (Gorton & Santomero, 1990). To be able to influence the market, agents should have either direct or indirect influence (Flannery, 2001). Agents can directly discipline banks when causing changes in market prices that increase funding costs to banks. Agents can indirectly discipline banks when influenced market prices have a signalling function to regulators, investors and policy makers regarding the condition of a bank.
Different agents can discipline banks. The most commonly mentioned are depositors. With the introduction of deposit guarantee schemes, the incentive to discipline is limited for depositors. In the European Union, deposits up to €100.000 are guaranteed. However, small depositors exerting market discipline is not necessarily desirable. Often, depositors do not consider all information available when choosing a bank, for instance because they care more about services offered by a bank than its solvency (Garten, 1995). Market discipline by depositors also implies high, inefficient, search costs. Moreover, depositor discipline can be destabilizing as depositors can create bank runs with large negative externalities, even when a bank is solvent. Therefore, depositors are not the preferred discipliners of banks.
The second group of discipliners, equity holders, are important discipliners as they have a stake in the bank and have the residual claim in case of default. They have the power to influence management, for example in the likelihood of managerial turnover (De Ceuster & Masschelein, 2003). On the other hand, they are also
incentivized to take on more risk, as they benefit the most from upside gains (Gropp, Vesala & Vulpes, 2006). For this reason, equity holders are not the preferred group of discipliners either.
Debt holders represent the third group of discipliners. Debt holders are focused on the downside risks of banks and are therefore much more aligned with the interests of bank supervisors and appropriate discipliners (Morgan & Stiroh, 2001) . Debt holders require different bond spreads when risk changes, thereby influencing borrowing costs and disciplining the banks.
Morgan and Stiroh (1999) argued that the bond market prices bank risk efficiently. They found evidence that suggests that investors even look beyond public measures of risk, such as the underlying portfolio of loans, in pricing a bank’s risk exposure. The research conclusions of Flannery (2000) and Evanoff and Wall (2001) on the effectiveness of the bond market are aligned. According to them, the bond market indicators are usually more accurate than supervisory assessments. However, Bliss (2001) showed that bond yield spreads are noisy measures of risk, even though the noise is not random. Moreover, Gropp et al. (2006) stressed that the debt market is hard to measure consistently due to varying maturities and tax treatments. When controlling for these measurement difficulties, the bond market is a good market to measure market discipline and is therefore the focus of this thesis.
2.2 Too Big to Fail
In the last two decades, the process of technological innovation and consolidation in the banking industry lead to larger banks. In Figure 2.1, asset growth of the top global systemically important banks as of November 2013 is shown (IMF, 2014). The banks all grew exponentially after 1997, even after the Global Financial Crisis. Only Citigroup did not grow larger after 2007. They suffered substantial losses in the crisis and received state support in 2008. Large banks can be ‘too big to fail’ because the damaging effects of failure are larger than that of other corporates or smaller banks; it brings along systemic risk. Systemic risk is the risk of financial instability caused by the collapse of a major part, or all of the financial system with potentially negative consequences for the real economy.
Systemic risk can arise in several ways. First, banks are subject to contagion risks. Contagion risk is the risk of many banks failing in response to one large bank failing, even if the other banks were solvent before. Banks are subject to this risk because they operate based on public trust (European Commission, 2014). If trust in a large bank is lost, this can result in the loss of confidence in other banks too. Moreover, banks are often highly interconnected. Cross-‐border banking claims have increased substantially, reaching more than 50% of global GDP in 2007 (Figure 2.2). This phenomenon was driven by worldwide deregulation and innovation of the banking
sector. Linkages have decreased to around 30% of global GDP in the process of bank deleveraging and restructuring since the crisis. Due to these linkages, a failing bank can financially harm other banks if they are exposed to each other. Second, systemic risk can arise because banks perform a critical function in the economy. They allocate savings to investments for both households and firms and they manage payment systems to various sectors in the economy and society. When a large bank fails, these functions are lost to the customers.
