An Examination of Covered Bonds

Bachelor Thesis in Banking and Finance Department of Banking and Finance University of Zurich

Prof. Dr. Kjell Nyborg

Assistant: Jiri Woschitz

Romano Gruber 09-724-915

Salviastrasse 8 7205 Zizers +41 79 406 11 42 [email protected]

June 8, 2012 Abstract This thesis explores Covered Bonds by providing a theoretical overview and investigating the development of the respective market in recent time. We collate jurisdictions and ﬁnd crucial distinctions in the frameworks between established and developing Covered Bond markets. By analyzing the outstanding amounts of 23,363 Covered Bonds we observe a strongly declining public and an expanding mortgage Covered Bond market. The average yield spreads calculated by country show an increase for Anglo-Saxon countries after the Lehman bankruptcy and a strong surge for euro area countries deep in debt.

Keywords: Structured Finance, Covered Bonds, Bank Funding, Pfandbriefe

JEL Classiﬁcation: G10, G12, G15, G21

I Bachelor Thesis in Banking and Finance (Focus Economics) by Romano Gruber

Romano Gruber Salviastrasse 8 7205 Zizers

An examination of Covered Bonds

Covered Bonds (for instance in Switzerland called Pfandbriefe) are bonds which are issued by specific firms by country. These bonds are backed by very liquid and cash flow generating assets such as public sector loans and mortgages. Compared to off-balance sheet instruments, called asset-backed securities, where these collaterals are put in separate ve- hicles (called Special Purpose Vehicle (SPV)) and refinanced separately, these bonds are remained on the balance sheet of the specific firm. This thesis focuses on Covered Bonds.

In a first part the thesis focuses on how Covered Bonds are treated in different countries. Second, you collect data from the (different) firms which issue Covered Bonds in each country and compare them by adequate descriptive statistics. A third section focuses on the question of how Covered Bonds are treated by rating agencies. In a fourth section, you establish the differences between Covered Bonds and asset-backed securities. In the last section you try to figure out, if there exists data on SPV performance over time. If this exists, it should be compared to the collected Covered Bond data.

The following paragraphs explain in more detail the diﬀerent potential parts:

The first part of the thesis focuses on how Covered Bonds are treated in different countries. Basically, this is a comparison of different aspects (for instance differences in law aspects, differences in requirements on the cash generating collaterals, which firms (are allowed to) issue Covered Bonds, etc.) in different countries1. Here the focus should lie on the question which aspects and differences could have an influence on the fact that Covered Bonds issued by different countries should be treated differently (for instance differences in the law of two countries on which assets may be used as cash flow generating collaterals may have an influence on the certainty of Covered Bonds).

1 The ECBC (European Covered Bond Council) has issued the ECBC Covered Bond Fact Book 2011. This can be downloaded after registration on http://ecbc.hypo.org/Content/Default.asp?PageID=501 and is about the Covered Bond market. The second part of the quite comprehensive reading is about all these aspects well-arranged by country. ECBC also provides a tool (comparative database ﬁndable on http://www.ecbc.eu) where this comparison can be done per aspect for two countries.

II In the second part you collect data in Datastream and/or Bloomberg on Covered Bonds performance in the past2. This should be done for every firm in all the countries which are listed in the ECBC Covered Bond Fact Book 2011 (c.f. www.ecbc.eu). After collecting the data (do not underestimate this task, it might be quite time consuming), Covered Bond performance in different countries (maybe for different issuers) should be compared by descriptive statistics. Do you see any parallels with respect to the aspects found in the first part of the thesis (for example, if the law on which assets can be used as collaterals is very limp in one country but very strict in the other, do you see that the Covered Bonds in the more strict country is less volatile and vice versa?). If you find that such differences in the law of countries might lead to different trends in historical Covered Bond prices, volume etc. it might be worth to establish further statistical analyzes which show these potential relationships.

The third part of the thesis focuses on how rating agencies treat covered debt? Maybe there is practical (or maybe even academic) literature on how rating agencies treat Cov- ered Bonds from different countries. Potential findings should be compared to findings in the first part of the thesis. Furthermore, do there exist CDS contracts on Covered Bonds? If yes, does the performance of such CDS contracts reflect what you expected in terms of the findings in the first section of the thesis?

The fourth part should focus on the differences between on-balance sheet (Covered Bonds) and off-balance sheet instruments (Asset-backed Securities). What are the consequences that follow from this simple difference that Covered Bonds are on balance sheet and ABS are off balance sheet in terms of risk (different sorts of risk such as liquidity risk, credit default risk, etc.) and return from a theoretical perspective?

The fifth and last part focuses on historical data for asset-backed-securities. Is it possible to find such data on Bloomberg or Datastream (or even other databases)? If yes, show a descripitve analysis (maybe the same analysis you did in the second part on the Covered Bond data). Can you find obvious differences?

This thesis is a quite heavy workload. Do not underestimate the different parts since some of them might be more time consuming than one thinks before starting. The goal of the thesis is not to answer every question in these five parts in every detail since time will be too short to do this. However, try to focus on each part at least a little bit and then decide on which parts you can add most insights. It might, for instance, be that there is no historical data findable on asset-backed securities in Bloomberg or Datastream. However, do at least figure out if it exists or not. It is not your fault if it does not exist.

2 Do also make a short list of other variables that can be found in the databases, e.g. volume outstanding, trading volume, etc.

III Course of your thesis:

• After completing the subdata 23 sheet you will have a ﬁrst meeting with your advisor to discuss in detail how to proceed in detail.

• If you have questions you can ask your advisor for an appointment. However, the goal is to acquire the knowledge on your own and try to solve problems on your own, but to discuss your own solution to a problem with your advisor.

Formal criteria and submission:

• You compose the thesis in English.

• Quality comes before quantity. Be brief and compact in writing. Your thesis should not be longer than 40 pages. Take journals as the JFE or JF as references how your work should look like.

• Pay attention on writing correctly and watch out for a scientifically, concise, but a fluent writing style. Do also pay attention on a correct use of citation (in the text not full information, only author and year) like commonly used in Journal of Finance or Journal of Financial Economics. Also in other formal aspects use the layout of these two Journals. Only use sufficiently described black-white graphs. Graphs and tables should be self-explanatory and understandable also without reading the whole thesis.

• At the beginning of the thesis write a short abstract (question, result) like in JFE or JF. This abstract is short and even more concise than an Executive Summary and comprises not more than 100 words.

• Create a maximally four-sided Executive Summary (question, procedure, results, and general evaluation), which is typically similar to the introduction of the thesis.

• You deliver three copies of your thesis to the dean’s oﬃce. The Executive Summary and the ordering are directly after the front page of the thesis. Furthermore, you submit 2 CD’s with the title of the thesis, your name, your address and your phone number on it: (1) One CD contains only the front page of your thesis (without your address and phone number), the Executive Summary separately, a combination of these two (front page without your address and phone number together with the Executive Summary) and also your full thesis. Deliver this in pdf and in word or preferably LaTeX; (2) A second CD with the whole thesis in word or LaTeX, the whole used data material, all computer codes to replicate your ﬁndings, and all the electronically saved references (for instance, papers in pdf format).

The Department of Banking and Finance has the rights of your thesis and may use part of it within the scope of science.

IV If you have questions or if you want to discuss your thesis you can contact Jiri Wos- chitz: [email protected]. We wish you good luck with the thesis.

Zurich, 31 January, 2012

Prof. Dr. Kjell Nyborg

V Executive Summary

Problem

Covered Bonds are one of the most important reﬁnancing instruments for banks in Europe and the respective market is one of the biggest private debt markets. Nevertheless, there is very little academic research on this topic. Since each country has adopted its own Covered Bond legislation, Covered Bonds might be treated very strict in one country but very limp in an other. Therefore, this paper compares several legal frameworks and overviews the Covered Bond topic from a theoretical perspective. Additionally, we analytically investigate the historical development of the Covered Bond market in terms of outstanding amount and riskiness.

Method

This thesis is split into three parts. The first part is based theoretical literature such as the Covered Bond Fact Book provided by the European Covered Bond Council (2011). In a first step, we propose a definition for Covered Bonds, then we introduce different issuance structures applied in different countries. Afterwards, based on existing literature, we state differences and convergences between Covered Bonds and mortgage-backed securities (MBS). In the last section we compare different legislations or contractual frameworks. In the second part, we compare the rating methodologies of the three big agencies Fitch, Moody’s and Standard & Poor’s and discuss the single procedures in detail. The last part of the thesis provides a market overview for Covered Bonds. In a first step, we analyze the outstanding amounts of Covered Bonds. Secondly, we analytically compare the development of average weekly Covered Bond prices and yield spreads for three different sub-samples: (i) by country, (ii) by country and collateral type and (iii) over all countries by collateral type. We examine data on 23,363 Covered Bonds which stem from Bloomberg and Datastream over a period from January 2005 until May 2012.

Results

In the first part, the developed definition consists of a list of requirements for Covered Bonds. We conclude, that a Covered Bond needs to be issued by a credit institution and that the size of the dynamic and monitored cover pool has to fit at least the outstanding bonds. Furthermore, the investors have a double recourse and their claims are ensured by protection mechanisms. Then, we inaugurate four different ways to issue Covered Bonds. The four structures are the Direct Issuance, the Pooling Model, the Issance by a specialized Funding Institution and the Issuance using a SPV. We do not find one structure superseding all other since we see advantages and disadvantages in every model. In the next section, we compare Covered Bonds and MBS. We state that MBS benefit from the independence of the issuer, the high transparency of the cover pool and lower market risk. Otherwise, Covered Bonds gain from the strict legislation, the higher quality of the cover pool and the double recourse. When comparing the legal frameworks, we find great distinctions between

VI the Anglo-Saxon countries and traditional European Covered Bond markets. Moreover, we observe clear distinctions between all legislations. We find some convergences when analyzing the methodologies of the rating agencies. All rating agencies rate Covered Bonds on a joint-default basis meaning that the issuer rating is both, the starting position and the lowest achievable rating. The uplift above the issuer rating is calculated differently. Fitch investigates the indipendence of the cover pool from the issuer. Moody’s focuses its analysis on the expected loss and on the probability of timely payments continuing after an issuer’s default and S&P’s approach is to analyze asset-liability mismatches. The first section of the last part examines the outstanding amount of Covered Bonds denominated in euros. Our investigation shows, that the outstanding amount of Covered Bonds backed by public sector loans has declined. The main driver behind that trend is Germany, where several legal amendments reduced the attractiveness to issue public Covered Bonds. On the other hand, the amount of outstanding mortgage Covered Bonds has increased in the last years. This trend is caused by the exploitation of new Covered Bond markets and the expanding amount of mortgage Covered Bonds in the established countries. The second purpose of the last part of the thesis is to explore the performance and the riskiness of the Covered Bond market. We find higher yield spreads for the U.S., Britain and Ireland immediately after the Lehman collapse compared to the other countries. With the beginning of the debt crisis, the spreads for countries deep in debt increased dramatically. For five countries we investigate differences between the certainty of public Covered Bonds and mortgage Covered Bonds. We find obvious differences only in countries which are struggling because of either a debt or a mortgage crisis. Additionally, for the differences of mortgage and public Covered Bonds we make a comparison over all countries. We find slightly higher spreads for mortgage Covered Bonds during the subprime crisis. With the beginning of the debt crisis we cannot observe any obvious differences.

Evaluation

When comparing the legal frameworks, we focus on the, in our view, most relevant criteria. We believe that the main differences between the legislation is caused by those properties. But there may be other differences not considered in this thesis which have a big influence as well. We discuss the procedures of the rating agencies in detail. But even after this study, the transparency of the methods is not fully assured. By making use of an application example, we realize that the assignment of a rating is not fully comprehensible. Since our analysis of the market development is based on analytical methods or descriptive statistics, our findings could be revised by statistical models.

VII Contents

I Introduction 1

II What is a Covered Bond? 3 A Deﬁnition ...... 3 B Comparison of the Covered Bond Structure Variations ...... 4 C ComparingCoveredBondsandMBS ...... 8 D LegalFrameworksindiﬀerentCountries ...... 9

III Covered Bond Rating Methods of the Rating Agencies 13 A ComparisonoftheRatingProcedures...... 13 B Fitch...... 14 C Moody’s ...... 18 D Standard&Poor’s ...... 21

IV Development of the Covered Bond Market 24 A Data ...... 24 B Development of the Outstanding Amount of Covered Bonds ...... 26 C DevelopmentoftheYieldSpreadsandPrices ...... 30

V Conclusion and further Research 36 A Conclusion...... 36 B FurtherResearch ...... 38

VI Appendix 41 A AppendixA-ListoftheFrameworks ...... 41 B Appendix B - Additional Information about the Data ...... 42 C AppendixC-AddtitionalFigures ...... 43

VIII List of Figures

1 DirectIssuancebyaCreditInstitution ...... 5 2 ThePoolingModel...... 6 3 Issuance by a Specialized Funding Institution ...... 7 4 IssuanceusingaSPV...... 8 5 SummaryofStandard&Poor’sRatingProcedure ...... 22 6 AmountOutstandinginEURdividedbyCoreCover ...... 27 7 Amount Outstanding in EUR divided by Core Cover and Countries .... 28 8 Amount Outstanding and New Issuances in EUR by Country ...... 29 9 AverageSpreadsdividedbyCountry ...... 31 10 AveragePricesbydividedbyCountry ...... 32 11 AverageSpreadsbyCoreCoverandCountry ...... 33 12 AveragePricesbyCoreCoverandCountry ...... 34 13 AveragePricesbyCoreCover...... 37 14 PriceDiﬀerencesofCovered Bonds by Country ...... 43 15 AveragePricesbyCoreCover...... 43

IX List of Tables

I Comparison of Legal Frameworks across diﬀerent Countries ...... 12 II ExampletoexplaintheRatingProcedures ...... 13 III OverviewoftheRatingSchemes ...... 15 IV Maximum achievable Covered Bond Rating on a PD Basis ...... 16 V MaximumNotchingaboveCoveredBond’sPDRating ...... 18 VI TPICaps...... 21 VII ExaminedCoveredBonds ...... 25 VIII Average Yield Spreads in the diﬀerent Periods ...... 35 IX ListoftheFrameworks ...... 41 X CompositionoftheSamplebyCountryandCurrency ...... 42

