Why do ABS CDOs Exist? Sun Young Parky Korea Advanced Institute of Science and Technology

Abstract This paper investigates the ABS CDOs (Asset-Backed Security Collateralized Debt Obligations) market between 2005 and 2007 and answers the question of why ABS CDOs exist. The dataset used in this paper contains 516 ABS CDOs, 4,023 CDO tranches, and 79,724 securities used as the collateral of ABS CDOs. Using detailed collateral information on ABS CDOs, I …nd that there is a large re-securitization rate discontinuity between AAA- and non-AAA-rated tranches. Less than 1% of AAA-rated subprime MBS tranches were used as collateral for cash ‡ow ABS CDOs. In contrast, more than half of non AAA-rated subprime tranches were re-securitized through cash ‡ow CDO. I also show that the reference portfolio of synthetic CDOs consists to a large degree of BBB-rated subprime mortgage backed securities. As a notional amount, $94 billion out of $172 billion, which was the total collateral amount of synthetic CDOs, was comprised of BBB-rated subprime MBS tranches. It is evident that the synthetic CDO market deteriorated due to short demand in the housing sector, but there is no evidence that the cash ‡ow ABS CDO market deteriorated between 2005 and 2007. (JEL G21, G24, G28)

I am indebted to Gary B. Gorton for his insightful guidance and encouragement. I deeply thank Frank J. Fabozzi and Edward Vytlacil for thoughtful suggestions and kind support. Participants in the Yale Finance Workshop provided thoughtful comments. Any remaining errors are my own. y Korea Advanced Institute of Science and Technology, E-mail Address: [email protected]

I.

Introduction

A collateralized debt obligation (CDO) is a structured …nancial product, which generates cash ‡ow from a portfolio of debt securities or a basket of credit default swaps (CDS contracts).1 CDOs hold debt securities such as bonds, bank loans, and asset backed securities as collateral to issue prioritized tranche notes.2 If a CDO invests in a …rst order securitization product, such as a subprime mortgage backed security, an alt-A mortgage backed security, a commercial mortgage backed security or other asset backed securities, such as those backed by student loans or credit card receivables, then it is called an asset backed security collateralized debt obligation (ABS CDO). ABS CDOs constitute the largest category in the CDO market by collateral type, and accounted for 65% of the CDO market in 2005. The signi…cant write-downs of …nancial institutions in 2007 were all related to ABS CDOs. While there are no o¢ cial statistics on losses during the …nancial crisis of 2007-2009, it is widely accepted that most subprime related losses were incurred by major players in the subprime mortgage securitization market, especially in the ABS CDO market.3 The largest losses hit big CDO arrangers such as Citigroup, Merrill Lynch, and UBS, and …nancial guarantors such as AIG, Ambac, and MBIA. According to Financial Crisis Inquiry Commission (FCIC), Citigroup and Merrill Lynch reported the most spectacular losses, largely because of their extensive collateralized debt obligation business, writing down a total of $23.8 billion and $24.7 billion respectively by the end of 2007. Billions 1

Longsta¤ and Rajan (2008). If a CDO holds bonds, it is called a collateralized bond obligation (CBO), and if it is backed by bank loans, it is called a collateralized loan obligation (CLO). If a CDO holds a second order securitization product such as a CDO or a CLO, it is called a CDO-squared. The fundamental technique is the same as for other securitization products: pooling and tranching. The di¤erence comes from the underlying collateral, and the credit risk of collateral is re‡ected in the design of the security. 3 To measure net derivatives exposure is challenging. Since …nancial institutions hedged their subprime related risk with credit derivatives, it is di¢ cult to track their exact loss amounts. For example, Goldman Sachs and Deutsche Bank were also the major players in the ABS CDO market but they succeeded to hedge its subprime related position by purchasing credit default swaps (CDSs). 2

1

more in losses were reported by large …nancial institutions such as Bank of America ($9.7 billion), Morgan Stanley ($10.3 billion), JP Morgan ($5.3 billion) and Bear Stearn ($2.6 billion). Even though ABS CDOs played an important role in the recent …nancial crisis, existing empirical studies on the ABS CDO market are limited (See Benmelech and Dlugosz (2009a), Gri¢ n and Tang (2009)).4 The literature focuses on the credit rating decision process and the factors which explain the downgrading of the CDO during the …nancial crisis. Since CDOs are mainly sold to private markets registered under 144A or Reg S, securities in CDOs do not need to be registered under the Securities Act of 1933. Furthermore, the CDO funding structure evolved with credit derivatives such as credit default swaps and synthetic CDOs. Due to the opacity of the CDO market and the complexity of CDO structure, it is challenging to fully comprehend what happened in the ABS CDO market before the crisis. ABS CDOs received more attention after the SEC’s fraud allegation against Goldman Sachs over its synthetic CDO, ABACUS 2007-1 deal.5 The popular perception is that …nancial intermediaries invented CDOs and other structured …nancial products to attract investors. Underwriting standards deteriorated because credit rating agencies (CRAs) were paid by …nancial intermediaries, and therefore had little incentive to construct a strict credit rating model. Thus, …nancial intermediaries were able to produce AAA-rated securities out of junk quality collateral, and sell them to investors who were fooled by an AAA rating.6 This paper analyzes the CDO market from a di¤erent perspective, namely, that ABS CDOs were produced to create liquidity rather than to attract investors. While the two 4

Coval, Jurek, and Sta¤ord (2009) did simulation studies how small change of parameters led to the dramatic change of CDO performance. Benmelech and Dlugosz (2009b) investigate collateralized loan obligation (CLO) and Longsta¤ and Rajan (2008) study the pricing of collateralized bond obligation (CBO). 5 At April 16, 2010, The Securities and Exchange Commission charged Goldman, Sachs & Co. and one of its vice presidents for defrauding investors by misstating and omitting key facts about a …nancial product tied to subprime mortgages. 6 “The whole concept of ABS CDOs had been an abomination,” Patrick Parkinson, currently the head of banking supervision and regulation at the Federal Reserve Board, told the FCIC. Also Peter Henning, a professor at Wayne State University Law School in Detroit and a former SEC enforcement lawyer told to Wall Street Journal at April 19, 2010, “To say these products were toxic waste is an insult to toxic waste.”

2

purposes are not mutually exclusive, the central idea of this paper is that securitization is a fundamental part of the credit intermediation process in the shadow banking system. Speci…cally, this paper investigates the ABS CDO market between 2005 and 2007 to answer the question of why ABS CDOs exist, or equivalently, why …nancial intermediaries produced ABS CDOs, which mostly ended up on their balance sheets. It also aims to provide detailed information on the ABS CDO collateral composition and deal structure by CDO type ((i) high grade cash ‡ow, (ii) mezzanine cash ‡ow, (iii) high grade hybrid/synthetic, and (iv) mezzanine hybrid/synthetic ABS CDOs) to analyze the evolution of the CDO market before the …nancial crisis. What is the bene…t of repackaging asset backed securities? According to Demarzo (2005), a …nancial intermediary can solve the adverse selection problem it faces when it has superior information about the value of assets by securitization. The intermediary can create a low-risk debt security, which is less sensitive to the intermediary’s private information by pooling assets and tranching liabilities, thus maximizing the return on its information.7 Securitization also facilitates trading by providing liquid securities in the shadow banking system. Dang, Gorton, and Homlstrom (2010) de…ne liquidity as the ability to trade a given amount quickly without the transaction moving prices, and without an uninformed party losing money to an informed party. According to them, debt is the optimal security for liquidity provision since it is the least information-sensitive. An information insensitive security is one where traders have no incentive to produce private information. Consequently, an information insensitive security is immune to adverse selection, and hence, is liquid. One of their key insights is that securitization is the creation of information-insensitive securities. To better understand the recent …nancial crisis, it is essential to understand the nature of the shadow banking system and its credit intermediation process. This is because the backdrop of the recent …nancial crisis was the shadow banking system. The ampli…cation 7

See also DeMarzo and Du¢ e(1999) and Leland and Plyle(1977). Asymmetric model of security design introduced by DeMarzo and Du¢ e(1999) and the signaling model of Leland and Plyle(1977). These papers develop signaling models of the sale of a security, in which the issuer signals a high value security by the issuer’s willingness to retaining a portion of the issue.

3

mechanism, which distributed risk from the subprime mortgage sector in the U.S. to the global …nancial system, was also the instrument of shadow credit intermediation.8 According to Pozar, Adrian, Ashcraft, and Boesky (2010), shadow banks are …nancial intermediaries that conduct maturity, credit, and liquidity transformation without access to central bank liquidity or public sector credit guarantees.9 The shadow banking system decomposes the simple process of deposit-funded, hold-to-maturity lending conducted by traditional banks, into a complex, wholesale-funded, securitization-based lending process that involves a range of shadow banks. Securitized products are commonly used as collateral in the repurchase agreement market (repo market) and the asset backed commercial paper (ABCP) market. The essence of shadow credit intermediation is a combination of the securitization product and secured lending technique. It is summarized as securitization-based credit intermediation. Financial institutions in the shadow banking system form a dense network through the shadow credit intermediation process. A few major players, especially in the derivatives market such as AIG, played a crucial role in the crisis. Financial intermediaries in the shadow banking system take advantage of specialization, comparative advantage, and innovation. The shadow banking system provides credit to the real economy just as traditional banking does. Since the system depends on wholesale funding, and there is no government backstop to these …nancial intermediaries, they are always exposed to the possibility of facing runs by wholesale funding providers. The core of the recent …nancial crisis is a problem with private money creation, according to Gorton (2009). It means that the securitization-based credit intermediation process in the shadow banking system experienced a problem due to the aggregate shock on the 8

The instruments of shadow credit intermediation include residential mortgage backed security (RMBS), asset backed commercial paper (ABCP), collateralized debt obligation (CDO) and CDO-squared. 9 According to their de…nition, credit intermediation involves maturity, credit, and liquidity transformation which can signi…cantly reduce the cost of credit relative to direct lending. Credit transformation refers to the enhancement of the credit quality of debt issued by he intermediary through the use of priority of claims. Maturity transformation refers to the use of short-term deposits to fund long-term loans, which creates liquidity for the saver but exposes the intermediary to rollover and duration risks. Liquidity transformation refers to the use of liquidity instruments to fund illiquid assets.