Figure 2.1 Total Assets of Large Banks Figure 2.2 Cross-Border Banking Linkages
(in billions of US Dollars) (as a Percentage of Global GDP)
Source: IMF Global Financial Stability Report (2014) Source: IMF Global Financial Stability Report (2014)
There are multiple indicators of systemic importance. The most significant indicator of systemic importance is size (Adrian & Brunnermeier, 2011). When a bank has a large amount of assets on its balance sheet, the bank is most likely systemically important. Deutsche Bank AG, a very systemically important bank, had 2.164,1 billion in assets at the beginning of 2012, relative to Banca Carige Italia SpA with 44,86 billion in assets at the beginning of 2012. Another measure of systemic importance is based on a list of systemically important banks developed by the Financial Stability Board (2017). The FSB determined systemic importance based on five indicators, all given an equal weight of 20%. The indicators are size, global (cross-‐jurisdictional) activity,
interconnectedness, substitutability and complexity. Systemically important banks were allocated to different buckets corresponding to the required level of additional capital buffers, ranging from no additional capital buffer (Caixabank SA) to 3,5% additional capital buffer (bucket 5). Deutsche Bank AG is the European bank with the highest
required additional capital buffer, namely 2% (bucket 3). The Financial Stability Board assigned only very large and systemically important banks to one of the buckets and thereby set a clear threshold. The European Central Bank (ECB) for instance set a much lower threshold. The ECB conducted a list with significant entities that are under the direct supervision of the ECB. This list is based on a bank’s size, economic importance,
cross-‐border activities and direct public financial assistance (Criteria for determining
significance section, para. 1). Whereas the list by the Financial Stability Board only consists of seven EU banks (excluding British banks), the list by the European Central Bank consists of 119 significant entities. To illustrate the difference, the European Central Bank did label Caixabank SA as a significant entity. The focus of this thesis is on ‘too big to fail’ in terms of size and the buckets of the FSB. The banks on the list of the ECB are not all ‘too big to fail’, as some of these banks were (supposed to be) wound up in the past under the supervision of national authorities. For this reason, the focus of this thesis is not on the list of the ECB.
The origins of the ‘too big to fail’ problem date back to 1972, when the Bank of the Commonwealth was bailed out for 1,2 billion dollars by the Federal Reserve (Nurisso & Prescott, 2017). The bailout finally lead to the belief that large banks were ‘too big to fail’ and governments would bail out these banks in case of insolvency. Prior to the Global Financial Crisis, policy makers tried to address this problem by acting ambiguous about the willingness to save failing banks (IMF, 2014). However, little uncertainty was left about the willingness to save failing banks when many large banks like ABN-‐Amro and Commerzbank were bailed out during the Global Financial Crisis. Even though the bailouts may have been necessary, they were costly for several reasons too.
First, the level of state support given to save the banks was extremely high in some countries and paid with taxpayers’ money at the expense of other possible public expenditures (European Commission, 2014). Second, the bailouts partially instigated the bank-‐sovereign nexus in some European countries. Large guarantees given to failing banks contributed to the build up of sovereign debt problems, in Europe eventually leading to the Sovereign Debt Crisis. High indebtedness of sovereigns creates
uncertainty if governments will be able to save banks in case of failure. The uncertainty jeopardizes bank stability and magnifies the anticipated need for and the cost of public intervention in the banking sector. This in turn further deteriorates the sovereign’s debt position, setting in motion a feedback loop between sovereign and bank stability (the bank-‐sovereign nexus). The Irish crisis of 2008 is a strong example of the mechanism of a bank-‐sovereign nexus (European Central Bank, 2017).
A more indirect cost of a bailout is related to market discipline. When banks are considered ‘too big to fail’ and a bailout is very likely, market discipline weakens
(Cubillas, Fernández & González, 2017; Gropp et al., 2006; Morgan & Stiroh, 1999). First, the expectations of state support are incorporated in the credit spreads of these banks (Acharya, Anginer & Warburton, 2013). Second, the costs of monitoring and taking action against a bank rise relative to the benefits, as creditors don’t lose (all) their money in case the bank fails. Subsequently, systemically important banks pay a lower price for risk than other banks or financial institutions and are not monitored as much. The perception of ‘too big to fail’ provides the banks with an implicit subsidy and encourages risk-‐taking, not because they are safer, but because they are bigger.