X I. Introduction

A Covered Bond is a corporate bond backed by high quality assets and issued by a credit institution for refinancing purposes. For the investors, a Covered Bond is regarded as a very secure investment in a long-term secured bank debt. To compensate the investor, the credit institution pays a certain amount of interest and reimburses the principal at maturity. Due to the covering of the debt, the default risk of a Covered Bond is lower than for general corporate debt. This evidence is supported by the higher ratings for Cov- ered Bonds compared with unsecured corporate bonds. The certainty of Covered Bonds is impressively demonstrated by the fact that over their entire history no Covered Bond has ever defaulted (see Tran (2011)). But where have these Covered Bonds come from? Schwarcz (2011) states in his paper, that Covered Bonds have their origin in Europe. In the 18th century, the first Covered Bonds (called Pfandbriefe) were issued in Prussia. Later on, Covered Bonds emerged in other European countries, namely Denmark, Poland and France. In the 19th century, nearly all European countries had adopted a Covered Bond system, which was fundamental to maintain the financial systems. From the mid of the 20th century, the importance of the Covered Bond market decreased and other financial instruments replaced Covered Bonds. After the issuance of the first Jumbo Pfandbrief 1 in Germany in 1995, the Covered Bond market experienced a revival and many European countries revitalized their Covered Bond system. In the recent history, other off-balance sheet funding instruments like mortgage- backed securities (MBS) were accused of being jointly responsible for the financial crisis. Thus, the banks need to find other funding instruments. One of the most named solutions are Covered Bonds (Bernanke, 2009). Today, the Covered Bond market is one of the biggest private debt markets. According to the European Covered Bond Council (ECBC)2 (2011) the amount of outstanding Covered Bonds was about EUR 2,500 bn at the end of 2010. The largest Covered Bond market is Germany followed by Spain, Denmark and France. In 2010, Covered Bonds with a volume of over EUR 600 bn were issued by about 300 different credit institutions. The announce- ment of the second Covered Bond Purchase Program (CBBP2) by the European Central Bank (ECB) to stabilize the financial sector points out the importance of Covered Bonds in Europe. The CBPP2 was launched due to the great success of the first CBPP3. With regard to the importance, in euro denominated Covered Bonds consequently account for the biggest share of the market. Covered Bonds are predominately bought by institutional investors such as credit institutions, pension funds or central banks, but there exists also a minority group of small private investors. The research in the area of Covered Bonds is dominated by banks or other non-academic

1 A Jumbo-Pfandbrief (also called Benchmark Covered Bond) is a Covered Bond with an issue size of at least EUR 1 bn. 2 The ECBC is a platform bringing Covered Bond issuer together. The ECBC represents over 95% of the Covered Bond issuers in the European Union (see http://www.ecbc.eu). 3 The CBPP1 had a volume of EUR 60 bn and started in July 2009 and persisted until June 2010. The CBPP2 with a volume of EUR 40 bn started in November 2011 and is expected to be completed in October 2012 (see Beirne et al. (2011)).

1 institutions. The ECBC publishes its Covered Bond Fact Book (often regarded as a standard reference) every year, starting in 2006. They provide extensive descriptions of the different Covered Bond legislations all over the world and present a market overview. Other reports with similar contents are often published by banks, for example by Deutsche Bank (2011) or J.P. Morgan (2011). Due to the size and the importance of the Covered Bond market, the lack of research in academia in this area is quite surprising. The academic research is mostly confined to the biggest Covered Bond markets in Western Europe such as Germany, France or Spain. For other countries, no or only very basic academic research is available. Packer, Stever and Upper (2007) compare the spreads of Covered Bonds across different countries and find that differences in the spreads are only weakly related to the differences in the legal frameworks. A newer research paper of Prokopczuk and Vonhoff (2012) states, that the differences in the legislative frameworks and the development of the real estate market in the single countries have only a small impact on Covered Bond spreads in stable periods, but during periods of economic distress, the effect is significant. Otherwise, general market conditions explain a huge fraction of the spread variation in both periods. Several studies examined the differences between Covered Bonds and securizations like MBS. Schwarcz (2011) finds that Covered Bonds are a valid alternative for securizations. He states that Covered Bonds have lower risk, but on the other hand, the costs of issuing Covered Bonds are higher compared to securizations. Carbo-Valverde, Rosen and Rodriguez-Fernandez (2011) conclude differently. Because Covered Bonds and MBS are issued for different purposes, they bring out that Covered Bonds are unsuitable as substi- tutes for MBS. Gambro et al. (2009) focus more on the U.S. market. They see a rapid development upcoming in the U.S. Covered Bond market because of the higher demand for liquidity for mortgage financing and the need for a higher diversity of funding sources. A few studies analyze the different issuance structures. Stöcker (2011) introduces five different issuance structure models and concludes, that there is not one model which fits all countries. Each country has to apply the model which fits best the country-specific requirements. Lassen (2005) made a similar study and concludes comparably. His analysis is already a couple of years old and cannot be fully taken into account due to the huge changes in the Covered Bond market over the past years. The methodologies of the rating agencies are analyzed by Forster, Heberlein and Sirotic (2012). They describe the procedures of the three big rating agencies and state differences and convergences in their paper. A short summary of the rating methods is also provided by the ECBC (2011) and Klein and Neuberg (2009). This paper proceeds as follows. Part II proposes a definition of Covered Bonds, explains different issuance structures, states differences and convergences between Covered Bonds and MBS and compares legal frameworks. Part III compares the rating methods of the big rating agencies. Each method is then described in more detail. In Part IV, the development of the Covered Bond market over the last years is discussed. Finally, Part V gives the conclusion and sets out proposals for further research.

2 II. What is a Covered Bond?

The purpose of this first part of the thesis is to provide an overview of the Covered Bond topic. To understand further analysis, the term Covered Bond has to be defined. This will be done at the first stage. Secondly, different issuance structures of Covered Bonds are compared and it is shown how several frameworks apply different structures. The third section elaborates differences of Covered Bonds and other structured finance products like MBS. Finally, the last section compares the Covered Bond legislations of several countries for a selection of criteria.

A. Deﬁnition

The next section provides a definition of Covered Bonds and an overview of the criteria a Covered Bond has to fulfill. There is not just one clear definition. The definition proposed in this paper is a compilation of the definitions introduced by Packer, Stever and Upper (2007), ECBC (2011) and Klein and Neuberg (2009). Covered Bonds belong to the category of structured finance, like asset-backed securities (ABS), MBS or collateralized debt obligations (CDO). In general, Covered Bonds are long- term debt securities backed by cash generating cover assets (called cover pool). The interest to the investor gets payed from the general cash flow. In general, the revenues from the cover assets are not passed through to the investors. The cover pool consists of the core cover 4 and some substitute assets. The core cover consists either of mortgage loans or public sector loans. In some European countries, ship and aircraft loans are additionally allowed. The substitute assets are restricted up to a certain limit of the nominal value of the cover pool and consist of various highly rated assets like senior MBS or exposures to credit institutions. The value of the cover pool needs to fit at least the value of the outstanding bonds, in nearly all legislations a certain amount of overcollateralization is required. If a credit institution has issued several Covered Bonds, there is not a single cover pool for every bond, but a large pool covering all outstanding bonds. To ensure the quality of the assets in the pool, the cover pool is dynamic, which means that non-performing assets can be removed from the cover pool and new assets can be added. The cover assets need to fulfill certain quality standards like specific loan-to- value 5 criteria (LTV criteria) for mortgage loans or a specific minimum rating for public sector loans. The geographical scope of eligible assets is often limited. To check the quality and the requirements of the cover pool by law or contract, a cover pool monitor is appointed. The principal duty of the supervisor6 is to check whether the cover pool is sufficient to repay the outstanding bonds and if all assets in the cover pool are eligible. A Covered Bond investor has a claim against the cover pool superior to other creditors (mostly established by a preferential claim by law). The cover pool should at

4 The term core cover is used to precisely deﬁne the asset class which builds the biggest part of the cover pool. We will use the term collateral synonymously. 5 The loan-to-value ratio is calculated by dividing the amount of the loan by the market value of the house. 6 In many countries, the banking supervision authority monitors the cover pool.

3 first be exclusively used to fulfill the claims of the Covered Bond holders and may later be transferred to the regulary insolvency estate. Additionally, Covered Bond holders have a claim against the regulary insolvency estate pari passu or even senior to the other creditors in case the cover assets are not sufficient to fulfill the claims of the investors. Furthermore, the protection mechanisms (often referred to as ring fencing) in case of an insolvency of the issuer need to be clearly specified. For example, whether a third-party management will be appointed to handle the cover pool or whether the assets are transferred to a special purpose vehicle (SPV) in case of an insolvency of the issuer. The issuer, or more precisely the originator 7, of a Covered Bond needs to be a credit institution. Depending on the applied issuance structure (see Part B), the issuer itself does not need to be a credit institution. The originator is either a universal or a completely specialized credit institution8. Depending on the legislation, the originator may issue Cov- ered Bonds itself, in other frameworks Covered Bonds need to be issued by a SPV or the assets need to be transferred to an other credit institution which has the permission to issue Covered Bonds. A specific legal framework (special law-based frameworks) builds in many countries the legal basis. In a few countries, there is no special legislation for Covered Bonds and the Covered Bond system is based on the general law (general law-based frameworks or structured Covered Bonds). We will discuss this issue more closely in part D.

In summary, a Covered Bond needs to fulﬁll the following requirements:

• The originator needs to be a credit institutions, the issuer may be the originator itself or an entity linked to the originator.

• The Covered Bond holders have a preferential claim against the cover pool and full recourse to the issuer (or to the underlying credit institution).

• The cover pool is dynamic and its quality and its compliance with the legal or contractual requirements are ensured by a monitor.

• The value of the cover pool ﬁts at least the value of the outstanding Covered Bonds.

• There are mechanisms which protect the claims of the Covered Bond investors against the cover pool from the claims of other creditors in case of an insolvency of the issuer.

B. Comparison of the Covered Bond Structure Variations

As mentioned in the ﬁrst part, there exist diﬀerent issuance structures which might have an impact on the certainty of Covered Bonds. The next section explains and compares

7 The originator of the assets is the institution which owns the cover assets. In this thesis, we will use the term issuer when we are referring to the credit institution behind the Covered Bond even if the issuer itself is e.g. a SPV. 8 The business area of a specialized credit institution is restricted to operations associated to Covered Bonds. The balance sheet of such an institution is almost exclusively consisting of Covered Bonds and the cover pool.

4 four particular types: The direct issuance by a credit institution (structure 1), the issuance by a specialized funding institution (structure 2), the pooling model (structure 3) and the issuance by a SPV (structure 4). Each structure is briefly described and illustrated by a simplified figure. We do not find a structure which supersedes all other structures. Each one of the four models has advantages and disadvantages.

B.1. Structure 1: Direct Issuance by a Credit Institution

This structure is the classical way to issue Covered Bonds. The direct structure is in particular applied by countries with a long Covered Bond history. For example, Austria (only the framework Fundierte Bankschuldverschreibungen)9, Germany or Spain use the direct structure. Apart from some country-speciﬁc peculiarities, the structure in Figure 1 is applied in general. According to St¨ocker (2011), the separation of the cover assets from the other assets is often problematic (especially whether the issuer is a universal credit institution). Often, in order to tackle this risk, a cover register consisting of every asset belonging to the cover pool clearly segregates the asset. In case of an issuer’s default, the interests of the investors are protected by several mechanisms. First, Covered Bond holders have a preferential claim against the cover pool. Second, there may be additional protection mechanisms like the appointment of a special cover pool administrator or the transfer of the cover assets to a SPV.

Credit Institution

Interest and Principal Investor Cover Pool Covered Bonds Purchase of Covered Bonds

Overcollateralization Cover Pool Monitor Pool Cover

Other Liabilities Other Assets and Equity

Figure 1. Direct Issuance by a Credit Institution. This Figure shows how a Covered Bond is issued directly by a credit institution. In this Figure, the issuer is a universal credit institution which also holds huge volumes of other assets on its balance sheet. Source: Own representation based on Deutsche Bank (2011).

B.2. Structure 2: The Pooling Model

The second structure is explained in Figure 2. According to Lassen (2005), this structure is used in Austria (only the framework Pfandbriefe), France (only the framework

9 A complete list of all considered frameworks can be found in Appendix A.

5 Caisse de Refinancement de l’Habitat (CRH)) and Switzerland (only the framework Swiss Pfandbriefe). ECBC (2008) states as the main difference to the direct structure, that several credit institutions (originators) do not issue Covered Bonds themselves, but they transfer10 the cover assets to a legally separate credit institution (centralized funding institution), whose only purpose is to issue Covered Bonds. In most cases, the cover assets remain on the balance sheet of the originator and the investors still have the dual recourse to the cover pool and the originator. Pooling models may lead to a more diverse cover pool and to a bigger issuance volume. A risk of this method is a potential monopolization and the political influence if there are only a few but very big issuers.

Participating Credit Institutions

Secured Liabilities Secured Liabilities Eligible Assets Eligible Assets

Overcollateralization Overcollateralization

Other Liabilities Other Assets Other Assets Other Liabilities and Equity and Equity

Secured Liabilities Secured Liabilities Eligible Assets Eligible Assets

Overcollateralization Overcollateralization

Other Liabilities Other Assets Other Assets Other Liabilities and Equity and Equity

Transfer of Cover Assets Centralized Funding Institution Interest and Principal Investor Cover Pool Covered Bonds Purchase of

Loan Monitor Pool Cover Covered Bonds

Figure 2. The Pooling Model. This Figure shows how a pooling model may work. The speciﬁc structure in this Figure is based on the Swiss Pfandbriefe framework. Source: Own representation based on Horath (2007).