4

housing sector in 2006. The link between the housing sector and the value of instrument of shadow credit intermediation triggered private information production; previously information insensitive securities became information sensitive and, thus, illiquid. The possibility of an adverse selection problem and uncertainty about the value of mortgage-related products led to a run on the repo and ABCP markets. The goal of this paper is to empirically test the hypothesis that securitization is the creation of an information insensitive security in the context of the ABS CDO market. In the …rst order securitization process, which produces mortgage backed securities from a pool of mortgages, the junior subprime MBS tranche is left as an information sensitive by-product after making AAA-rated tranche information insensitive.10 Therefore, if securitization is used to produce information insensitive securities, there must be more junior MBS tranches in the second and third order securitization process. To test this hypothesis, it is essential to distinguish cash ‡ow CDOs from hybrid/synthetic CDOs by funding structure. A cash ‡ow ABS CDO is structured to pay o¤ liabilities with the interest and principal payments of collateral. By contrast, a synthetic CDO is a credit derivative. It sells credit protection via credit default swaps rather than purchase cash assets. Hybrid CDOs use the funding structures of both cash and synthetic CDOs. Since synthetic CDOs can replicate a cash ‡ow CDO by choosing the underlying portfolio, the motivations for creating synthetic and cash ‡ow CDOs are distinct. In other words, a cash ‡ow CDO is used for capital formation or investment, but a synthetic CDO is an instrument for hedging risk or for speculating on changes in the price of a reference asset. To test whether securitization creates an information insensitive security, a key question is how cash ‡ow ABS CDOs are composed. Additional analysis of the collateral composition of synthetic ABS CDOs provides an overview of market development before the …nancial crisis. 10

In a companion paper (Park (2010a)), I show that the price of a BBB-rated tranche is sensitive to information about its collateral, deal structure and issuer, but an AAA-rated tranche is insensitive to any information. It is more di¢ cult to price a BBB-rated tranche than an AAA-rated tranche. This implies that market participants acknowledge the information sensitivity of junior tranches, and there is an incentive to produce private information; this property makes it di¢ cult to …nd buyers of junior tranches

5

The dataset used in this paper contains 516 ABS CDOs issued between 2005 and 2007, 4,023 CDO tranches, and 79,724 securities used as the collateral of ABS CDOs. Using detailed information on collateral composition and deal structure of ABS CDOs, my analysis yields three key results. First, I …nd that there is a large re-securitization rate distinction between AAA- and non-AAA-rated tranches. Less than 1% of AAA-rated subprime MBS tranches were used as collateral for cash ‡ow ABS CDOs between 2005 and 2007. By contrast, the range of re-securitization rates below AAA-rating is between 47% and 68%. Similarly, by analyzing the collateral information of CDO-squared, which is a type of CDO that invests in other CDOs, I …nd that only 5% of AAA-rated ABS CDO went into the third order securitization process through cash ‡ow CDO-squared. By contrast, the re-securitization rates of AA, A, and BBB-rated CDOs were 72%, 58%, and 37%, respectively. For example, 72% of AA-rated ABS CDOs were physically placed into the collateral of CDO-squared. This implies that CDO-squared was a major investor of ABS CDO tranches rated below AAA between 2005 and 2007. Synthetic CDOs can generate speci…c risk exposure depending on the choice of the reference portfolio. I con…rm that between 2005 and 2007, synthetic CDOs were mainly used for shorting the subprime mortgage sector. BBB-rated subprime MBS tranches referred by hybrid/synthetic CDO totaled $94.4 billion, which is equivalent to 234.9% of BBB-rated tranches issued between 2004 and 2007. This is possible because CDOs can refer the same security several times. From the reference count exercise, 990 securities were referred more than ten times. For example, the BBB-rated subprime MBS tranche issued by JP Morgan Mortgage Acquisition Corporation, JPMAC 2005-OPT1 M9, is referred 59 times. It is worthwhile to note that the estimated re-securitization rates of prime and alt-A MBS are substantially lower than subprime MBS. While 9% of subprime MBS were resecuritized when combining all rating groups, the re-securitization rates for prime and alt-A MBS were 4.24% and 2.51%, respectively. The di¤erence in re-securitization rates through synthetic CDOs is even larger. Prime and alt-A MBS, which were placed or referred on

6

hybrid/synthetic CDOs, were both less than 1%. However, 9.21% of subprime MBS were referred. This implies that if prime and alt-A MBS were re-securitized, they were mainly placed in cash ‡ow CDOs. It also suggests that subprime MBS was the most popular collateral type among all RMBS. This …nding is also related to collateral composition in the CLO market. There are no CLOs backed by investment grade loans, which are transactions of borrowers with senior unsecured debt ratings at or above BBB- by S&P. There are only CLOs backed by high yield loans, which are rated below BBB- by S&P. This fact suggests that there is more economic bene…t to repackaging illiquid and riskier assets, since there are more arbitrage opportunities in the securitization process. Second, I show that cash ‡ow CDO collateral composition was stable between 2005 and 2007. More importantly, there is no evidence of collateral deterioration. The asset class and credit rating composition were fairly constant over the sample period. The estimated weighted average rating factor (WARF) of high grade deals increased by only 3 points from 33 in 2005 to 36 in 2007. Among mezzanine deals, the estimated WARF even declined by 7 points. Note that a large WARF implies a high risk collateral pool. It is also worthwhile to note that 72.26% of cash ‡ow CDO were high grade type. It suggests that high grade ABS CDOs were demanded more than mezzanine ABS CDOs by investors in the cash ‡ow deals. In contrast to the cash ‡ow CDO market, I …nd that the synthetic CDO market experienced a dramatic change over the sample period. By all measures, the quality of reference portfolios severely deteriorated. The proportion of subprime MBS in the collateral of mezzanine deals increased by 18% from 53.39% in 2005 to 71.9% in 2007. Furthermore the proportion of BBB-rated securities increased by 20% from 55% in 2005 to 75.39% in 2007. This implies that the reference portfolios of synthetic CDO were concentrated in the riskiest part of the subprime mortgage sector. As a notional amount, $94 billion out of $172 billion, which was the total collateral amount of synthetic CDOs, was comprised of BBB-rated subprime MBS tranches. Note that the mezzanine type was dominant in hybrid/synthetic

7

CDOs. It accounted for 64% of the hybrid/synthetic CDO market, and market expansion between 2005 and 2006 was led by the growth of the mezzanine synthetic CDO. It suggests that mezzanine synthetic ABS CDOs were demanded more than high grade ABS CDOs by speculators or hedgers in the hybrid/synthetic deals. Third, average subordination level and spread increased with respect to riskiness of collateral pool. In cash ‡ow deals, the subordination level for AAA-rated tranches of high grade and mezzanine types are 5.9% and 24.6%, respectively. In synthetic ABS CDOs without super senior tranche, the subordination level for AAA-rated tranches of high grade and mezzanine are 14.2% and 25.7%, accordingly. Note that the average subordination level for the AAA-rated tranche in the subprime MBS is 20.7%. In addition, I show that mezzanine ABS CDOs paid higher interest than high grade ABS CDOs within the same rating group. In cash ‡ow deals, the mezzanine type paid 9 basis points more than the high grade type within the AAA rating. Note that synthetic deals did not pay di¤erent interest payments than cash ‡ow deals. Compared to the subprime MBS market, the AAA-rated tranche in the mezzanine cash ‡ow deal paid 24.5 basis points more than the AAA-rated subprime MBS tranche. The gap became larger for lower credit ratings. This is one of the reasons that demand for AAA and AA-rated CDO tranches grew before the …nancial crisis. The AAA-rated CDO tranche paid the highest interest among AAA-rated securities. I also …nd that the ABS CDO market price responded to the market disruption before the subprime MBS market. Issue spreads of BBB-rated CDO tranches responded around November 2006 when the ABX.HE BBB index spread started to jump. It is also an interesting exercise to compare the change in CDO issuing behaviors of key underwriters after November 2006. In conclusion, …nancial intermediaries created an information insensitive security through the securitization process, which is an essential part of the credit intermediation process. More than half of non AAA-rated subprime tranches were re-securitized. Between 2004 and 2007, by using securitization, the shadow banking system created more than $1,390 billion

8

of AAA-rated securities. It is evident that the synthetic CDO market deteriorated due to short demand in the housing sector, but there is no evidence that the cash ‡ow ABS CDO market deteriorated between 2005 and 2007. The rest of the paper is organized as follows. Section II provides a brief overview of the ABS CDO market evolution. Section III describes the data and presents summary statistics. Section IV presents the empirical results. Section VI concludes.

II. A.

ABS CDO market

History of CDO market11

I begin with some background on the CDO market. To understand the evolution of the CDO market, it is useful to study the motivation for producing ABS CDOs. I examine CDO market evolution by three categories: (i) collateral type, (ii) purpose, and (iii) funding structure. Evolution of Collateral Category: CBO, CLO, and ABS CDO.— According to the FCIC, Michael Milken’s Drexel Burnham Lambert assembled the …rst rated collateralized debt obligation in 1987 out of di¤erent companies’ junk bonds. During the 1990s, the collateral of CDOs was mainly corporate and emerging market bonds and bank loans. After the emerging market …nancial crisis of 1997, Prudential Securities found it pro…table to combine di¤erent kinds of asset-backed securities into one CDO. In the early 2000s, multisector ABS CDOs were common. But diversi…cation created di¢ culties in controlling CDO performance, since it was challenging to have expertise in all sectors. Thus, diversity in collateral decreased, with ABS CDOs focusing on mortgage assets with higher yields. ABS CDOs were limited in the early 2000s, but volume roughly doubled each year starting in 2003, until they virtually disappeared in the aftermath of the recent …nancial crisis. Table 18 reports the global CDO market issuance amount by underlying collateral from 2005 to the …rst quarter of 2007. It classi…es CDO issue amount into six categories by 11

I refer most of part here from FCIC sta¤ report: Credit Derivatives and FCIC …nal report.

9

collateral type: loans, bonds, structured …nance, mixed collateral, other swaps, and other.12 However, the major collateral types are loans, bonds, and structured …nance, which represent 26%, 2.6%, and 65% in 2005, respectively. CLOs and ABS CDOs constitute 91% of CDO market. Interestingly, there is no CLO backed by investment grade loans.13 In other words, all CLOs issued between 2005 and 2007 invest in high yield loans rated below BBB- from S&P. By contrast, the issuance amount of CBOs backed by investment grade bonds is larger than CBOs backed by high yield bonds. However, re-securitization rate for investment grade bonds and high yield bonds are 2.8% and 2.4%, respectively. The Purpose: Balance Sheet CDO and Arbirtage CDO.— By purpose, there are two types of CDOs: balance sheet CDOs and arbitrage CDOs. Balance sheet CDOs aim to remove assets or the risk of assets o¤ the balance sheet of the originator. Balance sheet CDOs can have either cash or synthetic funding structures. Balance sheet cash deals are used to reduce regulatory capital requirements by moving assets o¤ the balance sheet. The originator also transfers risk by purchasing CDSs via balance sheet synthetic CDOs. By contrast, arbitrage CDOs attempt to capture the mismatch between the yields of assets (CDO collateral) and the …nancing costs of the generally higher rated liabilities (CDO tranches). The motivation for arbitrage CDOs was not to transfer risk from balance sheets or capital relief. Instead, the motivations were (i) to generate fees for the dealer that arranged the structure and/or (ii) to create exposure to a certain portfolio for investors and/or (iii) to enable asset managers to raise assets under management and thereby earn fee income. Arbitrage CDOs became signi…cantly larger in aggregate volume than balance sheet CDOs starting in 2002, according to Asset-Backed Alert, and then became a dominant type 12

According to SIFMA classi…cation, if a CDO has 51% or more of a single collateral type, it goes into that bucket, otherwise it goes into ’Mixed Collateral’. Other Swaps are non-CDS which are collateral for a transaction. ’Other’ includes collateral such as funds, insurance receivables, cash, and assets that are not captured by the other categories shown above. 13 Investment grade loans are de…ned as loans with ratings at or above Baa3 from Moody’s or BBB- from S&P and high yield loans are de…ned as transactions of borrowers with senior unsecured debt ratings at …nancial close below Baa3 from Moody’s or BBB- from S&P. Investment grade bonds are de…ned as bonds with ratings equal to or above above Baa3 from Moody’s or BBB- from S&P. High Yield Bonds are de…ned as bonds with ratings below Baa3 from Moody’s or BBB- from S&P.

