Master’s thesis Business Economics: Finance
Lieselotte A. J. Bulsing
10266437
July 2016
The effects of announcements of post-financial crisis bank regulations and a
G-SIB declaration on Credit Default Swap Spreads through event studies
Thesis Supervisor: Dhr. Dr. J.E. Ligterink
Statement of Originality
This document is written by Student Lieselotte Bulsing who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is 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.
Acknowledgements
First of all, I would like to thank my supervisor for his support and feedback during the past months. He gave good and useful feedback and even the last two days before the deadline his
help was very accurate.
I would also like to thank Achmea for providing me with the opportunity of writing my thesis at their company. This was a great and educational experience.
Finally, I am grateful for the support from my parents, my sister and my boyfriend. They encouraged me during the entire process.
Abstract
This study investigates the effect of European bank regulations, introduced after the start of the financial crisis in 2009, on default probabilities for banks. Those regulations aimed to strengthen banks in order to decrease the probabilities that they would default. In order to measure whether the actual probabilities of default decreased, this study measures the effect of those regulations on the CDS Spread, as a proxy for bank risk. Two out of the four regulations displayed a negative effect on CDS Spreads, confirming the hypothesis that probabilities of default decreased. However, for two other regulations, this study finds increases in CDS Spreads. These results imply that the regulations do not necessarily decrease default probabilities for banks. Finally, the effect of a bank losing its Global Systemically Important Bank status is investigated. This bank’s CDS Spread decreased significantly after it lost its G-SIB status. The results of this study suggest that a ‘Too-Big-To-Rescue’ effect on pricing is present in the CDS market.
Table of Contents 1 Introduction 6
2 Theoretical Background 8
2.1 Empirical Effect of Earlier Regulations 8 2.2 CDS Spreads 10
2.3 Too-Big-To-Fail banks 11 2.4 Market Discipline on banks 12 2.5 Conclusion 15
3 Regulations after financial crisis 17 3.1 The BRRD 17
3.2 The LR, liquidity Ratios, the CRR and the CRDIV 17 3.3 The EMIR 18
3.4 The Liikanen Report 19 3.5 Hypotheses 19
4 Methodology 21
4.1 Compare CDS Spreads before and after the financial crisis 21 4.2 Event study for four regulations 25
4.3 Event study for a bank losing its G-SIB status 25 5 Data 27
5.1 Bank selection criteria 27 5.2 Variables 27
5.3 Time span 28
5.4 Summary Statistics and Correlation 29 6 Results 34
6.1 Regression (1): CDS Spreads before and after the crisis 34
6.2 Regression (2): CDS Spreads before and after the crisis using perc. changes 36 6.3 Regression (3): event studies for four regulations 37
6.3.1 the BRRD 38
6.3.2 the LR, liquidity Ratios, the CRR and the CRDIV 39 6.3.3 the EMIR 40
6.3.4 the Liikanen Report 43
6.4 Regression (4): event study for a bank losing its G-SIB status 44 6.5 Conclusion 47
7 Robustness 49 8 Discussion 51
8.1 Conclusion 51 8.2 Limitations 52
8.3 Suggestions for further research 53 9 Bibliography 54
1 Introduction
After the start of the financial crisis in 2009, authorities introduce several regulations in order to strengthen the banking sector. The purpose of those regulations is to reduce the probabilities that banks would default. A Credit Default Swap is a financial instrument that is used by investors as an insurance instrument for payments by banks. If the bank defaults, the investors still get their payments. The riskier the bank, the more expensive the Credit Default Swap (CDS). The CDS Spread is therefore considered a bank default indicator. For this reason, this study investigates whether probabilities of default actually decreased by investigating whether CDS Spreads changed after the announcements of the new bank regulations through several event studies. Moreover, literature on bank regulations often refers to the term Market Discipline. Market Discipline consists of two parts: first, investors monitor a company’s status and they set prices accordingly. Secondly, the company reacts to changing prices and changing investor opinions on the company’s status (Bliss and Flannery, 2002). The term market discipline is often used in combination with the term ‘Too-Big-To-Fail’. ‘Too-Big-To-Fail’ banks are banks that will be bailed out by the government if they fail. Some, like Balasubramnian and Cyree (2014), argue that financial products for banks that are considered ‘Too-Big-To-Fail’ do not entirely reflect the risks that the bank takes. Since investors get their money back anyway in case of default, there is no actual risk in investing in those companies. To this study this could imply that for ‘Too-Big-To-Fail’ banks, the CDS Spread could be a non-accurate reflection of the risks that a bank takes. During the sample period that this study uses, one bank officially changed from a Global Systemically Important Bank to a non-Global Systemically Important Bank (Financial Stability Board, 2012). For this bank, this study investigates whether CDS Spreads changed after it lost its G-SIB status through an event study.
The central question in this study that combines the two subjects mentioned in the previous paragraphs is therefore: ‘How do CDS Spreads respond to announcements of new regulations and to an announcement that a bank is no longer considered ‘Too-Big-To-Fail’?’. The results of this study indicate significant changes in CDS Spreads after the announcements of all regulations. This study expects a decrease in CDS Spreads after each of the announcements of the regulations. However, after the announcements of the EMIR and Basel III, CDS Spreads significantly increase, which implies that probabilities of default for banks increase. Furthermore, after the announcement of the BRRD and the Liikanen Report, CDS Spreads decrease significantly. This could imply that due to those regulations, probabilities that banks could default decrease. Besides, for the only bank in the sample that
lost its G-SIB status, CDS Spreads decrease significantly. The results of this study suggest the presence of a ‘Too-Big-To-Rescue’ effect in the banking sector. This is in line with a study by Balasubramnian and Cyree (2014), because they found that the ‘Too-Big-To-Fail’ effect in bond pricing decreased after the Dodd-Frank in the United States. Besides, Völz & Wedow (2011) also find evidence for a ‘Too-Big-To-Rescue’ effect in the banking sector.
The contribution of this research to existing literature is that this study investigates the effects of bank regulations on CDS Spreads for the first time. Only few studies investigate the effects of bank regulations on specific variables. There is almost no research available on the effect of the specific regulations discussed in this study. Besides, hardly any study investigates the effects of bank regulations on CDS Spreads in specific. Furthermore, Kleinow, Nell, Rogler & Horsch, (2014) studied banks officially becoming ‘Too-Big-To-Fail’. However, that study investigates the effect on stock prices, whereas this study investigates the effect on CDS Spreads. Besides, this study looks at the effect of a bank losing its G-SIB status, instead of gaining the status. The final contribution to existing literature is that this study finds suggestions for a ‘Too-Big-To-Rescue’ effect in the banking sector, just like Völz and Wedow (2011), however this study used official G-SIB declarations instead of variables for size of the company.
The next section discusses the effects of earlier regulations, theoretical background on CDS (spreads) and financial regulations. The section thereafter presents the financial regulations investigated in this study and the hypotheses that were used. Then, this paper elaborates on the method that was used, after which it discusses the data. Next, the results are discussed. In section 7, this study elaborates on the robustness of the regressions. Finally, the last section contains the conclusion and limitations of this study and recommendations for further research.