B.3. Structure 3: Issuance by a Specialized Funding Institution

This structure is applied, for example, in France (only the framework Obligations Fronciéres), Ireland or Norway. Stöcker (2011) characterizes this structure as follows: A credit institution (also called parent bank) transfers the cover assets to a specialized funding institution, which has a banking license and its only purpose is to issue Covered Bonds. Thus, the structure is quite comparable to the pooling model. The servicing of the cover assets is done by the parent bank and therefore, the issuer itself has nearly no staff. The separation

10 In this context, the term ’transfer’ is used as an umbrella term for every possible transfer technique, because the transfer mechanisms vary strongly across country.

6 of the cover assets from the regular assets is fundamental in this model. Because the issuer only holds cover assets and Covered Bonds on its balance sheet, the question arises whether the issuer is able to keep up the timely payments to the investors in case of a default of the issuer. The dual recourse to the cover assets and the issuer is still guaranteed.

Credit Institution

Secured Liabilities Eligible Assets

Overcollateralization Other Liabilities and Equity

Other Assets

Loan Transfer of Cover Assets Specialized Funding Institution Interest and Principal Investor Cover Pool Covered Bonds Purchase of

Cover Pool Monitor Pool Cover Covered Bonds

Figure 3. Issuance by a Specialized Funding Institution. This Figure shows how a possible issuance by a specialized funding institution may work. The issuer is a specialized institution with a banking license. In some frameworks, the issuer may hold some other assets on its balance sheet. Source: Own representation based on St¨ocker (2011).

B.4. Structure 4: Issuance using a SPV

Figure 4 shows how an issuance of a Covered Bond using a SPV works. The structure is applied in countries without a specific legal framework (e.g. Canada or U.S.) and in some European countries (e.g. Italy or UK) where this structure is stipulated by law. Stöcker (2011) states, that the attainment of insolvency segregation of the cover assets is the main reason to apply this structure. The cover assets will be transferred to a legally separate SPV (which has no banking license). According to ECBC (2011), Covered Bonds are still an on-balance sheet funding instrument, because the credit institution remains the originator of the cover assets. The Covered Bonds may be issued by the SPV (currently done by the U.S. and the second proposed structure by the Department of the Treasury (2008)) or the SPV guarantees the payments of the interest and the principal and the Covered Bonds remain on the balance sheet of the credit institution (see figure 4), like it is currently done in Italy or in the UK. The investor still has double recourse11 to the cover pool and the credit institution. According to Lassen (2005), advantages of this structure are the high degree of specialization and the easy segregation of the cover assets from the other assets. Disadvantages of this structure are, for example, that the SPV has no banking license and the low specialization of the underlying credit institution.

11 The current issued Covered Bond programs in the U.S. do not include the dual recourse.

7 Credit Institution Interest and Loan Principal Investor Outstanding Accounts Covered Bonds Purchase of Covered Bonds Overcollateralization

Other Liabilities Other Assets and Equity

Transfer of Cover Assets Special Purpose Vehicle

Guarantee Cover Pool Equity and Liabilities Cover Pool Monitor Pool Cover

Figure 4. Issuance using a SPV. This ﬁgure shows how an issuance using a SVP may work. In the structure shown here, the Covered Bonds itself remain on the balance sheet of the credit institution and the SPV only guarantees the payments like is currently applied in Italy or the UK. Source: Own representation based on St¨ocker (2011).

C. Comparing Covered Bonds and MBS

Part B describes the differences in the issuance structures. The fourth structure using a SPV to issue Covered Bonds, raises the question, in what properties a Covered Bond differs from an off-balance sheet instrument like a MBS. The following section tries to answer this question. If we look at the historical roots, we see that MBS emerged in the U.S. and the UK based on general law (see Deutsche Bank (2011)). Otherwise, Covered Bonds have their origin in Europe, in particular in Germany and Denmark and are often stipulated by specific legislations. A first important difference between Covered Bonds and MBS is the motivation of the issuer. According to Packer, Stever and Upper (2007), the motivation to issue Covered Bonds is refinancing. Otherwise, MBS are primarily used for risk or capital reduction. The risk and capital reduction is achieved through a true sale of the cover assets to a SPV, which is legally separate from the issuing bank as Carbo-Valverde, Rosen and Rodriguez- Fernandez (2011) state in their paper. Therefore, the assets are removed from the balance sheet of the issuer. On the other side for Covered Bonds, the cover assets remain in general on the balance sheet. The true sale of the cover assets to a SPV implies a greater indipendence from the issuing credit institution for MBS compared to Covered Bonds. Thus, according to ECBC (2011), the correlation between the credit quality of the cover pool and the credit quality of the credit institution is lower. Another implication is, that MBS generally have a pass-through structure unlike Covered Bonds, which apply mostly a fixed-rate bullet struc-

8 ture12. Consequently, the market risk for Covered Bonds is higher and therefore, a higher overcollateralization is required for Covered Bonds. Klein and Neuberg (2009) argue, that the greater indipence from the issuer has some negative implications: The cover pool of a MBS is static, which means that non-performing assets cannot be removed. Otherwise, the cover pool of Covered Bonds is dynamic and non-performing assets can be removed from the cover pool. As a result, on the one hand, the transparency of the assets in the cover pool is higher for MBS (the cover assets will remain in the pool), on the other hand, the quality of the cover pool of Covered Bonds is ensured by a monitor. Another implication of the dynamic cover pool is a typically longer maturity for Covered Bonds. The restrictions on the collateral are different for MBS and Covered Bonds. For MBS, the type and the quality of the cover assets is generally not restricted by law, for Covered Bond, in every legal framework (and even in the general-law based programs) there are some criteria which the cover assets have to fulfill. An interesting fact is, that (senior) MBS are often eligible as cover assets for Covered Bonds. According to Klein and Neuberg (2009), the payments to the investors of a MBS are only sourced by the assets in the cover pool. If these assets are insufficient to make the payments, the lowest tranches (junior tranches) start suffering losses. There is no recourse to the issuer of the MBS. Covered Bonds, on the other hand, may only become insolvent due to an issuer’s default. Only in this case the cover assets are needed to fulfill the claims of the investors. If these assets are insufficient to repay the investors, the Covered Bond holders still have recourse to the issuing credit institution. Summarizing, the advantages of MBS are the independence from the issuer, the lower market risk and the high transparency of the cover pool. The advantages of Covered Bonds are the dynamic cover pool, the double recourse and the strict monitoring.

D. Legal Frameworks in diﬀerent Countries

The next section provides an overview of the diﬀerent legal frameworks in a selection of countries. Firstly, it is explained, why these particular countries were chosen, secondly, the criteria listed in Table I will be explained. The description of the single frameworks is based on ECBC (2011).

D.1. Selection of Countries

The ﬁrst criterion of the selection of the countries is the importance of the covered bond sector in each country. Thus, we include countries like Germany, Austria or France. Due to the popularity of oﬀ-balance sheet structures like MBS in the U.S. or in the UK, the Covered Bond market began to grow only a few years ago in those countries. In order to take this into account, some well-established countries with a long tradition of Covered

12 A pass-through structure means, that the cash flows generated by the cover assets are distributed to the investors. A bullet structure, on the other side, stands for a fixed interest rate which does not depend on the cash flows of the cover assets.

9 Bonds are compared with countries with a very young or without Covered Bond legislation. The only countries in the sample without an adopted Covered Bond legislation are the U.S. and Canada. For Canada, there is a proposed legislation which can be taken into consideration. The facts about the U.S. Covered Bonds are based on the two already issued Covered Bond programs (U.S.1 in Table I). Besides this program, the proposed best practices for residential Covered Bonds for the U.S. published by the Department of the Treasury (2008) is also considered in the comparison (U.S.2 in Table I). In other countries, for example Denmark or Luxembourg, the amount of the issued Covered Bonds is very high compared to the size of the economy and therefore the Covered Bond market has systemic relevance for these countries. We include Switzerland in the analysis because it is the only country where only the pooling model is stipulated by law. Another interesting point to see is, how countries, which are currently struggling because of the debt crisis do handle Covered Bonds in their legal frameworks. For this reason, Italy, Greece and Portugal are included in the analysis. If there is more than one framework in a country, we choose the frameworks with the highest importance or frameworks which allow for mortgage loans as core cover. In some countries, the different stipulated frameworks differ only in the allowed core cover asset type (e.g. Luxembourg or Portugal). In other countries, the frameworks are completely different, for example in terms of issuance structure, LTV criteria or overcollateralization (e.g. Austria, France or Switzerland). There are many countries with an adopted legislation which are not considered in this brief comparison. In some countries, only very few or even no Covered Bonds has been issued by the banks (e.g. Turkey or New Zealand). Other countries are very small and the size of the Covered Bond market is limited (e.g. Latvia or Cyprus). Naturally, also a comparison of countries with a small Covered Bond market with other countries may be interesting, but the focus in this comparison lays on countries and frameworks with high importance.

D.2. Comparison of the Legal Frameworks

All criteria have been chosen with view to their effect on the difference of the legislations and their impact on the security of the Covered Bonds. There is a huge amount of other criteria which may be compared, but, in this paper the comparison is restricted to the criteria listed in the Table I. The next section analyzes, how the Covered Bond legislations differ across country. It can be read from Table I, that the issuance structures vary between the countries. In most of the European countries, structure 1 (see Part B) is applied. The U.S. and Canada along with Italy and Greece use structure 4 (the SPV structure). The best practices guide for the U.S. proposes either structure 1 or 4. Switzerland applies the pooling model (structure 2) and Structure 3 is stipulated by the French and the Norwegian Covered Bond law. In many countries, public sector loans are allowed as well as mortgages to build the core cover. Some legislation allow a hybrid core cover consisting of public sector loans and mortgages (e.g. France). The range of LTV criteria is quite wide across country (between

10 60% and 90%). The requirements on commercial mortgages (LTV cap of 60% in most of the jurisdictions) are generally higher than the requirements on residential mortgages. The LTV limits are calculated either by market value or by mortgage lending value. An interesting fact is that some legislation require removing a mortgage loan if it exceeds a certain LTV cap. Substitute assets are allowed in all legislations in the sample, except Switzerland. In the majority of the selected countries, exposures to credit institutions and/or (senior) ABS or MBS are permitted in the cover pool as substitute assets. The ABS/MBS and the credit institutions often need to have a certain minimum rating. In nearly all legislations the overcollateralization is required by law. In the UK, the overcollateralization is only required by contract, exactly like in the U.S. and Canada due to the fact that they do not have a legal framework. The minimum overcollateralization seems quite low, but it is only seen as a lower boundary in order of a higher level for individual Covered Bond programs implemented by contract. The monitoring of the cover pool is mostly performed by an independent monitor. In Finland and Denmark, the monitor is appointed by the credit institution itself and has to report to the financial services authority. The duties of the cover pool monitor vary greatly among the countries. Another important point is, how the protection of the Covered Bond holders against the claims of other creditors is ensured. In most instances, the Covered Bond investors have a preferential claim against the cover assets. To ensure the separation of the cover assets, a few legislations require an appointment of a specific cover pool administrator or the transfer of the cover assets to a SPV additionally. In every legal framework, except the U.S.13, the double recourse to the cover assets and the regulary insolvency estate (pari passu or even senior) is ensured. The framework of Switzerland does not need a protection against the claims of other creditors because the issuing institution does not have any other creditors. The table shows that there is not just one Covered Bond type in the world. Each single country has its own legislation. Some of the frameworks are very close (e.g. all Scandi- navian Countries or Luxembourg and Germany are quite similar), others differ in many specifications. Especially the difference between the Anglo-Saxon and the traditional Eu- ropean Covered Bond markets is remarkable. The question arises, whether a Covered Bond legislation at European level would be advantageous to increase the transparency. A Factor opposing this idea is the fact that the collateral assets are mostly located in the respective country and therefore the frameworks should be customized to meet the country-specific requirements. The minimum requirements of the Directive on Undertakings for Collective Investments in Transferable Securities14 (UCITS, Article 52(4)) is one step in this direction.

13 The best practices guidelines recommend a recourse to the insolvency estate pari passu to the other creditors. 14 The UCITS is an organization of the European Union. The goal of the organization is to standardize the regulation for transferable securities (see UCITS(2009).

11 Table I Comparison of Legal Frameworks across diﬀerent Countries

This Table compares diﬀerent legal frameworks. The categorization is based on the description of the legislations provided by the ECBC (2011). Source: Own representation.