10

of CDO market. According to SIFMA, the issuance amount of arbitrage CDOs was more than …ve times that of balance sheet CDOs in 2006. Funding Structure Evolution with Credit Derivatives: CDS, Indices of CDS and Synthetic CDOs.— The CDO market evolved with the development of the credit derivative, especially the credit default swap (CDS). Corporate CDSs …rst appeared in 1994 and this innovation led to the development of synthetic CDOs in which the underlying collateral is comprised of single-name or multi-name corporate CDSs, rather than loans or bonds. Individual dealers launched the …rst indices of corporate CDSs in 2001. The ABS CDO market evolved like the CBO market: …rst CDS on ABS were invented, then synthetic ABS CDOs, and …nally indices of CDS on ABS. The CDS on ABS have existed since at least 1998. In June 2005, ISDA introduced a template for CDS on ABS in order to replicate ABS cash ‡ows more closely during the life of the contract. In June 2006, an ISDA template for CDS on CDOs was introduced, which provided a standardized framework for parties to hedge or to speculate on CDO tranches. In January 2006, ABX.HE was launched. ABX.HE is an index of CDS on asset-backed securities, collateralized by subprime home equity loans. The index comprises …ve sub-indexes created by pooling like-rated tranches with ratings of AAA, AA,A, BBB, and BBB-. ABX.HE attracted greater liquidity and soon became a benchmark for subprime MBS market. Synthetic CDOs have two advantages over cash ‡ow CDOs. First, synthetic CDOs allow investors to gain long exposure to portfolios of CDS without owning the underlying reference obligations. This is particularly useful when it is di¢ cult to purchase collateral due to supply limitations. Second, because the underlying assets are credit default swaps rather than cash securities, the synthetic CDO issuer does not need to raise cash to complete the transaction. This allows a portion of the capital structure to be unfunded and CDO can therefore be issued with high leverage. This property led to the development of the synthetic structure with a super senior tranche.14 The …rst CDO to include the unfunded super-senior tranche 14

Super-senior refers to a tranche being senior in priority of payments to another tranche that has been publicly rated AAA/Aaa. The super-senior tends to be very large, whereas the AAA class subordinated to

11

was JP Morgan’s BISTRO (Broad Index Secured Trust O¤ering) program in 1997. The volume of synthetic and hybrid ABS CDO transactions exploded between 2005 and 2006. Several factors contributed to the growth of synthetic ABS CDOs. First, the short demand for mortgage related assets was growing between 2005 and 2006. Hybrid/synthetic ABS CDOs allow hedgers and speculators to create short positions. Second, there was a limited amount of subprime related collateral, as demand for the senior tranches of cash ABS CDOs grew. The hybrid/synthetic CDOs helped to …ll the gap between demand and supply. Third, market participants became more comfortable accepting this new asset class and market practices were standardized after introducing the ISDA templates. The introduction of ABX.HE index also helped by providing information on market sentiment.

III.

A.

Data and Summary Statistics

Sample Construction and Representativeness

An important contribution of this paper is to construct a new dataset to provide an overview of the evolution of the ABS CDO market between 2005 and 2007. I combine two datasets to investigate the ABS CDO market. I use The OpenSource Model from Pershing Square Capital Management to study collateral information on ABS CDO. The OpenSource Model contains detailed information on the collateral backing ABS CDOs, the type of ABS CDO, and the deal structure. Additionally, I collect information on deal and tranche from Bloomberg Professional by using CUSIP in the OpenSource Model. The …nal dataset contains 516 ABS CDOs issued between 2005 and 2007, 4,023 tranches in ABS CDOs, and 79,724 securities used as the collateral of ABS CDOs. There are no o¢ cial data on issue amount of ABS CDO. The best available statistics are provided by the Financial Crisis Inquiry Commission (FCIC). The FCIC estimates the it (often called the “mezz AAA”) is much smaller in size.

12

issue amount based on data provided by Citigroup and Moody’s. As can be seen in Figure 1, ABS CDO issuance exploded in 2004. The issuance trend is similar to private securitization activities in the MBS market. Table 5 indicates that the dataset constructed by this paper is representative to ABS CDO market during the sample period. There is a noticeable issuing pattern among ABS CDO types. For cash ‡ow ABS CDOs, the issuance amount of high grade ABS CDO is dominant. By contrast, for synthetic ABS CDOs, the issuance amount of the mezzanine CDO is dominant. The high grade cash ‡ow CDO amounts to $217.45 billion, and the mezzanine cash ‡ow CDO amounts to $67.70 billion between 2005 and 2007 in my sample. The high grade type accounts for 76% of cash ‡ow types. However, the opposite pattern is observed in the hybrid/synthetic type. The high grade synthetic CDO amounts to $22.4 billion and the mezzanine CDO amounts $114.53 billion. The expansion of hybrid/synthetic CDO is led by growth in the mezzanine class. The mezzanine hybrid/synthetic CDO issuance amounts increased ten times between 2005 and 2006. The information in the dataset can be classi…ed in three categories: (i) underlying collateral composition on the issue date, (ii) deal structure, and (iii) tranche characteristics. First, I decompose the collateral composition by using three categories: asset class, credit rating, and vintage. The asset class indicates whether the security is subprime MBS, Alt-A MBS, prime MBS, CMBS (commercial mortgage backed securities), HELOC (home equity line of credit), second lien MBS, other ABS such as credit card ABS, or auto loan ABS, CDO, CLO, or corporate bond. The credit rating represents the security rating given by credit rating agencies at the CDO issuance time. I use four rating categories: AAA, AA, A, and BBB/below BBB. The vintage indicates when the security was issued. I construct …ve vintage categories: year 2007, 2006, 2005, 2004 and year before 2004. I calculate the percentage of collateral composition for each category. The second category in my dataset includes information about deal structure. Deal structure includes the underwriter, the issue date, the type of ABS CDO, subordination

13

level for AAA-rated tranche, the number of tranches in the deal, the number of securities in the portfolio, deal size, and collateral size. I classify ABS CDOs by two categories: WARF (high grade or mezzanine) and structure (cash ‡ow or hybrid/synthetic). There are four …nal categories: (i) high grade cash ‡ow, (ii) mezzanine cash ‡ow, (iii) high grade hybrid/synthetic, and (iv) mezzanine hybrid/synthetic ABS CDOs. I additionally calculate the WARF, the total amount of collateral face value and the amount of unrated part in the deal. I also collect information on whether the deal includes the super senior tranche, and if it does, I calculate the subordination level for the super senior tranche. Third, my dataset includes information about tranche characteristics. I collect the tranche face value, initial credit rating, current credit rating, issue spread over the index, reset index, weighted average life, senior amount, subordination level, percentage of deal, and whether the tranche is write-o¤. B.

Illustration of ABS CDO types

I illustrate four types of ABS CDOs by using examples from my data sample. Table 2 presents an example of the deal structure and collateral characteristics of the KLORS 2006-3A deal issued by Merrill Lynch on September 26, 2006. It is a high grade cash ‡ow ABS CDO. There are six rated tranches and one unrated tranche. The deal totaled $1.986 billion and the total collateral amount was $2 billion. It consisted of 284 securities and the weighted average rating factor was 50.95. The weighted average rating factor is the score calculated by using the face value of the security and the corresponding credit score used by credit rating agencies.15 The subordination level for AAA rated tranches is 5.22%. In this deal, 36.37% of the collateral is subprime MBSs, and the total mortgage exposure of the deal is 61.39% when combining the percentage of subprime MBS, alt-A MBS, prime 15 For example, there are two bonds in the ABS CDO. One is rated AAA by two credit rating agencies, and the par value is $100. The other is rated BBB, and the par value is $900. The AAA- and BBB-rated bonds are given scores of 1 and 360, respectively. See table. Then the WARF is (1 100+360 900)/1000=324.1 and this deal is classi…ed as mezzanine. If the AAA rated bond is $900 and the BBB rated bond is $100, then the WARF is (1 900+360 100)/1000=36.9. Therefore, this deal is classi…ed as high grade.

14

MBS, CMBS, HELOC and second lien MBS. Furthermore, CDOs comprise 2.56% of the deal. If CDOs backing KLORS 2006-3A are backed by other MBSs, then the exposure is higher than 61.39%. By examining rating group composition, the deal includes 26.57% of AAA-rated bonds, and 34.84% of AA rated-bonds. There are only 3.27% of BBB-rated bonds. The majority of bonds were issued in 2006. Tranche A class paid 20 basis points over the 1-month LIBOR, and the most junior tranche class E, which was rated BB+, paid 625 basis points over the index. C.

Reference Counts

Table 4 presents the reference counts among 79,724 securities in 516 ABS CDO portfolios. I count how many times the security is placed into the collateral of ABS CDOs. There are two ways of being placed in the ABS CDO portfolio. If the security is physically purchased by ABS CDOs, then the reference count has to be one. If the security is referred as a CDS contract, then the security would be in the ABS CDO collateral more than once. I …nd that the portfolio of ABS CDOs consist of 24,249 distinct securities. For example, 15,642 tranches among 24,249 securities are placed once into the ABS CDO portfolio. Surprisingly, 990 tranches are referred more than ten times between 2005 and 2007. The maximum reference number is 59, which was reached by the subprime MBS tranche rated BBB, issued by JP Morgan Mortgage Acquisition Corporation and originated by Option One in 2005. The second table shows securities most referred by ABS CDOs between 2005 and 2007. As can be seen, they are all the most junior subprime MBS tranches.

IV. A.

Main Analysis

Evolution of Collateral Composition by ABS CDO type

Table 6 to table 8 provide summary statistics of average collateral compositoin for the 516 deals by ABS CDO type. Asset Type.— First, I present the average asset composition by ABS CDO type. High 15

grade ABS CDOs contained less subprime MBS than mezzanine ABS CDOs, regardless of funding structure. Among all CDO types, subprime MBS and CDO were the major asset components. Average exposure to the housing sector was 73% in cash ‡ow deals, and 80% in mezzanine hybrid/synthetic CDOs. See Figure 2 for average asset compositions by ABS CDO types. Second, I examine the average asset composition by year of issue. During the sample period, the collateral composition of cash ‡ow deals shows a stable pattern. In high grade cash ‡ow deals, the percentage of subprime MBS fell by 6%, from 47.75% in 2005 to 41.36% in 2007. By contrast, the percentage of CDOs increased from 18.54% to 27.83% over the same sample period. This …nding implies that growth in the CDO market, in turn, provided more raw materials for the ABS CDO market. Third, the asset composition of hybrid/synthetic deals changed substantially. In the high grade type, the percentage of subprime MBS fell by 40% from 62.39% in 2005 to 20.50% in 2007. This fall was accompanied by an increase in the percentage of CDOs from 19.41% to 70.93%. In synthetic ABS CDOs, subprime MBS and CDO were consisted of 85% of collateral. On average, 70.70% of underlying collateral of mezzanine hybrid/synthetic ABS CDO consisted of subprime MBSs, and 48.39% of collateral of high grade hybrid/synthetic CDO comprised CDOs. Credit Rating.— Table 7 provides the collateral composition by rating group. Cash ‡ow ABS CDO were composed of more highly rated collateral than synthetic ABS CDO in the same WARF categories. In other words, high grade cash ‡ow ABS CDOs consisted of 26.53% of AAA-rated collateral, but high grade synthetic ABS CDOs consisted of 16.40% of AAA-rated collateral. See Figure 3 for average credit rating compositions by ABS CDO types. (i) CF: Interestingly, the collateral composition by rating group shows fairly constant patterns over the sample period as in the case of asset type. This implies that the composition of the cash ‡ow ABS CDO market is stable. This …nding suggests that the cash ‡ow ABS