2 Theoretical Background
This section discusses subjects that are relevant for understanding the framework of this study. To start with, empirical effects of earlier regulations are discussed. It is important to see what the results of other studies are in order to form a solid expectation in this study. Besides, some researches study regulations that are similar to the ones investigated in this study, so knowing the results of those studies improves the understanding of the possible effects of the regulations discussed here.
Then, this section discusses the main determinants of CDS Spreads in order to be able to identify the right regressors in this study. Next, ‘Too-Big-To-Fail’ theories are discussed, since part of this study involves investigating the effect of a bank officially losing its ‘Too-Big-To-Fail’ (or G-SIB) status. Some theories argue that for those banks, there is no reflection of extra risk taking on bond or stock prices, since the government will bail out those banks anyway.
Finally, this section focuses on the concept market discipline. Like mentioned in the introduction, Bliss and Flannery (2002) separate market discipline into two parts: a part in which investors assess the company’s status and they adjust their prices accordingly and a part in which the company responds to the investors’ evaluations. Since some theories argue that ‘Too-Big-To-Fail’ banks are priced in a different way than non-‘Too-Big-To-Fail’ banks, it is worth it investigating previous research on this. This could help forming an expectation of what happens when a bank is officially no longer a Global Systemically Important Bank (G-SIB) (Financial Stability Board, 2012). An overview of the theories discussed in this section is presented in Table 1.
2.1 Empirical effects of earlier regulations
This study investigates the effect of banking regulations introduced after the start of the financial crisis. Therefore, this section discusses previous studies on regulations in the banking sector to see what they did to the probability of default. In particular, this section discusses a study on the Dodd-Frank Act in the United States, since the Dodd-Frank Act has similar features as the regulations introduced in Europe after the start of the financial crisis, such as rules for OTC trades and rules that aim to create more stability in financial markets (Dodd-Frank Wall Street Reform and Consumer Protection Act. H.R. 4173, 5th January 2010).
To start with, in an event study, Lutz (2016) found that after an announcement of the Fed that capital requirements would increase (like in Basel III), first abnormal stock returns,
Table 1: Overview of the theory
Theories discussed Authors
Empirical effects of earlier regulations
Increased capital requirements do not have an influence on subordinated bond pricing for banks
Lutz (2016)
Government guarantees are incorporated into debt prices Flannery & Sorescu (1996) Prices of unsubordinated bonds after the DFA decreased Balasubramnian & Cyree (2014)
CDS Spreads
CDS Spread changes are explained by the firm specific factors change in volatility, stock price return and the change in median CDS Spread in a class of companies where ratings are the same
Galil, Shapir, Amiram & Ben-Zion (2014)
The risk-free rate, leverage (credit risk regressors), bid-ask spread (liquidity regressor), term structure slope, swap spread, corporate bond spread, market return and market volatility (market regressors) have explanatory power in determining CDS Spreads in periods when there is credit risk.
Annaert, De Ceuster, van Roy & Vespro (2013)
Bank riskiness can be indicated by CDS Spreads. The usage of balance sheet variables after the financial crisis could have explanatory power.
Chiaramonte & Casu (2013)
Too-Big-To-Fail banks
The declaration of a bank being ‘Too-Big-To-Fail’ has a positive effect on the value of the firm and there is a possibility that ‘Too-Big-To-Fail’-banks can take on risks without having to pay higher interest rates.
O’Hara & Shaw (1990)
The discount in the yield spread for a bank being ‘Too-Big-To-Fail’ and the discount for the size of the bank decreased after the introduction of the DFA.
Balasubramnian & Cyree (2014)
De public does not incorporate G-SIB announcements information into the market Kleinow, Nell, Rogler & Rosch (2014)
Banks can not only be ‘Too-Big-To-Fail’, but they can also be ‘Too-Big-To-Rescue’
There is a negative relationship between the market to book ratio and the total assets of a bank. Those banks could potentially be ‘Too-Big-To-Rescue’.
Völz & Wedow (2011)
Demirgüç-Kunt and Huizinga (2013)
Market Discipline
There is no evidence that investors are able to influence companies when it comes to market discipline
Bliss & Flannery (2002)
A bank’s risk behavior influences the behavior of depositors. Moreover, they find that both small and large depositors exercise market discipline on banks, even though deposit guarantees are introduced.
Soledad Martinez Peria & Schmuckler (2001)
Market discipline decreased after the Long Term Capital Management bailout. Market discipline can only return to old levels if governments do not involve in financial institutions defaulting anymore.
Balasubramnian & Cyree (2011)
There is market discipline in the CDS market. Moreover, as the bank size increases, the CDS spread decreases.
for the companies that are in the S&P500 Financials Index, decreased in equity markets. However, after some days that effect disappeared again. Lutz also concluded that higher capital requirements did not influence bond rates. Since bond rates are not influenced after a regulation that partly resembles Basel III, this implies to this study that the market does not see an increase in risk for banks. Therefore, along this line of reasoning, CDS Spreads would not change either if capital requirements increase.
Secondly, Flannery and Sorescu (1996) performed another study on the incorporation of regulations into prices. They investigated the influence of government guarantees on market debt prices. They did this by regressing the yield spread on bank risk and other variables included for control. They found that the market incorporates government guarantees into debt prices. There is also a relation between the level of government guarantee and the degree to which this is reflected in debt market prices. When relating this to the current study: a deposit guarantee and a ‘Too-Big-To-Fail’ status (or a G-SIB status, which is investigated in this study) are comparable guarantees in the sense that investors are sure that they get their money back in case of bank default.
The studies mentioned in this subsection discuss the effects of regulations that resemble the regulations investigated in this study on several factors. One cannot draw direct conclusions to the regulations in this study, since the regulations are not entirely the same. For this study, this implies that whether regulations will have an effect on (CDS) pricing is still undetermined, but government guarantees are expected to influence the market (so official G-SIB statuses are expected to be incorporated as well). The next section discusses CDS Spreads’ determinants.
2.2 CDS Spreads
This section discusses Credit Default Swaps (CDSs). The CDS Spread is the price of an insurance against a default in a company’s payments. It is a reflection of the default probability of a bank (Annaert, De Ceuster, van Roy & Vespro, 2013). Therefore, the higher the default probability of the bank, the higher the insurance; the higher the CDS Spread. Since the regulations discussed in this study aim to strengthen banks, so the banks’ probabilities of default will decrease, the CDS Spread is investigated as a proxy for whether default probabilities actually decreased. The next paragraphs discuss determinants of CDS Spreads to get an overview of which regressors this study should take into account.
First of all, Galil, Shapir, Amiram & Ben-Zion (2014) find that CDS Spread changes are explained by the firm specific factors: change in volatility, stock price return and the change in median CDS Spread in a class of companies where ratings are the same. On top of
this, also market variables improve the model. They used time-series and multivariate regressions in order to perform their study.