1 2 Country AUT CAN DNK FIN FRA GER GRE ITA LUX NOR POR SPA SWE SUI UK U.S.1 U.S.2

Speciﬁc Legislation Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 3 Applied Structure 1/(2)4 1 1 3 1 4 4 1 3 1 1 1 2 4 4 1/4

Mortgages Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Core Cover 4 Public Sector Loans Yes S Yes Yes Yes Yes Yes S Yes Yes S Yes Yes Yes S

Residential in % 60 80/90 75/80 70 60/80 60 80 80 60 75 80 80 75 80 60/75/80 75 80 LTV Limits Commercial in % 60 60/70 60 60/80 60 60 60 60 60 60 60 60 60/75 5 Calculation Method MLV MV MV MV MLV MLV MV MV MLV MV MV MLV MV MV MV other n/a LTV cap which require the Mortgage to be removed? Yes Yes Yes Yes Yes Yes Yes Yes Yes

Substitute Exposure to Credit Institutions Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Assets ABS/MBS Yes Yes Yes Yes Yes Yes Yes Yes Yes Limit of Substitute Assets in % (nominal) 15 10 15 20 10 10 15 15 20 15 20 5 20 0 10 10 n/a

12 Overcollatera- Minimum in % 2 3.09 8 2 2 2 5.26 Yes 2 Yes 5.3 Yes Yes 11 10 Yes 5 lization required by Law Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Independent Cover Pool Monitor Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Duties Performing Audits of the Cover Pool Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes of the Performing Duties of Supervision Authority Yes Yes Yes Yes Yes Yes Yes Yes Yes Monitor Veriﬁcation of the Coverage Tests Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Protection Transfer of Assets to SPV Yes Yes Yes 6 6 against Claims Preferential Claim Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes other other Yes of other Speciﬁc Cover Pool Administrator Yes Yes Yes Yes Yes Yes 7 Investors Recourse to the Insolvency Estate PP PP SEN PP SEN PP PP SEN PP PP PP PP PP PP PP Yes

1 For countries with several frameworks, the following legislations are considered: AUT: Pfandbriefe, FRA: Obligations Fonci´eres, LUX: Lettres de Gages, POR: Mortgage Covered Bonds, SUI: Swiss Pfandbriefe, US1: currently applied structure, U.S.2: recommended framework by the Department of the Treasury. 2 The content in this Table is based on the Best Practice Guide for Residential Covered Bonds published by the Department of the Treasury (2008). The abbreviation ’n/a’ stands for not available (if this criteria is not explicitly speciﬁed in the draft). 3 The numbers in this row refer to the numbering of the issuance structures in Part B. 4 The abbreviation ’S’ stands for Substitute Asset. 5 The abbreviations ’MV’ and ’MLV’ stand for Market Value and Mortgage Lending Value. The U.S. applies an other method to valuate mortgage. 6 Switzerland applies the pooling model, where the separation of the cover assets and the cover pool administrator is assigned before insolvency. The U.S. does not apply any of the protection mechanisms. 7 The abbreviations ’SEN’ and ’PP’ denote senior and pari passu to the other creditors. III. Covered Bond Rating Methods of the Rating Agencies

In the following sections, the rating methods of the three big rating agencies Fitch, Moody’s and Standard & Poor’s (S&P) will be compared and discussed in detail. To illustrate and explain those methods, the same simple and ﬁctive example which is described in Table II will be used. For reasons of simplicity, only a few important features of a Covered Bond are presented in the example.

Table II Example to explain the Rating Procedures

This Table shows a ﬁctive example of a Covered Bond. Only a few very important properties are listed in the Table. This example is later used to illustrate how the rating agencies proceed in their rating methodology. Source: Own representation.

Criteria Description IssuingBank GermanMortgageBank,SpecializedBank Issuer Rating A by Fitch and S&P, A1 by Moody’s LegalFramework GermanPfandbriefact CoreCover CommercialandResidentialMortgages Overcollateralization 2%byLaw,10%byContract LTVLimits 60%forCommercialandResidentialMortgages Geographical Scope Germany exclusively PreferentialClaimbyLaw Yes Special Cover Pool Administrator Yes Additional Protection Mechanisms Cover Register, No Transfer to SPV

A. Comparison of the Rating Procedures

The next short section compares the rating procedures of the rating agencies15. After dis- cussing the evolution in the rating methods, both, the convergences and the differences in the methodologies are discussed. In the past years, some big changes in terms of rating methodologies were made. If we take a look at the rating procedure of S&P for German Pfandbriefe from 199716, the only criterion determing the rating is the cover pool. The analysis of the cash flows, interest rate risks and the overcollateralization determine the rating. A concrete link between the issuer and the Covered Bond is not considered. This Structured Finance approach was especially used by S&P’s. Moody’s changed earlier to a joint-default approach by considering the issuer rating. For the Structured Finance approach an issuer can improve the rating by upgrading the quality of the cover pool or by eliminating asset-liability mismatches. The biggest influence in the joint-default approach has the issuer default rating. Therefore, the rating may be improved by lowering the general risk or by increasing the capital. Nowadays, all big rating agencies assign a strong

15 A detailed study on the rating agencies has been done by White (2010) or de Haan and Amtenbrink (2011). 16 The report Covered Bonds: Criteria For Rating German Pfandbriefe is available online on www.sandardandpoors.com.

13 linkage between the issuer and their respective Covered Bonds (see Golin (2006)). At first sight, the rating methodologies of the three big agencies seem very different. But there are certain similarities in the three approaches. One common aspect is, that all agencies assume that there is no loss to the investors if the issuer remains solvent. Another common property is the connection between the issuer rating and the rating of the Covered Bonds. All three agencies assign no rating below the (senior unsecured) credit rating of the issuer. Furthermore, all rating agencies calculate in some way a rating uplift of a certain number of notches for each program. Therefore, a boundary is defined, where the final rating will be stated. The way how they compute the rating uplift is quite different. Fitch’s approach is to analyze the dependence of the Covered Bond from the issuer. A high indipendence of the probability of default (PD) of the Covered Bond from the issuer’s creditworthiness is considered positive. The rating is then influenced by the sufficiency of the overcollateralization and the recovery prospects of the collateral. Moody’s on the other hand, focuses mainly on the analysis of the expected loss which is determined by the credit strength of the issuer and the value of the cover pool. The second step in their rating process is comparable with Fitch’s first step, since they analyze the probability of the continuation of the payments to the investors after an issuer’s default. The third agency is S&P. Their first step is quite similar to Fitch’s first and Moody’s second step because they analyze the separation of the Covered Bond from the issuer and assume, that the cover assets need to be sufficient to satisfy the claims of the investors. This assessment is done by their analysis of the asset-liability mismatch. A special point in S&P’s procedure is their assumption, that a government may intervene in case of an issuer’s default to protect their economy whether the Covered Bond market in the respective country is of systemic relevance. Furthermore, it is interesting to see how stringent the single rating approaches are. Ac- cording to ECBC (2011), S&P provides the greatest uplift to the issuer rating, Moody’s the smallest. For example, all triple-A rated Covered Bonds by Moody’s and Fitch have an issuer default rating of at least A1 and A+, respectively. On the other side, S&P may assign AAA for a Covered Bond of a A- rated issuer. To explain the ratings which may be assessed, Table III provides an overview of a view important rating examples and their qualitative and quantitative meaning.

B. Fitch

The following analysis of Fitch’s rating methodology is based on their report Covered Bond Rating Criteria. For the rating of Covered Bonds Fitch mainly analyzes the PD and loss given default. The rating process can be illustrated as a three step process. In the ﬁrst step, the agency calculates the risk of an interruption of payments upon an issuer’s default to determine the maximum achievable rating on a PD basis. Secondly, Fitch performs a stress testing of the overcollateralization which may reduce the rating reached in the ﬁrst step. In the last step, the recovery prospects of the cover pool are measured. This may

14 Table III Overview of the Rating Schemes

This Table shows of the meaning of diﬀerent important ratings. Only Moody’s assigns explicitly a number to the ratings in terms of expected loss and PD. In this Table, the idealized cumulative PD and expected loss are stated for ﬁve years. The qualitative description is taken from Fitch (2011), but it is also valid for the other agencies. Source: Own representation based on Fitch (2011), Moody’s (2010) and S&P (2009).

Fitch Qualitative Moody’sIdealized Moody’sIdealized and Moody’s Description CumulativePD CumulativeEL S&P (in%,5years) (in%,5years) AAA Aaa Highestcreditquality 0.003 0.002 AA Aa2 Veryhighcreditquality 0.068 0.037 A A2 Highcreditquality 0.467 0.257 BBB Baa2 Goodcreditquality 1.580 0.869 BB Ba2 Speculative 8.410 4.626 B B2 Highlyspeculative 20.710 11.391 CCC Caa2 Substantialcreditrisk 48.750 26.813

aﬀect the rating in a positive or negative way. However, a Covered Bond program achieves at least the issuer default rating (IDR). The next section focuses on each step in more detail.

B.1. Step 1: Analysis of the Discontinuity Factor (D-Factor)

The D-Factor measures the risk of an interruption of payments to the investors after the issuer got insolvent. The D-Factor is measured on a scale from 0% to 100%. A D-Factor of 0% expresses complete indipendence of the PD of the Covered Bond from the issuer’s creditworthiness. On the other side, a Discontinuity-Factor of 100% stands for an automatic default of the Covered Bond upon an issuer’s default. Table IV shows how Fitch applies the D-Factor in combination with the IDR to determine the maximum achievable rating on a PD basis. Fitch names four key factors which inﬂuence the continuity of Covered Bond payments given the default of the issuer. The D-Factor is calculated as a weighted average of the following four criteria:

• Asset Segregation (45% weight) Fitch analyzes the strength of the asset segregation mechanisms. On the one hand, Fitch measures how effectively the cover assets are separated from the claims of the other creditors and if the excess overcollateralization is immune to claims of other creditors. On the other hand, Fitch investigates, if assets in the cover pool are subject to potential claw-back risks or if the cover pool’s cash flows mix up with the issuer’s other cash flows.

• Liquidity Gap (35% weight) Only in limited circumstances, the cash ﬂows of the cover assets do exactly match the Covered Bond payments on a timely basis. The agency compares the time needed to sell the underlying cover assets in a stress

15 Table IV Maximum achievable Covered Bond Rating on a PD Basis

This table shows how Fitch indicates the maximum rating achievable on a PD basis through the combination of the likelihood of default with the relevant IDR and the D-Factor for a given Covered Bond program. The ﬁrst column indicates the IDR and the ﬁrst row shows a selection of potential D-Factors. Source: Fitch (2011).

D-Factors IDR 100% 70% 60% 50% 40% 30% 20% 10% 0% AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA AA+ AA+ AA+ AAA AAA AAA AAA AAA AAA AAA AA AA AA+ AA+ AA+ AA+ AAA AAA AAA AAA AA- AA- AA AA AA+ AA+ AA+ AAA AAA AAA A+ A+ AA- AA- AA- AA AA AA+ AAA AAA A A A+ AA- AA- AA AA AA+ AAA AAA A- A- A A+ A+ AA- AA- AA AA+ AAA BBB+ BBB+ A- A A+ A+ AA- AA AA+ AAA BBB BBB BBB+ BBB+ A- A A+ AA- AA AAA BBB- BBB- BBB BBB BBB BBB BBB+ A AA- AAA BB+ BB+ BBB- BBB- BBB- BBB BBB BBB+ A AAA BB BB BB+ BB+ BBB- BBB- BBB BBB A- AAA BB- BB- BB BB BB+ BB+ BBB- BBB BBB+ AAA B+ B+ BB- BB BB BB+ BB+ BBB- BBB AAA B B B+ BB- BB- BB BB+ BBB- BBB AAA B- B- B B+ BB- BB- BB BB+ BBB- AAA CCC(+)CCC(+)B- B B+ BB- BB- BB BBB- AAA

situation (macro level) to the time granted by the protection mechanisms of the program (cover pool level).

• Alternative Management (15% weight) The claims of the Covered Bond investors will be processed by a third-party manager upon an insolvency of the issuer. Fitch studies the legal framework or the contractual clauses to evaluate (i) how long it would take to appoint a substitute manager, (ii) if there are any interest conﬂicts of the administrator (e.g. if there is not only a single administrator for the Cov- ered Bond investors but an administrator for both, the secured and the unsecured creditors) and (iii) if the substitute manager has all the powers to take the neces- sary actions (e.g. sell the cover pool or borrow in order to make timely payments to investors).

• Covered Bond Oversight (5% weight) The domestic banking authorities may play an important role in the Covered Bond monitoring. Especially if the Covered Bond market is systematically important, Fitch beliefs that the supervision authorities will monitor the Covered Bond market more closely. The rating agency analyzes the monitoring of the Covered Bond issuers through the regulation authority.

For the example from the beginning of this section (see Table II), Fitch would assign a low D-Factor (e.g. 10%), because of the strong asset segregation mechanism (preferential claim by law and the cover register), the clearly deﬁned third-party management and the

16 enormous size of the German Covered Bond market. In combination with the issuer rating of A, the maximum rating on a PD basis would be a AAA.

B.2. Step 2: Stress-Testing Overcollateralization

The Covered Bond rating on a PD basis from the first step results in a range of possible PD ratings for the Covered Bonds. In the second step Fitch selects the PD rating corresponding to the highest level of stress that the cover pool (including overcollateralization) can withstand. The overcollateralization is seen as a buffer to protect the Covered Bond holders against credit and market risk. To test how sufficient the overcollateralization can avoid a default of the Covered Bonds, the agency analyzes the quality of the cover assets regarding the asset’s credit risk, the maturity mismatch, the interest rate and currency risk. Additionally, the agency tests how much of voluntary overcollateralization above the level required by law can be taken into account. Fitch may not give full credit to the available overcollateralization if there are no contractual commitments or public statements. If the issuer has a high short-term rating (at least F2 ), Fitch considers the lowest level of overcollateralization observed during the preceding 12 months. Below this rating, the agency takes only the legal minimum overcollateralization into consideration. Fitch performs an iterative stress testing of the overcollateralization, which starts from the assumption of the issuer’s default. The cash flow model of the rating agency compares the cash flows from the cover pool with the required payments to the Covered Bond holders in a wind-down scenario. Fitch assumes, that the cover pool is administrated by an alternative management, that no further assets can enter the cover pool and that no further Covered Bonds are issued. The assumed cost of an alternative manager is also taken into account. The first stress scenario runs on the highest achievable PD-rating (determined in step 1). In case of an insufficient overcollateralization, the analysis is reiterated on the next lower rating level, down to the IDR. After this second step, the Covered Bond program will reach a rating between the IDR and the maximum achievable rating on a PD basis. For the example stated in Table II, the contractual overcollateralization of 10% would be taken into account and the first stress scenario runs on the AAA level. For illustrative reasons, we assume that the overcollateralization resists the stress testing on the AA level.

B.3. Step 3: Calculation of the Recovery Uplift

In case of an issuer’s default, the Covered Bond program may beneﬁt from recoveries of the cover pool. If the Covered Bond is rated at least investment grade (BBB) after step 2, there may be a potential uplift up to two notches and if the bond is rated below BBB, the uplift is limited to three notches (see table V). The recovery prospects of the cover pool are scaled into six categories, from poor to outstanding. If the recovery prospects are rated poor or below average the rating on a PD basis can be reduced. After this third step the ﬁnal rating is determined. We assume for illustrative purposes, that for the example in table II Fitch assigns a recov-

17 ery rating of RR2, which implies a rating uplift of one notch and therefore, the ﬁnal rating of this program would be AAA.