16

CDO market was for producing information insensitive securities. There is no evidence of the deterioration of collateral composition for cash ‡ow ABS CDOs. Less than 1% of BBBrated securities were in the collateral of high grade cash ‡ow ABS CDOs, but 60n% of the collateral of mezzanine ABS CDO was composed of BBB and below BBB-rated collateral. (ii) Hybrid/Synthetic: Compared with cash ‡ow ABS CDOs, the rating composition changed dramatically in the hybrid/synthetic ABS CDOs. The lower rating group composition increased between 2005 and 2007. For high grade synthetic ABS CDOs, the percentage of AAA-rated collateral fell by 36% from 41.41% in 2005 to 5.17% in 2007. This fall was accompanied by an increase in the percentage of A-rated collateral from 23.07% to 75.80%. Due to the collateral eligibility limitation, the percentage of BBB and below BBB-rated collateral is less than 5% during the sample period. For mezzanine synthetic ABS CDOs, the percentage of BBB and below BBB-rated collateral increased by 20% from 55.02% to 75.39%. These …ndings con…rm that the qualities of the collateral or the reference pool of hybrid/synthetic ABS CDOs deteriorated. It implies that the synthetic CDO was created to provide speculating opportunities. Vintage.— Table 8 shows the vintage composition by ABS CDO types. There are three main …ndings. First, due to the massive issue volume between 2005 and 2006, all ABS CDO types were comprised of collateral issued between 2005 and 2006. The percentage of collateral issued between 2005 and 2006 was 76.53n% on average. Second, since cash ‡ow ABS CDOs have to physically purchase collateral, they are likely to contain collateral issued in the same year. However, when market disruptions began at the end of 2006, private mortgage securitization activities started decreasing. As a result, the ABS CDOs issued in 2007 used the collateral issued in 2006 to reduce inventories in the warehouse. Third, the vintage composition of synthetic ABS CDOs displays a similar pattern. Synthetic CDOs contain or refer collateral issued in the same or previous year. From the vintage composition, it is not possible to conclude that the qualities of collateral deteriorated. The vintage composition is largely determined by the eligible collateral supply and short demand on the speci…c

17

collateral. B.

Change of Deal Structure

Table 9 represents the average deal characteristics by ABS CDO types. I calculate the WARF of each deal by using the credit rating of collateral and the rating score used by credit rating agencies. In high grade ABS CDOs, the WARF of the cash ‡ow ABS CDO was 36 points below that of the synthetic ABS CDO. By contrast, for mezzanine categories, the WARFs of cash ‡ow and synthetic ABS CDO were 322 on average. Note that a large WARF implies a higher risk of underlying collateral. The average WARF of high grade cash ‡ow ABS CDOs is fairly constant over the sample period. The average WARF of high grade synthetic ABS CDOs increased by 50 points from 45.46 in 2005 to 97.36 in 2007. The average WARF of mezzanine cash ‡ow ABS CDOs decreased by 7 points, implying that the rating composition did not deteriorate. By contrast, the average WARF of mezzanine synthetic ABS CDOs increased by 109 points from 235.77 in 2005 to 344.90 in 2007. Figure 10 presents the WARF histogram. It shows that the issuers of ABS CDOs focus on the target WARF according to ABS CDO types. On average, one ABS CDO consisted of 151 securities The minimum number of securities in a CDO portfolio was 44 and the maximum was 990. The deal size of high grade cash ‡ow ABS CDOs was the largest among four ABS CDO types. The average deal size of high grade cash ‡ow deals was $ 1.4 billion. On average, the deal size of the mezzanine cash ‡ow ABS CDO was the smallest at $461 million. The average deal sizes of synthetic ABS CDO were $605 million and $651 million for high grade and mezzanine, respectively. Over the sample period, the sizes of ABS CDOs increased in all types, and sizes increased more in the synthetic category. In particular, the deal size of the mezzanine synthetic ABS CDO more than doubled from $348 million to $801 million. The collateral size was larger than the deal size due to the existence of unrated tranches. For example, the …rst loss tranche, equity tranche and super senior tranches were unrated in

18

synthetic ABS CDOs. By di¤erencing collateral amounts to deal size, it is possible to back out the unrated deal size. For example, there were $70 million of unrated tranches in high grade cash ‡ow deals on average, which is equivalent to 5% of deal size. Due to the existence of the super senior tranche in synthetic ABS CDOs, the proportion of unrated tranches was much higher. For instance, the unrated proportions were 65% and 62% of the rated deal for high grade and mezzanine synthetic ABS CDOs, respectively. The average number of rated tranches was 7.79. As noted earlier, the number of rated tranches for cash ‡ow ABS CDOs was fairly constant over the sample period. The average was 8 and 7 for high grade and mezzanine, respectively. By contrast, the average number of rated tranches in synthetic ABS CDOs increased between 2005 and 2007. For high grade, the average number increased from 5.6 in 2005 to 8.0 in 2007. For mezzanine, the average number of tranches in one deal increased from 6.26 to 9.15 during the sample period. This is further evidence that the synthetic ABS CDO market experienced more changes between 2005 and 2007. C.

Change of Tranche Characteristics

Table 10 presents summary statistics of average tranche characteristics for 4,013 tranches by ABS CDO type and rating group. I classify the tranche by using the S&P credit rating, and split tranches into …ve categories: AAA, AA, A, BBB, and BB. For example, I combine AA+, AA, and AA- rated tranches into the AA group for the analysis. Figure 12 and Figure 13 present issue spread over its index by rating groups over the sample period. As can be seen, the issue spread became tighter in all credit rating groups between 2005 and 2006. As noted earlier, the issue spread of BB responded as early as November 2006, and issue spreads for all rating groups jumped as bad news arrived sequentially. Note that the variation in issue spread is small. The issue spread is more standardized than spreads in the subprime MBS market. Above BBB rating, the issue spread followed a trend similar to the subprime MBS

19

market. On average, AAA-rated high grade deals paid less than mezzanine deals, regardless of structure. Mezzanine deals need more subordination level for AAA-rated tranches, and pay a higher interest to compensate risk. By contrast, synthetic deals did not pay more than cash ‡ow deals. The average issue spread of an AAA-rated tranche in high grade cash ‡ow was 36.37 basis points over the 3-month LIBOR. BB-rated tranches paid 658.33 basis points over the 3-month LIBOR in 2005. In 2005, since the 3-month LIBOR was between 3.2% and 4.7%, investors of BB-rated tranches would receive about 10% interest. When comparing with the average issue spread of subprime MBSs, the di¤erence increased for junior tranches. For AAA-rated tranches, ABS CDOs paid 10 basis points more than subprime MBSs. The di¤erences for AA-rated, A-rated, and BBB-rated groups were 20 bps, 95 bps, and 180 bps, respectively, implying that risk compensation increased exponentially, since default risk was concentrated in the junior tranches. Super Senior Tranche Characteristics.— Among 516 ABS CDOs, there were 56 deals containing super senior tranches. The average super senior tranche size was $736 million. The average size increased from $720 million in 2005 to $793 million in 2006, and then decreased to $581 million in 2007. The subordination of the super senior tranche increased over the sample period. The average subordination level increased from 29% to 40% between 2005 and 2007. In other words, the percentage of the super senior tranche in the total collateral size decreased from 71% to 60% over the same period. to 60% over the same period.

V.

Collateral Analysis

What is the motivation for re-securitization? In the subprime MBS securitization process, AAA-rated subprime MBS tranches became the least information sensitive. Junior tranches, which protect AAA-rated tranche, are sensitive to the private information of the issuer. As noted earlier, if securitization is used to create an information insensitive security, then there must be more junior tranches in the re-securitization process through 20

ABS CDOs. This is because an AAA-rated tranche does not need to go through other credit and liquidity transformations via a second securitization process. Note that I calculate the issue amounts of subprime MBS tranches issued between 2004 and 2007 by rating groups. I calculate the total amount of subprime MBS used as CDO collateral by rating groups and CDO types. Then I back out the percentage of subprime MBS in each ABS CDO type. A.

Subprime MBSs in ABS CDOs

A key …nding of this paper is the discontinuity between the proportion of AAA- and non-AAA-rated tranches in the portfolio of ABS CDOs. There is less than 1% of AAA-rated subprime MBS tranche in the ABS CDO portfolio. By contrast, at least 50% of non-AAArated tranches were placed into the ABS CDO portfolio, and 65% of subprime BBB rated tranches were physically sold to cash ‡ow CDOs. In particular, the majority of AAA-rated subprime MBS went into high grade cash ‡ow CDOs. The most junior tranche, which was rated BBB, was mainly located in the collateral of mezzanine cash ‡ow CDOs. This …nding con…rms that ABS CDOs were the major investor of junior tranches of subprime MBSs. Through the re-securitization process, an information sensitive security was transformed into an information insensitive security. As can be seen in the average ABS CDO deal structure, di¤erent types of ABS CDOs have di¤erent subordination levels, according to their credit risk. Various credit enhancement mechanisms are inserted in the deal structure to protect AAA-rated tranches from the credit risk of underlying collateral. As table 12 demonstrates, there are monotonically increasing amounts of junior tranches in hybrid/synthetic CDOs. By contrast, there are 0.30% of AAA-rated, 8.74% of AA-rated, and 45.6% of A-rated in hybrid/synthetic CDOs. Surprisingly, there are 234.9% of BBBrated subprime MBS tranches in the ABS CDO collateral. The dominance of the BBB-rated subprime MBS tranche is due to the dominance of mezzanine hybrid/synthetic ABS CDOs. Mezzanine hybrid/synthetic ABS CDOs alone contain 234.33% of BBB rated subprime MBS tranche. An important implication of this …nding is that hybrid/synthetic CDOs mainly use

21

the BBB-rated tranche for its reference security. This was the main reason for early market disruption due to the poor performance of CDOs. It shows that the hybrid/synthetic CDO market is driven by di¤erent factors, which led the cash ‡ow CDO market. B.

CDOs in ABS CDOs

CDOs also invest previously issued CDOs. I …nd a discontinuity between the proportion of AAA- and non-AAA-rated CDOs in cash ‡ow CDOs. Speci…cally, 5.24% of AAA-rated CDOs went into the third order securitization process. Despite the percentage being slightly higher than that of the AAA-rated subprime MBS tranche, it is much less than the percentage of junior CDO tranches in the collateral of cash ‡ow CDOs. Interestingly, a majority of AA and A-rated CDOs went into CDO portfolios. For example, 72.3% of AA-rated CDOs were sold to other CDOs. The proportion sold to CDO-squareds was decreasing below BBBrating, implying that the credit risk of the most junior tranche in CDOs was higher than the credit risk of junior tranches in MBS; thus, it was less demanded by underwriters in the third order securitization market. CDOs in hybrid/synthetic ABS CDOs show a pattern similar to the subprime MBS in hybrid/synthetic ABS CDOs. There were 1.07% of AAA-rated ABS CDO tranches. The re-securitization rates increased from AAA-rated to A-rated CDO tranches. The resecuritization rates were decreasing below BBB-rating. The re-securitization rate peaked at A-rating in the hybrid/synthetic class, in contrast to the cash ‡ow ABS CDO category.