Secondly, Annaert, De Ceuster, van Roy, & Vespro (2013) performed another study on the determinants of CDS Spreads using multivariate panel regressions. They found that in crisis periods, the factors: risk-free rate, leverage (credit risk regressors), bid-ask spread (liquidity regressor), term structure slope, swap spread and market return (market regressors) have explanatory power in periods when there is credit risk. In periods when there is not much credit risk, there is not much explanatory power.
The fact that CDS Spreads reflect the risk of a bank, is also confirmed by Chiaramonte and Casu (2013). They investigated CDS Spreads before, during and after the crisis and found that balance sheet variables explained CDS Spreads during and after the financial crisis. However, before the financial crisis they did not have much explanatory power. According to them this is due to the fact that the CDS Spreads were somewhat flat before the crisis.
In conclusion, changes in volatility, stock price returns and the change in median CDS Spread are important determinants of CDS Spreads. Besides, there are multiple factors that display high explanatory power in crisis periods, but in non-crisis periods this explanatory power deteriorates. Finally, Balance Sheet variables only had explanatory power during and after the financial crisis. Those are the most important determinants from the studies mentioned before. When comparing those determinants, it is surprising that they do not mention the same determinants. This study will take this information into account when determining the regressors in this study. In the next section, this study pays attention to ‘Too-Big-To-Fail’ banks.
2.3 Too-Big-To-Fail banks
Sometimes it is, implicitly or explicitly, obvious that some banks are considered ‘Too-Big-To-Fail’ in the economy. According to O’Hara and Shaw (1990), deposit insurance could bring stability to the economy, because investors would not need to run to the bank anymore for their money if they are afraid that the bank will default. They argue that when there is full deposit insurance, there is no actual bankruptcy risk anymore for those banks, since they cannot fail (they are ‘Too-Big-To-Fail’). This could imply that the bank does not need to pay a risk premium in order to compensate for its risks. This could create wealth effects for Big-To-Fail’ banks. In the paragraphs hereafter, the empirical effects of a bank being a ‘Too-Big-To-Fail’ bank are discussed.
First of all, O’Hara and Shaw (1990) empirically find that as soon as a financial institution is declared ‘Too-Big-To-Fail’, this has a positive effect on the value of the firm on the announcement date when compared to firms that are not declared ‘Too-Big-To-Fail’. If a firm is not declared ‘Too-Big-To-Fail’ this has a negative effect on the value of the firm. They investigated this by measuring the residual returns around the announcement date. They also mention the possibility that banks could be taking on risk without having to pay a higher interest rate in order to compensate. This way, they could benefit from the fact that they are Big-To-Fail’. This could also be important for this study, since it could be that ‘Too-Big-To-Fail’ banks’ lower probability of default is not only reflected in firm value, but also in CDS Spreads, since CDS Spreads are a reflection of the probability of default for a company.
Balasubramnian and Cyree (2014) investigated the yield spreads of ‘Too-big-to-Fail’ banks after the introduction of the Dodd-Frank Wall-street Reform and Consumer Protection Act (DFA) in the United States. They define yield spreads as the difference between the yield on a subordinated note or debenture minus the yield on a comparable riskless note or debenture. They found that before the introduction of the DFA, there was a discount in the yield spreads for subordinated notes and debentures for the size of the bank and for the fact whether it was considered ‘Too-big-To-Fail’. After the introduction of the DFA, the discount for a bank being ‘Too-Big-To-Fail’ and the discount for the size of the bank decreased. When the discount in the yield spread decreases, the yield increases and as a consequence the bond price decreases. They investigated this by regressing the yield spread on vectors of possible influential variables. To this study, this implies that the yields on the subordinated notes or debentures were too low before the DFA, the bond prices too high. The fact that the discounts in the yield spreads decreased, could imply that market discipline improved, since investors base their pricing less on the size of the bank or on the fact that it is considered ‘Too-big-to-Fail’ and more on their view of how the company performs.
Kleinow, Nell, Rogler and Rosch (2014) studied what happened when several banks were declared G-SIBs, while at the same time new stricter regulations were introduced. They investigate three moments on which banks are declared ‘Too-Big-To-Fail’, but they find that the public does not incorporate that information in the market. They mention two possible explanations: on the one hand it could be that they had already expected that those banks would become ‘Too-Big-To-Fail’. On the other hand it could be that the public does not trust declarations that banks become ‘Too-Big-To-Fail’. To this study, this implies that a bank becoming a G-SIB does not necessarily imply a pricing adjustment in the market for those
banks. This is contradictory to the study mentioned by O’Hara and Shaw (1990). Therefore, this study investigates what the direction of the CDS Spreads will be after a bank lost its G-SIB status. If the market does not incorporate G-G-SIB information, then it should not make a difference in CDS Spreads.
Apart from banks being To-Fail’, banks could also be considered Rescue’. Völz and Wedow (2011) find that some banks are considered ‘Too-Big-To-Rescue’ due to the high costs of bailout for such banks. They find this by regressing the CDS Spread on variables that indicate the size of the company. They argue that the presence of banks that are ‘Too-Big-To-Rescue’ are dangerous for the stability of the economy. Demirgüç-Kunt and Huizinga (2013) also mention the possibility of banks possibly being ‘Too-Big-To-Rescue’. They find that there is a negative relationship between the market to book ratio of a bank and its total assets. This implies: the larger the bank, the lower the market value of the bank.
To summarize, there is no agreement in the literature on whether the fact that a bank is considered Big-To-fail’ is incorporated into the market. Next to the concept ‘Too-Big-To-fail’, Völz and Wedow (2011) and Demirgüç-Kunt and Huizinga (2013) argue that banks can also be considered ‘Too-Big-To-Rescue’. Those theories play a large role in forming expectations about a bank losing its G-SIB status, which is investigated in one of the next sections of this study. The next section will discuss market discipline on banks.
2.4 Market Discipline on banks
Bliss and Flannery (2002) separate the concept ‘Market Discipline’ into market monitoring and market influence. They distinguish two situations: a situation in which investors watch a company’s performance and incorporate this performance evaluation into prices and a situation in which companies actually respond to the changing investor views and changing prices. For example, if a bank engages in risky operations, investors could require higher bond yields to be willing to invest in it. This is the first part of market discipline. The second part comes in when the bank actually responds to the price changes by for example adjusting their risk taking. Both variants of market discipline play a role in this study. The degree to which investors monitor a bank and adjust their prices to it, determines to what degree prices are an accurate reflection of for example risk taking by banks. The degree to which banks respond do those investor signals, determines whether investors actually have the power to influence actions, such as for example risk taking. This study presents empirical findings on market discipline in the next paragraphs.