Table V Maximum Notching above Covered Bond’s PD Rating This table shows how Fitch indicates the maximum notching above the Covered Bond’s rating on a PD basis. Each rating is assigned with certain recovery prospects and a historical recovery range. This results in a maximum notching above or below the rating on a PD basis for both, Covered Bonds rated below or above investment grade. Source: Fitch (2011).

RecoveryRatings RecoveryProspects Range InvestmentGrade Speculative Grade RR1 Outstanding 91-100 2 3 RR2 Superior 71-90 1 2 RR3 Good 51-70 1 1 RR4 Average 31-50 - - RR5 BelowAverage 10-30 -1 -1 RR6 Poor 0-10 -1/-2 -2/-3

C. Moody’s

Moody’s applies a two-step process to rate Covered Bonds. In the ﬁrst step, the agency uses a quantitative model called Moody’s EL Model (EL stands for expected loss). Moody’s assumes that there will be no loss to the investors as long as the issuer meets its payment obligations. In case of an issuer’s default, the model starts to investigate the value of the cover pool. The second step of the analysis focuses on the probability of timely payments continuing on the Covered Bonds following an issuer’s default. Every Covered Bond program will be assigned with a timely payment indicator (TPI) cap, which determines the maximum achievable rating. Most of the programs are solely rated by Moody’s EL Model because the TPI cap for most programs is Aaa.

C.1. Step 1: Moody’s EL Model

The model calculates on a monthly basis the expected loss to the Covered Bond investors by multiplying the probability of an issuer’s default with the Covered Bond loss upon such a default. The credit strength of the issuer and the value of the cover pool are the major determinants of this ﬁrst step and are discussed further below.

C.1.1. The Credit Strength of the Issuer Moody’s assumes, that the probability of a default on the Covered Bond is not higher than the PD of the issuer. The PD is based on the issuer’s senior unsecured rating. Additionally, issuer-related beneﬁts are taken into account (e.g. a dynamic cover pool managed to the beneﬁt of the investors). If the issuer has no public rating from Moody’s (e.g. if the issuer is a SPV), the issuer rating is based

18 on the rating of the originator (owner bank) and the strength of the linkage between the issuer and the originator17.

C.1.2. Value of the Cover Pool Moody’s names three key factors which determine the value of the cover pool. These three factors are discussed below.

C.1.2.1. Credit Quality of the Cover Pool In Moody’s view, the credit quality of the cover pool is determined by the amount of losses on the cover assets that will accrue after a default of the issuer. The agency measures the quality of the cover pool by the collateral score. The collateral score describes how much credit enhancement is needed in order to reach a Aaa rating. The lower the collateral score the better the quality of the cover assets. For mortgage-backed Covered Bonds the collateral score is determined by the aﬀordability underwriting guidelines, the LTV criteria and the quality of the valuation. For public sector Covered Bonds, Moody’s names the credit strength of the public sector entities and the concentration in the cover pool (both, geographical and borrower concentration) as key criteria determining the collateral score. As Moody’s recalculates the collateral score for most programs on a quarterly basis, changes in the quality of the cover pool caused by adding or removing cover assets are monitored.

C.1.2.2. Refinancing the Cover Pool Because the maturity of the cover assets is often longer than the maturity of the Covered Bonds, the repayment of the principal may rely on funds being raised against the cover pool in case of an issuer’s default, if the natural amortization is not sufficient to repay the principal. In most cases, the issuer has to allow for a discount to the assumed value which creates refinancing risk. The agency names the refinancing margin (the level of discount), the portion of the cover pool exposed to refinancing risk (normally 50% is considered as minimum) and the average life of refinancing risk (average remaining life of the cover pool at the time of issuer default is five years) as the three components which determine the refinancing risk of the cover pool.

C.1.2.3. Interest Rate and Currency Mismatches After a default of the issuer, Covered Bond investors may be exposed to interest rate and currency mismatches which are caused by diﬀerences in the duration or in the payment promises made on the cover pool and the Covered Bonds. Moody’s determines the level of the interest rate and the currency risk by the following three criteria: (i) the size of the interest rate or currency movement, (ii) the portion of assets with interest rate or currency mismatches and (iii) the average life of the mismatch (only in case of interest risk). The collateral score, the reﬁnancing risk and the interest rate and currency mismatches are summed up. The resulting number (in percent) is used to calculate the expected loss.

If we apply the ﬁrst step of Moody’s Covered Bond rating procedure on the example

17 There is another report of Moody’s called Moody’s Approach to Rating Financial Entities Specialized in Issuing Covered Bonds which focuses more on this point and can be found on wwww.moodys.com.

19 in Table II, Moody’s takes first the rating of the issuer into account (A1 ). Secondly, the agency analyzes the value of the cover pool. The strict LTV criteria, the high overcollateralization and the non-existent currency risk affect the rating positive. A high geographical concentration of the cover assets has a negative effect on the rating. We assume a rating based on the EL Model of Aa1.

C.2. Step 2: Moody’s Timely Payment Indicators

The TPIs measure the likelihood that the timely payment of the interest and the principal would continue after an issuer’s default. Every Covered Bond may reach a TPI between Very Improbable and Very High. As long as the issuer is solvent, the timely payments should be made on the Covered Bonds. In case of an issuer’s default, the timely payments to the Covered Bond investors rely no more on the issuer but on external support, liquidity or the legal/contractual framework. Moody’s names the refinancing risk (which is described above) as most important risk to timely payments. But also some other potential risks to timely payments are named. The continuity of servicing and cash management, the risk that Covered Bonds will accelerate after an issuer’s default or the uncertainty whether the Covered Bond law is sufficient to guarantee timely payments are examples for potential risks. Moody’s mentions, that the TPIs vary from program to program, but the following five drivers behind the TPIs are generally consistent across jurisdictions. The first factor determines the base TPI which will be adjusted by the following four factors. These five factors are considered in turn below:

• Factor 1 Strength of the Legislation/Contract Moody’s evaluates the legislation or the contractual clauses on which the Covered Bond program is based. Various properties of a legislation are considered, e.g.: requirements for NPV tests, structure of the bond (bullet or pass-through), rights of the cover pool administrator or the minimum reﬁnancing period.

• Factor 2 Hedging In Moody’s view, hedging can aﬀect the Covered Bond program positive or negative. From a timely payment perspective, a swap may be positive because it may act as a timely payment buﬀer for the program. On the other side, swaps may not survive an issuer default or the swap counterparty could prevent the sale of the assets or liabilities. Moody’s may also consider a swap provided by the issuer (or an entity in the issuer’s group) as negative for TPI purposes.

• Factor 3 Types of Assets The main criteria to assess the quality of the cover assets is how likely the cover assets are expected to be sold in the future upon an issuer default. Whether Moody’s believes that the assets cannot or only highly unlikely be sold in between a certain time period, they assess a TPI of Very Improbable irrespective of the other features.

• Factor 4 Nature of Liabilities This fourth factor analyzes which kind of bond is

20 issued. In Moody’s view, a pass-through bond is less risky than a bullet bond because the reﬁnancing risk is more limited.

• Factor 5 Other Factors Moody’s takes also some other factors into account to assess the TPIs. For example, they consider the sovereign rating, informal timely payment arrangements, the correlation between the performance of the issuer and the cover pool, an additional overcollateralization and whether the issuer is rated sub-investment or low investment grade.

Table VI TPI Caps This table shows how Moody’s indicates the maximum rating achievable (TPI cap) based on a combination of the issuer rating and the timely payment indicators (discussed in step 2). The TPI cap for an issuer rated above A1 is always AAA. Source: Moody’s (2010).

Timely Payment Indicators Issuer Ratings Very Improbable Improbable Probable Probable-High High Very High A1 Aaa Aaa Aaa Aaa Aaa Aaa A2 Aa1 Aa1 Aaa Aaa Aaa Aaa A3 Aa2 Aa2 Aaa Aaa Aaa Aaa Baa1 Aa3 Aa3 Aa1 Aa1 Aaa Aaa Baa2 A1 A1 Aa2 Aa2 Aa1 Aaa Baa3 A3 A2 A1 Aa3 Aa2 Aa1 Ba1 Baa3 Baa2 Baa1 A3 A2 A1 Ba2 Baa3 Baa2 Baa1 A3 A2 A1 Ba3 Baa3 Baa2 Baa1 A3 A2 A1 B1 Ba3 Ba2 Ba1 Baa3 Baa2 Baa1 B2 Ba3 Ba2 Ba1 Baa3 Baa2 Baa1 B3 Ba3 Ba2 Ba1 Baa3 Baa2 Baa1

For the example in Table II the assigned TPI cap would be AAA irrespective of the assigned TPI, because for an issuer rated A1 (see table VI). We assume the assigned TPI to be Probable-High. The strong legislation and the high quality of the assets affect the TPI positive. The bullet-structure of the Covered Bond has a negative effect on the assigned TPI. The final rating of Moody’s is A1, which is the rating calculated in step 1, because the second step does not influence the rating.

D. Standard & Poor’s

The rating criteria of Standard & Poor’s is based on its report called Covered Bonds: Re- vised Methodology and Assumptions for assessing Asset-Liability Mismatch Risk in Covered Bonds published in December 2009. In S&P’s view, the main criteria to rate Covered Bond programs is to analyze asset-liability mismatches (ALMM), which are caused by the fact, that the maturities of the cover assets (greater than 10 years) are generally longer than the maturities of the Covered Bonds (on average three to seven years). The ALMM risk,

21 the program categorization and some other criteria determine the maximum rating uplift the Covered Bond program can reach over the issuer credit rating. To arrive at the ﬁnal Covered Bond rating S&P applies a ﬁve step process which is described below and in Figure 5.

Five Key Areas of Standard & Poor’s Covered Bond Rating Analysis

Operational Asset Risk Cash Flow Legal Risk and Counterparty Risk Administrative Risk Risk

Zero STEP 1: ALMM Classi!cation Low Moderate High

Category Category Category STEP 2: Program Categorization 1 2 3

Category ALMM Risk 1 2 3 STEP 3: The maximum potential Zero Unrestricted Covered Bond Rating Low 7 6 5 Max. Potential Moderate 6 5 4 Ratings High 5 4 3 Uplift

Determine target credit STEP 4: Cash Flow and Market enhancement to achieve Value Analysis maximum potential rating uplift

Compare target credit STEP 5: The Covered Bond enhancement level with available Rating credit enhancement

Figure 5. Summary of Standard & Poor’s Rating Procedure. This Figure shows how Standard & Poor’s applies their ﬁve step rating process for assessing ALMM in Covered Bonds. In the left column, the single steps are listed. Each step is described shortly on the right. Source: Standard & Poor’s (2009).

D.1. Step 1: ALMM Classiﬁcation

The goal of this first step is to calculate the ALMM exposure of the program and then to classify the ALMM risk. To calculate the ALMM exposure, S&P stress tests the cash flows to cover assets credit risk to address risks associated with the cover pool. Structural features of the program such as default risk of assets, operational risk or derivative counterparty risk are considered. S&P does also make some adjustments to the cash flows that may increase or decrease the ALMM, for example a particular level of prepayments, Covered Bonds with maturity extension features, scheduled asset amortization or liquidity facilities. The agency then considers the timing of mismatch. They differentiate between short term exposures and exposures in the medium or long run. In S&P’s view, a mismatch in the near future is more dangerous because it would imply to sell assets in the near future. After these calculations, S&P assigns a risk classification (low, moderate or high) to every Covered Bond program18. This classification determines the maximum potential number

18 The exact ALMM percentages for the three categories are 0% to 15% for Low, 15% to 30% for Moderate and bigger than 30% for High.

22 of notches uplift. For example, in case of a moderate ALMM risk, the maximum uplift is between four and seven notches (see Figure 5 for more details). If there is no ALMM risk at all, a program can be rated on a de-linked basis from the issuer.

D.2. Step 2: Program Categorization

Secondly, S&P segments the program into three different categories (from Category 1 to Category 3 ) which are predominantly driven by country. The categorization may differ within a country if there are different programs with different features. To assess the programs, two factors are considered by S&P. (i) Range of funding options: In S&P’s view, a greater range of funding options is positive for the program. Covered Bond programs should have three possible funding options: Repo transaction with a national central bank (or the ECB), third-party loan facility and the ability to sell assets. (ii) Differentiating the strength of the funding sources: This second factor analyzes the importance of the Covered Bond sector in a country. S&P believes, if the Covered Bonds are systemically important and the cost of a failed Covered Bond program on the economy would be extremely high, that the government or the central bank will do whatever they can to protect their economy. Additionally, in such a country the range of investors is broad and the Covered Bond market is active which benefits the program. After analyzing these factors, every jurisdiction is segmented in a category between Cat- egory 1 and Category 3. S&P assigns Category 1 a maximum potential uplift of five to seven notches (e.g. Germany, Denmark or Spain) and Category 3 a maximum uplift of three to five notches (e.g. U.S. or Greece) from the issuer credit rating.

D.3. Step 3: The Maximum Potential Covered Bond Rating

The third step of the analysis combines the assessment of the ALMM exposure (step 1) and the ability to cover its funding needs (step 2) to a maximum degree to which a program’s rating may exceed the IDR. The outcome of every possible combination can be found in the Figure 5. The maximum potential rating of a Covered Bond program is then determined by the issuer credit rating increased by the appropriate number of notches. In case of a direct issuance by the bank, the issuer rating is equal to the rating from S&P. If the bank uses a SPV to issue Covered Bonds, S&P applies other criteria19 to determine the relevant issuer credit rating.

D.4. Step 4: Cash Flow and Market Value Analysis

In the fourth step, S&P evaluates the target credit enhancement that corresponds to the maximum rating uplift (mentioned in step 2). The rating agency applies market value stresses to the collateral pool and analyzes the cash ﬂows in a situation where asset- liability mismatches occur. To apply the stress testing, S&P calculates the net present

19 see S&P’s criteria for Group Methodology, to ﬁnd on www.standardandpoors.com for further information.