2.3 The Single Resolution Board
In response to the many bailouts during the Global Financial Crisis and the negative externalities that followed, policymakers brought about financial reforms. Amongst the reforms are higher capital requirements and strengthened supervision of banks to limit the probability and cost of failure and contagion, but also a stronger resolution
framework (IMF, 2014). In the European Union, the Bank Recovery and Resolution
Directive (BRRD) and the Single Resolution Mechanism Regulation (SRMR) were adopted in 2014. The BRRD provides authorities with arrangements to deal with failing banks at national level and to tackle cross-‐border banking failures. Also it requires banks to prepare recovery plans to deal with financial distress (Bank recovery and resolution section, para. 2). The SRMR constitutes the resolution framework of banks in EU countries participating in the banking union (Single Resolution Mechanism section, para. 1). The Single Resolution Board (SRB) has responsibility for the enforcement of the SRMR and has been operational as an independent European Union Agency since
January 2015.
Bank resolution is a controlled approach to bank failure and promotes market functioning. Instead of managing resolution by national authorities, the higher
integration and interconnectedness of banks in the European Union called for a common framework in the EU (European Commission, 2014). In this framework, the SRB gets full responsibility for the resolution of a bank when the European Central Bank decides a bank is “failing or likely to fail”. The aim of the SRB is to avoid future bailouts and to place the costs of a bank resolution on the shareholders and debt holders and thereby limit the costs to taxpayers and the real economy (Single Resolution Mechanism section, para. 2). When resolution is effective, both shareholders and debt holders will suffer losses in case of default (European Commission, 2014). Losses strengthen the incentive
to discipline banks and reduce moral hazard. Subsequently, risk should be priced correctly in the markets.
If a bank is placed under resolution, the SRB has four resolution tools. The tools are intended to secure public interests, including the continuity of a bank’s critical functions and financial stability, all at minimal cost to taxpayers. Before any resolution tool is set into motion, a bank’s capital has to be written down or converted. The first resolution tool is the sale of business. This entails the total or partial sale of the bank to another entity. Assets stay in the banking system so this causes little disruption. The second resolution tool is the bridge institution. Part of a bank is then transferred to a temporary (completely or partially) publicly owned entity. ‘Bad assets’ are separately liquidated. Third, the SRB can use the asset separation tool. Assets, rights and liabilities are transferred to an asset management vehicle, also completely or partially publicly owned. The last and most important tool is the bail-‐in tool, which is provided for in the Bank Recovery and Resolution Directive. Equity and debt are then written down or converted, placing the costs on shareholders and debt holders. The BRRD clearly states which liabilities are eligible for bail-‐in and in what order. However, the BRRD also states that ‘exceptional circumstances’ exist for instruments. The resolution authority then has the discretion to decide if an instrument is eligible for bail-‐in. Which instruments are exceptional is not defined however, creating uncertainty and room for different bail-‐in procedures among member states. Huertas (2016) discussed the different
implementations of the BRRD among member states and suggested the convergence towards one creditor hierarchy for banks across the EU or at least EU-‐wide standards for what circumstances an instrument can be ‘exceptional’. Phillipon and Salord (2017) recognized the same problem and proposed reduced discretion in the application of the bail-‐in exceptions. Another problem that Huertas mentioned is the legal enforceability of the bail-‐in tool. Systemically important banks often operate globally. When instruments are issued in third countries (not under the BRRD), investors may seek protection from a court in the third country to prevent the bank from bailing in the investor’s liability. The BRRD tackles the problem by requiring banks to give all new liabilities eligible for bail-‐in the right contractual terms that indicate if the instrument is subject to write-‐ down or conversion. In the future, this requirement will give the resolution authorities the rights to bail in every eligible liability.