To begin with, Bliss and Flannery (2002) empirically study market discipline for Bank Holding Companies in the US. Through both parametric and non-parametric tests using stock prices and bond prices they find no strong empirical evidence that investors are able to influence companies’ behavior. In other words, banks do not respond to price changes by changing their behavior. Since investors cannot influence those banks, they conclude that supervisory institutions still have to keep exercising their influence on companies. To the current study, this could imply that for example extra risk taking by banks could be reflected in the pricing of CDS Spreads, however this need not imply that those price changes influence managerial behavior inside banks.
Secondly, Soledad Martinez Peria and Schmuckler (2001) investigate market discipline by depositors on banks in Mexico, Argentina and Chile by regressing interest rates and the change in deposits on a vector of bank fundamentals variables, time specific variables and bank specific variables. They find that a bank’s risk behavior influences the behavior of depositors. Moreover, they find that both small and large depositors exercise market discipline on banks, even though deposit guarantees are introduced. This implies that both small and large depositors react to banks behavior by withdrawing their deposits or by demanding higher interest rates. To this study, this implies that depositors can exercise market discipline. In other words: the prices that depositors pay or ask when lending or depositing are a function of the bank’s risk taking. Since some theories argue that for ‘Too-Big-To-Fail’ banks prices are not a reflection of the actual risks that they are taking, it is important to note that in some markets market discipline does exist (however, this research was not performed in Europe, but this study assumes that depositor behavior is similar). According to the definition of market discipline by Bliss and Flannery (2002), the kind of market discipline that Soledad Martinez Peria and Schmuckler (2001) describe is market monitoring, not market influence.
Balasubramnian and Cyree (2011) argue that the government’s behavior plays an important role in the existence of market discipline on banks. By regressing the yield spread on vectors of possible influential variables, they find strong evidence for the existence of a ‘Too-Big-To-Fail’ discount in yield spreads amongst investors in the market for bonds after the bailout of Long Term Capital Management by the Federal Reserve Board in 1998. The discount on yield spreads for the size of bank became twice as large after the bailout had taken place. In other words, if a discount on a yield spread increases, the yield spread decreases. This implies that the bond price increases, after the government had bailed out Long Term Capital Management. The bailout was a signal from the government that implied
that there would be government guarantees. According to them, market discipline can only return to old levels if governments no longer involve themselves in financial institutions defaulting. To the current study, this implies that there is evidence for the deteriorating effect of government guarantees on market discipline, because the presence of a ‘Too-Big-To-Fail’ discount shows that investors do not fully incorporate their assessment of a bank’s performance into pricing, but they also consider other factors, such as the fact that a bank could be considered ‘Too-Bog-To-Fail’.
When relating the concept market discipline to CDS Spreads, the question arises to what degree market discipline exists in the CDS market. In other words, to what degree do investors assess banks and adjust prices and to what degree are investors able to influence banks’ behavior? Völz & Wedow (2011) investigate this by regressing the CDS Spreads on a variable that indicates the probability of default, amongst others. Since the coefficient on this variable turned out to be significantly positive, they conclude that there is market discipline in the CDS market. Moreover, they investigate the size of the bank in relation to the CDS Spreads. They find that there is a negative relationship between the two: as the bank size increases, the CDS spread decreases. This is due to the increase in the chance that the company is bailed out if a bank becomes larger in size. This confirms the ‘Too-Big-To-Fail’ phenomenon discussed in section 2.3. When relating this to the concept market discipline as described by Bliss and Flannery (2002), this implies that market monitoring is present in this market. Market influence is not investigated by Völz and Wedow (2011).
In conclusion, there is agreement in the literature that depositors adjust prices to their evaluation of the bank’s performance. However, Bliss and Flannery (2002) find no strong evidence that banks in turn also respond to those price changes. Furthermore, research shows that as the government bails out companies, prices reflect a ‘Too-Big-To-Fail’ effect. Finally, there is evidence that specifically in the CDS market investors monitor banks and they adjust their prices accordingly.
2.5 Conclusion
In conclusion, the past section discussed the effect of past regulations on several variables. There exist no studies on the same regulations as those that are investigated in this research. Therefore, this study cannot make predictions based on those empirical findings. However, the studies do give an idea of how the market reacts to regulations that resemble the regulations that are investigated in this study. This section mentioned several determinants of CDS Spreads, which will be taken into account when determining the regressors in this study. Thirdly, there is no agreement in existing literature on whether the
market incorporates the fact that a bank is considered ‘To-Big-To-Fail’ into pricing. Therefore, this study investigates what happens in this study when a situation is investigated in which a G-SIB loses its G-SIB status. Finally, there is agreement in the literature that investors monitor companies and react to this by changing their prices, however there is no evidence that the company in turn reacts to the changes in prices. This is something that could become clear after performing the current study. The next section presents the four financial regulations that this study investigates. After this, it presents the hypotheses.
3 Regulations after financial crisis
The research question of this study is ‘How do CDS Spreads respond to announcements of new regulations and to an announcement that a bank is no longer considered ‘Too-Big-To-Fail’?’. The first part of this research question concerns announcements for regulations. The four regulations are taken from a report published by the European Banking Authority (European Banking Authority, 2015). This section discusses each regulation. Some regulations consist of multiple regulations, so an aggregated announcement effect applies here. Table 2 mentions the dates on which the new regulations are announced. At the end of this section, this study presents the hypotheses.
Table 2: The announcement dates of the regulations of interest
Regulation Announcement date
BRRD 15 November 2008 (European Commission, 2014; European
Commission, 2008). Basel III, including LR and Liquidity Ratios,
CRR, CRDIV
2 April 2009 (Regulation (EU) No 575/2013, 26th June 2013)
EMIR 24 and 25 September 2009 (University of Toronto G20 Information Centre, 2009).
Liikanen Report 2 October 2012 (European Commission, 2015)
This table presents the announcement dates of each of the regulations. Basel III introduces the LR, Liquidity Ratios, CRR and CRDIV at the same time. Therefore, one single announcement date applies.
3.1 The BRRD
The first regulation that this paper discusses is the EU Bank Recovery and Resolution Directive (BRRD). After the start of the financial crisis, the need for this directive existed, since several governments had to bail out banks, using tax money. The objective of this directive is to increase stability in the banking sector and to control the failure of a bank in order to make sure that the government no longer needs to bail out banks using public money. This directive obliges banks and resolution authorities to anticipate a situation in which the bank will or is about to default. Supervisors intervene in an early stage of instability. Resolution authorities are empowered to separate viable activities from non-viable activities and to split or merge the bank. Resolution authorities have the authority to oblige shareholders and creditors the cover debts partly (European Commission, 2014).
3.2 The LR, liquidity Ratios, the CRR and the CRDIV
Basel III consist of multiple regulations (Regulation (EU) No 575/2013, 26th June 2013). Therefore, this paragraph contains information on the Capital Requirements
Regulation (CRR), the Capital Requirements Directive (CRDIV), the rules about the Leverage Ratio (LR) and the rules on Liquidity Ratios. The aim of Basel III is to make the banking market safer and more transparent (Bank for International Settlements). The next paragraphs presents each regulation in short.