23 value (NPV) of the expected cash ﬂow using a stressed discount rate (spread shock). This stressed discount rate exceeds the relevant funding rate (e.g. EURIBOR). The size of the asset spread widening depends on the types of assets in the cover pool and on the location of those assets. For example, the target asset spread widening is larger for residential mortgages located in Spain than for AAA-rated public sector debt. If a program is able to sell enough assets to meet such mismatches in a stressed situation and is still able to service the remaining debt, it can reach the maximum potential rating. Credit and structural risks of the assets and the program structure are also taken into account to analyze whether the credit enhancement provided is commensurate with the maximum achievable rating.

D.5. Step 5: The Covered Bond Rating

After the determination of the target credit enhancement, S&P calculates the actual credit enhancement which is appropriate for the cover pool. If the actual credit enhancement is higher or equal to the target credit enhancement, the maximum potential rating is assigned to the Covered Bond program. Otherwise, the rating of the Covered Bond will be lower than the maximum potential rating. In this case, the agency lifts the rating up, as long as the credit enhancement is able to cover the market value exposure and the amount actually provided. Additionally to the program rating, S&P assigns an outlook to every Covered Bond program, which is categorized as positive, stable, developing or negative. Those outlooks provide S&P’s view for a rating change and its direction.

To illustrate the rating process, the example introduced above is discussed for S&P. The first step classifies the Covered Bond in four possible categories. We assume, the exemplary program is classified in the category low. The German Covered Bond program fits into program category 1, which results in a maximum potential rating uplift of 7 notches. We assume, that the Covered Bond can withstand the stress testing in step 4 on the highest level, which would result in a final Covered Bond rating of AAA.

IV. Development of the Covered Bond Market

A. Data

In the next section, we analyze the development of the Covered Bond market. Firstly, we provide information on the examined data. In the second stage, we show how the size of the Covered Bond market has changed overall, in single countries and distinguished by collateral type in the single countries. In the last part of this section, we investigate the development of the yield spreads and the prices of Covered Bonds in recent time. For our analysis we use data from Bloomberg and Datastream. The sample period is Jan- uary 1, 2005 to May 1, 2012. We want to include a long period in our analysis to find differences in the behavior of Covered Bonds in stable periods as well as in periods of economic distress. We divide the sample period into three sub-periods. The first period, called

24 the Pre-Crisis-Period, lasts from the start of the sample until June 2007 and is declared as the stable period. The second period is named Crisis-to-Lehman-Period. This second period starts in July 2007 and ends on the day of the bankruptcy of Lehman Brothers (September 15, 2008). The last period is September 15, 2008 to May 1, 2012 and is called the Post-Lehman-Period.

Table VII Examined Covered Bonds

The ﬁrst table shows how the number of issued Covered Bonds in the dataset is distributed among countries and core cover. The classiﬁcation of the core cover and the country is based on Bloomberg. Examples for other collateral types are Property, ABS and Loans.

Core Cover Type Country PublicLoans Mortgages Other NotAvailable Total Austria 696 469 0 34 1,199 Britain 18 225 0 27 270 Denmark 0 1,273 1 7 1,281 France 387 655 5 94 1,141 Germany 9,547 5,517 3 384 15,451 Ireland 317 64 1 11 393 Luxembourg 401 17 0 18 436 Norway 11 276 1 31 319 Spain 56 463 1 9 529 Sweden 7 698 3 60 768 Switzerland 13 387 0 13 413 U.S. 0 701 8 Other 70 834 0 251 1,155 Total 11,523 10,885 15 940 23,363

The second part of the table shows the Covered Bonds used in the analysis arranged by coupon type and embedded option. The classification is based on data provided by Bloomberg. The first column shows a selection of embedded options, the first row contains different coupon types.

Coupon Types Embedded Option Fixed Floating StepCoupon Variable ZeroCoupon Other Total AtMaturity 12,674 3,899 206 578 398 380 18,135 Extendible 119 193 0 5 0 6 323 Callable 1,395 156 1,819 452 244 7 4,073 Sinkable 455 34 2 6 4 1 502 Callable/Extendible 13 17 11 9 0 0 50 Callable/Putable 6 3 0 32 0 0 41 Callable/Sinkable 140 17 0 18 0 0 175 Other 27 28 4 4 0 1 64 Total 14,829 4,347 2,042 1,104 646 395 23,363

To get a wide sample of Covered Bonds, we consider all bonds classiﬁed as Covered Bond, Pfandbrief or Jumbo-Pfandbrief in Bloomberg irrespective of their detailed speci- ﬁcations. Our basic sample contains information for 23,363 Covered Bonds. The average maturity in the sample is roughly 8.5 years, the average outstanding amount of a Covered

25 Bond denominated in EUR is about 550 million EUR. Covered Bonds denominated in EUR account for about 70% of all examined Covered Bonds. Beside the EUR U.S. Dollars, Swiss Francs and the Scandinavian currencies (e.g. Danish Krone or Swedish Krona) constitute a large part of the Covered Bonds in the sample. The number of Covered Bonds issued by the same credit institution varies greatly. There are some banks with many issuances (e.g. Landesbank Baden-Würtenberg (GER) with 1,517 issuances or the Eurohypo AG (GER) with 658 issuances) but also many institutions which issued only one or two Covered Bonds. Our sample contains over 400 different Cov- ered Bond issuing credit institutions. We use Bloomberg to find all static determinants such as currency, maturity, coupon type or embedded options. Table VII shows the distribution of the collateral type across country. The number of issued bonds in each category varies greatly among countries. All in all, the number of public Covered Bonds is nearly equal to the number of mortgage Covered Bonds. Germany, Austria, Ireland and Luxembourg are the biggest markets for public Covered Bonds. On the other side, the biggest markets for mortgage-backed Cov- ered Bonds are Germany, Denmark, France and Spain. There is a very small proportion of Covered Bonds which is backed by other assets, for example by ABS or property. For some of the examined bonds, Bloomberg provides no information on the core cover type (named as Not Available in the table). In the lower part of the table we distinguished between embedded option and coupon type. Most of the Covered Bonds in the sample have no embedded option (At Maturity) and a fixed or a floating coupon. Only a minority has special embedded options or an exotic coupon type. We use the ISIN Code from Bloomberg to find the corresponding bond in Datastream to get the weekly prices and the weekly yield spreads of the Covered Bonds from this data source. We find prices for 18,198 Covered Bonds and yield spreads for 16,345 Covered Bonds. Afterwards, we merge the dataset containing the time series from Datastream and the dataset with the static information from Bloomberg to one big data set containing time series of prices for 18,198 Bonds, yield spreads for 16,345 Covered Bonds and static information for 23,363 Covered Bonds. This data set is used for the following analysis.

B. Development of the Outstanding Amount of Covered Bonds

For the analysis of the nominal outstanding amount in the sample we used only Covered Bonds denominated in EUR. Of the 23,363 Covered Bonds we have to drop 6,798 Covered Bonds because they are issued in other currencies. We drop another 421 Covered Bonds because we ﬁnd no information on their maturity and another 305 for which we have no information on the amount outstanding. We make no restrictions on coupon type or embedded options. As a result, we have 15,839 Covered Bonds included in the analysis of the outstanding amounts. We take the amount outstanding at issuance and assume that these amounts will remain constant until the maturity day. The calculation of the outstanding amount is based on weekly data.

26 1,800

1,600

Mortgages 1,400 Public Loans Total

1,200

1,000

Amount Outstanding in EUR bn 800

600

400

2005 2006 2007 2008 2009 2010 2011 2012 Date

Figure 6. Amount Outstanding in EUR divided by Core Cover. This Figure shows how the outstanding amount of Covered Bonds denominated in EUR has changed over the last years. The period is January 1, 2005 to May 1, 2012. Source: Own representation.

Figure 6 shows the development of the outstanding amount denominated in EUR divided by collateral type. We exclude core cover types other than mortgages and public loans because their market share is only very limited. Thus, we divide the whole sample into two groups and calculate the total amount outstanding on a weekly basis for each group. Additionally, we add the total amount outstanding as a benchmark. The Figure shows clearly that the outstanding amount of public sector-backed Covered Bonds is strongly declining. The main driver behind that trend is Germany, which is by far the biggest public Covered Bond market. The amount of outstanding public Covered Bonds is dramatically decreasing in this country We will discuss this more closely in the next section. On the other hand, the amount outstanding of mortgage-backed Covered Bonds denominated in EUR is strongly increasing. The reason for that are the recently adopted Covered Bond legislations of countries like the U.S., Canada and some Eastern Europe countries, which had no Covered Bond law at all. In these countries, Covered Bonds are mainly used for mortgage refinancing. This trend is also supported by increasing size of the French and Spanish mortgage Covered Bond market. Another reason could be a substitution effect as a many banks might use their mortgage loans to issue Covered Bonds instead of MBS. The total amount of outstanding Covered Bonds in euro in the sample grows over the last years from EUR 1,400 bn to EUR 1,600 bn. The ECBC (2011) states, that the total amount of outstanding Covered Bond denominated in EUR at the end of 2010 was EUR 1,650 bn. In our sample, the outstanding amount at the end of 2010 is about EUR 1,500 bn. Thus, our sample is a quite appropriate reflection of the Covered Bond market. In Figure 7, we compare the amount outstanding denominated in EUR in our sample for a selection of some important countries distinguished by collateral type. We choose countries, in which the issuance of Covered Bonds backed by both, public loans and mortgage

27 400 300

300 FRA−Mortgages 200 FRA−Public Loans 200 SPA−Mortgages SPA−Public Loans 100 100

0 0

Amount Outstanding in EUR bn 2005 2006 2007 2008 2009 2010 2011 2012 2005 2006 2007 2008 2009 2010 2011 2012 Date Date 40 60

AUT−Mortgages 30 AUT−Public Loans 40 20 IRL−Mortgages 20 IRL−Public Loans 10

0 0

Amount Outstanding in EUR bn 2005 2006 2007 2008 2009 2010 2011 2012 2005 2006 2007 2008 2009 2010 2011 2012 Date Date 1’000

800 GER−Mortgages GER−Public Loans 600

400

200

0 2005 2006 2007 2008 2009 2010 2011 2012

Amount Outstanding in EUR bn Date

Figure 7. Amount Outstanding in EUR divided by Core Cover and Countries. This Figure shows how the nominal amount outstanding of Covered Bonds denominated in EUR in the sample divided by core cover has changed over the last years for a selection of countries. The period is January 1, 2005 to May 1, 2012. Source: Own representation.

loans, have a significant share of the total market. Therefore, we restricted the analysis on Austria, France, Germany, Ireland and Spain. The Figure shows that in all examined countries the amount of mortgage Covered Bond has risen over the last years. Especially in France and Spain, we can observe a strong increase of outstanding Covered Bonds backed by mortgages. The situation for public Covered Bonds is differently. In Germany, the amount outstanding is sharply declining and in the other countries, the amount of outstanding Covered Bonds backed by public sector loans remains more or less constant. Due to the strong declining of public sector backed Covered Bonds in Germany, we have to look more closely at the situation in this country. At the beginning of our investigation in 2005, the German public Covered Bond market is by far the biggest Covered Bond market with an outstanding amount of over 800 bn EUR. In the examined period, the outstanding amount has more than halved and is below EUR 400 bn in May 2012. There are two main reasons for this evolution. Firstly, the German state banks (Landesbanken) made changes in their business model. The public sector lending lost importance and therefore less eligible assets are available to issue public Covered Bonds. The second reason for the declining is legal amendments. According to Herpfer (2007), the German government abolished the guarantee obligation (Gewährträgerhaftung) and unlimited state support (Anstaltslast) for German state banks. The law was established between 2001 and 2005. Previously, exposures to those credit institutions were eligible as cover assets for public Covered Bonds due to the state guarantee. This has changed with the legislative amendment. Therefore, the scope of eligible assets decreased even more. Another implication of the abolishment of these government guarantees were rating downgrades of Covered Bonds issued by state banks which reduced the attractiveness to issue Covered Bonds.

28 Amount Outstanding in EUR

1’800

1’600

1’400

Austria 1’200 France Germany Ireland 1’000 Italy Luxembourg Spain Sweden 800 Total

Amount Outstanding in EUR bn 600

400

200

0 2005 2006 2007 2008 2009 2010 2011 2012 Date

New Issuances in EUR

20 France 15

Issuances in EUR bn 0 2005 2006 2007 2008 2009 2010 2011 2012 Date

20 Germany 15

Issuances in EUR bn 0 2005 2006 2007 2008 2009 2010 2011 2012 Date

20 Spain 15

Issuances in EUR bn 0 2005 2006 2007 2008 2009 2010 2011 2012 Date

Figure 8. Amount Outstanding and Issuances in EUR by Country. The upper Figure shows how the outstanding amount of Covered Bonds denominated in EUR divided by country has changed over the last years. The period is January 1, 2005 to May 1, 2012. The lower Figure shows the new issuances denominated in EUR of France, Germany and Spain over the same period on a weekly basis. The two peaks in the lower chart are two big issuances of callable mortgage Covered Bonds. The first peak is caused by Banques Populaires Caisses d’Epargne. The reason for the second peak is a mortgage Covered Bond series issued by SociétéGéneral in June 2011. Source: Own representation.

Figure 8 shows the development of the outstanding amount of Covered Bonds denominated in EUR by country. It becomes clear, that Germany, Spain and France dominate the Euro Covered Bond market20, the other countries in the graph have only a very small market

20 These four country are also among the top four in the total market. The fourth big player would be Denmark, but the biggest part of Danish Covered Bonds is denominated in Danish Krone.

29 share. It should be remembered, that Sweden has a big share of its issuances denominated in Swedish Krona. The lower part of the chart shows the new issuances of the three big markets. In Germany, new issuances are clearly decreasing in terms of volume and in terms of the frequency of new issuances. On the other hand, we see that there are a lot more new issuances in France in the second half of the period. The new issuances in Spain remain more or less constant during the considered time period.

C. Development of the Yield Spreads and Prices

In the next section, we analyze the yield spreads and the prices of the Covered Bonds over time. The purpose of this part is to investigate, how the yield spreads and the prices developed during the single periods. Furthermore, we compare the evolution of the spreads and the prices across country. In a next step, we will investigate if there are obvious diﬀerences in the behavior of public and mortgage Covered Bond within a country. Finally, we analyze if there are visible diﬀerences between the two collateral types.

C.1. Examined Data

The yield spread is calculated as the difference of the yield (to maturity) of a Covered Bond and the yield of a government bond with the same maturity denominated in the currency of the Covered Bond. If the maturity of the Covered Bond does not exactly match the maturity of the government bond, the yield of the government bond is estimated by lin- ear interpolation (see Thomson Financial (2005)). As we restrict our analysis on Covered Bonds denominated in EUR, the benchmark government bond is either a German or a French bond. Fabozzi (2007) states, the yield spread (or credit spread) is a measure for the credit risk of a Bond. A high yield spread expresses a high credit risk. Other factors influencing the yield spread are the presence of embedded options, liquidity risk, the maturity and the taxability of the interest. From the 16,345 Covered Bonds with available yield spreads in Datastream, we exclude 2,776 which are denominated in other currencies. Another 1,863 drop out due to their embedded option and 503 drop out due to their coupon type as we only include fixed and floating coupons. Thus, we have 11,203 Covered Bonds left over for the analysis of the yield spreads. We calculated the average yield spreads on a weekly basis. For the investigations of the price development of the Covered Bonds we started again from the total database. Due to huge differences in the pricing, we dropped 3,319 Bonds because they have special embedded option (e.g. callable or sinkable Covered Bonds). We only include Covered Bonds with fixed or floating coupon type and drop another 1,084 Covered Bonds due to their coupon type (e.g. variable coupon or step coupon). One Bond drops out because we do not have information on its maturity. Finally, we result in a sub database of 13,794 Covered Bonds. For the prices and the yield spreads we excluded obvious data errors.

30 C.2. Development of the Covered Bond Market across Countries

If we compare the yield spreads for different countries, we can observe some clear patterns. We selected countries according to their importance or countries which might show interesting differences. We distinguish between countries which are often declared as stable (Austria, Britain, France, Germany, Sweden and the United States) and countries often determined as unstable (Greece, Ireland, Italy, Portugal, Spain or see Figure 9). The credit spreads are on a very low level during the first period for both, the stable and the unstable sample. The spreads are clearly below 100 basispoints (bp), which confirms the assumption, that a Covered Bond is an investment with a high degree of certainty. As we will see later on, this is not fully recognized by market players in unstable periods. With the beginning of the financial crisis in July 2007, the spreads begin to increase in all countries. After the collapse of Lehman Brothers, the spreads rise sharply. Due to the fact that Covered Bonds are bank debt, the insolvency of Lehman Brothers seems to have a huge impact on the certainty of Covered Bonds. If we look at the two samples of countries, we cannot observe a big difference between the countries in the upper and the countries in the lower sample (one has to be aware of the scales in the two figures). The spreads for countries in both samples are on a comparably high level. But there is a difference between the Anglo-Saxon countries (Ireland, UK and U.S.) compared to the other European countries. This could be partly due to the fact, that the bankruptcy of Lehman Brothers had a bigger effect on those countries.

400 Austria Britain 300 France Germany Sweden 200 Total UnitedStates

100

0 2005 2006 2007 2008 2009 2010 2011 2012

2000 Greece Ireland Yield Spreads in Basispoints 1500 Italy Portugal Spain 1000 Total

500

0 2005 2006 2007 2008 2009 2010 2011 2012 Date Figure 9. Average Spreads divided by Country. These Figures show the development of the average Covered Bond yield spreads among countries. The upper Figure contains the stable countries, the lower Figure the unstable countries. The period is January 1, 2005 to May 1, 2012, the vertical lines separate the three periods. Source: Own representation.

After the collapse of Lehman, the spreads remain for a short period on a very high level. Afterwards, we notice decreasing yield spreads in all countries. The reason for that could be the general recovery of the economy but also the CBPP1 announced in May 2009

31 by the ECB could have an effect on the declining spreads. After the period of recovery, the spreads begin to rise again. In this phase, the spreads begin to differ more across country. We observe, that the recovery continues in the stable countries. In the unstable countries, the spreads rise to a level above the peak after the Lehman collapse. Especially the development in Greek market is extreme. On the highest point, the risk premium reaches a level of nearly 2000 bp. The enormous level of the yield spreads is quite surprising, if we remember the legal framework of Greece (see Table I). The development in Ireland, Italy, Portugal and Spain is comparable on a lower level. Also in the upper sample, the spreads begin to increase but the level of the spreads is very low compared to the countries in the lower sample. In Figure 10 we made the calculation for prices corresponding to Figure 9. The development of the prices reflect the observations from the spreads. For both samples, the prices are very solid during the stable period. In the Crisis-to-Lehman period we see a declining of all prices and after the Lehman Brothers collapse. Afterwards, the prices recovered in the stable sample. Again, we notice the big differences between the upper and the lower sample with the beginning of the debt crisis. It is rather remarkable, that the average prices of Covered Bonds in Greece and Portugal fall below 70%.

120

110

100 Austria 90 Britain France 80 Germany Sweden 70 Total UnitedStates 60 2005 2006 2007 2008 2009 2010 2011 2012

120

110

100 Greece 90 Ireland

Price in % Italy 80 Portugal Spain 70 Total 60 2005 2006 2007 2008 2009 2010 2011 2012 Date Figure 10. Average Prices divided by Country. These Figures show the development of the average Covered Bond prices among countries. The upper Figure contains the stable countries, the lower Figure the unstable countries. The period is January 1, 2005 to May 1, 2012, the vertical lines separate the three periods. Source: Own representation.

In summary, we can only observe small differences between countries in both, the yield spreads and the prices in stable periods. After the Lehman Brother collapse, the yield spreads increase more in the Anglo-Saxon countries than in the other countries in the sample. With the beginning of the debt crisis, the spread and price differences between the stable and unstable countries become larger. We conclude, that the behavior of Covered Bonds is similar during stable periods, but in times of crisis there are large country specific differences.

32 C.3. Development of the Covered Bond Market divided by Collateral Type across Country

In Figure 11 we calculate again the weekly average yield spreads divided by collateral type among countries. We choose ﬁve countries for this analysis. We analyze countries in which both core cover types have signiﬁcant share of the market (see Figure 7). Thus, we analyze the development in Austria, France, Germany, Ireland and Spain. We will compare the development of both Covered Bond types for each country.

200 200

AUT−Mortgages 150 150 AUT−Public Loans

100 100 Yield Spread in BP 50 Yield Spread in BP 50 GER−Mortgages GER−Public Loans 0 0

2005 2006 2007 2008 2009 2010 2011 2012 2005 2006 2007 2008 2009 2010 2011 2012

500 500

400 400 FRA−Mortgages FRA−Public Loans SPA−Mortgages 300 300 SPA−Public Loans

200 200 Yield Spread in BP Yield Spread in BP 100 100

0 0 2005 2006 2007 2008 2009 2010 2011 2012 2005 2006 2007 2008 2009 2010 2011 2012

1200

1000

IRL−Mortgages 800 IRL−Public Loans

600

400 Yield Spread in BP

200

0 2005 2006 2007 2008 2009 2010 2011 2012

Figure 11. Average Spreads by Core Cover and Country. These Figures show the development of the average Covered Bond yield spreads by country and core cover. The period is January 1, 2005 to May 1, 2012. Source: Own representation.

Firstly, we will analyze Germany and Austria, because the behavior of the spreads is similar in these countries. In the pre-crisis period, we cannot observe a difference between mortgage and public Covered Bonds. During the financial crisis, we see, that the spreads for mortgage-backed Covered Bonds are on a slightly higher level. This observation goes in line with the rating agencies, which assign in general higher ratings for public Covered Bonds compared to mortgage Covered Bonds due to the higher liquidity and the lower credit risk of public sector loans compared to mortgages loans (see Part III or Moody’s (2010)). The spreads for mortgage-backed Covered Bonds remain also higher for the period after the Lehman collapse. Afterwards, we cannot observe a clear difference anymore. This may be caused by increasing spreads of public Covered Bonds due to the debt crisis in the Euro Area.

33 In France during the stable period, we can observe slightly lower spreads for Covered Bonds backed by public loans. During the Crisis-to-Lehman period, there are no clear diﬀerences between the spreads of the two Covered Bond types anymore. The situation changes in 2010. The spreads for public Covered Bonds increase and are much higher until the end of our sample period. This fact is mainly caused by the Dexia Bank. The bank is a big issuer for public Covered Bonds and is in big troubles. If we exclude the Dexia Bank from the analysis, the spreads for mortgage and public Covered Bonds are roughly on the same level. We conclude, that the debt crisis has increased the spreads for public Covered Bonds, but not to the extent stated in the Figure.

105 105

100 100 Prices in % Prices in %

AUT−Mortgages GER−Mortgages AUT−Public Loans GER−Public Loans 95 95 2005 2006 2007 2008 2009 2010 2011 2012 2005 2006 2007 2008 2009 2010 2011 2012

110 110

SPA−Mortgages 105 105 SPA−Public Loans

100 100 Prices in % Prices in % 95 95 FRA−Mortgages FRA−Public Loans 90 90 2005 2006 2007 2008 2009 2010 2011 2012 2005 2006 2007 2008 2009 2010 2011 2012 110

105

100

90 Prices in % IRL−Mortgages 85 IRL−Public Loans

2005 2006 2007 2008 2009 2010 2011 2012 Figure 12. Average Prices by Core Cover and Country. These Figures show the development of the average Covered Bond prices by country and core cover. The period is January 1, 2005 to May 1, 2012. Source: Own representation.

The situation in Ireland is quite different. The first mortgage Covered Bond was issued only in 2006, but the amount of outstanding Covered Bonds backed by mortgages increased rapidly (see Figure 7). The development of the spreads until the beginning of the financial crisis is equivalent to the evolution in the other countries. During the financial crisis and after the bankruptcy of Lehman Brothers, the two Covered Bond types behave very similar. In the middle of 2010, the spreads for mortgage Covered Bond rise dramatically.

34 The reason could be the mortgage crisis in Ireland which reduced the quality of the cover pools and therefore increased the risk of mortgage Covered Bonds.

Table VIII Average Yield Spreads in the diﬀerent Periods

This Table shows the weekly average yield spreads in basispoints, the standard deviations, the maximum and the minimum value in each period and the number of observations for each country in the diﬀerent periods. The observations for Irish mortgage Covered Bonds start in the middle of the ﬁrst period. Therefore, we excluded the whole panel from the investigation. Source: Own representation.

France Germany Ireland Spain Mortgage Public Mortgages Public Mortgages Public Mortgages Public Panel A Pre-Crisis (01/01/2005 - 07/01/2007) Mean 14.39 13.09 20.68 20.15 - 12.05 20.08 12.73 StdDev 3.32 2.74 2.83 3.34 - 1.57 5.13 1.75 min 9.20 8.65 9.2 7.80 - 8.80 12.55 8.30 max 20.55 18.75 27.70 27.30 - 17.05 28.90 16.55 n 130 130 130 130 -130.00 130 130 Panel B Crisis-to-Lehman (07/01/2007 - 09/15/2008) Mean 49.22 45.85 63.08 60.01 72.98 46.67 73.92 61.31 StdDev 14.97 13.00 14.81 14.27 27.04 13.01 23.66 21.64 min 20.25 18.60 22.4 22.40 27.04 14.90 28.90 9.60 max 78.80 71.90 87.55 85.10 123.80 75.55 117.80 99.10 n 63 63 63 63 63 63 63 63 Panel C Post-Lehman-Period (09/15/2008 - 05/01/2012) Mean 116.10 148.77 96.84 90.79 534.43 279.73 247.88 457.40 StdDev 41.35 81.85 27.91 26.71 331.56 77.30 105.91 100.72 min 60.80 66.30 61.95 55.90 132.45 84.95 100.50 72.00 max 219.40 409.00 165.00 154.50 1207.05 457.40 480.40 444.90 n 189 189 189 189 189 189 189 189 Panel D Overall Period (01/01/2007 - 05/01/2012) Mean 70.45 85.62 65.35 61.68 419.07 150.20 141.67 115.68 StdDev 55.36 85.85 39.95 37.39 350.15 139.95 130.51 114.51 min 9.20 8.65 9.20 7.80 20.65 8.80 12.55 8.30 max 219.40 409.00 165.00 154.50 1207.05 457.40 480.40 444.90 n 382 382 382 382 252 382 382 382

If we compare the situation for mortgage and public Covered Bonds in Spain, we get a different picture. During the whole examined period, we see that higher spreads for Covered Bonds backed by mortgages loans compared to public sector Covered Bonds. In the pre- crisis period, the difference is small. In the Post-Lehman period, the spread widens. At first sight, this seems very surprising due to the huge debt problem of the Spanish state. But there is not only a debt problem in Spain, but also a mortgage crisis and a highly volatile mortgage market which might increases the spreads on mortgage Covered Bonds. In Figure 12 we did the same calculation of the average weekly prices as above. The prices in Austria and Germany are very robust to the crisis. We observe only small changes in those countries. In France, we see the sharp fall of the public Covered Bonds at the end of our sample period. In Spain and Ireland, the declining of the mortgage Covered Bonds is clearly visible.

35 In Table VIII we compare the average spreads for each country divided by collateral type for each period, as well as over the whole sample period. This confirms our statements from the analysis of the charts. Across all countries, we observe a high correlation between both collateral types. Especially in stable periods, there is no big difference in terms of yield spreads between the two categories. In periods of economic distress, the influence of the core cover type on the certainty of the Covered Bond seems to be stronger, especially if there is either a debt or a mortgage crisis in the single countries.