VI.

Conclusion: Liquidity Creation through Securitization

As Bernanke said, “in discussing the cause of the crisis, it is essential to distinguish between triggers and vulnerabilities in the economy.”The shadow banking system is as important as the traditional banking system in providing a substantial amount of credit to the real economy. Financial intermediaries in the shadow banking system are interconnected and interdependent through the securitization-based credit intermediation process. Securi22

tization has more than thirty years of history and has helped to lower the cost of funding, to improve the availability of credit, and to reduce the volatility of the …nancial system as a whole. It is natural to interpret securitization and its products as parts of the shadow credit intermediation process. The securitization technique is pooling assets, tranching liabilities, and delinking the credit risk of the asset from the originator by using a special purpose vehicle. When information is asymmetric, the …nancial intermediary can solve the adverse selection problem through securitization. This technique makes it possible to produce information insensitive securities, to maximize the return on information for underwriters, and maximize trading. Information insensitive securities have provided liquidity as the shadow banking system has grown during past decade. The system needed to create liquidity to support its growth. Between 2004 and 2007, by using securitization, the shadow banking system created more than $1,390 billion of AAA-rated securities. Evidently, between 2005 and 2006, credit derivatives including mezzanine synthetic CDOs played an important role in triggering the …nancial crisis, and contributed to the major write-downs and collapse of large …nancial institutions. However, the unprecedented magnitude of the recent crisis was also attributed to vulnerabilities in the private and public sectors. After investigating the subprime MBS and ABS CDO markets, it is not possible to conclude that the subprime MBS and cash ‡ow ABS CDO market deteriorated, as is the popular perception.

23

REFERENCES Adrian, Tobias , Boesky, Hayley , Ashcraft, Adam, and Zoltan Pozsar. 2010, “Shadow Banking.”Federal Reserve Bank of New York Sta¤ Reports No. 458. Benmelech, E¢ , and Jennifer Dlugosz. 2009. “The Credit Rating Crisis.”NBER Working Paper No.w15045. Benmelech, E¢ , and Jennifer Dlugosz. 2009. “The alchemy of CDO credit ratings.” Journal of Monetary Economics, 56: 617-634. Bernanke, Ben S. 2010. Causes of the Recent Financial and Economic Crisis. Testimony in front of the Financial Crisis Inquiry Commission. Carey, Mark, Correa, Ricardo, and Kotter, Jason. 2009. “Revenge of the Steamroller: ABCP as a Window on Risk Choices.” Board of Governors of Federal Reserve System, Working Paper. Coval, Joshua D., Jurek, Jakub, and Sta¤ord, Erik. 2009. “Economic Catastrophe Bonds.”Americal Economic Review, 99(3): 628-666. Coval, Joshua D., Jurek, Jakub, and Sta¤ord, Erik. 2009. “The Economics of Strucrued Finance.”Journal of Economic Perspective, 23(1): 3-25. Covitz, Daniel M., Liang, Nellie, and Suarez, Gustavo. 2009. “The Evolution of a Financial Crisis: Panic in the Asset-Backed Commercial Paper Market.”Board of Governors of Federal Reserve System, Working Paper No. 2009-36. Financial Crisis Inquiry Commission. 2010a. “Credit Derivarives and Mortgage-Related Derivatives.” Preliminary Sta¤ Report. Financial Crisis Inquiry Commission. 2010b. “Overview on Derivatives.” Preliminary Sta¤ Report. Financial Crisis Inquiry Commission. 2011. The Financial Crisis Inquiry Report. Public A¤airs Report. Goodman, Laurie S., Shumin Li, Dougals J. Lucas, Thomas A. Zimmerman, and Frank J. Fabozzi. 2008. Subprime Mortgage Credit Derivatives. NJ: Wiley. Goodman, Laurie S., Dougals J. Lucas, Thomas A. Zimmerman, and Frank J. Fabozzi. 2006. Collateralized Debt Obligations: Structures and Analysis. NJ: Wiley. 24

Dang, Tri Vi, Gary B. Gorton, and Bengt Holmstrom. 2010. “Opacity and the Optimality of Debt for Liquidity.”Working Paper. Demarzo, Peter M. “The pooling and tranching of securities: a model od informed intermediation.”Review of Financial Studies, 18(1): 1-35. Demarzo, Peter M., and Du¢ e, Darrell. 1999. “A Liquidity-based Model of Security Design.”Econometrica, 67(1): 65-99. Gorton, Gary B. 2008. “The Panic of 2007.” Proceedings of the 2008 Jackson Hole Conference, Federal Reseve Bank of Kansas City Symposium. Gorton, Gary B. 2010. Slapped by the Invisible Hand, the Panic of 2007. New York: Oxford University Press. Gorton, Gary B., and Andrew Metrick. 2009a. “Securitized Banking and the Run on the Repo.”Forthcoming in the Journal of Financial Economics. Gorton, Gary B., and Andrew Metrick. 2009b. “Haircuts.” Yale University School of Management Working Paper. Gorton, Gary B., and George Pennacchi. 1990. “Financial Intermediaries and Liquidity Creation.”Journal of Finance, 45(1): 49-72. Gorton, Gary B., and Nicholas S. Souleles. 2005. “Special Purpose Veichles and Securitization.”In The Risks of Financial Institutions, ed. Rene Stulz and Mark Carey, Chicago: University of Chicago Press. Gri¢ n, John M., and Tang, Dragon Yongjun. 2010. “Did Subjectivity Play a Role in CDO Credit Ratings?”McCombs Research Paper Series No. FIN-04-10. Longsta¤, Francis A., and Rajan, Arvind. 2008. “An Empirical Analysis of the Pricing of Collateralized Debt Obligation.”Journal of Finance, 63(2): 529-563. Mortgage Banker’s Association. 2009. The 2009 Mortgage Market Statistical Annual: Volune I and II. Bethesda: Inside Mortgage Finance Publication. Tavakoli, Janet M. Structured Finance and Collateralized Debt Obligations. 2008. NJ: Wiley. Raynes, Sylvain, and Rutledge, Ann. 2003. The Analysis of Structured Securities. New York: Oxford University Press.

25

A

ABS CDO, Shadow Credit Intermediation and the Financial Crisis

Bear Stearn’s Hedge Fund Failure and Repo Lending.— In July 2007, two Bear Stearns hedge funds, the High-Grade Structured Credit Fund and the High-Grade Structured Credit Enhanced Leveraged Fund managed by Ralph Cio¤ were failed. Ralph Cio¢ was managing 11 CDOs with $18.3 billion in assets and 2 hedge funds with 18 billion in assets by the end of 2006. He …nanced mortgage related asset purchases by borrowing in the repo markets. The ability to borrow using the AAA and AA tranches of CDOs as repo collateral facilitated demand for those securities. CDO tranches pay an interest rate over the cost of borrowing. If the market value of the collateral fell, the repo lenders could and would demand more collateral from the hedge fund to back the repo loan. In summer of 2007 the subprime mortgage market had experienced substantial increases in delinquencies from homeowners, which caused sharp decreases in the market values of subprime mortgage related bonds. Subsequently, repo lenders required two hedge fund to provide additional cash on their loans since their collateral value fell. This type of repo run contributed to the sale of Bear Stearns to JP Morgan Chase on March 2008. At this time, repo run was triggered by the uncertainty about the solvency of Bear Stearns. Citigroup CDO, Liquidity Put and ABCP market.— Financial Crisis Inquiry Commission investigators conclude a primary cause of Citigroup’s 2008 bailout was the use of “liquidity puts”on its ABCP. ABCP is composed of a bankruptcy-remote special purpose vehicle (SPV), or conduit, that issues commercial paper (CP) and uses the proceeds of such issuance primarily to obtain interests in various types of assets. Instead of issuing the triple-A tranches of the CDOs as long-term debt, Citigroup structured them as short-term asset-backed commercial paper. Asset-backed commercial paper was a cheap form of funding at the time, and it had a large base of potential investors, particularly among money market mutual funds. To mitigate the liquidity risk and to ensure that the rating agencies would give it their top ratings, Citibank provided assurances to investors, in the form of liquidity puts. In selling the liquidity put, for an ongoing fee the bank would be on the hook to step in and buy the commercial paper if there were no buyers when it matured or if the cost of funding rose by a predetermined amount. Citi would write liquidity puts on $25 billion of commercial paper issued by CDOs, more than any other company. Citigroup was forced to buy back $25 billion of CDO in 2007 and became a receiver of the biggest taxpayer bailout for a U.S. bank in 2008. In August alone 2007, outstanding U.S. ABCP plummeted almost $200 billion. At the end of 2007, the amount outstanding had dropped nearly $400 billion from the peak of 1,250 billion on August 4th of 2007 beacuse of question about the asset quality.

26

AIG, CDS on CDO, Collateral Call, and Private Information Production.— AIG’s business of o¤ering credit protection on assets of many sorts, including mortgagebacked securities and CDOs, grew from $20 billion in 2002 to $211 billion in 2005 and $533 billion in 2007. AIG Financial Products became a major over-the counter derivatives dealer, eventually having a portfolio of 2.7 trillion in notional amount. Among other derivatives activities, the unit issued credit default swaps guaranteeing debt obligations held by …nancial institutions and other investors. In exchange for a stream of premium-like payments, AIG Financial Products agreed to reimburse the investor in such a debt obligation in the event of any default.In 2004 and 2005, AIG sold protection on super-senior CDO tranches valued at $54 billion, up from just $2 billion in 2004. It had written more than $ 6 billion of liquidity puts on commercial paper issued by CDOs. AIG did not post any collateral when it wrote these contracts; but unlike monoline insurers, AIG Financial Products agreed to post collateral if the value of the underlying securities dropped, or if the rating agencies downgraded AIG’s long-term debt ratings. Beginning on July 26, 2007, Goldman Sachs requested a …rst collateral call which ask to put more collateral to back credit default swaps amount of $1.8 billion to AIG. Despite the limited market transparency in the summer of 2007, it was extremely di¢ cult to estimate the reasonable price for subprime related securities. The marks by di¤erent securities …rms varied widely, from as little as 55% of the bonds’originaly value to virtually full value. For example, Goldman valued one CDO, the Dunhill CDO, at 75% of par, whereas Merrill valued it at 95% of par; the Orient Point CDO was valued at 60% of par by Goldman but at 95% of par by Merrill. When it is di¢ cult to evaluate the value of troubled assets, the possibility of private information production can lead to the loss to the uninformed party. Goldman Sachs Presence in the Mortgage Securitization Market.— Studying the role of Goldman Sachs provides an insight to understand the shadow credit intermediation process. From 2004 through 2006, the company provided billions of dollars in loans to mortgage lenders; most went to the subprime lenders Ameriquest, Long Beach, Fremont, New Century, and Countrywide through warehouse lines of credit, often in the form of repos. During the same period, Goldman acquired $ 53 billion of loans from these and other subprime loan originators, which it securitized and sold to investors. From 2004 to 2006, Goldman issued 318 mortgage securitizations totaling $184 billion (about a quarter were subprime), and 63 CDOs totaling $32 billion; Goldman also issued 22 synthetic or hybrid CDOs with a face value of 35 billion between 2004 and June 2006. Goldman Sachs was a warehouse lender to subprime mortgage originators, a major underwriter of subprime MBSs and ABS CDOs.