Authorities introduced the CRR and the CRDIV because the main difference why some companies defaulted during the financial crisis and why others did not, was that the ones who did survive kept more and higher quality capital. Besides, the Council of the European Union and the European Parliament felt the need for more cooperation on fiscal, supervisory and monetary areas between countries in order to track developments between countries. The CRR introduces a ‘Single Rule Book’ that includes universal rules for all banks in the European Union. The CRDIV enhances risk management within companies and transparency. The new rules require banks to increase the highest quality capital that they hold (TIER 1) from 2 to 4.5%. On top of this, five new capital buffers are introduced (European Commission, 2013).
Basel III introduces leverage ratio rules, because the build-up of excessive leverage played an important role in the development of the financial crisis. The leverage ratio rules aim to prevent banks from building up this excessive leverage (Bank for International Settlements, 2014). The rules imply that the ratio of TIER 1 capital over the Total Leverage Ratio Exposure non-Risk Weighted Assets should exceed 3 per cent (European Banking Authority, 2015).
The last topic included in Basel III which is discussed here is Liquidity Ratios requirements. The first liquidity ratio requirement is the 30-day Liquidity Coverage Ratio (LCR). In the beginning of the financial crisis, banks came into short-term liquidity problems. Liquidity vanished quickly and banks came into trouble. Therefore, the LCR requires banks to hold enough high quality leverage, so they can respond to crisis situations (Bank for International Settlements, 2013). The second regulation concerning Liquidity Ratio requirements is the Net Stable Funding Ratio (NSFR). This regulation requires banks to have a stable construction for their funding in case their regular sources of funding are disturbed. This should decrease the default probability of the bank. The ratio of available over required stable funding should be larger than or equal to 100%. The amount of required liquidity depends on the characteristics of the banks’ assets (Bank for International Settlements, 2014). 3.3 The EMIR
Excessive risk taking and poor supervision played an important role in the development of the financial crisis. To prevent this from happening again, the G20-top in
Pittsburgh announced that at the end of 2012 several rules and requirements would be set in the Over-The-Counter derivatives market. The aim of those rules would be to strengthen the international financial regulatory system in order to make improvements to the Over-The-Counter (OTC) derivatives market (University of Toronto G20 Information Centre, 2009). This announcement laid the grounds for what was later introduced as the European Market Infrastructure Regulation (EMIR). This regulation determines risk and clearing requirements for OTC derivatives trades, it requires that derivatives trades are reported and it sets requirements for Central Clearing Parties (Regulation (EU) No. 648/2012, 2012, July 4th 2012).
3.4 The Liikanen Report
Another reform that The European Banking Authority (European Banking Authority, 2015) mentions is the Liikanen Report. This report aims to prevent that ‘Too-Big-To-Fail’ banks would default, since this could bring a lot of damage to the financial system. The aim of this regulation is to decrease bank default probabilities, which could create more stability for the banking sector. The Liikanen Report prohibits speculative activities inside a bank. Besides, it lays the ground for a possible separation of trading that involves high risks by banks (European Commission, 2014).
3.5 Hypotheses
This section discusses three hypotheses. The first paragraph states the hypothesis for a general investigation whether CDS Spreads significantly changed after the crisis compared to before the crisis. Then, this section discusses the hypotheses concerning the event studies on the regulations. Finally, the hypothesis concerning a bank losing its G-SIB status is discussed.
To start with, this study makes a comparison between CDS Spread before and after the financial crisis. The reason for this is that all regulations that this study investigates, are introduced as a consequence of the financial crisis. If default probabilities decreased after the financial crisis, then this should be reflected in a decrease in CDS Spreads. Since all regulations aim to make the banking market safer and therefore to reduce default probabilities, this study expects the CDS Spreads to decrease after the crisis, compared to before the crisis.
Hypothesis 1: CDS Spreads decreased after the financial crisis, compared to before the financial crisis.
All regulations mentioned in sections 3.1 until 3.4, serve to strengthen the banking sector. They aim to increase transparency and to reduce the risk that banks default. Therefore, the expected effect of all those regulations is that probabilities of default decrease. This study measures the change in CDS Spreads for each regulation by performing an event study containing observations from half a year before the announcement date of the new regulations until half a year after this date. This study expects default probabilities to decrease after each of the regulations. This study expects that this is reflected in CDS Spreads. Therefore, the hypothesis is:
Hypothesis 2: All announcements of the regulations mentioned in this paragraph have a negative effect on CDS Spreads.
Finally, this study performs an event study on the effect of a bank becoming a G-SIB during the sample period. ‘Too-Big-To-Fail’ (or G-SIB) banks are banks for which prices do not accurately reflect the risks that they take. Therefore, it is interesting to see what happens with the market’s view on the default probability of that one bank, once it officially loses its G-SIB status. Since the government will save the bank if it is about to default, the probability that investors will get their money back increases once the bank is considered ‘Too-Big-To-Fail’. If the bank does not do much damage to the economy if it defaults, it could be that the government does not save it. This increases the default probability from the viewpoint of investors. Therefore, this study expects that the CDS Spread will increase after the announcement that the bank becomes a G-SIB. This research performs the event study on observations around the announcement date, half a year before the announcement until half a year after the announcement.
Hypothesis 3: After losing a G-SIB status for a bank, the CDS Spread increases.
In conclusion, this section presented the four regulations that are investigated in this study. Besides, it presented the hypotheses for a comparison of CDS Spreads before and after the crisis, the hypotheses on the event studies on four regulations and the hypothesis on a bank losing its G-SIB status. The next section discusses the methodology used in this study.
4 Methodology
This section presents this study’s method. The first regression serves to give an overview on whether CDS Spreads significantly changed after the financial crisis. Secondly, the method that is used to investigate the effect of each of the four bank regulations on CDS Spreads is discussed. Finally, this section presents the method used in order to measure the effect of a bank losing its G-SIB status.
4.1 Compare CDS Spreads before and after the financial crisis
In order to get a view of how CDS Spreads compare before and after the crisis, regression (1) combines the weekly observations from 2008 and 2015. A dummy variable that indicates whether the observation is from before or after the crisis should indicate whether the financial crisis had a significant effect on CDS Spreads.
𝐶𝐷𝑆 𝑠𝑝𝑟𝑒𝑎𝑑(𝑖,𝑡)= 𝛼 + 𝛽𝐹,𝑘∗ 𝐹𝑖𝑟𝑚𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐(𝑖,𝑡)+ 𝛽𝑀,𝑘∗ 𝑀𝑎𝑟𝑘𝑒𝑡𝑡+ 𝛽𝐿,𝑘∗ 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑡+ 𝛽5∗ 𝑃𝑜𝑠𝑡𝑐𝑟𝑖𝑠𝑖𝑠 + 𝛽6∗
𝑃𝑜𝑠𝑡𝑐𝑟𝑖𝑠𝑖𝑠 ∗ 𝐹𝑖𝑟𝑚𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐(𝑖,𝑡)+ 𝜀 (𝑖,𝑡) (1)
Secondly, this study performs regression (2) in order to measure the weekly change in CDS Spreads. In this regression, for all variables, percentage growth variables were used, except for ‘postcrisis’. This is a dummy variable, so measuring growth does not have any meaning. In all cases, a company fixed effects regression with robust standard errors was used.