In the EU resolution framework, state is still possible if it qualifies as a precautionary measure. Although the Single Resolution Board is not involved in precautionary recapitalizations, Philippon and Salord (2017) argued that this can undermine the credibility of the SRM. They suggested reduced discretion as the
condition of granting is too broadly defined. The current condition of granting public funds is “in order to remedy a serious disturbance in the economy of a Member State and preserve financial stability”. They proposed stricter conditions, including
quantitative and qualitative requirements.
Several event studies were done to see if the implementation of the resolution framework alone had effect on market discipline. The IMF found that the European Commission’s proposal for (not the adoption of) the Single Resolution Mechanism did not have any significant effect on the markets, whereas the proposal for the Bank
Recovery and Resolution Directive did.
Schäfer, Schnabel and Weder di Mauro (2016)
did find significant differences in stock prices after the formal agreement of the Single Resolution Mechanism. They analyzed stock returns and CDS spreads of EU banks after bail-‐in events and the implementation of the Single Resolution Mechanism. They found widespread negative effects on stock returns after the agreement of the SRM, with a stronger effect on systemically important banks. However, the market response was much smaller than for the response to actual bank failures. They concluded that the success of the Single Resolution Board crucially depends on the handling of next resolution cases. From these findings, no conclusions can be drawn upon the expected effect of the establishment of the Single Resolution Board on market discipline.
2.4 Resolution Cases
In the United States, the ‘too big to fail’ problem was and is present as well. Acharya et al. (2013) found that bondholders of major financial institutions in the United States did not accurately price risk because they expected government support in case of default. The United States government acted on the expectation of public support in the market and passed the Dodd-‐Frank Wall Street Reform and Consumer Protection Act (‘The Dodd-‐Frank Act’) in 2010, of which avoiding future bailouts is one of the main objectives. Acharya et al. (2013) did an event study on the Dodd-‐Frank Act to see if it strengthened market discipline. They found that the passing of the Act did not eliminate expectations of government support and did not strengthen market discipline. However, Acharya et al. (2013) also found that following the rescue of Bear Stearns in March 2008, before the Dodd-‐Frank Act was introduced, larger financial institutions experienced a greater decrease in their credit spreads than smaller institutions. By contrast, after the collapse of Lehman Brothers in September 2008, larger financial institutions experienced a greater increase in their credit spreads than smaller
institutions. These market responses show that the expectation of government support in the United States influenced market discipline for large institutions. This could
indicate that the passing of the Dodd-‐Frank Act did not in itself strengthen market discipline because credible commitment is critical, which was possibly missing. In figure 2.3, level of implicit subsidies to systemically important banks in the US is shown. The findings of Acharya et al. (2013) do not rule out the second hypothesis of this thesis, providing that the Single Resolution Board credibly commits to its objective.
So far, the Single Resolution Board has been involved in three resolution cases.
The European Central Bank decided on June 6th 2017 that Banco Popular Español S.A.
was failing or likely to fail (Resolution Cases section, para. 2). The next day, the SRB first wrote down the shares of Banco Popular Español S.A. and converted junior debt to shares. Thereafter, the SRB transferred all shares to Banco Santander S.A.. Because of this, Banco Popular Español S.A. could continue to operate under normal business conditions. The second resolution case concerned Banca Popolare di Vicenza SpA and Veneto Banca SpA in June 2017. The SRB decided that resolution could take place under Italian authority (Resolution Cases section, para. 3). Under their authority the banks were supposed to be wound up. However, Italian authorities decided that the winding up of the banks would have a serious impact on the real economy and asked the European Commission for approval to use national funds to facilitate the liquidation. The Commission approved, partially because shareholders and subordinated debt holders contributed to the costs. The approval resulted in cash injections of about €4.785 billion and state guarantees of about €12 billion (European Commission, 2017). Even though the SRB did not decide on state aid, the fact that the banks received state aid does not contribute to the resolution objectives of the SRB. Third, the European Central Bank decided on February 23rd 2018 that ABLV Bank, AS and its subsidiary
ABLV Bank Luxembourg S.A were failing or likely to fail (Resolution Cases section, para. 5). The SRB decided that ABLV Bank, AS and its subsidiary should be wound up under national authority. The bank is currently in the process of liquidation. Moreover, in July 2017, Banca Monte dei Paschi di Siena was precautionary recapitalized by the Italian government. Philippon and Salord (2017) argued that this can undermine the credibility of the SRB too, even though the SRB is not directly involved.
2.5 Quantifications of Implicit Subsidies
Different models have been tested and used to quantify the effects of ‘too big to fail’. The International Monetary Fund (IMF) approached three of them (IMF, 2014). The IMF used the same sample for all three approaches, consisting of the list of G-‐SIBs identified by the Financial Stability Board and the three largest banks (by asset size) in each country (if not already in the list of G-‐SIBs). The first approach is measurement of the
bond spread differential between systemically important banks and other banks. The IMF used this method and found the results shown in figure 2.3. Since the Global Financial Crisis, the implicit subsidies have fallen significantly for the United States and to a lesser extent in Japan as well. For Europe, the implicit subsidy has been positive and rising since the Global Financial Crisis until around 2013, reaching a peak of over 500 basis points. From then on, it has been falling again. According to the IMF, the high subsidies in the Euro Area largely reflect the Sovereign Debt Crisis. The shortcoming of the bond spread differential approach is that it does not consider risk characteristics of a bank. When controlling for these characteristics of banks in the United States, results prove that systemically important banks did enjoy funding advantages after the crisis (IMF, 2014).
Figure 2.3 Bond Spread Differential between Systemically Important Banks and Other Banks (in basis points)
Source: IMF Global Financial Stability Report (2014)
The second approach is the contingent claims analysis approach, for which Credit Default Swap Prices are used to measure implicit subsidies. The advantage of this approach is that it controls for bank characteristics. However, there is only limited CDS data available and it may not be reliable in case of a market turmoil. Figure 2.4 shows a positive trend for the Euro Area too, but the implicit subsidy estimations are
substantially lower. Since 2012, implicit subsidies in the Euro Area have averaged around 90 basis points, whereas the results of the first approach show a substantially higher average subsidy. The results of the second approach even show negative implicit subsidies between 2011 and 2012.
Figure 2.4 Mean Implicit Subsidy for Systemically Important Banks in the Euro Area Estimated with the Contingent Claims Analysis Approach (in basis points)
Source: IMF Global Financial Stability Report (2014)
The ratings-‐based approach is the third approach. In this approach, the IMF made use of the fact that credit rating agencies generally take a government’s implicit guarantee into account and translated the rating uplift to bond spreads. Like the second approach, this approach includes bank characteristics. A shortcoming is the dependency on credit rating agencies and views among credit rating agencies are not always similar. Also, credit ratings adjust very slowly. Figure 2.5 shows the results for the Euro Area. The results again show positive but declining implicit subsidies for systemically important banks since the peak of the Global Financial Crisis. This applies to both solvent systemically important banks and systemically important banks that are already in distress (banks with positive support). Implicit subsidies never turn negative.
Figure 2.5 Mean Implicit Subsidies in the Euro Area Derived from Credit Ratings (in basis points)
Source: IMF Global Financial Stability Report (2014)
Even though the results of the IMF show different values for the different approaches, the results of the three approaches suggest positive average subsidies from
2005 to 2014 for systemically important banks in the Euro Area. The results therefore support the first hypothesis of this thesis.
3. Methodology
3.1 Model 1: Too Big to Fail
In this thesis, market discipline in the banking sector is measured by examining how systemic importance (‘too big to fail’) affects the spread of a bank’s bonds. Ordinary Least Squares regressions are done using two different models, following the first quantification approach discussed in paragraph 2.4. The first model is based on the empirical model by Acharya et al. (2013):
𝑆𝑝𝑟𝑒𝑎𝑑!,!,! = 𝛼 + 𝛽!𝑇𝐵𝑇𝐹!,!+ 𝛽!𝐵𝑜𝑛𝑑 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!,!,!+ 𝛽!𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!,!
+ 𝛽!𝑀𝑎𝑐𝑟𝑜 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!,!+ 𝜖!,!,!
The subscripts i, b and t denote banks, bonds and time (in months) respectively. Spread is a specific bond spread, which is the difference between the yield on the bond and the yield on the corresponding maturity-‐matched and country-‐specific Treasury bond. When there is no Treasury bond exactly matching in maturity, the closest match is used.
TBTF stands for ‘too big to fail’ and represents the systemic importance of a bank. As
explained in the literature review, there are several ways to estimate systemic
importance. In this thesis, assets (measured in logarithms) and the buckets developed by the Financial Stability Board are used to measure the effect of systemic importance on bond spreads. The TBTF measure in terms of assets shows a continuous range in systemic importance and does not include a certain threshold separating systemic and non-‐systemic banks. The Financial Stability Board assigned only very large and
systemically important banks to one of the buckets and thereby set a clear threshold. The list by the Financial Stability Board only consists of seven banks in the European Union (excluding British banks). Both measures of ‘too big to fail’ are tested to ensure the final conclusions are robust.
Bond Controls are bond specific control variables, including the time to maturity
in years and the seniority of the issue. The time to maturity and the seniority of a bond determine the risk of a bond and thereby the bond spread. Therefore, these bond controls are taken into account. Contrary to the approach of the International Monetary Fund, risk characteristics of a bank are included in this model. The Bank Controls are the following bank specific control variables: debt-‐to-‐equity ratio, return on assets, price-‐to-‐ book ratio of equity and the maturity mismatch. The bank characteristics influence the
likelihood of distress and therefore the risk to bondholders. Macro Controls include the market risk premium, which is the difference between the expected return on a risky market portfolio and the risk-‐free rate. The other two macro controls are the yield spread between long-‐term (10 year) treasury bonds and the short-‐term (three-‐month) treasuries as a measure for the term structure, and the BAA-‐AAA corporate bond spread as a proxy for default risk. Finally, 𝛼 represents a constant and 𝜖!,!,! the error term. More details on the variables are given in Appendix A.
To measure if the establishment of the Single Resolution Board improved market discipline, a second test is done. In this test, regression results of the first model are compared, one with data from before July 2014 and the other with data from July 2014 onwards. The adoption of the Single Resolution Mechanism Regulation is used as the establishment date instead of the actual establishment of the Single Resolution Board in 2015. The adoption of the SRMR meant the Single Resolution Board was going to come with certainty and it gave clarification on the tools of the Single Resolution Board. It is expected that agents anticipated on the effect of an operational Single Resolution Board once the Single Resolution Mechanism Regulation was adopted. In the regressions, the coefficients of the two measures of TBTF are compared to see if the effect of systemic importance on market discipline has decreased since the
establishment of the SRB.
3.2 Model 2: Too Big to Fail and the Bank-‐Sovereign nexus
In case of a highly indebted sovereign, a bank guarantee is less credible and effective. Bondholders may see their position as more risky and subsequently might ask for a higher return on their bonds. The European bank-‐sovereign nexus may therefore have strengthened market discipline regardless of the Single Resolution Mechanism
Regulation. Including sovereign debt positions into a second model can possibly improve the first model, especially considering the countries in the dataset, ranging from Italy with a weak fiscal position to the Netherlands with a strong fiscal position. The regressions of the second model test for these effects and at the same time test for the validity of the conclusions that are drawn from the regressions based on the first model.
The second model including sovereign debt positions is constructed as follows:
𝑆𝑝𝑟𝑒𝑎𝑑!,!,! = 𝛼 + 𝛽!𝑇𝐵𝑇𝐹!,!+ 𝛽!𝐵𝑜𝑛𝑑 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!,!,!+ 𝛽!𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!,!
+ 𝛽!𝑆𝑜𝑣𝑒𝑟𝑒𝑖𝑔𝑛 𝑑𝑒𝑏𝑡 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛!,!+ 𝑀𝑎𝑐𝑟𝑜 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!,!+ 𝜖!,!,!
The first regressions include only government debt as a percentage of GDP. Debt-‐to-‐GDP reflects a government’s fiscal position and thereby the likelihood a government is able to provide subsidies to banks. In a second regression series, the primary surplus as a percentage of GDP is included. In 2011, every government had a budget deficit, and for each country the budget deficit decreased over the years, some budget deficits even turning into budget surpluses. Because of the gradual improvement of government balances and the differentiation between countries, the course of the crises in Europe is reflected in the variable. This is important as the dataset includes data from 2011 onwards, which is in the middle of the crisis. The third regression series include both the Debt-‐to-‐GDP ratio and the Primary Surplus.
3.3 Sample
The sample of banks used in this thesis consists of banks under the authority of the European Central Bank. In order to gauge the effect of systemic importance on market discipline, the sample includes both systemically important banks and other banks. A summary of the banks in the sample is shown in Appendix B. Data of the banks is collected from Thomson One, Datastream and Bankfocus. The sample includes data of banks from 2011 onwards, as Bankfocus is a considerably new database only covering data since 2011.
3.4 Preparation of the Dataset
Before the regressions are done, several modifications are made to the data. First, to ensure the bond spreads are comparable, bonds with derivative features, floating interest rates and warrants are excluded from the sample. Also, bonds with a time to maturity of less than one year are excluded. Second, Bankfocus only provides annual accounting data. Therefore, the end-‐of-‐year accounting data is linearly interpolated to monthly data. For the price-‐to-‐book ratio, monthly market prices are used. Third, the data on the spreads, bank controls and macro controls contain a number of extreme observations. To ensure that outliers do not influence the results, the data is winsorized to the outer one percentile. In other words, all observations for the spread, bank
controls and macro controls with values lower than the 1st percentile and higher than
the 99th percentile of all values are set to the values of the 1st and 99th percentile. Fourth,
the dataset includes bonds with a perpetual character. The time to maturity for those bonds is set to 40 years, the highest value of Ttm in the dataset. Fifth, standard errors are clustered on bank level. Last, a simple correlation test is done to test for linearity among pairs of variables and to test for multi-‐collinearity, the variance inflation factors
are estimated. The critical value for the variance inflation factor is set at 10 (Hair, Black, Babin & Anderson, 2010).
4. Results
4.1 Descriptive Statistics
An overview of the variables is found in table 4.1. The dataset consists of 14.246 unique observations for nineteen different banks from seven different countries, further specified in Appendix A.
Table 4.1: Descriptive Statistics
Mean STD N Min Max
Spread 2,608054 1,85068 14246 0,001 8,554 TBTF in assets 20,13058 1,584658 14246 15,14161 21,46195 TBTF in buckets 1,027657 0,9912606 14246 0 3 Time to maturity 10,87571 10,48534 14246 0 40 Seniority 3,871528 1,346201 14221 1 7 Debt-to-equity ratio 17,98212 7,493638 14246 4,136131 43,49978 Return on assets 0,2747156 0,3904113 14246 -4,155679 8,011952 Price-to-book ratio 0,6728791 0,275852 13511 0,168 1,565 Maturity mismatch 1,188695 2,772 14246 0,0876894 16,15222
Default risk premium 0,9184859 0,2536132 14246 0,55 1,49
Term structure premium 1,835729 0,4909172 14246 1,01 2,97
Market risk premium 6,990429 0,94975 14246 5,19 12,06
4.2 Too Big to Fail and Market Discipline
The results of the first OLS regression with cluster robust standard errors are shown in table 4.2. The first two regressions still include the debt-‐to-‐equity ratio. However, when testing for correlation with the other variables, the debt-‐to-‐equity ratio shows to have high correlation with the other bank specific control variables. The variable is
insignificant for all two regressions and it has a negative sign for one regression, which is not as expected. Furthermore, the R2 does not improve when including the debt-‐to-‐
equity ratio. Therefore, the debt-‐to-‐equity ratio is not considered in further regressions and the analysis is focused on the last regressions that do not include the debt-‐to-‐ equality ratio. No (multi-‐) collinearity is detected in the last two regressions; the variance inflation factor has a mean of 1.32 and 1.31 for regressions three and four respectively.
For both systemic importance (TBTF) measures, the coefficient is significant and