27

Table 1 CDO Top Underwriters (Unit: $ million)

1 Citigroup 2 Merrill Lynch 3 UBS 4 Goldman Sachs 5 Barclays Capital 6 Calyon 7 Bank of America 8 Credit Suisse 9 Lehman Brothers 10 Deutsche Bank Total

2007 Volume 28,300 28,100 20,700 11,400 7,900 7,400 6,900 5,470 5,000 4,990

N

2006 Volume

16 27 21 11 4 4 5 7 5 6

26,300 40,900 22,940 25,400 9,200 12,500 7,340 11,900 7,300 12,700

N

2005 Volume

N

26 38 24 19 8 9 11 15 10 15

11,100 14,000 9,270 12,600 5,420 2,950 3,270 3,340 3,590 4,190

13 19 11 11 5 3 5 7 5 10

168,000 153 247,000 236 110,000 127

Note: Securities …rms approved the selection of collateral, structured the notes into tranches, and were responsible for selling them to investors. Three …rms— Merrill Lynch, Goldman Sachs, and the securities arm of Citigroup— accounted for more than 30% of CDOs structured from 2004 to 2007.

Number of Deals by CDO Type HG CF Mezz CF HG Hyb/Syn Mezz Hyb/Syn Total Citigroup Merrill Lynch UBS Goldman Sachs Barclays Caylon Bank of America Credit Suisse Lehman Brothers Deutsche Bank

14 37 7 14 8 9 10 6 4 2

15 32 16 6 1 2 6 10 8 13

5 3 2 8 3 0 1 1 0 0

28

21 12 31 13 5 5 4 12 8 16

55 84 56 41 17 16 21 29 20 31

Table 2 An Example of a High Grade Cash Flow CDO Deal: Merrill Lynch KLREROS 2006-3 Deal and Tranche Structure (Issue Date 2006/9/26) This table represents an example of a high grade cash ‡ow CDO in the sample. There are 6 tranches in the deal and the total deal amount is $ 1,994.2 million.

Tranche

Rating

A1 A2 B C D E

AAA AAA AA A BBB BB+

Amount($MM) Spread Sub(%) Current Rating 1,800 90 54 9.8 25.8 6

Prefrence share Unrated

25 44 51 130 315 625

9.3 4.8 2.1 1.6 0.3 0.0

CC CC CC CC CC CC

E¤ective Date 7/24/2008 7/24/2008 3/18/2008 1/11/2008 1/11/2008 1/11/2008

8.6

This table below represents the deal and collateral characteristics of KLEROS 2006-3 Deal.

Deal and Average Collateral Characteristics # of Securities WARF Collateral Value($MM) Subprime AltA 284

51 AAA

2,006.12 AA

A

61.4%

BBB 2007

26.6% 34.8% 30.1% 3.3%

29

0%

2006

Prime

CDO

9.8%

32.6%

12.8% 2005

2004

83.7% 5.5% 3.7%

Table 3 An Example of a HG Synthetic CDO Deal: Goldman Sachs Abacus 2005-2 Deal and Tranche Structure (Issue Date 2006/11/30) This table represents an example of a high grade synthetic CDO in the sample. There are 8 tranches in the deal.

Tranche

Rating Amount($MM) Spread Sub

WAL Current Rating E¤ective Date

Super Senior N/A A1 AAA A2 AAA A3 AAA B AAC A D BBB E BB

1,000.000 72.500 40.000 20.000 39.375 28.125 12.500 21.875

N/A 60 70 85 115 160 350 N/A

20.00% 14.20 11.00 9.40 6.25 4.00 3.00 1.25

N/A 7.4 7.5 7.6 7.7 7.8 7.9 11.3

N/A CCCCCCCC D D D N/A

F

15.625

N/A

0.00

12.7

NR

NR

7/8/2010 1/21/2010 9/13/2010 6/16/2010 6/16/2010 2/25/2010

This table below represents the collateral information backing Abacus 2005-2 Deal.

Deal and Average Collateral Characteristics # of Securities WARF Collateral Value($MM) Subprime CDO 99

145 AAA 0%

AA

1200 A

57.37%

BBB

2007

5.65% 70.52% 17.5%

30

0

2006

CMBS

ABS

15.22% 12.64% 14.75% 2005

2004

4.6% 52.72% 38.45%

Reference Portfolio Constraints of ABACUS 2005-2 Determination of compliance with a Reference Portfolio Constraint shall be determined by rounding the current portfolio value (whether expressed as a percentage, without units or in units such as years) to one decimal place (e.g., 20.0% rather than 20.03% and 7.2 years rather than 7.17 years) and comparing the rounded value with the constraint level. Each constraint shall be determined as of the date of inclusion of each Reference Obligation in the Reference Portfolio.

Reference Obligation Notional Amount Weighted Average Life Wrapped Reference Obligations Non-U.S. Reference Obligations Non-U.S. Dollar Denominated Reference Obligations RMBS Reference Obligations CMBS Reference Obligations CDO Reference Obligations High-Grade Structured Product CDO Reference Obligations CLO and Euro CLO Reference Obligations CDO Reference Obligations (other than CLO and Euro CLO) CMBS Large Loan Reference Obligations Reference Obligations with an Eligibility Rating of “BBB+” or below

Maximum Maximum Maximum Maximum Maximum

$12,500,000 8.0 years 10.0% 20.0% 10.0%

Maximum Maximum Maximum Maximum

60.0% in the aggregate 25.0% in the aggregate 25.0% in the aggregate 3.0% in the aggregate

Maximum 10.0% in the aggregate Maximum 20.0% in the aggregate Maximum 5.0% Maximum 20.0% in the aggregate

Note: WAL is declining by approximately 0.5 years each year starting in June 2006 to June 2009. All Wrapped Reference Obligations must have an explicit rating of “AAA”by S&P or “Aaa” by Moody’s. All non-US dollar denominated Reference Obligations must be denominated in either Euros or Sterling. all CMBS Large Loan Reference Obligations must have an explicit rating of at least “A-”by S&P or “A3”by Moody’s. There is also a single servicer concentration consantraint.

(Source: Abacus 2005-2 Term Sheet)

31

Table 4 Reference Counts of Tranches in ABS CDO Collateral Reference Count Number of Tranches Reference Count Number of Tranches 1 2 3 4 5 6 7 8 9 10

15,642 6,273 3,281 1,840 1,070 603 381 249 192 145

11 12 13 14 15 16 17 18 19 20

Reference Count Issuer of Tranche 47 47 47 49 50 53 54 54 59

108 95 91 66 55 54 52 41 32 25 Tranche Name

Morgan Stanley ABS MSABS 2005-HE5 Merrill Lynch Home Equity Loan Trust MLMI 2006-HE1 B2A Home Equity Asset Trust HEAT 2005-8 M8 Soundview Home Loan Trust SVHE 2006-OPT5 M8 New Century Mortgage Securities LLC NCHE 2005-4 M8 ACE Securities Corp Home Equity Loan Trust ACE 2006-NC1 M9 ACE Securities Corp Home Equity Loan Trust ACE 2005-HE7 M8 ACE Securities Corp Home Equity Loan Trust ACE 2006-NC1 M8 JP Morgan Mortgage Acquisition Corp JPMAC 2005-OPT1 M9 Table 4 presents the reference counts among 79,724 securities in 516 ABS CDO portfolios.

I count how many times the security is placed into the collateral of ABS CDOs. There are two ways of being placed in the ABS CDO portfolio. If the security is physically purchased by ABS CDOs, then the reference count has to be one. If the security is referred as a CDS contract, then the security would be in the ABS CDO collateral more than once. I …nd that the portfolio of ABS CDOs consist of 24,249 distinct securities. For example, 15,642 tranches among 24,249 securities are placed once into the ABS CDO portfolio. Surprisingly, 990 tranches are referred more than ten times between 2005 and 2007. The maximum reference number is 59, which was reached by the subprime MBS tranche rated BBB, issued by JP Morgan Mortgage Acquisition Corporation and originated by Option One in 2005. The second table shows securities most referred by ABS CDOs between 2005 and 2007. As can be seen, they are all the most junior subprime MBS tranches.

32

Table 5 ABS CDO Issue Amount and FCIC Estimated Issue Amount (Unit: $ Billion) 2004

2005

2006 Total

FCIC estimates CF Hybrid/Synthetic

71 35

100 117

63 99

234 251

135.5 65.6

80 64.1

285.7 137.4

85.5 82.9

295.6 229.1

Deal Size CF Hybrid/Synthetic

70.1 7.33

Collateral Size CF 79.7 130.7 Hybrid/Synthetic 30.24 116.3

Number of Deals by CDO Types Number of Deals 2005 2006 2007 Total HG CF Mezz CF HG Synthetic Mezz Synthetic

43 60 5 19

70 59 16 91

43 28 16 66

156 147 37 176

Total

127

236

153

516

Percentage of HG/Mezz Type in CF/Synthetic Deals % of Market HG CF Mezz CF HG Synthetic Mezz Synthetic

2004 2005 2006 Total 68.1 31.9 18.7 81.3

33

79.6 20.4 18.8 81.2

83.0 17.0 21.1 78.9

77.5 22.5 19.7 80.3

Table 6 Collateral Composition Summary Statistics by Asset Class (Unit: %) type

Subprime Alt A Prime CMBS CDO Other ABS Other

HG CF Mezz CF HG Synthetic Mezz Synthetic

44.26 52.46 36.21 70.70

10.18 5.21 2.09 1.54

13.45 10.09 3.85 2.75

2.65 3.10 2.59 4.56

21.57 14.59 48.39 9.17

3.10 5.59 2.96 3.00

4.76 8.82 3.91 8.26

Total

55.04

5.24

8.15

3.43

17.28

3.76

7.04

Subprime Alt A Prime CMBS CDO Other ABS Other HG CF 2005 2006 2007

47.75 43.90 41.36

7.94 11.97 9.51

13.38 15.16 10.71

2.00 1.92 4.49

18.54 19.59 27.83

4.98 2.44 2.31

5.41 4.98 3.76

Total

44.26

10.18

13.45

2.65

21.57

3.10

4.76

Mezz CF 2005 2006 2007

51.40 53.25 53.13

4.92 5.64 4.95

12.14 7.37 11.34

4.55 1.85 2.62

12.10 14.21 20.69

8.18 4.69 1.91

6.71 12.66 5.36

Total

52.46

5.21

10.09

3.10

14.59

5.59

8.82

19.41 34.91 70.93 48.39

4.79 4.11 1.24 2.96

5.86 5.07 2.14 3.91

HG Synthetic 2005 2006 2007 Total

62.39 43.74 20.50 36.21

0.52 3.45 1.22 2.09

2.65 5.19 2.88 3.85

4.39 3.53 1.09 2.59

Mezz Synthetic 2005 2006 2007

53.39 73.43 71.92

1.19 1.33 1.94

3.91 2.26 3.10

4.18 4.44 4.84

9.87 8.16 10.36

8.96 2.27 2.29

18.49 8.09 5.54

Total

70.70

1.54

2.75

4.56

9.17

3.00

8.26

34

Table 7 Collateral Composition Summary Statistics by Credit Rating (Unit: %) type