𝛥𝐶𝐷𝑆 𝑠𝑝𝑟𝑒𝑎𝑑(𝑖,𝑡)= 𝛼 + 𝛽𝐹,𝑘∗ 𝛥𝐹𝑖𝑟𝑚𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐(𝑖,𝑡)+ 𝛽𝑀,𝑘∗ 𝛥𝑀𝑎𝑟𝑘𝑒𝑡𝑡+ 𝛽𝐿,𝑘∗ 𝛥𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑡
+ 𝛽𝑘∗ 𝑃𝑜𝑠𝑡𝑟𝑖𝑠𝑖𝑠 + 𝛽𝑘∗ 𝑃𝑜𝑠𝑡𝑐𝑟𝑖𝑠𝑖𝑠 ∗ 𝛥𝐹𝑖𝑟𝑚𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐(𝑖,𝑡)+ 𝜀 (𝑖,𝑡) (2)
All regressions in this study are partly based on a study by Balasubramnian and Cyree (2014, p. 159) and applied to the European 2009 Financial crisis. All four regression equations regress the CDS spread on a vector of liquidity variables, a vector of market variables and a vector of firm specific variables, just like in the research by Balasubramnian and Cyree (2014, p. 159). However, the regressions do not include all variables mentioned by Balasubramnian and Cyree (2014), the variables were chosen according to what was considered necessary and available for this study and other new variables were added. Table 3 presents an overview of this study’s variables and the expected direction of their coefficients in the regressions. The reason that this study based its regression on the study by Balasubramnian and Cyree (2014), is that they investigated the effect of a regulation on market discipline. The idea behind this is that market discipline is reflected in the yield spreads: as market discipline increases, market monitoring increases and therefore yield
spreads adjust to changes in a firm’s risk. This study investigates the effect of regulations on default probabilities through CDS Spreads. As regulations require banks to become safer and more transparent, default probabilities change and therefore CDS Spreads adjust to those changes. Since this line of reasoning is relatively similar, this study used Balasubramnian and Cyree’s (2014) regression equations as a basis. The next paragraph elaborates on the specifications of the variables.
To start with, the dependent variable in this study is the CDS Spread, not the Yield spread as is used by Balasubramnian and Cyree (2014). This study uses the CDS spread, because it is a default probability indicator. This study uses the CDS Spread as a proxy for the default probability, whereas the yield spread represents the differences between yields (this is a relative measure).
This study uses three out of the five market variables (European equivalents) that Balasubramnian and Cyree (2014) used: the FTSE100, the ECB benchmark bond yield and the difference between the 30 year and the 2 year bond yield (Gov Bond yield 30yr - 2yr). Besides, one variable is added (the European CDS Index). Those market variables serve to control for market movements, such as economic up or downturns, specific events such as the Greek sovereign debt crisis and other events that took place during the sample period. In the case of for example the Greek government debt crisis, it could be that the unrest in the market influenced CDS Spreads. However, this study tries to control for this by adding for example the European CDS Index and the FTSE100 to the regression to reflect general market movements. The ECB ten-year government bond yield reflects the risk free rate, which could also be affected by several events such as for example the Greek sovereign debt crisis. Chauvet & Potter (2005) argue that the higher this ten-year yield, the higher expectations of short-term interest rates. This implies that a high ten-year yield could indicate that the economy will be in an upturn. The difference between the thirty-year and the two-year bond yield indicates the slope of the yield curve. This serves as a crisis indicator: if the curve slopes downward, this could indicate the investors expect interest rates to drop in the long term, for example because they expect an economic downturn. As a consequence, the yield curve slopes downward (Chauvet & Potter, 2005). The reverse logic applies for an up sloping yield curve. Therefore, this variable serves to indicate the ‘mood’ of the market. Balasubramnian and Cyree (2014) did not include the European CDS index as a market variable, but this study added it to control for general movements in the CDS market. Due to data availability, the volatility index for European stocks and the 30-day Eurodollar – T-bill rate could not be added.
Since Balasubramnian and Cyree (2014) investigated the bond market by measuring yield spreads, they added the characteristics of the bond such as for example maturity and bond age as liquidity variables. However, the current study does not investigate the bond market, but the CDS market. Therefore, adding those variables would not make sense. Instead, this study added the average bid-ask spread of the top 10 non-banks in terms of market value from the AEX (Euronext, 2016) to control for movements in liquidity. This study took the market values from DataStream.
All firm specific variables that Balasubramnian and Cyree (2014) used, are used in this study as well, except for the ratio Trust Preferred Securities over Total Assets, the Z-score and the coupon rate as an indicator for taxes. For the other variables, the European equivalents are applicable in this study (not a ‘Dodd-Frank’ indicator but a ‘postcrisis’ indicator). Since trust preferred securities are not available for the firms in this sample, this variable is not used. The Z-score by Altman (1968) requires Working Capital for banks, which is not available in DataStream for the banks in the current sample. Instead, this study adds Return on Equity as a performance indicator. The RoE indicates the reverse of the Z-score: the Z-score is a default indicator whereas the RoE is a profitability indicator (Dewenter & Malatesta, 2001). Since this study investigates European regulations that exist in several countries with different tax systems, controlling for taxes is not applicable here.
Balasubramnian and Cyree (2014) added a vector of control variables with dummies for each of the banks in the sample. This study used company fixed-effects, so using a dummy for each bank is not necessary. Another difference between the study by Balasubramnian and Cyree (2014) and this study is that in their study for the firm-specific variables, lags are included instead of the variables at the time. This study tries this and concludes that this does not improve the fit of the model. Section 7 on Robustness provides more detailed information on this. The last difference between the current study and the study by Balasubramnian and Cyree (2014) is that they used the Generalized Method of Moments. Using this in the current study would be beyond the scope of this research and therefore it is not included. Instead, a panel data OLS regression using company fixed-effects is used.
Looking at the literature discussed on the determinants of CDS Spreads in section 2.2, one can see that almost all variables mentioned by Galil, Shapir, Amiram & Ben-Zion (2014) are included in the regressions here. The only variable that is not included is the change in volatility, since here volatility is already used by Balasubramnian and Cyree (2014). Since this study uses Balasubramnian and Cyree’s (2014) model, the volatility is used instead of the change in volatility. The determinants found by Annaert, De Ceuster, van Roy, & Vespro
(2013) were risk-free rate, leverage (credit risk regressors), bid-ask spread (liquidity regressor), term structure slope, swap spread and market return (market regressors) during crisis periods. Leverage regressors such as for example the leverage ratio are used in this study. However, some regressions omitted this variable, because it is a balance sheet variable that has yearly observations. For the regressions that do no omit this variable, it is kept in the model. For the regressions that omit the leverage ratio, it is left out. This study assumes that the expected effect of the leverage ratio in Table 3 could be both positive and negative, since until some point extra debt over equity increases profitability, whereas after that point it could bring the bank in problems if it takes on too much debt. Besides, Annaert, De Ceuster, van Roy & Vespro (2013) include the swap spread (as an indicator for the stability of the banking sector) in their regressions, however, the euro interest rate swaps that are necessary to calculate this are not available. Besides, other market variables in this study are also able to correct for banking market movements, so the assumption is made here that including those could be sufficient. Chiaramonte and Casu (2013) argued that balance sheet variables were able to explain CDS Spreads during and after the crisis. Since part of this study investigates observations also from before the crisis, those balance sheet variables are not included.