C.4. Development of the Covered Bond Market divided by Collateral Type

In this last part of the market analysis, we compare the development of the two different core cover types. We divide our database of weekly yield spreads into two samples. The first sample contains all Covered Bonds backed by mortgages, the second sample all public Covered Bonds. Again, we calculate the weekly average for the spreads. In Figure 13 the evolution of the spreads is illustrated. We observe a strong correlation between the two samples over the whole sample period. If we look more closely at the development in the single periods, we cannot observe clear differences between Covered Bonds backed by mortgages and those backed by public loans until the bankruptcy of Lehman Brothers. After this event, we see considerably lower spreads for public Covered Bonds compared to mortgage Covered Bonds. This could be caused by the subprime crisis in the U.S. In this crisis, a huge amount of MBS defaulted which might have stirred up fears that the same happens to mortgage-backed Covered Bonds. The difference disappears by the end of 2010. The reason for that could be the upcoming debt crisis in the Euro countries.

180 Mortgages Public Loans

160

140

120

100

80 Yield Spread in BP

0 2005 2006 2007 2008 2009 2010 2011 2012 Date

Figure 13. Average Yield Spreads by Core Cover. This Figure shows the development of the average Covered Bond yield spreads among core cover. The period is January 1, 2005 to May 1, 2012. Source: Own representation.

36 V. Conclusion and further Research

A. Conclusion

This thesis consists of three parts. In the first part, we examined Covered Bonds from a theoretical point of view. At first, we tried to find a suitable definition for Covered Bonds. We provide a list of requirements on the issuer, the cover pool and the ring fencing mechanisms a Covered Bond program has to fulfill. Then in a second step, we introduced four different structures used to issue Covered Bonds. Those four structures cover all structure variations in the different frameworks. Thirdly, we differentiated between Covered Bonds and off-balance sheet instruments like mortgage-backed securities. We conclude, that MBS are more independent from the issuer, have a lower market risk and their cover pool are more transparent. On the other side, Covered Bonds are supervised more closely, the investors have a double recourse against the cover pool and the issuer and the quality of the collateral pool is higher due to its dynamic structure. In the last step of the first part, we searched for differences in the legal frameworks across countries and find that the biggest differences occur between the Anglo-Saxon countries and the traditional European Covered Bond markets. In the second part of the thesis, we looked into the methodologies of the rating agencies Fitch, Moody’s and S&P. We found, that by now all three agencies base their analysis on the unsecured rating of the issuer and calculate a certain amount of uplift. To calculate the level of uplift, Fitch investigates the independence of the collateral pool from the issuer, stress- tests the overcollateralization and assigns a rating uplift for outstanding recovery prospects of the cover assets. Moody’s calculates in the expected loss in case of a default of the issuer and might reduce the rating of the first step in case the probability of timely payments continuing after an issuer’s default is low. S&P fixes the maximum achievable rating by combining the asset-liability mismatches and the quality of the respective legislation. The rating is then adjusted by the level of required credit enhancement. In the last part, we analyzed the historical development of the outstanding amounts as well as the evolution of the yield spreads and the Covered Bond prices over the period from January 2005 until May 2012. We observed a strong decrease of amount outstanding of German public Covered Bonds and therefore reduction of not only the whole German Covered Bond market, but also the public Covered Bond market. In the other countries, we saw an increasing outstanding amount of mortgage Covered Bonds. We observed very low yield spreads for Covered Bonds in the period before the financial crisis irrespective of the country. After the bankruptcy of Lehman Brothers we discovered higher spreads for the Anglo-Saxon countries and extremely high yield spreads for the highly indebted countries during the debt crisis. The spread development of Covered Bonds collateralized by mortgages and by public loans is highly correlated. Only in periods of economic distress, we observe considerable differences.

37 B. Further Research

Due to the paucity of academic studies in this area, there is a huge amount of topics for further research. In this section, we will provide an overview of possible research areas based on our investigations. Based on Part II, an investigation of different legislations within a country would be interesting. For example, in France there are four different frameworks, of which some are based on specific law, others are based on general law. An empirical research may show, from which specifications a Covered Bond program benefits in terms of security or return. A similar study could be done for Switzerland, where the two big banks (Credit Suisse and UBS) started their own Covered Bond program in 2010 (Credit Suisse) and 2009 (UBS), respectively. An empirical comparison may find interesting evidences due to the completely different structures of these programs. From our point of view, an empirical cross-country investigation of the influence of the distinction in the legislations on the certainty of Cov- ered Bonds is hardly possible The differences in the specifications of MBS and Covered Bonds could result in differences in the behavior or in the performance. Especially investigations within countries or within a single issuer might have interesting results. For example, one could test the influence of a change of the issuer rating on Covered Bonds and MBS. The hypothesis based on our investigation would suggest a bigger influence of the issuer rating on Covered Bonds. Out of Part III arise some new fields of investigation. It may be interesting to investigate if the rating procedures of the agencies are applicable for all frameworks. Some of the agencies provide specific reports for several countries. There might be interesting differences. A second possible research topic in this area could be the influence of rating changes on Covered Bonds. An empirical study could investigate both, the influence of the Covered Bond ratings and the influence of the ratings of the issuer on the Covered Bonds. The differences between public and mortgage Covered Bonds in terms of credit risk could be investigated in more detail. For example, an investigation on the influence of the mortgage market on mortgage-backed Covered Bonds would be possible or similarly, the impact of the performance of government bonds on public Covered Bonds. Another exciting field for an investigation could be to investigate differences in the credit risk between unsecured corporate bonds and Covered Bonds of the same credit institution. With this method, one could measure the quantitative influence of the covering on the credit risk of the Covered Bond.

38 References

Beirne, John, Lars Dalitz, Jacob Ejsing, Magdalena Grothe, Simone Manganelli, Fernando Monar, Benjamin Sahel, Matjaz Susec, Jens Tapking, and Tana Vong, 2011, The Impact of the Eurosystem’s Covered Bond Purchase Programme on the Primary and Secondary Markets, ECB Occasional Paper 122, 1–36.

Bernanke, Ben S., 2009, The future of mortgage ﬁnance in the United States, The B.E. Journal of Economic Analysis & Policy 9(3.2), 1–9.

Carbo-Valverde, Santiago, Richard J. Rosen, and Francisco Rodriguez-Fernandez, 2011, Are Covered Bonds a Substitute for Mortgage-Backed Securities?, FRB of Chicago Work- ing Paper 2011-14, 1–35. de Haan, Jakob, and Fabian Amtenbrink, 2011, Credit Rating Agencies, Working Paper, Netherlands Central Bank.

Deutsche Bank, 2011, Market Guide: Covered Bond Market - Set for further structural Changes, to ﬁnd on http://www.euromoneyconferences.com/downloads/SFF11/ Mar- ketGuide2011.pdf (04/20/2012).

European Covered Bond Council (ECBC), 2008, European Covered Bond Fact Book (ECBC: Brussel).

, 2011, European Covered Bond Fact Book (ECBC: Brussel).

Fabozzi, Frank J., 2007, Fixed Income Analysis (CFA Investment Series).

Fitch Ratings, 2011, Covered Bonds Rating Criteria, to ﬁnd on http://www.ﬁtchratings. com/creditdesk/reports/report frame.cfm?rpt id=648551 (03/12/2012).

Forster, Yehudah, Helene Heberlein, and Alla Sirotic, 2012, Covered Bond Ratings, in James R. Tanenbaum, and Anna T. Pinedo, ed.: Covered Bond Handbook . chap. 7 (Practising Law Institute) 2 edn.

Gambro, Michael S., Anna H. Glick, Frank Polverino, Patrick T. Quinn, and Jordan M. Schwartz, 2009, What are Covered Bonds and why should anyone care?, The Real Estate Finance Journal Winter 2009, 59–63.

Golin, Jonathan, 2006, Covered Bonds and Pfandbriefe: Beyond Pfandbriefe: Innovations, Investment and Structured Alternatives (Euromoney Institutional Investor PLC).

Herpfer, Jens, 2007, Der Pfandbrief - Pfandbriefgesetz vs. Novellierung des HBG (GRIN Verlag).

Horath, Robert, 2007, Die besicherte Reﬁnanzierung des Hypothekargesch¨aftes: Der Schweizer Pfandbrief und seine Entwicklungschancen (Haupt Verlag AG).

J. P. Morgan, 2011, The J.P. Morgan Covered Bond Handbook 2011 (J. P. Morgan).

39 Klein, Bruce, and Susan E. D. Neuberg, 2009, Covered bonds, in Talcott J. Franklin, and Thomas F. Nealon, ed.: Mortgage and Asset Backed Securities Litigation Handbook . chap. 13 (Thomson Reuters/West) 2 edn.

Lassen, Tim, 2005, Specialization of Covered Bond Issuers in Europe, Housing Finance International December 2005, 3–12.

Moody’s, 2010, Moody’s Approach to Rating Covered Bonds, to ﬁnd on http://www.moodys.com/researchdocumentcontentpage.aspx?docid=PBS SF191950 (03/12/2012).

Packer, Frank, Ryan Stever, and Christian Upper, 2007, The Covered Bond Market, BIS Quarterly Review 34(3), 43–55.

Prokopczuk, Marcel, and Volker Vonhoﬀ, 2012, Risk Premia in Covered Bond Markets, ICMA Centre Discussion Papers in Finance DP2012-03.

Schwarcz, Steven L., 2005, The Conundrum of Covered Bonds, The Business Lawyer 66(3), 561–586.

Standard & Poor’s, 2009, Covered Bonds: Revised Methodology And Assump- tions For Assessing Asset-Liability Mismatch Risk In Covered Bonds, to ﬁnd on http://www.standardandpoors.com/ratings/articles/en/eu/?articleType=HTML&asset ID=1245199921184 (03/20/2012).

St¨ocker, Othmar, 2011, Covered Bond Models in Europe: Fundamentals on legal Struc- tures, Housing Finance International Winter 2011, 32–40.

The Department of the Treasury, 2008, Best Practices for Residential Covered Bonds, to ﬁnd on http://www.treasury.gov/about/organizational-structure/oﬃces/General- Counsel/Documents/USCoveredBondBestPractices.pdf (04/22/2012).

Thomson Financial, 2005, Infostream, to ﬁnd on http://extranet.datastream.com/News Events/newweb/MarApr 2005/Infostream MarApr05.pdf (06/01/2012).

Tran, Quoc D., 2011, Analyse von Risikopr¨amien im europ¨aischen Covered Bond Markt (GRIN Verlag: Norderstedt).

UCITS, 2009, Directive 2009/65/EC of the European Parliament and of the Council, to ﬁnd on http://eur-lex.europa.eu/LexUriServ/ LexUriS- erv.do?uri=OJ:L:2009:302:0032:0032:EN:PDF (06/02/2012).

White, Lawrence J., 2010, Markets: The Credit Rating Agencies, Journal of Economic Perspectives 24, 211 – 26.

40 VI. Appendix

A. Appendix A - List of the Frameworks

Table IX List of the Frameworks

This Table shows all considered frameworks in diﬀerent countries. The categorization is adopted from ECBC (2011).

Country Name of the Framework Fundierte Bankschuldverschreibungen Austria Pfandbriefe Belgium Belgic CBs Bulgaria Bulgarian CBs Canada Canadian CBs Cyprus Cyprus CBs Særligt Dækkede Obligationer - SDO Denmark Særligt Dækkede Realkreditobligationer - SDRO Realkreditobligationer - RO Finland Finish CBs CRH General Law Based CBs France Obligations Foncières Obligations àl’Habitat Germany Pfandbriefe Greece Greek CBs Hungary Hungarian CBs Ireland Asset Covered Securities Italy Obbligazioni Bancarie Garantite Lettres de Gage publiques Luxembourg Lettres de Gage publiques Lettres de Gage mobilières Netherlands Dutch registered CBs programs Norway Norgay CBs Poland Polish CBs Obriga¸cões sobre o Sector Público Portugal Obriga¸cões Hipotecárias Russia Mortgage Obligations Slovakia Slovakish CBs Cédulas Hipotecarias Portugal Cédulas Hipotecarias Territoriales Sweden Swedish Covered Bonds UBS CB Switzerland Swiss Pfandbriefe Credit Suisse CB United Kingdom Regulated Covered Bonds United States U.S. Covered Bond

41 B. Appendix B - Additional Information about the Data

Table X Composition of the Sample by Country and Currency

This Table shows how the currencies of the Covered Bond sample are distributed among countries. Most of the Covered Bonds are denominated in euro. The classiﬁcation is based on Bloomberg.

Currency Country EUR USD GBP DKK NOK SEK CHF Other Total Austria 973 0 0 0 0 0 48 178 1199 Britain 143 15 93 0 9 2 3 5 270 Denmark 236 0 0 1000 15 16 13 1 1281 France 773 155 17 2 10 5 111 68 1141 Germany 12729 291 21 6 4 6 207 2187 15451 Ireland 203 108 15 0 6 5 12 44 393 Luxembourg 198 117 14 0 4 0 90 13 436 Norway 61 19 0 0 215 5 16 3 319 Spain 521 5 0 0 0 0 0 3 529 Sweden 120 29 3 0 44 472 58 42 768 Switzerland 10 8 0 0 0 0 394 1 413 U.S. 62000000 8 Other 592 77 3 1 20 1 44 417 1155 Total 16565 826 166 1009 327 512 996 2962 23363

42 C. Appendix C - Addtitional Figures

AUT−Differences FRA−Differences GER−Differences IRL−Differences 10 SPA−Differences

−5

−10 Differences between Mortgage and Public Covered Bond Prices of each Country in %

−15 2005 2006 2007 2008 2009 2010 2011 2012 Date

Figure 14. Price Diﬀerences of Covered Bonds by Country. This Figure shows the development of the diﬀerence of the Covered Bonds backed by Mortgages and the Covered Bonds backed by Public Loans by country. The period is 01/2005 to 05/2005. Source: Own representation.

105 Public Loans Mortgages Total 104

103

102

101

100 Average Price in %

96 2005 2006 2007 2008 2009 2010 2011 2012 Date

Figure 15. Average Prices by Core Cover. This Figure shows the development of the average Covered Bond prices by core cover. The average prices are calculated as a weekly mean of all Covered Bonds of each core cover type. Source: Own representation.