AAA

AA

A

BBB

HG CF 26.53 42.59 22.26 0.84 Mezz CF 2.23 4.60 16.87 61.99 HG Synthetic 16.40 24.57 53.76 0.94 Mezz Synthetic 2.16 1.85 13.74 71.65 Total

10.59 16.60 20.08 42.38

AAA

A

BBB

26.46 41.43 26.12 42.28 27.27 44.25

21.12 22.93 22.27

0.38 1.24 0.63

Total 26.53 42.59

22.25

0.84

2005 2006 2007

AA HG CF

Mezz CF 2005 2006 2007

2.16 1.53 3.83

5.11 3.30 6.19

17.69 15.55 17.82

61.37 63.36 60.46

Total

2.23

4.60

16.87

61.99

HG Synthetic 2005 41.41 25.92 2006 19.80 32.58 2007 5.17 16.12 Total 16.39 24.56

23.07 41.30 75.80 53.76

4.16 0.87 0 0.94

Mezz Synthetic 2005 2006 2007

1.17 2.35 2.18

5.61 1.35 1.45

Total

2.16

1.84

35

11.76 14.15 13.73

55.02 72.40 75.39

13.73883 71.65

Table 8 Collateral Composition Summary Statistics by Vintage (Unit: %) type

2004

2005

2006

2007

Other

HG CF 6.06 35.12 42.37 12.94 Mezz CF 14.64 41.68 30.27 5.93 HG Synthetic 5.37 30.83 52.13 8.53 Mezz Synthetic 7.69 35.57 42.56 6.14 Total

9.00

2004

36.82 39.71

2005 2006 HG CF

2007

8.31

Other

2005 2006 2007

17.05 60.76 0 0 2.45 35.01 56.75 0 0.97 9.67 52.44 34.69

8.07 1.44 2.22

Total

6.07

3.48

35.12 42.38 12.95 Mezz CF

2005 2006 2007

27.28 49.42 0 0 6.67 45.18 38.44 0 4.06 17.89 58.37 17.24

Total 14.64 41.69 30.28

9.65 7.62 2.45

5.93

7.46

2005 2006 2007

15.21 52.17 0 0 6.34 38.22 46.90 0 1.33 16.77 66.11 14.00

2.76 4.57 1.79

Total

5.37

8.54

3.12

2005 2006 2007

31.46 35.08 0 0 7.63 44.71 39.50 0 0.94 23.11 57.04 13.24

25.28 6.13 5.67

Total

7.70

8.02

HG Synthetic

30.83 52.14 Mezz Synthetic

35.57 42.57

36

6.14

3.48 7.46 3.12 8.02 6.13

Table 9 Deal Summary Statistics WARF Subordination # of Bond Collateral Amount Deal Size % ($MM) ($MM) HG CF Mezz CF HG Synthetic Mezz Synthetic

37.11 322.54 73.63 322.45

5.92 24.64 14.21 25.71

193.91 137.72 115.27 132.35

1,470 453 1,220 1,050

1,400 461 605 651

Total

218.16

18.62

151.26

1,020

819

WARF Subordination # of Bond Collateral Amount Deal Size (%) ($MM) ($MM) HG CF 2005 2006 2007

33.96 39.15 36.93

6.01 5.84 5.96

181.98 200.63 194.91

1,260 1,480 1,650

1,010 1,540 1,540

Total

37.11

5.92

193.91

1,470

1,400

Mezz CF 2005 2006 2007

324.39 323.22 317.12

24.51 23.48 27.18

150.60 133.56 118.86

424 452 518

442 465 491

Total

322.54

24.64

137.72

453

461

HG Synthetic 2005 2006 2007

45.46 58.71 97.36

. 14.10 14.25

113.40 123.75 107.38

1,130 1,370 1,090

144 656 699

Total

73.63

14.21

115.27

1,220

605

Mezz Synthetic 2005 2006 2007

235.77 324.26 344.90

29.54 23.46 24.22

129.05 138.46 124.86

1,290 1,040 992

348 605 801

Total

322.45

25.71

132.35

1,050

651

37

Table 10 Tranche Summary Statistics: Issue Spread over 1 or 3 Month LIBOR (Unit: basis points) HG CF Mezz CF HG Synthetic Mezz Synthetic Subprime MBS AAA AA A BBB BB

36.37 72.01 188.11 385.07 697.37

46.90 79.03 202.14 432.44 653.89 Basis

41.56 77.75 199.38 393.06 592.92 2005

45.01 75.39 194.81 426.89 659.89

2006 2007 HG CF

All

AAA 35.94 31.84 44.88 36.37 AA 60.10 54.51 120.77 72.01 A 148.06 139.78 317.63 188.11 BBB 288.24 323.97 597.11 385.07 BB 658.33 618.18 895.00 697.37 Mezz CF AAA 39.71 38.44 60.91 46.90 AA 66.96 61.34 117.27 79.03 A 150.90 158.14 297.56 202.14 BBB 293.92 353.13 591.71 432.44 BB 550.00 585.39 838.89 653.89 HG Synthetic AAA 46.38 34.19 AA 67.55 55.58 A 150.59 143.30 BBB 296.00 75.40 BB 700.00 313.36

46.50 41.56 103.59 77.75 275.94 199.38 527.61 393.06 650.00 592.92

Mezz Synthetic AAA 41.02 40.66 54.02 45.01 AA 68.03 59.28 105.03 75.39 A 150.17 149.36 288.68 194.81 BBB 308.75 344.84 572.26 426.89 BB 637.50 638.46 713.08 659.89 38

23.29 48.19 86.64 185.45 N/A

Table 11 Re-Securitization rate of Subprime MBS through CDO by Rating (Unit: %) CF AAA 0.90 AA 47.03 A 68.38 BBB 65.80

Synthetic

All

0.30 8.74 45.60 234.90

1.20 55.78 113.98 300.70

HG CF Mezz CF HG Synthetic Mezz Synthetic AAA AA A BBB

0.87 45.45 54.96 2.61

0.03 1.58 13.43 63.18

0.25 7.00 14.47 0.57

0.04 1.74 31.13 234.33

Subprime MBS Re-securitized Amount (Unit: $ million) HG CF Mezz CF HG Synthetic Mezz Synthetic Subprime MBS Issue Amount AAA AA A BBB

9,490 55,000 31,600 1,050

321 1,910 7,720 25,400

2,760 8,470 8,320 230

462 2,110 17,900 94,200

39

1,090,000 121,000 57,500 40,200

Table 12 Re-Securitization rate of CDO through CDO-squared by Rating (Unit: %) AAA AA A BBB BB B

CF

Synthetic

All

5.24 72.30 58.05 37.33 21.37 5.69

1.07 14.07 93.89 57.90 10.66 0.00

6.30 86.37 151.95 95.23 32.04 5.69

HG CF Mezz CF HG Synthetic Mezz Synthetic AAA AA A BBB BB B

5.17 70.55 47.72 1.04 0.00 0.00

0.07 1.76 10.34 36.30 21.37 5.69

0.84 11.67 74.50 0.49 0.00 0.00

0.23 2.40 19.40 57.41 10.66 0.00

ABS CDO Re-securitized Amount (Unit: $ million) HG CF Mezz CF HG Synthetic Mezz Synthetic CDO issue amount AAA AA A BBB BB B

15,500 19,400 7,110 140 0 0

206 483 1,540 4,900 451 12

2,510 3,210 11,100 66.6 0 0

695 659 2,890 7,750 225 0

40

300,000 27,500 14,900 13,500 2,110 211

2,000 1,500 ($ billion) 1,000 500 0

Prime

altA

Mortgage Origination Amount in CDO

Subprime MBS Issue Amount

Table 13 Origination, Total Amount in MBS and CDO by Mortgage Type (Unit: $ billion) Origination MBS Issue Amount Amount Placed in CDO Prime Alt A Subprime

1,912 1,245 1,956

911 1,106 1,478

45 32 270

Securitization Rate and Re-securitization Rate between 2004 and 2007 (Unit: %) Securitization Rate Resecuritization Rate Prime Subprime

47.66 88.85 75.55

4.97 2.85 18.27

Resecuritization in CF Resecuritization in Syn Prime Alt A Subprime

4.24 2.51 9.06

0.73 0.35 9.21 41

Table 14 Asset & Rating Decomposition of CDO Collateral (Unit: $ million) HG CF

AAA

AA

A

BBB

Subprime 9,490 55,000 31,600 1,050 CDO 15,500 19,400 7,110 140 alt A 15,400 5,220 2,210 12 Prime 17,100 8,200 3,190 61 CMBS 2,740 1,880 947 52 HELOC 36 148 78 0 CLO 639 1,050 524 14 Other ABS 1,840 683 539 7 Corporate 5 0 0 0 Second Lien 257 2,870 1,470 36 Unknown 286 86 15 0 Total

63,293 94,537 47,683 1,371

unknown

Total

541 4,220 1,270 3,170 1,090 45 69 4,490 438 621 994

97,681 46,370 24,112 31,721 6,709 307 2,296 7,559 443 5,254 1,381

16,948

223,832

Mezz CF

AAA

AA

A

BBB

Unknown

Total

Subprime CDO alt A Prime CMBS HELOC CLO Other ABS Corporate Second Lien Unknown

321 206 270 382 242 0 7 107 0 83 0

1,910 483 404 445 188 4 12 81 0 138 0

7,720 1,540 657 1,150 321 16 247 291 0 244 0

25,400 5,360 1,605 3,508 1,229 102 1,761 1,013 0 960 2

895 809 102 468 160 25 391 120 0 122 0

36,246 8,398 3,038 5,953 2,140 147 2,418 1,611 0 1,547 2

3,092

61,501

Total

1,618 3,665 12,186 40,940

42

Table-continued Asset & Rating Decomposition of CDO Collateral (Unit: $ Million) HG Synthetic

AAA

AA

A

Subprime CDO alt A Prime CMBS HELOC CLO Other ABS Corporate Second Lien Unknown

2,760 2,510 606 881 637 0 447 259 0 0 0

8,470 3,210 354 823 14 0 472 30 0 132 0

8,320 11,100 33 238 665 18 346 269 0 1 0

230 67 0 0 9 0 48 38 0 0 0

43 346 26 90 159 0 29 815 21 0 486

19,823 17,233 1,018 2,032 1,484 18 1,342 1,411 21 133 486

8,100 13,505 20,990

391

2,014

44,999

BBB

Unknown

Total

94,200 7,970 1,394 2,496 4,241 95 918 1,818 0 677 12

1,220 477 175 249 483 35 66 1,940 1,070 91 15,400

115,892 12,691 2,810 4,607 8,864 141 1,145 4,358 1,070 903 20,053

10,642 3,907 22,958 113,821

21,205

172,534

Total Mezz Syn

AAA

Subprime CDO alt A Prime CMBS HELOC CLO Other ABS Corporate Second Lien Unknown

462 695 296 531 3,650 0 37 320 0 21 4,630

Total

AA

A

2,110 17,900 659 2,890 343 602 516 815 189 301 3 8 5 119 61 219 0 0 22 92 0 12

43

BBB Unknown

Total

Table 16 Citigroup ABS CDO Issuing Behavior after September 2006

Deal Name

Issue Date

CDO type WARF Size($MM) Subprime % BBB %

ESPF 2006-1A JACKS 2006-4A CETUS 2006-2A MUGLO 2006-1A CETUS 2006-4A LCERT 2006-1A SINGA 2006-1A FAB 2006-1A OCTAN 2006-3A HSPI 2006-1A RAFF 2006-2A TOPG 2006-2A TSHP 2006-1A COOKS 2007-8A CLSVF 2007-3A COOKS 2007-6A OCTON 2007-1A PALMR 2007-1A ADMSQ 2007-2A PLETT 2007-1A EIGHT 2007-1A LSECA 2007-1A ARMTG 2007-1A STAK 2007-1A HSPI 2007-2A RDGW 2007-2A CVSC 2007-4A PPEAK 2007-1A BONF 2007-1A

9/7/2006 9/21/2006 9/27/2006 11/14/2006 11/15/2006 11/29/2006 11/29/2006 11/30/2006 12/6/2006 12/12/2006 12/13/2006 12/14/2006 12/14/2006 2/15/2007 2/28/2007 3/5/2007 3/6/2007 3/7/2007 3/8/2007 3/8/2007 3/15/2007 3/28/2007 3/29/2007 5/3/2007 6/14/2007 6/27/2007 6/29/2007 7/3/2007 7/27/2007

HG Syn Mezz Syn Mezz Syn Mezz Syn Mezz Syn Mezz Syn HG CF Mezz CF Mezz Syn Mezz Syn HG CF Mezz Syn Mezz Syn Mezz Syn HG Syn HG Syn Mezz Syn Mezz Syn Mezz Syn Mezz Syn HG Syn Mezz Syn HG CF Mezz Syn Mezz Syn HG CF Mezz CF HG Syn HG CF

46.4 305.6 364.2 333.9 362.1 360.7 46.7 78.9 316.8 390.6 53.3 384.2 255.5 49.3 118.0 71.3 388.5 332.7 453.5 345.8 115.5 363.1 37.1 383.8 367.8 51.1 466.6 20.6 30.4

Note: Table represents the issuing behavior of Citigroup.

44

780 35 300 170 450 2,000 1,000 410 300 240 1,000 1,000 1,500 0 1,000 21 1,000 300 1,000 200 1,000 500 3,000 1,500 810 3,000 410 1,500 2,200

21.6 25.0 27.5 19.9 29.9 20.8 22.3 0.0 26.5 0.3 22.8 35.0 31.0 4.2 0.0 32.9 49.9 29.0 45.8 34.0 0.0 21.5 18.2 40.2 0.0 17.7 0.0 12.0 27.2

2.5 72.5 81.8 71.7 81.9 79.6 0.0 12.5 69.5 90.5 0.0 82.9 56.6 8.8 0.0 0.0 81.4 73.7 93.0 75.8 0.0 87.7 0.0 78.3 70.5 0.0 73.9 0.0 0.4

Table 17 Goldman Sachs ABS CDO Issuing Behavior after September 2006

Deal Name ABAC 2006-11A ALTS 2006-3A LOCH 2006-1A ABAC 2006-15A ABAC 2006-14A ABAC 2006-14A HUDHG 2006-1A ABAC 2006-HG1A DVSQ 2006-7A HUDMZ 2006-1A FORTS 2006-2A WAYF 2006-2A GSCSF 2006-3GA ABAC 2006-9A HUDMZ 2006-2A FORTD 2007-1A CAMBR 7A ANDY 2007-1A TWOLF 2007-1A ABAC 2007-18A ALTS 2007-4A WICKP 2007-1A DVSQ 2005-5A

Issue Date CDO Type WARF Size($MM) Subprime % BBB% 9/20/2006 Mezz Syn 240.1 730 65.8 58.5 9/28/2006 HG CF 29.3 1,900 17.2 0 10/4/2006 HG Syn 106.5 1,200 41.6 0 10/11/2006 HG Syn 126.0 2,200 62.2 0 10/19/2006 Mezz Syn 261.0 830 67.7 63.8 10/19/2006 Mezz Syn 261.0 830 67.7 63.8 11/1/2006 HG CF 42.1 1,500 59.4 0 11/30/2006 HG Syn 126.6 890 0 0 12/5/2006 HG CF 55.8 2,000 58.1 0 12/5/2006 Mezz Syn 333.5 2,000 94.6 83.7 12/7/2006 Mezz CF 438.6 510 60.8 70.2 12/21/2006 Mezz Syn 0.0 1,000 0 0 1/18/2007 HG CF 97.5 1,600 47.0 0 1/29/2007 HG Syn 140.6 1,200 59.7 0 2/8/2007 Mezz Syn 328.1 400 100 85 2/14/2007 Mezz CF 307.9 490 52.3 83.4 2/28/2007 Mezz Syn 322.2 890 94.8 74.9 3/20/2007 Mezz Syn 394.3 310 98.4 88.5 3/27/2007 HG Syn 126.4 1,000 0 0 5/15/2007 Mezz Syn 683.0 1,000 0 70.6 5/31/2007 HG CF 28.7 1,500 2.4 0 7/12/2007 Mezz Syn 0.0 990 0 0 8/10/2007 HG CF 52.4 2,000 73.7 0

Note: Table represents the issuing behavior of Goldman Sachs.

45

Table 17 CDO Asset Managers 1. ACA Management L.L.C. 2. AIG Global Investment 3. Aladdin Asset Management 4. Allianz AG 5. American Express Asset Management 6. Anthracite Capital 7. Ares Management 8. AXA Investment Managers 9. Bear Stearns Asset Management 10. BlackRock Financial Management 11. C-BASS 12. Cambridge Place Investment Management 13. Capital Trust 14. Carlyle Investment 15. Cheyne Capital Management 16. Church Tavern Advisors 17. Clinton Advisory 18. David L. Babson 19. Ellington Management Group 20. Fidelity Management 21. Fortress Investment Group

22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43.

GMAC Institutional Advisors Golden Tree Asset Management Greenwich Street Capital Highland Capital Management INVESCO Advisors JPMorgan Investment Management Katonah Capital Lennar Partners MKP Capital Management LLC New York Life Investment Management Newcastle Investment Corp NIB Capital Bank Oak Hill PIMCO Princeton Asset Management Putnam Investments Sankaty Advisors State Street Global State Street Res earch TCW Vanderbilt Capital Western Asset Management

(Source: Abacus 2005-2 Term Sheet) Note: The CDO manager is alos a important market participant. The role of the CDO manager was to select the collateral, such as mortgage backed securities, and in some cases manage the portfolio on an ongoing basis.

46

200000 ($ million) 100000 150000 50000 0

Investment Grade Bonds Structured Finance

0

500

($ Billion) 1,000

1,500

High Yield Loans High Yield Bonds Other

Investment Grade Bonds

High Yield Bonds

Bond Issuance

CBO Issuance

Table 18 CDO Market Issuance by Underlying Collateral (Unit: $ Million) Investment High Yield Investment High Yield Structured Finance Grade Loans Grade Bonds Loans Bonds

Other

2005 2006 2007 Q1

0.0 0.0 0.0

71,018.2 178,288.2 43,783.1

4,044.9 39,565.3 20,425.0

3,074.4 2,668.6 400.0

178,040.8 316,369.4 93,626.1

16,847.1 13,388.1 194.5

Total

0.0

293,089.5

64,035.2

6,143.0

588,036.3

30,429.7

Note: Re-securitization rate for investment grade bonds and high yield bonds are 2.8% and 2.4%, respectively. (Source: SIFMA (2007))

47

250 200 ($ billion) 100 150 50 0

2005

2006 HG CF HG Hyb/Syn

48

2007 Mezz CF Mezz Hyb/Syn

100 80 60 (%) 40 20 0

HG CF

Mezz CF

HG Hyb/Syn alt A housing related

prime other

0

20

40

(%)

60

80

100

subprime CDO

Mezz Hyb/Syn

HG CF AAA

HG Hyb/Syn AA

Mezz CF A

49

BBB

Mezz Hyb/Syn unknown

100 80 60 (%) 40 20 0

2005

2006 altA housing related

prime other

0

20

40

(%)

60

80

100

subprime CDO

2007

2005

2006

subprime CDO

altA housing related

50

2007 prime other

100 80 60 (%) 40 20 0

2005 AA

A

2007 BBB

unknown

0

20

40

(%)

60

80

100

AAA

2006

2005 AAA

2006 AA

A

51

2007 BBB

unknown

100 80 60 (%) 40 20 0

2005

2006 altA housing related

prime other

0

20

40

(%)

60

80

100

subprime CDO

2007

2005 AAA

2006 AA

A

52

2007 BBB

unknown

100 80 Frequency 40 60 20 0

200 400 600 estimated WARF (Weighted Average Rating Factor)

800

0

Subordination level (%) 20 40

60

0

0

200

400 estimated WARF HG CF

53

Mezz CF

600

800

250 200 (bps) 100 150 50 0

01jan2005

01jan2006

01jan2007

01jan2008

Issue Date Mezz CF

HG Hyb/Syn

Mezz Hyb/Syn

Median spline

0

500

(bps)

1000

1500

HG CF

01jan2005

01jan2006

01jan2007

01jan2008

Issue Date HG CF

Mezz CF

HG Hyb/Syn

54

Mezz Hyb/Syn

Median spline

100

76.5

85.9

73.7

40

(%)

60

80

94.2

26.3

20

23.5 14.1

0

5.8

High CF

Mezz CF

High Hyb/Syn

85.9

76.0

AAA-rated tranche

73.7

66.9

40

(%)

60

80

100

subordination for AAA

Mezz Hyb/Syn

21.5 26.3

14.1

13.9

HG Hyb/Syn

HG HYb/Syn w/ SS

11.7

0

20

10.19

subordination for AAA

Mezz Hyb/Syn Mezz Hyb/Syn w/ SS AAA-rated

55

Super Senior

56

AAA AA+ AA AA- A A- BBB+ BBB BBB- BB+ BB BB- B+ B BCCC NR AAA 0.09 0.20 0.08 0.05 0.01 0.01 0.34 0.06 0.50 0.13 0.15 0.58 0.25 0.57 0.20 66.96 29.90 AA+ . . . . . . . . . . . . . . . 81.86 18.26 AA . 0.04 . 0.30 . . 0.04 . . . . . 0.65 0.38 0.47 59.33 38.94 AA. . . . . . . . . . . . . . . 59.27 40.73 A+ . . . . . . . . . . . . . . . 70.83 29.17 A . . . . 0.20 . . . . 0.07 0.24 . . . . 62.30 37.28 A. . . . . 3.63 . . . . . . . . 1.45 53.32 41.76 BBB+ . . . . . . . . . . . . . . . 40.03 59.97 BBB . . . . . . . . . 0.81 . . . 0.09 . 62.33 36.76 BBB. . . . . . . . . . . . . . . 41.64 58.22 BB+ . . . . . . . . . . . . . . . 65.27 34.64 BB . . . . . . . . . . . . 1.27 . . 54.03 44.55 BB. . . . . . . . . . . . . . . 92.48 7.46 B+ . . . . . . . . . . . . . . . . 100.00 B . . . . . . . . . . . . . . . 100.00 .

Table 19 ABS CDO Credit Rating Transition Matrix : Current Rating as of February, 2011 (Unit: %)