Table 3: Variables used
Variables Indicator for Expected Direction
Market variables
FTSE 100 (ftse100)
Euro Area 10 year benchmark bond yield, ECB (tenyear)
Gov Bond yield 30yr - 2yr (yield302) DataStream Europe Other Financial 5 Year Credit Default Swap Index (cdseurope)
Market stock price movements Risk free rate
Yield curve slope CDS market movements - + - + Liquidity Variable
Average bid-ask spread 10 largest companies AEX based on market value (baspread)
Changes in liquidity + Firm-specific variables
Stock Price Return (stockpricereturn) Non-performing loans/total assets (nplta) Market Value (marketvalue)
Volatility (pricevol) Crisis Indicator (postcrisis) Return on Equity (roe) Leverage Ratio (leverageratio)
Stock price movements Exposure
Size of the bank
Underlying volatility of the bank Effect of the financial crisis Profitability
Proportion debt to equity
- + - + + - +/-
Regression (2) uses percentage changes. The RoE, Price Volatility and NPL/TA are measured yearly in DataStream. Therefore, instead of weekly percentage changes, yearly
percentage changes are used in the regression. This is to prevent that one column of zeros is created.
4.2 Event study for four regulations
This section presents the method that this study uses in order to do an event study for each of the four regulations. The regression equation that is used in order to investigate the announcement effects of each of the regulations is:
𝐶𝐷𝑆 𝑠𝑝𝑟𝑒𝑎𝑑(𝑖,𝑡)= 𝛼 + 𝛽𝐹,𝑘∗ 𝐹𝑖𝑟𝑚𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐(𝑖,𝑡)+ 𝛽𝑀,𝑘∗ 𝑀𝑎𝑟𝑘𝑒𝑡𝑡+ 𝛽𝐿,𝑘∗ 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑡
+ 𝛽4∗ 𝑑𝑢𝑚𝑚𝑦 𝑎𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡 𝑒𝑓𝑓𝑒𝑐𝑡 + 𝛽5∗ 𝐷𝑢𝑚𝑚𝑦 𝑎𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡 𝑒𝑓𝑓𝑒𝑐𝑡 ∗ 𝐹𝑖𝑟𝑚𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐(𝑖,𝑡)
+ 𝜀 (𝑖,𝑡) (3)
For each regulation, weekly observations from half a year before until half a year after the announcement of the new regulation are investigated. Table 2 in Section 3 presents the announcement date for each of the regulations. This study assumes that it is better to measure starting from the announcement date then starting from the enactment date, since on the enactment date all companies have to fulfill obligations already so it is hard to measure an actual change. The same variables as in the previous section will be used, described in Table 3. Figure 1 provides a time line representing the period that is investigated for each regulation.
4.3 Event study for a bank losing
its G-SIB status
It is not possible to use the same event study as the one shown in regression (3), because there multiple banks were investigated so yearly observations for certain variables did not cause any problems. However, since only one bank’s G-SIB status changes during the sample period, another test needs to be performed. Therefore, another event study method is applied here. For the observations before the G-SIB declaration, this study runs the following regression, using relative CDS Spread changes in a market return model, based on an article by Andres, Betzer & Doumet (2016):
Figure 1: Timeline representing the period investigated for each regulation
0.5 year before 0.5 year after
𝐶𝐷𝑆 𝑠𝑝𝑟𝑒𝑎𝑑 (𝑡)=∝ +𝛽 ∗ 𝐸𝑢𝑟𝑜𝑝𝑒𝑎𝑛 𝐶𝐷𝑆 𝑖𝑛𝑑𝑒𝑥 (4)
For the observations after the G-SIB declaration, this study performs Stata’s sign test to test whether:
ln(CDS Spread) = ln(∝ +β ∗ European CDS index)
These regressions use a time span of one year before until one year after the G-SIB declaration. The reason that one year is used instead of half a year, as in the previous regressions, is that this regression concerns only one bank. Therefore, in order to have a larger sample of observations, one year before until one year after the declaration is used.
1 year before 1 year after
G-SIB declaration
5 Data
This section presents a description of the data used. First, this section discusses the criteria for the selection of banks. The paragraph thereafter presents a description of the data for the variables. Thereafter, the time span is explained. Finally, the last paragraph presents the summary statistics and the correlation matrix.
5.1 Bank selection criteria
This study gathered all data on CDS Spreads for banks from DataStream. The sample of European banks’ Credit Default Swaps is included DataStream’s CDS Spreads list: L#431027 ‘European Bank CDS spreads’. After exporting those firms’ weekly CDS Spreads over time, they are placed in the format of panel data. After exporting the list, the sample consisted of CDS Spreads for 42 banks. This list consists of senior unsecured and subordinated unsecured CDSs. During this study, when collecting other variables, several banks dropped when variables were not available for them. This study left out three outliers as well, see section 7 on Robustness. In the end, 29 banks were left. The list of banks that are used in this study is provided in Table 1 in the Appendix.
This study uses weekly data instead of daily data as used by Balasubramnian and Cyree (2014). The reason for this is that this study expects that looking at daily CDS Spread movements would also reflect daily price movements instead of longer term movements that display the long term effect of the regulations on CDS Spreads.
5.2 Variables
All weekly data on variables are gathered from DataStream, except for the G-SIBs and the data on the 10 year, the 2 year and the 30 year benchmark government bond yield. The Statistical Data Warehouse from the European Central Bank (2016) provided the government bond data. All variables are continuously compounded yields (percentage per year).
Since for all CDS products, only the codes of the products are available, manually the mnemonic codes for the names of the respective banks are looked up in DataStream. For some banks, multiple codes are available if the stock is traded on multiple exchanges. If this is the case, this study chose a code in a country for which the stock is traded in euros. This way, the data are all in the same units. This study used this list of bank codes in order to extract the firm specific variables for each bank from DataStream.
The Financial Stability Board (2012) provides data on which banks are official G-SIBs. Table 2 in the Appendix provides a list of the G-SIBs included in the sample. The first publication of the G-SIB qualification takes place in November 2011. This study assumes
that the banks that are considered G-SIB at that point, could also have been a G-SIB before, since systemically important banks cannot become systemically important within a short time period. Updates by the Financial Stability Board are released in 2013 and 2014. However, for European banks there is only one bank for which the G-SIB status changes during the sample period, which is Lloyds Bank (Financial Stability Board, 2102; Financial Stability Board, 2013; Financial Stability Board, 2014). When processing the data in Stata, all lines in the data dropped if an observation is not available. The most recent date for which all observations are available is 31 December 2015.
5.3 Time span
As the starting point of the financial crisis 1 January 2009 was used, since 2009 is the first year in which Europe experienced negative growth (World Bank). Besides, data on CDS Spreads in DataStream are only available starting from December 2007, so a period far before the crisis could not have been chosen. However, for further research on whether CDS Spreads significantly changed after the crisis compared to before the crisis, it could be a suggestion to use observations from a period far before the financial crisis. In the current situation, it could be that American unrest in financial markets already play a role in European CDS Spreads in 2008, for example Lehman Brothers defaulted in September 2008 (Baba & Packer, 2009), however data for periods much further before the crisis are not available. Besides, measuring a change between 2008 and 2015 still can have a meaning, since it could be that in 2015 CDS Spreads became more stable, since this is after the crisis, whereas 2008 is before or at the beginning of the crisis.
For the first regression, that investigates whether CDS spreads significantly changed after the crisis compared to before the crisis, this study uses weekly data from 2008 for the ‘pre-crisis’ period. The starting point for the ‘postcrisis’ period is the most recent year before this study was performed. Many variables are only available until 31 December 2015, so that date is the last date in this sample. Therefore, data from 2015 are used for the ‘postcrisis’ period.
For the event studies on the four regulations that this research performs, the periods before and after the announcement of the new regulation (not the enactment) must be taken into account in order to measure the announcement effect. This study uses a time span of six months before the announcement until six months after the announcement of the new regulations. Balasubramnian and Cyree (2014) did not use this time span in their research: they investigated half a year before the announcement and half a year after the enactment of the regulation. However, the current research performs an event study, therefore it uses half a
year before until half a year after the announcement date. Besides, this study investigates several regulations that were introduced shortly after each other. If this study used observations for each regulation half a year before the announcement and half a year after the enactment, then it could be that a mixture of effects is measured instead of only one regulation’s effect. This risk is already present when using the current method, however when using Balasubramnian and Cyree’s (2014) method, the effect of this mixture of effects could become enlarged.
5.4 Summary statistics and correlation
This section presents the summary statistics and the correlation of the data in this study. Table 4 presents the summary statistics of the entire sample. Table 5 presents the summary statistics of 2008 compared to 2015 and Table 6 presents the correlation matrix.
To start with, for the entire period 2008-2015 there are over 10,000 observations. However, most regressions use only part of that dataset. The reason for this is that many event studies only require one year of weekly data around the actual date of the event.
Table 4: Summary Statistics for the period 2008 until 2015 Variable Obs Mean Std. Dev. Min Max CDS Spread (bp) 10,100 200.53080 226.65370 30.50000 2,044.09000 FTSE100 10,100 5,817.80800 785.02490 3,542.40000 7,103.98000 10yr ECB bond yield 10,100 0.03218 0.01084 0.00852 0.04815 CDS Index Europe 10,100 239.69190 136.46130 79.29700 775.58800 NPL/TA 10,100 0.04400 0.05284 0.00151 0.30186 30yr-2yr ECB yield 10,100 0.02249 0.00602 0.00140 0.03315 Bid-Ask Spread 10,100 0.29642 0.13954 0.06840 1.02630 Stock Price Return 10,071 0.86350 9.13836 -0.99600 448.09740 Market Value (mil) 10,100 23,200 19,400 23,2 101,000 Price Volatility 10,100 31.72378 9.65813 15.16000 72.71000 RoE 10,100 1.37910 16.95478 -149.70000 25.54000 Leverage Ratio 10,100 820.08470 756.32800 175.99000 5,895.16000
This table contains the summary statistics for each variable used in this regression.
There are several notable facts in Table 4. For example, the mean CDS Spread is 201 basis points. This implies that in order to insure payment from a bank, on average 2.01 per cent was paid. The maximum value here was 2044 basis points, which implied an insurance payment of more than 20 per cent. Another factor that stands out is the 30 year - 2 year government bond yield. This variable indicates whether the yield curve on average sloped up or downward in the sample. According to the results, there is a difference of 0.0225 on average between those two. This implies that the 30 year yield is on average 2 basis points
higher than the 2 year yield, which could indicate an up sloping yield curve. An up sloping yield curve could indicate that this study investigates a period of economic upturn (Chauvet & Potter, 2005), whereas in practice the period 2008-2015 contained a crisis. Therefore, this result is counterintuitive.
The first regression in this study investigates whether CDS Spreads changed between 2008 and 2015. Therefore, that regression is performed on observations from 2008 and 2015. Table 5 provides the means of the variables from those two periods separately and compares those using Stata’s t-test (mean comparison test) for unequal variances using Welch’s approximation. This test is used, because the samples from 2008 and 2015 have unequal sizes and the variances are not known and they are not necessarily equal (Ruxton, 2006). When comparing the summary statistics of the sample of observations from 2008 to the sample of observations from 2015, the t-tests point out that for each variable in the sample, the means are different in 2008 than in 2015. This implies that the datasets of 2008 and 2015 contain significantly different observations.
Table 5: Summary statistics 2008 compared to 2015 Variable Observations 2008 Mean 2008 Observations 2015 Mean 2015
Sign different? Alpha 5%
CDS Spreads (bp) 1,079 100.4802 1377.0000 97.0938 yes
FTSE100 1,079 5299.3740 1377.0000 6581.1830 yes
10yr ECB bond yield 1,079 0.0436 1377.0000 0.0128 yes
CDS Index Europe 1,079 297.9494 1377.0000 108.1265 yes
NPL/TA 1,079 0.0184 1377.0000 0.0556 yes
30yr-2yr ECB yield 1,079 0.0115 1377.0000 0.0159 yes
Bid-Ask Spread 1,079 0.2695 1377.0000 0.3419 yes
Stock Price Return
(%) 1,077 5.8793 1375.0000 0.2038 yes
Market Value (mil) 1,079 22800 1377.0000 33100 yes
Price Volatility (%) 1,079 27.3324 1377.0000 30.5026 yes
RoE 1,079 4.6944 1377.0000 5.9493 yes
Leverage Ratio 1,079 1316.5530 1377.0000 472.8474 yes
In this table, the variable means from the period 2008 are compared to the variable means from the period 2015. Using Stata’s t-test (mean comparison test) for unequal variances using Welch’s approximation at an alpha of 5%, this study
concludes that the means for each variable are significantly different in 2008 than in 2015.
In the correlation matrix in Table 6, several numbers stand out. The coefficients that are discussed, are mentioned in bold in Table 6. To start with, the CDS Spread is positively correlated with the ratio Non-Performing Loans over Total Assets. This implies that the larger the ratio Non-Performing Loans over Total Assets, the larger the CDS Spread. At first sight, this result might seem counterintuitive, however there could be an explanation for this:
